This research paper needs to analyze  “Does Race Affect Access to Government Services”. 

you can use outsource but also need to use the given material to analyze the question too. 

Please use MLA format and need 15 pages includes the citation page. 

This paper should also include an Abstract, Introduction. I include the example in the file for format reference.

Does Race Affect Access to Government Services?
An Experiment Exploring Street-Level Bureaucrats
and Access to Public Housing

Katherine Levine Einstein Boston University
David M. Glick Boston University

Abstract: While experimental studies of local election officials have found evidence of racial discrimination, we know littl

e

about whether these biases manifest in bureaucracies that provide access to valuable government programs and are less tied
to politics. We address these issues in the context of affordable housing programs using a randomized field experiment. We
explore responsiveness to putative white, black, and Hispanic requests for aid in the housing application process. In contrast
to prior findings, public housing officials respond at equal rates to black and white email requests. We do, however, find
limited evidence of responsiveness discrimination toward Hispanics. Moreover, we observe substantial differences in email
tone. Hispanic housing applicants were 20 percentage points less likely to be greeted by name than were their black and
white counterparts. This disparity in tone is somewhat more muted in more diverse locations, but it does not depend on
whether a housing official is Hispanic.

Replication Materials: The data, code, and any additional materials required to replicate all analyses in this arti-
cle are available on the American Journal of Political Science Dataverse within the Harvard Dataverse Network, at:
http://dx.doi.org/10.7910/DVN/1HOVTU.

S
cholars across the social sciences have used ex-
periments to test for racial bias in a wide range
of arenas, including patient evaluations, job ap-

plications, and professors’ responses to student requests
(Bertrand and Mullainathan 2004; Doleac and Stein 2013;
Milkman, Akinola, and Chugh 2014; Pager, Western, and
Bonikowski 2009; Schulman et al. 1999). In political sci-
ence, audit-style experiments have focused on state leg-
islators’ and local election officials’ responses to con-
stituent requests about voting (Butler and Broockman
2011; McClendon 2016; White, Nathan, and Faller 2015).
It goes without saying that understanding elected and
election officials is critical and valuable, and that any
bias that affects access to voting is an important po-
litical science matter. Without dismissing the substan-

Katherine Levine Einstein is Assistant Professor, Department of Political Science, Boston University, Boston, MA 02215 (kleinst@bu.edu).
David M. Glick is Assistant Professor, Department of Political Science, Boston University, Boston, MA 02215 (dmglick@bu.edu).

Authors’ names are listed alphabetically. Einstein is the corresponding author. Thanks to David Broockman, Dan Butler, Joe Doherty,
Christian Grose, Ariel White, participants at the 2015 University of Southern California SoCLASS conference and the 2015 SPSA conference,
and the three anonymous reviewers for helpful comments. Additional thanks to Brianna Bloodgood, Rob Pressel, and Ramya Ravindrababu
for outstanding research assistance. Any errors are our own.

1One political science study, to our knowledge, does experimentally evaluate these issues (Ernst, Nguyen, and Taylor 2013). Because it was
an in-person study of interactions at local welfare offices, though, it was necessarily much smaller in scale, and thus likely less generalizable
(n=54).

tive and normative implications of biased responses to
queries about voting procedures, we argue that this line
of scholarship misses a crucial quantity of interest far
more relevant to the day-to-day lives of people who
reach out to government: access to tangible benefits and
programs.1

The supply of social services—particularly those tar-
geted to low-income individuals—is limited. Moreover,
accessing many of these programs is complex. For the

se

reasons, potential beneficiaries will need to rely on street-
level bureaucrats who are positioned to adversely affect
access if they discriminate against potential beneficiaries
in ways consistent with findings from similar contexts
(Butler and Broockman 2011; McClendon 2016; White,
Nathan, and Faller 2015). Indeed, an ample literature

American Journal of Political Science, Vol. 61, No. 1, January 2017, Pp.

100

–116

C©2016, Midwest Political Science Association DOI: 10.1111/ajps.12252

100

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 101

suggests that, in the absence of clear rules designed to
preclude discrimination, bureaucrats with discretion can
act according to their own biases (Brodkin 1997; Jones
et al. 1977; Katznelson 2005; Lieberman 1998; Lipsky
1980).

We use the case of affordable housing and our own
audit-style experiment of over 1,000 housing authorities
to test whether street-level bureaucrats discriminate when
citizens attempt to access substantive programs and ser-
vices (see, e.g., Grose 2014 on using field experiments
to study institutions, including a call for studies of bu-
reaucrats). Public housing is scarce and in high demand.
Roughly 1.2 million individuals currently reside in public
housing managed by over 3,300 public housing author-
ities (Department of Housing and Urban Development
[HUD] 2015a). Recipients represent a small fraction of
the total population in need of such programs. For exam-
ple, in October 2014, the city of Chicago opened its wait
lists for the first time in 4 years. It received 80,000 applica-
tions in one day (Bowean 2014). Street-level bureaucrats
play an important role in helping public housing seek-
ers navigate the at-times byzantine application process.
Baltimore is illustrative: Applicants were only allowed to
apply for a Section 8 voucher (which subsidizes private
market rentals) October 22–30, 2014, and they were re-
quired to apply online.

Paralleling previous field experiments in political sci-
ence and other fields, we e-mail public housing officials
with putative constituent service requests using identifi-
ably white, black, and Hispanic names. While it is cer-
tainly reasonable to ask whether an additional field ex-
periment using this design can contribute substantively
important findings, we believe our study offers several
advantages. First and foremost, public housing officials
constitute a more generalizable test of bureaucratic dis-
crimination than do the election officials who were the
subject of a recent, and impressive, audit study that con-
cluded that bureaucrats discriminate (White, Nathan,
and Faller 2015). While voter registration is important,
it is inherently political. Barring nefarious activity, it is
also an abundant resource. Public housing is neither.
Instead, housing agencies are precisely the sort of au-
tonomous bureaucracies—featuring standardized federal
procedures with many functions devolved to the local
level—highlighted in theories of street-level bureaucrats’
behavior. In addition, we believe that our outcome vari-
ables offer a useful mix of familiarity and novelty. While
we measure responsiveness in familiar ways, we also in-
clude an additional dependent variable, tone, that has
received much less attention. We argue that tone can af-
fect the ultimate distribution of benefits in subtle but
important and underappreciated ways.

Moreover, while finding more evidence of bias holds
obvious interest, results that fail to show bias—as is the
case with some of ours—are also important. Given the
well-deserved prominence of recent experimental studies
of racial discrimination, and the gravity of the subject
and the results, findings that question the generalizability
of existing cases are especially important (see Arceneaux
and Butler 2016 for more on the importance of reporting
null results from well-designed experiments). Against an
abundance of results documenting discrimination, find-
ings of its absence are vital for suggesting pathways by
which it might be curbed. Thus, any null results may
both serve as an important brake on overgeneralization
in the literature and provide valuable practical lessons for
countering discriminatory tendencies.

Our results reveal a mix of striking patterns and non-
patterns. First, we do not find evidence of responsiveness
discrimination in general. In fact, we find evidence contra
other well-designed experiments’ findings of discrimina-
tion against blacks. Second, when focusing on tone, we
do observe meaningful differences. Ostensibly

Hispanic

e-mailers were about 20 percentage points less likely to be
addressed by name at the beginning of responses than
were their black and white counterparts. Further analysis
focused on the demographics of officials and commu-
nities provides some tentative, but mixed, support for
theories of representative bureaucracy, familiarity bias,
and contact theory. We conclude by discussing possible
explanations for the relative lack of racial bias in our study,
including the racial demographics of public housing, fair
housing legislation, and bureaucratic professionalism.

Racial Bias and Bureaucracy

We consider two potential types of harm that could
emerge if bureaucrats discriminate in their responses.
The first is straightforward and well trod in previous
field experiments: the failure to provide relevant, factual
information. By not responding, and thus not offering
information about the complicated application process, a
bureaucrat likely increases the challenges associated with
applying for public housing and reduces one’s likelihood
of completing the process. The second is somewhat more
subtle and less widely used.2 A bureaucrat could respond,
but her tone could be less friendly and encouraging to
members of particular groups. Recipients of unfriendly
communications might lose confidence in their chances

2White, Nathan, and Faller (2015) feature friendliness as one of
their dependent variables, though it does not comprise a central
component of their analysis.

102 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

of obtaining the resources they are pursuing or lower their
trust in the organization. They might also make inferences
about the effort an official will exert on their behalf over
an inherently long process. All of these mechanisms could
make applicants less likely to follow up with additional in-
quiries or diminish the likelihood they obtain the desired
service for other reasons.

Both individual and institutional factors point to po-
tential discrimination in social service provision. Sadly, a
strong reason to foresee discrimination is the fact that hu-
man beings administer these programs. Previous research
in political science and a variety of related fields suggests
that we should observe fewer and less friendly responses
to black and Hispanic constituents. Racial stereotypes re-
main widespread (Bobo 2001) and a powerful influence
on American public opinion (Kinder and Kam 2009).
These biases have manifested in a wide variety of settings.
All else equal, blacks and Hispanics making queries about
voting procedures are less likely to receive responses from
state legislators and local election officials than whites
(Butler and Broockman 2011; White, Nathan, and Faller
2015). Outside of political outcomes, blacks and His-
panics receive fewer callbacks for low-wage jobs (Pager,
Western, and Bonikowski 2009) and fewer e-mails
from faculty members concerning research opportuni-
ties (Milkman, Akinola, and Chugh 2014) than whites
do. They are also told about and shown fewer available
homes and apartments (Turner et al. 2013), receive fewer
e-mails and lower bids when selling goods online (Doleac
and Stein 2013), and told higher prices when buying used
cars (Ayres and Siegelman 1995).3

Quite simply, one could predict bias by assuming
that the tendencies and psychology that affect other gov-
ernment officials, professors, realtors, and online sell-
ers also affect social service employees. Moreover, social
resources, notably public housing, are generally scarce.
Such scarcity may create especially strong incentives for
discrimination. There is also some empirical evidence
to support the intuition that discrimination manifests
among social service bureaucrats (Davis, Livermore, and
Lim 2011; Ernst, Nguyen, and Taylor 2013; Fording, Soss,
and Schram 2007; Keiser, Mueser, and Choi 2004; Schram
et al. 2009). Devolution of welfare policy has yielded
significant variation in autonomous welfare offices’ sanc-
tions of participants. This variation includes more ex-
plicit sanctions and negative treatment for racial minori-
ties (Ernst, Nguyen, and Taylor 2013; Fording, Soss, and
Schram 2007; Keiser, Mueser, and Choi 2004). On balance,
this evidence provides strong reasons to expect at least as

3The Doleac and Stein (2013) and Ayres and Siegelman (1995)
studies focus exclusively on black-white discrimination.

much discrimination in social welfare service provision as
other scholars have found in other arenas. Similar institu-
tional factors related to devolution and frontline officials’
roles apply to some of the cases in the literature and to
public housing. Taking these lines of thinking in concert
brings us to the first hypothesis:

H1: All else equal, bureaucrats will be less responsive
and less friendly to blacks and Hispanics than
they are to whites.

On the other hand, there are also reasons to expect
mechanisms unique to bureaucratic social service pro-
vision to mitigate broad discriminatory tendencies. One
prominent line of scholarship uses the racial classifica-
tion model to explain discrimination specifically in the
social welfare context (Schram et al. 2009; Soss, Fording,
and Schram 2008); this research predicts that we should
observe racially discriminatory sanctions when minor-
ity clients are perceived as less motivated or responsible.
So Hypothesis 1’s predictions may not hold up in the
absence of cues signaling a lack of client reliability. How-
ever, because the primary interest of this research is in
sanctioning and disciplinary action—rather than the re-
sponsiveness of central interest in our study and other
audit-style analyses—it is unclear how relevant these the-
oretical predictions will be to our study.

In addition, the theory of representative bureau-
cracy also implies that we may observe differences in
responsiveness to blacks and Hispanics, and that both
groups may receive better service than postulated in
Hypothesis 1. Scholars have long been concerned about
whether bureaucracies, as unelected bodies tasked with
critical aspects of policy implementation, pose a threat to
democratic representation. One line of research in pub-
lic administration argues that these fears may be over-
stated. Rather than acting as unrepresentative bodies of
unelected officials guided by individual or institutional
interests, bureaucracies can represent the interests of their
constituents. These scholars argue that bureaucrats whose
traits coincide with a population’s demographic diversity
are more apt to actively represent constituents’ interests
(Krislov 1974; Mosher 1968), particularly in the case of
otherwise disadvantaged minority groups (Bradbury and
Kellough 2011; Coleman, Brudney, and Kellough 1998;
Krislov 1974; Meier 1993; Meier, Wrinkle, and Polinard
1999; Sowa and Selden 2003; though see Ernst, Nguyen,
and Taylor 2013). A similar, but related, literature finds fa-
vorable interactions when citizens and bureaucrats share
race, ethnicity, or gender (Epp, Maynard-Moody, and
Haider-Markel 2013; Riccucci, Van Ryzin, and Lavena
2014). This line of scholarship brings us to our second
hypothesis:

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 103

H2: All else equal, bureaucrats will be more re-
sponsive and friendly to members of their own
racial/ethnic groups.

While some of the representative bureaucracy schol-
arship (Meier, Wrinkle, and Polinard 1999) contends that
demographic diversity among bureaucrats will lift the
fortunes of all of a bureaucracy’s constituents—not just
coethnic constituents—for simplicity’s sake, we focus on
the most basic version of representative bureaucracy.

A third closely linked hypothesis stems from the-
ories of intergroup contact (Allport 1954; Pettrigrew
1998) and familiarity bias (Tversky and Kahneman 1972).
Putting aside bureaucrats’ demographic characteristics,
we might anticipate an inclination to aid members of
groups with which they are most familiar. Contact may
diminish discriminatory biases, or psychological mecha-
nisms may prompt individuals to favor the familiar over
the novel. Given the power differential between casework-
ers and clients and the inherent community-based nature
of housing, we suspect that the important contact will take
place in the local community outside of a social service
agency. This leads us to our third hypothesis:

H3: All else equal, bureaucrats exposed to more black
and Hispanic constituents will be more respon-
sive and friendlier to blacks and Hispanics, re-
spectively.

The Case of Public Housing

To evaluate these predictions, we selected the case of pub-
lic housing. We opted for it over other plausible candidates
because it is a federal program that is heavily devolved to
the local level. Common federal guidelines make local
housing authorities simultaneously institutionally com-
parable and highly autonomous. We should not expect,
for example, housing authorities in the South to be radi-
cally different (institutionally, at least) from those in the
East, or for variations in state laws to affect outcomes. In-
deed, all public housing authorities are administering the
same federal programs: subsidized housing choice vouch-
ers under the Section 8 program and/or conventional
public housing. The former allow recipients to use subsi-
dies to obtain market-rate private-sector apartments. The
latter offers subsidized accommodations in government-
owned units.

There is sufficient devolution to the local level, how-
ever, that, aligning with the definition of a “street-level
bureaucrat” (Lipsky 1980), local offices and officials have
discretion over policies they did not themselves design. In

order to encourage the creation of mixed-income hous-
ing developments, the federal government dramatically
increased public housing authorities’ discretion with the
passage of the 1998 Quality Housing and Work Respon-
sibility Act. This legislation permitted public housing of-
ficials far greater authority in setting the rules for se-
lecting tenants and Section 8 voucher recipients so long
as they abided by federal fair housing guidelines (Lazio
1998; Schwartz 2010; Vale 2000). By allowing for local
autonomy and the proliferation of varying selection cri-
teria, devolution creates a more complicated informa-
tional environment. This environment increases the im-
portance of public housing authority officials in helping
people navigate the application process. Moreover, it of-
fers street-level bureaucrats greater opportunities to ex-
ercise individual authority and discretion (Fording, Soss,
and Schram 2007).

While no one bureaucracy can possibly stand in as
a representative for all, we believe that public housing
officials represent a body of bureaucrats distinct from
election officials who administer inherently political and
less scarce programs. Indeed, as nonpolitical social service
providers, public housing officials likely share important
characteristics with many counterparts administering so-
cial programs in the Department of Health and Human
Services (HHS), for example. While not necessarily in-
dicative of local staffing, HUD and HHS have virtually
identical agency prefere federal level (Clinton and Lewis
2008). Thus, public housing bureaucrats permit us to
test discrimination in a substantively different, and more
broadly representative, context than those used in prior
research.

Data and Methods

We e-mailed public housing authorities at their publicly
available e-mail addresses (or online contact forms) us-
ing an audit study design. Each housing authority re-
ceived an e-mail on one of two days in the same week
during 2014.4 We e-mailed all public housing authori-
ties that could plausibly be matched with a core city in
a metropolitan or micropolitan area (n = 1,017). This
was to ensure that the public housing authorities we se-
lected could easily be matched with census demographic
data (many public housing authorities covered regions
that would be difficult to match with census geocodes).
Our sample comprises an enormous range of places. It

4The e-mails were sent on two separate days as a consequence of
Google Mail batching limits.

104 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

includes communities with 10,000 residents alongside the
largest cities in the country.

Each housing authority was randomly assigned (via
random number generator) to receive an e-mail from one
of six different accounts with putatively white, Hispanic,
and black names (Table SI1 in the supporting infor-
mation shows balance in the average demographics for
the communities assigned to each treatment). For each
racial/ethnic group, we chose one male and one female
name to address any possible gender interactions emerg-
ing from the disproportionately female-headed house-
hold composition of public housing.5 Using male and
female names reduces the number of observations for
each treatment. Nevertheless, we felt it was important to
have both, given the prevalence of females in public hous-
ing and the preponderance of analyses that use only male
names in the audit study literature. Below, we analyze the
data with males and females separated and consolidated
by race.

The six names were as follows: Brett Smith,

Emily

Smith (white); Tyrone Johnson, Shanice Johnson (black);
and Santiago Martinez, Gabriela Martinez (Hispanic).
For the white and black e-mailers, we chose names that
were among the 20 most distinctively black and white
names in Levitt and Dubner (2010). In a similar audit
study published in a leading economics journal, Brett and
Tyrone predicted their respective races at a rate of greater
than 90% (Bertrand and Mullainathan 2004). Observa-
tional data from California similarly bolster our choices
of female names: 97% of children named Shanice from
1989 to 2000 are black, and more than 98% of their coun-
terparts named Emily are white (Fryer and Levitt 2004).

Our selection of Hispanic names relied more heavily
on the surname (Martinez) being distinctively Hispanic.
While the black and white surnames we used are rela-
tively common among both racial groups (Word et al.
2000),6 Martinez is strongly linked with Hispanic ethnic-
ity. Nearly 92% of individuals with the surname Martinez
are Hispanic (Word et al. 2000). More generally, sur-
name is reliably and widely used in both political science
and other disciplines as a strong predictor of Hispanic
ethnicity (Barreto, Segura, and Woods 2004; Henderson,
Sekhon, and Titiunik 2015; Wei et al. 2006). The U.S. Cen-
sus Bureau provides a list of Spanish surnames (including
Martinez) that correctly identifies 93.6% of all Hispan-

5Roughly three-quarters of families in virtually all forms of pub-
lic housing are female-headed households (National Low Income
Housing Coalition 2012).

6Although Johnson appears at a slightly higher than expected rate
among blacks (33.5%), neither surname is meant to signal racial
identity; instead, we used widely validated first names for this
purpose.

ics; just as importantly, fewer than 5% of those identified
are false (Barreto, Segura, and Woods 2004; Word and
Perkins 1996). Combining Martinez with two first names
drawn from an online list of the 100 most popular His-
panic names (BabyCenter en Espanol 2011), these two
names powerfully signal Hispanic ethnicity.

While we generally followed previous studies’ prac-
tices in selecting names, we made a couple of important
adjustments. We were, when possible, attentive to the
age that names implied. One name prominently used in
a recent audit study—Deshawn—came into use almost
exclusively after 1970 according to data from the Social
Security Administration. Thus, Butler and Broockman’s
(2011) comparison of responses to e-mails from Deshawn
and Jake—a name that has been in use with varying
prominence since the turn of the 20th century (Watten-
berg 2005)—may actually be estimating the causal effect
of being a young black male relative to being a white male
of unclear age. This could inflate the amount of mea-
sured bias. The age distributions for Tyrone and Brett
largely overlap, with peaks in the 1960s and 1970s. Sim-
ilarly, Emily’s and Shanice’s distributions largely over-
lap, with peaks in the 1980s.7 We were unable to match
Hispanic names’ age distributions with their black and
white counterparts, however; virtually all distinctively
Hispanic names—including the two we chose,

Santiago

and Gabriela—peak in the 1990s and 2000s in Social Se-
curity Administration data, likely due to recent trends in
Hispanic migration.

Our e-mail text was the following:

Hello,
My name is X and I’m trying to figure out how
to apply for public housing. I believe I may be
eligible.
Can you direct me to information about applying
for public housing here? I also heard there might
be a wait list for this program. How long is it?
Thanks, X

We used a fairly generic request for “public
housing”—rather than specifying a particular program—
for several reasons. First, we wanted to ensure that our
e-mail would be equally applicable and reasonable at all
of the public housing authorities in our sample. Since
there are significant variations in the scale and type of
programs offered, we avoided incorporating particular
programs in our query for help. Second, we eschewed
referencing specific programs because we did not want to

7While we ideally wanted to use “older” female names, analogous
to Tyrone and Brett, none of the distinctively black female names
rose to prominence until after the 1970s.

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 105

signal an overly sophisticated applicant. Doing so could
affect bureaucratic responsiveness in a variety of unin-
tended ways.

Finally, we note that e-mail is not the only, or per-
haps even primary, way that many reach out to housing
authorities—though the same could be said for elected
representatives, election officials, and others who have
been audited via e-mail. Nevertheless, our data suggest
that many housing authorities are used to corresponding
and connecting with potential residents online. Fifty-five
percent of the housing authorities we contacted either
provided easily accessible e-mail addresses or “contact
us” web forms. In the other 45%, we could not easily
find e-mail contact directly from the housing authority.
In these instances, we used the e-mail contact informa-
tion available via HUD’s website. Importantly, these less
web-friendly housing authorities were evenly distributed
across the three racial/ethnic groups (43% (black), 45%
(white), and 47% (Hispanic)). While these cases suggest
that e-mailing a housing authority may not always be the
best way to get information from it, even 49% of our e-
mails to the harder-to-find addresses received responses.
E-mailing even the less web-friendly housing authorities
to seek information thus does not appear to be especially
unusual. To further document the web-friendliness of
housing authorities, we took a random sample of 50 (5%
of our total sample) and visited their websites to look
for information targeted at potential applicants. Sixty-
six percent provided easily accessible information about
things such as the application process, the status of waiting
lists, eligibility criteria, or housing stock. In some other
cases, housing authorities did not have a discernible web
presence. In the statistical models below, we include a
variable indicating whether we had to use the HUD web-
site to find an e-mail address (the Hard Email variable).

Ethical Considerations

Our e-mail text more generally reflects important eth-
ical considerations that are prominent and well dis-
cussed in similar field experiments (e.g., Butler and
Broockman 2011). First, we contacted government em-
ployees in their professional capacity. Second, our ex-
perimental treatments were not designed to alter their
behavior but rather to measure it. Third, we were atten-
tive to minimizing the amount of time workers devoted to
requests for information from fictitious constituents. We
were especially attuned to this issue since one of us actu-
ally worked in public housing prior to entering academia.
The “can you direct me” portion of the e-mail was de-
signed to encourage housing officials to either send a
link to a webpage or to copy and paste standard direc-

tions. The bulk of the e-mails we received do, in fact, fea-
ture these sorts of responses. The “how long is [the wait
list]” query similarly elicited fast answers that required
no more than a few words. Moreover, we did not engage
with housing officials at all after the first e-mail, even if
our putative request for assistance was met with a follow-
up question. While some burden is necessary in order
to gain insight into how bureaucrats allocate their finite
time (Hall 1996), we believe that our minimal interven-
tion did not substantially distract housing officials from
serving their constituents. Finally, we did use deception,
consistent with all other prominent audit studies cited
in this article. This deception is necessary to experimen-
tally test whether bureaucrats exhibit racial biases in their
responses to constituents. Without the random assign-
ment of race and gender, we would be unable to measure
this important quantity of interest. We, of course, took
anonymity very seriously, and all of our analyses reflect
comparisons across groups of housing officials.

Key Variables

We focus on two ways bias could manifest: (1) re-
sponsiveness and (2) friendliness. The former is di-
rectly analogous to the main dependent variable in ex-
isting racial bias research. We calculate responsiveness
rates and assess the timeliness and completeness of the
responses.

Friendliness is a less widely used variable, and quan-
tifying it is somewhat more challenging. We use what we
believe to be the most easily comparable (and least sub-
jective) measure across e-mails: whether the e-mailer is
addressed by proper name. We were lenient in coding
“yes.” A named salutation could be as casual as “Hi Brett”
or as formal as “Dear Ms. Martinez.”8 An ample literature
in psychology and public opinion suggests that named
salutations are surprisingly important to recipients. In
particular, they dramatically boost survey responses
(Heerwegh 2005; Joinson and Reips 2007). Moreover, this
effect (in some studies) is stronger when the sender is seen
as powerful (Joinson and Reips 2007), as may be the case
for housing officials e-mailing public housing applicants.
There are many potential mechanisms at play here. Per-
haps most prominently, a wide array of psychological
research reveals that individuals have a powerful affinity

8This measure is somewhat analogous to that featured in White,
Nathan, and Faller (2015). They code friendliness using both salu-
tation by proper name and the use of “explicitly friendly language”
like “Let us know if you have any more questions.” We viewed
the “explicitly friendly language” as somewhat more challenging to
accurately code, and thus limited ourselves to the more clear-cut
proper name salutation.

106 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

for their own name. This manifests in ways such as more
carefully examining resumes and advertised brands with
names similar to one’s own (Howard and Kerin 2011).
We anticipate that named salutations, then, may lead an
applicant to feel more warmly and better served. In ad-
dition, a named salutation might also signal the effort a
caseworker will put forward in helping a client through
the housing application process. Because we do not fol-
low up on our initial e-mail correspondence, though, we
cannot measure the ensuing interactions.

To evaluate Hypothesis 2, we need information about
workers’ demographics. Unfortunately, public housing
authorities do not publish data on the racial and ethnic
breakdowns of their staffs. Thus, we attempt to at least
roughly assess Hypothesis 2 by incorporating, when pos-
sible, the ethnicity of the housing authority official who
responds to the e-mail. We do so by coding Hispanic eth-
nicity based on the responding official’s e-mail address
name.

Note that we are only measuring whether the re-
sponding housing official is Hispanic or not. In an ideal
world, we would use the responder’s name to classify him
or her as black or white in order to better correspond with
our treatment groups. As above, while ethnically distinc-
tive surnames are reliable and widely used predictors of
Hispanic ethnicity, there are no census lists of black and
white surnames that provide anywhere near the same ac-
curacy that the Hispanic surname list does (Henderson,
Sekhon, and Titiunik 2015).

We are also only capturing the ethnicity of those indi-
viduals who actually respond to our e-mails. Information
about those who received e-mail requests but did not reply
is unfortunately unavailable due to the nature of housing
authority e-mail addresses, which rarely feature an indi-
vidual’s name. Instead, they are typically a generic address
like “citynamepublichousing@cityname.org.” Responses
then typically come from an individual’s named account.
Thus, we can evaluate Hypothesis 2 only for those obser-
vations for which we received e-mail responses. Since the
treatments were randomly assigned, in expectation, our
e-mails should have reached Hispanic officials at equal
rates in each condition, allowing us to compare those that
were returned. If a disproportionate number of Hispanic
officials received the e-mails in the Hispanic conditions
due to random chance, it would undermine these infer-
ences. While we cannot know who the e-mails reached,
we do know that the average Hispanic populations in the
communities were similar across the three racial treat-
ment groups (14.0% in the white group, 14.5% in the
black group, and 13.8% in the Hispanic group). Lastly, we
note that this issue applies to only part of our Hypothesis 2
analysis—the portion exploring bias in responsiveness—

and not the rest of the results for Hypothesis 2 or our
other hypotheses.

To evaluate Hypothesis 3—which postulated that bu-
reaucrats in communities with more black and Hispanic
residents would be more responsive to black and Hispanic
constituents, respectively—we include community racial
demographics. We obtain the proportion white, black,9

and Hispanic at the “place” level from the American Com-
munity Survey’s 2012 5-year estimates. In addition, we in-
corporate other relevant demographic controls from the
census into our statistical models: city poverty rate and
population. The poverty rate is likely related to the de-
mand for housing and the overall number of inquiries to
which a bureaucrat must respond. Similarly, larger cities
may have more professional housing authorities or more
resources along with different public housing applicant
pools, for example.

Results

We begin by evaluating Hypothesis 1 and testing whether
the race of the inquirer affects whether housing officials
are more likely to respond and the speed and compre-
hensiveness of their responses. In Figure 1, we report the
raw and unmodeled data—the proportion of inquiries
that received e-mail responses (with 95% confidence in-
tervals). On the left, we separate out all six conditions
(e.g., white male [Brett]). On the right, we consolidate by
race. Overall, we do not find strong evidence of differ-
ent responsiveness by race or gender (� 2(5) = 5.92, pr =
.31 for the six treatment category tabulation). The range
across the six conditions is a mere 10 percentage points,
from a low of 53% to a high of 63%.

Given the theoretical emphasis on discrimination
against minorities, we primarily care about comparisons
between whites and blacks and whites and Hispanics.
The former comparison offers findings contrary to those
in other studies. Our findings do not merely fail to find
significant evidence of discrimination due to insufficient
statistical power. To the extent we do get variation, it is
in a direction counter to other findings of discrimina-
tion against blacks. Blacks actually received the highest
response rates: 60.7% versus 57.5% for whites (p = .41).
Even comparing the extreme ends of the white and black
95% confidence intervals yields a maximum plausible
discrimination of only 8 percentage points.

The evidence concerning Hispanics is more mixed.
It does not offer especially strong support for or
against the discrimination hypothesis. The two Hispanic

9We calculate the proportion non-Hispanic white and black.

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 107

FIGURE 1 Proportion of Inquiries That Received a Response

By race/ethnicity crossed with sex

Tyrone

Shanice

Santiago

Gabriela

Brett

Emily

0
.2

.4
.6

.8
1

P
ro

po
rt

io
n

G
et

tin
g

a
R

es
po

ns
e

Treatment Group

By race/ethnicity (collapsed by sex)

Black

Hispanic

White

0
.2
.4
.6
.8
1
P
ro

p
o

rt
io

n
G

e
tt

in
g

a
R

e
sp

o
n

se
Treatment Group

treatment conditions did receive the two lowest response
rates. The difference between whites and Hispanics is 4.8
percentage points (p = .11). On the one hand, attribut-
ing systematic discrimination on the basis of this fairly
small and borderline significant (with a one-tailed test)
difference would constitute overly extrapolating from
the given data. On the other hand, the Hispanic find-
ings do not provide the evidence against the discrimina-
tion hypothesis that the black name results do. In fact,
the substantively small magnitude we find is similar to
the anti-Hispanic responsiveness difference uncovered in
White, Nathan, and Faller (2015) with more observa-
tions (and thus smaller standard errors). This could be
taken as evidence bolstering the conclusion of discrim-
ination against Hispanics. It could also be taken as evi-
dence that responsiveness bias against Hispanics is very
small in magnitude and is only significant when the Ns
are sufficiently large. Importantly, having all three racial
groups in our study allows us to compare antiblack to
anti-Hispanic discrimination. This comparison, as we
discuss later, shows that the two types of antiminority
discrimination are not the same and may derive from (or
get reduced by) different mechanisms. The difference be-
tween blacks and Hispanics is about 8 percentage points
(p = .02).

We utilize multiple logit model specifications to test
for effects more rigorously. In Table 1, we depict the
effects of race on our dependent variables of interest.
Here, in accordance with our theoretical predictions, we
consolidate the treatments by race/ethnicity (males and
females together). These models include control variables
that may be associated with responsiveness: percent black
and percent Hispanic in the community, poverty rate (re-

lated to the demand for housing and perhaps the number
of inquiries), population (logged), and a variable that
indicates that a housing authority’s online contact ad-
dress was difficult to obtain (Hard Email). Similar mod-
els that do not consolidate the treatments by race pro-
duce very similar results (see Table SI2 in the supporting
information).

The model in the first column reports the results us-
ing the responsiveness variable. Contrary to Hypothesis 1,
it provides little to no evidence of racial discrimination.
It does show less responsiveness in poorer areas (perhaps
they have less housing available or bureaucratic capac-
ity to respond). It also shows less responsiveness from
housing authorities that did not provide easily accessible
electronic contact information. While these variations are
worthy of further exploration, the main implications con-
cern race. Our failure to reject the null is not the same
as demonstrating no effect. Nonetheless, we believe these
results in conjunction with the next set of findings on
response speed are strongly suggestive of an absence of
racial discrimination. The likelihood ratio test comparing
Model 1 to the same model without the race indicators is
insignificant (� 2 = 4.49, p = .106). As with the summary
statistics, this model provides some suggestive evidence
of small anti-Hispanic bias, but none that reaches con-
ventional significance levels.10

10We note that the main results of all three models in Table 1 are
unchanged (except for the Hard Email control variable becom-
ing insignifiant in Model 3) when including the HUD assessment
scores that we discuss below as a proxy for a housing authority’s
professionalism and capacity. (Because these scores are only avail-
able for about two-thirds of the housing authorities, we do not
include them in the models we report).

108 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

TABLE 1 Logit models for the three Dependent
Variables, with All Three Racial Groups
and Other Demographic Controls
Variables

(2) (3)
(1) Response in Proper Name

Response 24 Hours Greeting

Black Name 0.13 0.29 −0.21
(0.17) (0.26) (0.22)

Hispanic Name −0.21 0.10 −0.82∗∗
(0.16) (0.25) (0.22)

Percent Black −0.31 −0.17 0.71
(0.52) (0.87) (0.74)

Percent Hispanic −0.32 −0.92 1.33∗
(0.43) (0.69) (0.63)

Poverty Rate −1.65 1.28 −.93
(1.21) (2.00) (1.71)

Log Population 0.13 −0.01 −0.04
(0.07) (0.10) (0.09)

Hard Email −.46∗∗ −0.14 −0.42∗
(0.14) (0.22) (0.19)

Constant −0.50 1.33 0.88
(0.75) (1.16) (0.99)

Observations 978 551 549
Log Likelihood −650.7 −286.8 −367.5
Note: Standard errors in parentheses. Base category is white. De-
mographic variables from 2012 American Community Survey.
∗∗p<.01, ∗p<.05.

Finally, we address the possibility that aggregate bal-
ance is masking offsetting preferential treatment (an is-
sue that becomes even more important in our discus-
sion of representative bureaucracy below). Butler and
Broockman (2011) find that among Democrats, black
state legislators are more likely to reply to black con-
stituent requests than to their white counterparts. One
way to roughly estimate the underlying distribution of
housing officials is to assume they are drawn from the
populations in which the housing authorities are located.
The average community in our sample is 66.5% white,
13.5% black, and 14.5% Hispanic. Thus, even if blacks
(or Hispanics) are disproportionately represented in gov-
ernment jobs (about 19% of blacks, 14% of whites, and
10% of Hispanics work in the public sector; Department
of Labor, 2012a, 2012b), it is very unlikely they are over-
represented enough to account for the overall balance we
observe.

To supplement these findings, we considered two
other indicators of responsiveness. First, we assessed
speed by checking whether those who did respond did

so within 24 hours (essentially by the end of the next
working day, given the timing of our e-mails.) We con-
ducted all of the same analyses as we did with the basic
responsiveness dependent variable (see Figure SI1 and
Table SI3 in the supporting information and Model 2
in Table 1). These results largely mirror those above.
In fact, they offer even less evidence of discrimination,
mainly because the suggestive anti-Hispanic bias disap-
pears. Brett, the white male, actually obtained the lowest
24-hour response percentage and had the highest median
response time. Finally, responsiveness bias might manifest
in the quality of information received. Only 48% of re-
sponses included wait list times—a query we included in
the e-mails. Of Hispanics who obtained a response, 43%
received information about wait list times, compared to
50% for blacks and 49% for whites. These results fall
well short of conventional levels of statistical significance
(p = .427, � 2 test).

Evidence of Discrimination: Tone

We also postulated that racial bias might manifest in offi-
cials’ friendliness when they do respond. Using salutation
by proper name as our measure of tone (see above), we
report the proportion of responses that began with a per-
sonalized greeting in Figure 2 by the six treatments and
three racial groups.

Here, we do see evidence of racial bias, but only
against Hispanics. Sixty-one percent of the messages to
the white e-mailers began with a named salutation. Only
41% to Hispanics did. This difference is highly signifi-
cant (p = .000). Only 37% of responses to Santiago began
with a named salutation, compared with over 60% for
Tyrone, Brett, and Emily.

The model we report in the text, in the third column
of Table 1, further demonstrates the negative impact a
Hispanic name has on receiving a named greeting. The
Hispanic name indicator variable is substantial, nega-
tive, and highly significant. Substantively, the change in
the predicted probability of getting a named salutation
as a function of changing from a white name to a His-
panic one (with other variables held at their means) is
–20.0 percentage points, with 95% confidence interval
[−30.5, −10.0]. Moreover, the percent Hispanic variable
is positive and significant, suggesting that named greet-
ings, perhaps especially for Hispanics, are more likely in
areas with higher Hispanic populations. (Similar mod-
els that include all six treatments separately [Table SI4]
bolster these results.) These results differ from those in
White, Nathan, and Faller (2015), as they find discrimina-
tion in responsiveness but not tone. Part of this variation

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 109

FIGURE 2 Proportion of Responses That Begin with a Personalized Greeting

By race/ethnicity crossed with sex
Tyrone
Shanice
Santiago
Gabriela

Brett Emily

0
.2
.4
.6
.8
1
P
ro
po
rt
io
n
G
et
tin
g
a
R
es
po

ns
e

U
si

n

g
a

P
ro

pe
r

N
am

e
Treatment Group
By race/ethnicity (collapsed by sex)
Black
Hispanic
White
0
.2
.4
.6
.8
1
P
ro
p
o
rt
io
n
G
e
tt
in
g
a
R
e
sp
o
n

se
U

si
n

g
a
P
ro

p
e

r
N

a
m

e
Treatment Group

in results, though, may be attributable to slight differences
in how friendliness in tone was coded.11

Community and Officials’ Demographics

Representative bureaucracy, contact theory, and familiar-
ity bias may explain some of the variation between our
results and findings of bias in other settings. We thus as-
sess the interaction of the race of the putative constituent
and (1) the demographics of public housing officials and
(2) the traits of the community in which she is applying.

Because we were only able to determine a hous-
ing official’s ethnicity when he or she responded to our
e-mails, we cannot observe the denominator. We can only
estimate responsiveness by assuming that randomization
balanced the odds of an e-mail arriving on the screen of a
Hispanic official in the first place. Doing so suggests that
responsiveness is not driven by officials favoring respon-
dents of their own race. Of all of the responses for whites,
blacks, and Hispanics, 6.2%, 6.1%, and 9.0% respectively
came from ostensibly Hispanic officials (p = .453). Sim-
ilarly, conditional on replying, Hispanic officials are not
more likely to respond in 24 hours to Hispanic appli-
cants (white: 81.8%, black: 83.2%, Hispanic: 88.2%; p =
.88). Again, there is little evidence to support the conclu-
sion that Hispanic e-mailers did appreciably better when
interacting with Hispanic housing officials.

In Figure 3, we show the proportion of friendly re-
sponses coming from Hispanic officials. We find that

11As above, we also did this analysis coding all nonresponses as
unfriendly responses to address the truncation problem. This does
not affect the results.

FIGURE 3 Proportion of Responses from
Hispanic Housing Officials That
Begin with a Personalized Greeting

Black
Hispanic
White
0
.2
.4
.6
.8
1
P
ro

p
o
rt

io
n
G

e
tt
in

g
a

R
e
sp

o
n
se

U
si

n
g
a

P
ro

p
e
r

N
a
m

e
Treatment Group

Hispanic officials were no more likely than their non-
Hispanic counterparts to address Hispanic constituents
by name. Forty-seven percent of responses from His-
panic housing officials addressed Hispanic constituent re-
quests by name, compared with 40% from non-Hispanic
housing officials—a difference that was not statistically
significant in our sample. Furthermore, the lower rate
at which Hispanics receive named greetings (displayed
in Figure 2) appears to hold even when we only look
at responses from Hispanic housing officials. Though
the results are no longer statistically significant due to
large standard errors, Hispanics continue to receive the

110 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

lowest share of named greetings, even when the hous-
ing official responding to their request is also Hispanic.12

Thus, our admittedly crude efforts to test Hypothesis 2
(representative bureaucracy) suggest that it is an unlikely
candidate to explain why we did not find the same re-
sponsiveness bias against minorities that other scholars
have. These results come with important caveats though.
The cell sizes are small. More importantly, we are unable
to test for analogous effects for black and white housing
officials.

The effects of community traits do provide some sup-
port for Hypothesis 3. The first way we move beyond
average effects is to create variables capturing whether
each housing authority is in a community that is in the
top, middle, or bottom third of the distribution for black
population and for Hispanic population. Because these
distributions are so skewed, the top thirds for black and
Hispanic percentages start at around 12% black and His-
panic, respectively. Communities in the bottom thirds
have essentially no black or Hispanic residents. We chose
to convert continuous demographic variables into cate-
gorical ones both to reduce the impact of places with 80%
or 90% black (or Hispanic) populations and because the
theoretical questions are better thought of in terms of high
or low populations rather than ones about the marginal
effect of a one-unit increase in the population of a minor-
ity group.

We analyze the effects of these 1/3 variables by split-
ting our sample into the three racial/ethnic treatment
groups and estimating separate models. We report the
results of these logit models for the responsiveness and
friendliness dependent variables in the supporting infor-
mation (Tables SI5 and SI6). These models include con-
trols for city population (logged), poverty rate, and our
“hard email” indicator. The main effects of interest are
the 1/3 variables. To focus on substantive effects, in Fig-
ure 4, we plotted the predicted change in the probability
of getting (1) a response and (2) a named greeting. Here,
the point estimates capture the effects of moving from a
lowest-third population (black/Hispanic) community to
a top-third community, with all other variables held at
their means. For example, the very first point depicts the
extent to which e-mailing a housing authority in a high-
black population community changes the probability of
a white e-mailer obtaining a response relative to one in a
low-black community.

12This finding helps us to rule out the possibility that officials are
not responding to Hispanics by name because of a lack of famil-
iarity with Hispanic names. Indeed, were this the case, we would
expect Hispanic officials to address putative Hispanic constituents
by name at higher rates.

The left panel in Figure 4 shows that the overall non-
bias in responsiveness findings we reported earlier does
not vary in important ways with community demograph-
ics. Whites’, blacks’, and Hispanics’ likelihoods of receiving
responses are always similar irrespective of whether they
are e-mailing in a community in which they are a relatively
large or small fraction of the population. This means that
the aggregate results are not masking extremely high re-
sponsiveness in black communities and discrimination in
others, for example. These null results are robust to using
other population cut points (top versus bottom 1/4s and
1/5s instead of 1/3s).

The named salutation models (right panel) do offer
some suggestive evidence that a community’s population
affects the likelihood that blacks and Hispanics receive
friendlier responses. Black names may be less likely (just
outside the p = .1 level) to receive a named greeting in a
high-Hispanic area. Hispanic names, on the other hand,
may be more likely (p = .08) to receive friendly responses
in areas in the top third of the Hispanic population dis-
tribution compared to those in the lowest third. While we
report the more conservative (and insignificant) results
in which we split community racial demographics into
thirds, the Hispanic results increase in magnitude and
significance when comparing the top 1/4 to the bottom
1/4 (p = .06) and the top 1/5 to the bottom 1/5 (p = .01)
of the percent Hispanic distributions.

To conclude our analysis, we zero in on substan-
tively and theoretically interesting permutations of racial
demographics. Now, we create variables to distinguish
four special types of communities: “high white” (top 1/3
white, bottom 1/3 black, bottom 1/3 Hispanic), “high
black” (top 1/3 black, bottom 1/3 white, bottom 1/3 His-
panic), “high Hispanic” (top 1/3 Hispanic, bottom 1/3
white, bottom 1/3 black), and “high black and Hispanic”
(top 1/3 black, top 1/3 Hispanic, bottom 1/3 white). Fo-
cusing on these permutations directly speaks to ques-
tions of how responsiveness varies in communities that
are dominated by one group, and whether racial minori-
ties do better or worse depending on whether they are
concentrated (alongside other nonwhites) in a commu-
nity, or whether they are the primary minority group.
All communities in the data that do not fit into one of
these four categories are lumped together in the base-
line. As before, we report the full model results (again
splitting the sample by e-mailer race) in the supporting
information (Tables SI7 and SI8). We plot substantive
effects of interest in Figure 5. Here, the point estimates
capture the differences in the predicted probability of get-
ting a response (or named greeting) as a consequence of
moving from one community type to another. For white
names, the main question of interest—as articulated in

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 111

FIGURE 4 Change in Predicted Probability of Getting a Response and Getting a Named Greeting as a
Function of Moving from a Community in the Lowest Third of Percent Black/Hispanic to
Moving to One in the Highest Third

Note: Estimates based on models described in the text and reported in Tables SI5 and SI6 in the supporting information.

FIGURE 5 Change in Predicted Probability of Moving from a Town of Type X to a Town of Type Y

Note: Demographic variables from 2012 American Community Survey. Town type indicated by x-axis labels. Mainly White = top 1/3
in white population, bottom 1/3 in black and Hispanic; High Black (Hispanic) = top 1/3 black (Hispanic) population, bottom 1/3
white and Hispanic (black); High Black and Hispanic = top 1/3 black and top 1/3 Hispanic population. Probabilities are outputs from
the models in Tables SI7 and SI8 in the supporting information.

Hypothesis 3—is responsiveness in very white areas com-
pared to highly diverse ones. For blacks (and Hispanics),
three comparisons are theoretically important: (1) the dif-
ference between a very white community and a very black
(or Hispanic) one (one’s own group is disproportionately
represented), (2) the difference between a very white area
and a very black and Hispanic one (minorities, including,

but not exclusively, one’s own group, are disproportion-
ately represented), and (3) the difference between a very
Hispanic and a very black area (one’s own group versus a
different minority group).

This analysis again shows little variation in basic re-
sponse rates. Consistent with Hypothesis 3, it provides
some suggestive evidence, as above, that Hispanics obtain

112 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

fewer unfriendly/formal responses in higher minority ar-
eas than they do elsewhere. Notably, the largest (and only
significant effect at the .05 level) is the difference between
mainly white and high minority (Hispanic and black)
areas for named responses to the e-mails from Hispan-
ics. This suggests that the underlying mechanism is more
about day-to-day exposure to minorities in the commu-
nity than to differential treatment by group. Hispanics do
not do appreciably better in high Hispanic areas than they
do in high black areas. As before, the effects are similarly
signed and more significant (p < .05) when substituting a 1/4 cut point that compares more extreme places to each other.

Discussion

Our findings offer a mixed and nuanced portrait of dis-
crimination, and its absence, in bureaucratic responsive-
ness. We find some support for the bias against minorities
hypothesis (Hypothesis 1) in the context of e-mail tone.
Our further exploration of Hypotheses 2 and 3, how-
ever, suggests little support for Hypothesis 2. Hispanic
housing officials did not provide more friendly responses
to Hispanics’ constituent service requests than they did
to others’. We are cautious in extrapolating too much
from these results, however, given our inability to test
Hypothesis 2 for white and black bureaucrats. Finally, we
do find modest and suggestive support for Hypothesis 3
in the context of Hispanic e-mail tone.

Perhaps the most striking feature of our analysis—in
contrast with a wealth of audit studies—is the absence of
antiblack discrimination. Below, we discuss several possi-
ble systematic explanations for this important main dif-
ference. Before delving into these more substantively in-
teresting possibilities, we address whether the key results
may actually be a consequence of class signaling in our de-
sign. Perhaps the polished (though informal) English in
our e-mails suggested a high socioeconomic status appli-
cant and muted potential discrimination. Bertrand and
Mullainathan’s (2004) seminal experimental study pro-
vides helpful evidence to counter this concern. They find
that higher social class does not mitigate antiblack labor
discrimination. Similarly, using a telephone audit to study
housing discrimination, Massey and Lundy (2001) find
that racial discrimination persists regardless of class cues,
though lower-class status does exacerbate antiblack dis-
crimination. On balance, prior scholarship suggests that
sending a favorable class cue is unlikely to have induced a
null black discrimination result.

Moreover, it is possible that these findings are largely
a confirmation of the racial classification model (Schram

et al. 2009; Soss, Fording, and Schram 2008). In the ab-
sence of discrediting information about minority clients,
caseworkers may have been disinclined to engage in
overt discrimination. The fact that we do find evidence
of discrimination—in friendliness toward Hispanics—
despite the absence of discrediting information in that
treatment condition makes this possibility less likely.

Blacks, Familiarity Bias, and Representative
Bureaucracy

Demographic data from the National Low Income Hous-
ing Coalition (2012) and the American Community
Survey (United States Census Bureau 2013) reveal that
blacks are significantly overrepresented in America’s
public housing. While blacks compose just over 10%
of the total population, and under 20% of total renters,
they represent over 40% of public housing residents and
voucher recipients. Whites, conversely, are significantly
underrepresented, and Hispanics are proportionately rep-
resented. It could be, then, that the absence of discrim-
ination against blacks is a consequence of housing au-
thority officials’ disproportionate familiarity with black
constituents in the public housing context.

Because we are sampling at the housing authority level,
however, we suspect that this mechanism may contribute
to, but not completely explain, our results. America’s
racial geography suggests that the overrepresentation of
blacks in public housing is likely most stark in large cities.
However, each large city is only one observation in our
study. Consequently, the bottom third of our distribution
of housing authorities comes from communities with es-
sentially zero black residents. Thus, it is unlikely that the
bulk of our housing authorities are disproportionately
accustomed to black clients. Nevertheless, it is possi-
ble that these mechanisms affect bigger cities and help
explain why the population variable often had a positive
effect on responsiveness to blacks, even when controlling
for overall black population rates (e.g., Table SI5).

Racial Coding of Poverty

A second, and related, potential explanation stems
from research by Gilens (1999). This line of scholar-
ship contends that, courtesy of the disproportionate use
of images of blacks in negative stories about poverty,
Americans largely oppose redistributive welfare initiatives
because they associate blacks with the undeserving poor.
It could be that, as a consequence of their jobs, housing
authority employees are more likely to have had personal

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 113

encounters with poor blacks and whites alike. Therefore,
they are not receiving biased information about the racial
composition of the poor from the media. Instead, their
views of race and the deserving or undeserving poor may
be based on their direct experience. Applying Gilens’s
theory of mass attitudes about welfare to public housing
officials offers a potential explanation for our findings.
Experience working directly with the poor may dimin-
ish the racial coding of poverty, or lead to a different one,
which could in turn curb the discriminatory impulses that
manifest in other populations with different perceptions
of the connections between race and poverty.

Fair Housing Legislation

Finally, in their research on voting officials, White,
Nathan, and Faller (2015) find lower levels of discrim-
ination in locales that were covered by the Voting Rights
Act (VRA). They cite research from Pager and Shepherd
(2008) suggesting that organizational awareness, moni-
toring, and procedures can make potential discriminators
cognizant of and reduce the risk of bias. Much of what
public housing administrators do is governed extensively
by the Fair Housing Act. Moreover, because of the Fair
Housing Act, the Department of Housing and Urban De-
velopment makes an effort to regularly measure private
market housing discrimination with frequent audit stud-
ies centered on residential steering and other forms of
discriminatory behavior (Turner et al. 2002, 2013). Hous-
ing bureaucrats may therefore be highly knowledgeable
about the potential for discrimination and about the pro-
cedures the federal government employs to mitigate it.
This could make housing officials—like voting officials in
VRA-regulated locales—less apt to discriminate. The Fair
Housing Act emerged in 1968 in the context of national
concern about black poverty and residential segregation.
It may thus raise awareness about antiblack, but not anti-
Hispanic, discrimination. As with the Voting Rights Act,
the Fair Housing Act is not randomly assigned to par-
ticular bureaucracies or locations, so we cannot assess its
causal impact. Nonetheless, the absence of discrimination
against blacks in particular suggests at least some efficacy
in addressing housing discrimination. It may even point,
especially in conjunction with the Voting Rights Act find-
ings, to more broadly applicable interventions for reduc-
ing discrimination in other areas and suggest why housing
bureaucracies differ from other contexts in which we do
observe discrimination.

It is also plausible that the prominence of the Fair
Housing Act made respondents suspicious that they were
being studied by academics or the government. While we
cannot disprove this possibility, we believe that several

pieces of evidence suggest that it is unlikely. First, while
discrimination in the private housing market has been
extensively studied (and audited), public housing author-
ities themselves have, to our knowledge, never been au-
dited. Second, many of the responses we received seemed
genuine and informal in nature—not the sorts of replies
you would anticipate from a bureaucrat aware she was
being audited. Here is a sampling of some of these replies:
“How old are you?” “If you would like to contact my of-
fice, I would explain it to you better. My office number
is [office phone number]” “[Housing authority website]
has all the info u need.” Third, the presence of an anti-
Hispanic bias in friendliness further indicates that our
experiment participants were unaware of our study; we
should expect no bias to manifest if respondents were
aware of the audit.

Bureaucratic Professionalism

The discussion of the Fair Housing Act prompts a broader,
related question: Does bureaucratic professionalism more
generally mitigate discrimination? Variations in bureau-
cratic professionalism cannot explain our article’s experi-
mental nonresults on responsiveness; the randomization
in design ensures that differences (or non differences) in
responsiveness across places are not driven by jurisdic-
tional characteristics. However, heterogeneity in discrim-
ination by bureaucratic professionalism might point to
interesting policy prescriptions for mitigating racial bias.
Highly professional bureaucracies may be better at ensur-
ing that they reply to outreach. They may also be more
sensitive to issues of discrimination in housing in general
and may be more likely to prevent differential treatment.
To assess this possibility, we use the Public Housing As-
sessment Scores (PHAS) that the Department of Housing
and Urban Development assigns to each housing author-
ity. HUD calculates these scores based on four compo-
nents: (1) physical inspection, (2) financial assessment,
(3) management operations, and (4) use of the Capital
Fund (a source of housing authority funding that can be
used for a variety of operations). HUD collects the data
that compose these ratings through self-reports and au-
dits (Department of Housing and Urban Development
2015b).

We split the sample by HUD assessment scores.13

The median assessment score in our sample was 89 out
of 100. This actually corresponded to the cut point be-
tween “high performers” and others in the qualitative

13Not all housing authorities had assessment scores available (n =
671 for all analyses).

114 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

designations that accompany the numerical scores. The
data suggest that professionalism is positively associated
with responsiveness to both whites (19 percentage point
difference between high performers and others, p = .01)
and blacks (16 percentage point difference, p = .02), but
not Hispanics (difference of less than 1 percentage point).
In these housing authorities, the black response rates ex-
ceed those of Hispanics by a whopping 15 percentage
points (p = .024). Conversely, whites, blacks, and His-
panics all obtain about the same response rates from the
lower-performing housing authorities. While these gen-
eral trends hold when we use different cut points, the stark
black and Hispanic patterns soften. Thus, we cannot draw
any strong conclusions.

Conclusion

By comparing bias toward blacks and Hispanics—a juxta-
position most studies eschew to preserve power (though
see Schram et al. 2009)—we are able to uncover a pre-
viously unspecified type of bureaucratic bias that im-
pacts one group (Hispanics) but not the other. Indeed,
our findings suggest that Hispanics are marginally less
likely to receive responses from housing bureaucrats, and,
when they do receive responses, they are far less likely to
be friendly ones. Furthermore, our results likely repre-
sent a lower bound on discrimination against Hispanic
constituents. In our study, putative Hispanic constituents
are sending e-mails with polished grammar, which may
provide counter-stereotypical cues to caseworkers. This
anti-Hispanic bias has important policy implications. In
her in-depth exploration of local bureaucracies, Marrow
(2011) found that these organizations—including schools
and social service organizations—play a critical role in in-
corporating (or failing to incorporate) new Hispanic im-
migrants. Unfriendliness toward Hispanics seeking pub-
lic housing, a low-income population that likely already
feels stigmatized, has the potential to be quite alienat-
ing. These types of interactions with government agen-
cies could harm prospects for broader political and social
incorporation.

Moreover, our findings reveal that we may observe
different discriminatory biases depending upon the po-
litical and economic arena. This does not necessarily mean
that researchers and policy makers should take our arti-
cle as evidence that discrimination is more limited than
previously feared, but it should at least serve as a check
against overgeneralizing from the excellent existing work
that does find discrimination. At a minimum, our find-
ings suggest that bureaucratic discrimination may be con-
tingent on the demographics of the place in which a bu-

reaucracy is situated. More generally, we suspect that the
level of bias depends on the structure of the bureaucracy.
A future research agenda that compares different bureau-
cracies, and bureaucratic oversight, might help scholars
and policy makers identify policy solutions to the long-
thorny problem of racial discrimination in the public and
private sectors.

References

Allport, Gordon W. 1954. The Nature of Prejudice. Reading, MA:
Addison-Wesley.

Arceneaux, Kevin, and Daniel M. Butler. 2016. “How Not to
Increase Participation in Local Government: The Advantages
of Experiments When Testing Policy Interventions.”Public
Administration Review 76(1): 131–139.

Ayres, Ian, and Peter Siegelman. 1995. “Race and Gender Dis-
crimination in Bargaining for a New Car.” American Eco-
nomic Review 85(3): 304–21.

BabyCenter en Espanol. 2011. “100 Most Popular Hispanic Baby
Names of 2011.” http://www.babycenter.com/0_100-most-
popular-hispanic-baby-names-of-2011_10363639.bc.

Barreto, Matthew A., Gary M. Segura, and Nathan D. Woods.
2004. “The Mobilizing Effect of Majority-Minority Districts
on Latino Turnout.” American Political Science Review 98(1):
65–75.

Bertrand, Marianne, and Sendhil Mullainathan. 2004. “Are
Emily and Greg More Employable Than Lakisha and
Jamal? A Field Experiment on Labor Market Discrimina-
tion.” American Economic Review 94(4): 991–1013.

Bobo, Lawrence. 2001. “Racial Attitudes and Relations at the
Close of the Twentieth Century.” In America Becoming:
Racial Trends and Their Consequences, ed. Neil J. Smelser,
William Julius Wilson, and Faith Mitchell. Washington, DC:
National Academy Press, 264–301.

Bowean, Lolly. 2014. “Chicago Housing Authority Opens
Wait Lists for Public Housing Vouchers.” http://www.
chicagotribune.com/news/ct-cha-waiting-list-met-1028-
20141027-story.html.

Bradbury, Mark, and J. Edward Kellough. 2011. “Representa-
tive Bureaucracy: Assessing the Evidence on Active Repre-
sentation.” American Review of Public Administration 2(2):
157–67.

Brodkin, Evelyn Z. 1997. “Inside the Welfare Contract: Discre-
tion and Accountability in State Welfare Administration.”
Social Service Review 71(1): 1–33.

Butler, Daniel M., and David E. Broockman. 2011. “Do Politi-
cians Racially Discriminate Against Constituents? A Field
Experiment on State Legislators.” American Journal of Polit-
ical Science 55(3): 463–477.

Clinton, Joshua D., and David E. Lewis. 2008. “Expert Opinion,
Agency Characteristics, and Agency Preferences.” Political
Analysis 16(1): 3–20.

Coleman, Sally, Jeffrey L. Brudney, and J. Edward Kellough.
1998. “Bureaucracy as a Representative Institution: Toward a
Reconciliation of Bureaucratic Government and Democratic
Theory.” American Journal of Political Science 42(3): 717–44.

DOES RACE AFFECT ACCESS TO GOVERNMENT SERVICES? 115

Davis, Belinda Creel, Michelle Livermore, and Younghee Lim.
2011. “The Extended Reach of Minority Political Power:
The Interaction of Descriptive Representation, Managerial
Networking, and Race.” Journal of Politics 73(2): 494–507.

Department of Housing and Urban Development. 2015a.
“HUD’s Public Housing Program.” http://portal.hud.
gov/hudportal/HUD?src=/topics/rental_assistance/phprog.

Department of Housing and Urban Development. 2015b.
“Integrated Assessment Subsystem—Public Housing
Assessment System (NASS-PHAS).” http://portal.hud.gov/
hudportal/HUD?src=/program_offices/public_indian_
housing/reac/products/prodphasintrule.

Department of Labor. 2012a. “The African-American Labor
in the Recovery.” http://www.dol.gov/_sec/media/reports/
BlackLaborForce/BlackLaborForce .

Department of Labor. 2012b. “The Latino Labor Force
at a Glance.” http://www.dol.gov/_sec/media/reports/
HispanicLaborForce/HispanicLaborForce .

Doleac, Jennifer L., and Luke C. D. Stein. 2013. “The Visible
Hand: Race and Online Market Outcomes.” Economic Jour-
nal 2013: F469–92.

Epp, Charles R., Steven Maynard-Moody, and Donald P.
Haider-Markel. 2013. Pulled Over: How Police Stops Define
Race and Citizenship. Chicago: University of Chicago Press.

Ernst, Rose, Linda Nguyen, and Kamilah C. Taylor. 2013. “Citi-
zen Control: Race at the Welfare Office.” Social Science Quar-
terly 94(5): 1283–1307.

Fording, Richard C., Joe Soss, and Sanford F. Schram. 2007.
“Devolution, Discretion, and the Effect of Local Political
Values on TANF Sanctioning.” Social Service Review 81: 285–
316.

Fryer, Roland G., and Steven D. Levitt. 2004. “The Causes and
Consequences of Distinctively Black Names.” Quarterly Jour-
nal of Economics 119(3): 767–805.

Gilens, Martin. 1999. Why Americans Hate Welfare: Race, Media,
and the Politics of Anti-Poverty Policy. Chicago: University of
Chicago Press.

Grose, Christian R. 2014. “Field Experimental Work on Political
Institutions.” Annual Review of Political Science 17: 355–70.

Hall, Richard L. 1996. Participation in Congress. New Haven,
CT: Yale University Press.

Heerwegh, Dirk. 2005. “Effects of Personal Salutations in E-mail
Invitations to Participate in a Web Survey.” Public Opinion
Quarterly 4(4): 588–98.

Henderson, John A., Jasjeet S. Sekhon, and Rocio Titiunik.
2015. “Cause or Effect? Turnout in Hispanic Majority-
Minority Districts.” Working paper, University of California
at Berkeley and University of Michigan. http://www-
personal.umich.edu/titiunik/papers/HispanicTurnout .

Howard, Daniel J., and Roger A. Kerin. 2011. “The Effects of
Name Similarity on Message Processing and Persuasion.”
Journal of Experimental Social Psychology 46(1): 63–71.

Joinson, Adam N., and Ulf-Dietrich Reips. 2007. “Personalized
Salutation, Power of Sender and Response Rates to Web-
Based Surveys.” Computers in Human Behavior 23(3): 1372–
83.

Jones, Bryan D., Saadia Greenberg, Clifford Kaufman, and
Joseph Drew. 1977. “Bureaucratic Response to Citizen-

Initiated Contacts: Environmental Enforcement in Detroit.”
American Political Science Review 71(1): 148–65.

Katznelson, Ira. 2005. When Affirmative Action Was White. New
York: W. W. Norton.

Keiser, Lael R., Peter R. Mueser, and Seung-Whan Choi. 2004.
“Race, Bureaucratic Discretion, and the Implementation of
Welfare Reform.” American Journal of Political Science 48(2):
314–27.

Kinder, Donald R., and Cindy D. Kam. 2009. Us Against Them:
The Ethnocentric Foundations of American Opinion. Chicago:
University of Chicago Press.

Krislov, Samuel. 1974. Representative Bureaucracy. Englewood
Cliffs, NJ: Prentice-Hall.

Lazio, Rick. 1998. “Title V of the FY 99 VA/HUD Ap-
propriations Conference Report: ‘The Quality Hous-
ing and Work Responsibility Act of 1998.” http://www.
phada.org/qhwra98.html.

Levitt, Steven D., and Stephen J. Dubner. 2010. Freakonomics.
New York: Harper-Perennial.

Lieberman, Robert C. 1998. Shifting the Color Line: Race and the
American Welfare State. Cambridge, MA: Harvard University
Press.

Lipsky, Michael. 1980. Street-Level Bureaucracy. New York:
Russell Sage Foundation.

Marrow, Helen B. 2011. New Destination Dreaming: Immigra-
tion, Race, and Legal Status in the Rural American South. Palo
Alto, CA: Stanford University Press.

Massey, Douglas S., and Garvey Lundy. 2001. “Use of Black En-
glish and Racial Discrimination in Urban Housing Markets:
New Methods and Findings.” Urban Affairs Review 36(4):
452–69.

McClendon, Gwyneth H. 2016. “Race and Responsiveness: An
Experiment with South African Politicians.” Journal of Ex-
perimental Political Science. Forthcoming. https://docs.
google.com/viewer?a=v&pid=sites&srcid=
ZGVmYXVsdGRvbWFpbnxnd3luZXRobWNjbGVuZG9uf
Gd4OjY4OGNlOWM5NTMyMzlkOGE.

Meier, Kenneth J. 1993. “Latinos and Representative Bureau-
cracy: Testing the Thompson and Henderson Hypotheses.”
Journal of Public Administration Research and Theory 3(4):
393–414.

Meier, Kenneth J., Robert D. Wrinkle, and J. L. Polinard. 1999.
“Representative Bureaucracy and Distributional Equity: Ad-
dressing the Hard Question.” Journal of Politics 61(4): 1025–
39.

Milkman, Katherine L., Modupe Akinola, and Dolly Chugh.
2014. “What Happens Before? A Field Experiment Ex-
ploring How Pay and Representation Differentially Shape
Bias on the Pathway into Organizations.” Working pa-
per. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=
2063742.

Mosher, Frederick C. 1968. Democracy and the Public Service.
New York: Oxford University Press.

National Low Income Housing Coalition. 2012. “Who
Lives in Federally Assisted Housing?” http://nlihc.org/sites/
default/files/HousingSpotlight2-2 .

Pager, Devah, and Hana Shepherd. 2008. “The Sociology
of Discrimination: Racial Discrimination in Employment,

116 KATHERINE LEVINE EINSTEIN AND DAVID M. GLICK

Housing, Credit, and Consumer Markets.” Annual Review
of Sociology 34: 181–209.

Pager, Devah, Bruce Western, and Bart Bonikowski. 2009. “Dis-
crimination in a Low-Wage Labor Market: A Field Experi-
ment.” American Sociological Review 2009 74(5): 777–99.

Pettrigrew, Thomas F. 1998. “Intergroup Contact Theory.” An-
nual Review of Psychology 49: 65–85.

Riccucci, Norma M., Gregg G. Van Ryzin, and Cecilia F. Lavena.
2014. “Representative Bureaucracy in Policing: Does It In-
crease Perceived Legitimacy?” Journal of Public Administra-
tion Theory 24(3): 537–51.

Schram, Sanford F., Joe Soss, Richard Fording, and Linda
Houser. 2009. “Deciding to Discipline: Race, Choice, and
Punishment.” American Sociological Review 74(3): 398–422.

Schulman, Kevin A., Jesse A. Berlin, William Harless, Jon F.
Kerner, Shyrl Sistrunk, Bernard J. Gersh, Christopher K.
Taleghani, Ross Dubé, Jennifer E. Burke, Sankey Williams,
John M. Eisenberg, William Ayers, and José J. Escarce. 1999.
“The Effect of Race and Sex on Physicians’ Recommenda-
tions for Cardiac Catheterization.” New England Journal of
Medicine 340: 618–26.

Schwartz, Alex F. 2010. Housing Policy in the United States. New
York: Routledge.

Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2008.
“The Color of Devolution: The Politics of Local Punishment
in an Era of Neoliberal Welfare Reform.” American Journal
of Political Science 52(3): 536–53.

Sowa, Jessica E., and Sally Coleman Selden. 2003. “Administra-
tive Discretion and Active Representation: An Expansion of
the Theory of Representative Bureaucracy.” Public Adminis-
tration Review 63(6): 700–710.

Turner, Margery Austin, Diane K. Levy, Doug Wissoker,
Claudia Aranda, Rob Pitingolo, and Rob Santos. 2013.
“Housing Discrimination Against Racial and Ethnic
Minorities in 2012.” Urban Institute. http://www.huduser.
org/portal/publications/fairhsg/hsg_discrimination_2012.
html.

Turner, Margery Austin, Stephen L. Ross, George C. Gal-
ster, and John Yinger. 2002. “Discrimination in Metropoli-
tan Housing Markets: National Results from Phase I
HDS 2000.” Urban Institute. http://www.huduser.org/portal
/Publications/pdf/Phase1_Report .

Tversky, Amos, and Daniel Kahneman. 1972. “Availability:
Heuristic for Judging Frequency and Probability.” Cognitive
Psychology 4(2): 207–32.

United States Census Bureau. 2013. “American Community
Survey 5-Year Estimates, 2007–2012.” http://factfinder.
census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh
=t.

Vale, Lawrence. 2000. From Puritans to the Projects: Public Hous-
ing and Public Neighbors. Cambridge, MA: Harvard Univer-
sity Press.

Wattenberg, Martin. 2005. “Baby Names, Visualization,
and Social Data Analysis.” Information Symposium
on Information Visualization. http://ieeexplore.ieee.org/
stamp/stamp.jsp?tp=&arnumber=1532122.

Wei, Iris I., Beth A. Virnig, Dolly A. John, and Robert O’Morgan.
2006. “Using a Spanish Surname Match to Improve Iden-

tification of Hispanic Women in Medicare Administrative
Data.” Health Services Research 41(4): 1469–81.

White, Ariel R., Noah L. Nathan, and Julie K. Faller. 2015. “What
Do I Need to Vote? Bureaucratic Discretion and Discrimina-
tion by Local Election Officials.” American Political Science
Review 109(1): 129–42.

Word, David L., Charles D. Coleman, Robert Nunziata,
and Robert Kominski. 2000. “Demographic Aspects of
Surnames from Census 2000.” http://www2.census.gov/
topics/genealogy/2000surnames/surnames .

Word, David L., and R. Colby Perkins. 1996. “Building a Spanish
Surname List for the 1990’s—A New Approach to an Old
Problem.” Technical Working Paper No. 13. https://www.
census.gov/population/documentation/twpno13 .

Supporting Information

Additional Supporting Information may be found in the
online version of this article at the publisher’s website:

These tables (some of which are referred to in the text)
provide supplemental information about our data and
models.
Table SI1: Average community demographics by treat-
ment
Table SI2: Logit models for the dependent variable cap-
turing whether housing officials re-sponded. All six treat-
ments included separately
Table SI3: Logit models for the dependent variable cap-
turing whether housing officials responded in 24 hours if
they responded.
Table SI4: Logit models for the dependent variable cap-
turing whether housing officials began responses with a
named greeting if they responded.
Table SI5: Split sample (by race of emailer) logit models
for the email response dependent variable with cities
broken into thirds based on the percent of the population
that is black and the percent Hispanic
Table SI6: Split sample (by race of emailer) logit models
for the named greeting dependent variable with cities
broken into thirds based on the percent of the population
that is black and the percent Hispanic
Table SI7: Split sample (by race of emailer) logit models
for getting a response with indicator variables for special
types of towns (e.g. mainly white or high Hispanic).
Table SI8: Split sample (by race of emailer) logit mod-
els for getting a named greeting with indicator variables
for special types of towns (e.g. mainly white or high
Hispanic).
Figure SI1: Proportion of responses received in 24 hours
(excluding non-responses)

Inequality in Police Service Provision:

Evidence from Residential Burglary Investigations ∗

Rebecca Goldstein†

July

9

,

2

0

20

∗I am grateful to Laura Vittorio, Captain John Leavitt, and Sergeant Aaron Wine of the Tucson Police De-
partment for invaluable assistance with this project. I am thankful for comments from Abhay Aneja, Steve
Ansolabehere, Pamela Ban, David Deming, Ryan Enos, Jonathan Gould, Jennifer Hochschild, Kaneesha John-
son, Mayya Komisarchik, Jens Ludwig, Devah Pager, Deepak Premkumar, Jim Snyder, Jessica Trounstine,
Chris Warshaw, Bruce Western, Hye Young You, and workshop and panel participants at the 20

1

8

American
Political Science Association Annual Meeting, Harvard, Boston University, and UC Berkeley, and for the ex-
cellent research assistance of Jack Demuth, Whitney Driver, and Sean Gerhart. This project received support
from the Pershing Square Fund for Research on the Foundations of Human Behavior and the Multidisci-
plinary Program on Inequality and Social Policy. This manuscript has not been peer-reviewed; please do not
cite without author’s permission

.

†Assistant Professor, UC Berkeley School of Law (Jurisprudence & Social Policy Program). Email: rgold-

stein@berkeley.edu. Website: https://rebeccasgoldstein.com.

1

mailto:rgoldstein@berkeley.edu

mailto:rgoldstein@berkeley.edu

https://rebeccasgoldstein.com

Abstract

When a crime victim calls the police for help, what type of response do they re-

ceive? While scholars have extensively documented racial inequalities in the police’s

punitive functions, this paper considers police as service providers. It leverages uniquely

granular data on over 2,

5

00 residential burglary investigations in Tucson, Arizona to

consider the predictors of investigative thoroughness. Contrary to conventional wis-

dom about police behavior, demographics of victims or officers do not predict inves-

tigative thoroughness. Instead, the most important predictor of investigative thor-

oughness is whether the burglary featured a forced entry into the residence, since

forced entry cases feature more evidence and thus provide greater likelihood of case

clearance. However, the probability of forced entry differs significantly by neighbor-

hood, meaning that the seemingly neutral decision to maximize clearance rates has

unequal consequences.

2

1.

Introduction

When a victim of a crime calls the police for help, what type of response do they receive?

Does the quality of police service provision vary depending on the race, gender, income,

age, or neighborhood of the victim? Scholars have extensively documented racial inequal-

ities in the police’s punitive functions – pursuing, searching, and arresting criminal sus-

pects (Knowles, Persico, and Todd 2001; Beckett, Nyrop, and Pfingst 200

6

; Persico and

Todd 2008; Golub, Johnson, and Dunlap 200

7

; Antonovics and Knight 2009; Gelman, Fa-

gan, and Kiss 2007; Mitchell and Caudy 20

17

) – and traffic ticketing (Baumgartner et al.

2017; West 20

18

; Goncalves and Mello 2020). Much less scholarly attention has been de-

voted to inequalities in police services rendered to victims of crimes or disturbances, de-

spite the fact that there are many more crime victims than there are arrests every year

(Truman and Morgan 20

16

; DOJ 2016).

This paper is the first to use incident-level data to consider whether there are inequal-

ities – based on race, gender, income, age, or neighborhood – in the quality of policing

services rendered to victims of crimes. It finds that demographic characteristics do not

dictate what sorts of burglary investigations victims receive. The main determinant of

investigative thoroughness is instead whether the burglary featured a forced entry into

the residence. Forced entry burglaries, however, are disproportionately concentrated in

better-off neighborhoods of the city, and so even though these neighborhoods’ residents

are not directly discriminated against conditional on their neighborhood, unconditional

social inequalities in service provision remain.

Scholarship analyzing inequality in police service provision is rare in part because

incident-level data is usually difficult to access. In this paper, I leverage a novel and

uniquely granular data set from the Tucson Police Department (TPD). By partnering with

the TPD, I obtained data on every residential burglary in Tucson in calendar year 2016

(over 2,500 burglaries), including information on the nature of the incident; the races and

genders of each civilian and police officer present at each burglary scene; and the activ-

ities officers undertook at the scene. I also use the address information of each incident

to link incidents to American Community Survey data on neighborhood demographic

characteristics. I then compare three measures of service quality for residential burglary

1

investigations – the amount of time spent at the scene, whether or not officers dusted for

fingerprints, and whether or not a burglary detective was ever assigned to the case – to all

of these incident- and neighborhood-level characteristics.

Contrary to conventional wisdom, burglary victims of different races and genders do

not receive different levels of thoroughness in their investigations, and officers of different

races and genders do not provide different levels of investigative thoroughness. Instead,

I find that, conditional on a host of contextual variables, the main determinant of inves-

tigative thoroughness is whether the burglary featured a forced entry into the residence.

Officers spend

11

more minutes (just over 20% more time) at the scene of forced entry

burglaries, they are 66% more likely to dust for fingerprints, and over

4

5% more likely to

assign a detective to the case, conditional on relevant situational and demographic vari-

ables.

Making a causal claim about the relationship between forced entry and investigative

thoroughness is challenging because forced entries are not randomly assigned to resi-

dences. To overcome the potential for selection bias, I employ three sets of fixed effects:

date-level (to account for crime seasonality, police resource differences on weekends and

holidays, and any other unobserved time-varying differences), hour-level (to account for

any differences between daytime and nighttime incidents, between officer work shifts, and

any other unobserved patterns related to time of day), and Census block group-level (to

account for any cross-sectional confounding at the level of the Census block group, such as

neighborhood socioeconomic characteristics). These regressions also include a full battery

of controls for incident-level confounders, including level of urgency at the time of the 911

call (indexed by a dispatcher-assigned priority level), the number and demographics of

the victims present, and the number and demographics of the officers who responded. In

the

Appendix

, I carry out additional tests to address the possibility that racial minorities

call the police only in relatively serious incidents and I find that they do not.

Even though racial minorities are not discriminated against in the quality of investi-

gation that they receive, thoroughly investigated cases are not distributed equally among

Tucson’s neighborhoods. The probability of a forced entry differs significantly by type of

neighborhood and type of residence, and is particularly low in high-poverty areas and

2

multi-unit residences. Forced entry is

10

% less likely in high poverty (

40

% poverty or

more) versus low poverty neighborhoods (p < 0.01), and

15

% less likely in an apartment

versus in a single-family residence (p < 0.01). Unconditional on any background vari-

ables, officers spend 7 fewer minutes (about 8% less time) at residential burglaries that

take place in high poverty compared to low poverty neighborhoods (p = 0.0

3

). Inequal-

ities do exist, then, because of the distribution of forced- and unforced-entry burglaries

across different types of residences.

These findings show how formally neutral criteria can create inequalities in service

provision. Even though TPD officers are using a criterion (forced entry) that is facially

neutral with respect to race and socioeconomic class to decide where to focus their inves-

tigation efforts, unconditional social inequalities are nonetheless present.

Although these results run contrary to a large amount of research about the racially

discriminatory way police behave in their punitive roles (Gelman, Fagan, and Kiss (2007);

Braun, Rosenthal, and Therrian (2018)), they are consistent with the theoretical predic-

tion that inequalities in public service provision will emerge only for substitutable public

goods (Besley and Coate

19

91). Furthermore, although economic analysis of crime has

traditionally focused on the incentives of would-be offenders (Becker 1968; Freeman 1999;

Pinotti 2017), these results are consistent with research showing that police officers, too,

respond to incentives (in their case, the performance-based incentives used to evaluate

police officer performance (Mas 2006; Carpenter 20

14

)).

This paper makes several contributions to the research literature. First, it contributes

to the scholarly understanding of racial bias in policing, which has so far focused entirely

on the police’s punitive activities and roles, even though crime victims significantly out-

number arrestees (Truman and Morgan 2016; DOJ 2016). Police not behaving in a racially

discriminatory manner in their service provision role considerably complicates our un-

derstanding of race and policing in light the large literatures on racial bias in traffic stops,

pedestrian stops, and arrests (Golub, Johnson, and Dunlap 2007; Goel et al. 2016; Ritter

2017; Braun, Rosenthal, and Therrian 2018). Second, the fact that there are not condition-

ally worse quality burglary investigation services provided to low-income, heavily mi-

nority neighborhoods provides empirical support for the classic Besley and Coate (1991)

3

model, which predicts that inequalities in public service provision will emerge only for

substitutable public goods. Third, the unconditional inequality in thoroughness of bur-

glary investigations provided in poorer and wealthier neighborhoods illustrates how ser-

vice provision inequalities can emerge out of a socioeconomically neutral bureaucratic

incentive structure. Finally, the fact that police officers exercise most effort on cases that

they are most likely to solve provides evidence that street-level bureaucrats, just like other

bureaucrats, are focused on maximizing their performance on observable and quantifiable

measures which influence their reputations, and in consequence, their future professional

prospects (Holmström 1999; Carpenter 2014) – in short, that police officers respond to in-

centives in choosing where to focus their efforts, just as members of the public respond to

incentives in choosing whether, when, and where to commit crimes (Di Tella and Schar-

grodsky 2004).

The remainder of this paper proceeds in four parts. First, I discuss what existing theo-

retical and empirical research on public service provision, bureaucratic behavior, inequal-

ity, and policing would lead us to expect about where police are likely to direct inves-

tigative resources. Second, I describe the policing context in Tucson and burglary inves-

tigations in general. Third, I present results from the analysis of over 2,500 residential

burglaries. Finally, I draw conclusions and suggest directions for future research.

2.

Policing: Public goods, bureaucratic behavior, and inequality

What does the literature on public goods provision suggest about how police are likely to

allocate investigative resources? The classic Besley and Coate (1991) model predicts that

inequalities in public services will be greatest where it is possible for wealthy citizens to

purchase higher-quality private goods to substitute for lower-quality public goods. In-

deed, urban public services, such as education and housing, feature extreme racial and

socioeconomic inequalities in part for this reason (Duncan and Murnane 2011; Krivo and

Kaufman 2004), and police-provided services which are privately substitutable, such as

community security services, are sometimes replaced by private services in wealthy neigh-

borhoods (Trounstine 2015). But it is hard to imagine a private substitute for burglary

investigation services, and so the Besley and Coate model would not predict racial or

4

socioeconomic inequalities to emerge in their provision. Indeed, other non-substitutable

police services, such as emergency response, in fact feature much greater resources de-

voted to high-crime, high-poverty neighborhoods (Walker, Spohn, and DeLone 20

12

; Ci-

han, Zhang, and Hoover 2012).

The scholarly literature on crime as a market-driven phenomenon focuses primarily

on the incentives of would-be offenders (Ehrlich 1973; Kelly 2000; Pinotti 2017). But the

level of crime is driven not just by the behavior of would-be offenders – police officer

behavior also shapes the equilibrium level of crime (Persico 2002; Chalfin and McCrary

2018; McCrary and Premkumar 2019). Existing economic analysis of officer incentives

has attempted to explain well-known patterns of police bias against racial minorities in

the context of arrests, pedestrian stops, and traffic stops (Persico 2002; West 2018; Braun,

Rosenthal, and Therrian 2018; Goncalves and Mello 2020). A smaller literature describes

the incentives of police officers, often with a focus on opportunities that the police occa-

sionally have to directly collect money for their agencies through seizures of cash from

criminal enterprises (Benson, Rasmussen, and Sollars 1995; Mast, Benson, and Rasmussen

2000; DeAngelo, Gittings, and Ross 2018).

But these analyses examine the police only in their role as pursuers of criminal sus-

pects, not in their role as service providers to witnesses and victims. Accusations of po-

lice bias against racial minorities, though, often include descriptions of police who do

not serve minority community members effectively when those community members are

crime victims (Natapoff 2006; Leovy 2015). Just as scholars have aimed to detect the

presence of racial bias in police stopping and arresting activity (Knowles, Persico, and

Todd 2001; Grogger and Ridgeway 2006; Persico and Todd 2006; Mitchell and Caudy 2017;

Goncalves and Mello 2020), this paper aims to detect any such parallel bias in police ser-

vice provision activity.

In exploring this question, a foundational premise is that police are incentivized – just

as other bureaucrats are – to be seen as high-performing in the eyes of their superiors

(Lipsky 1980). Theory and empirical evidence on incentives in bureaucracies would pre-

dict that police officers will be incentivized to focus their efforts on activities which con-

tribute to achievement on the performance measures on which they are primarily eval-

5

uated (Niskanen 1971; Sigelman 1986; Carpenter and Krause 2012; Lemos and Minzner

2014). The most basic measure of police performance is the clearance rate, the number of

crimes that are “cleared” divided by the total number of crimes (Mas 2006; Rayman 20

13

).

The advent of data-driven policing that is directly targeted at lowering the rate of crime

and increasing the rate of crime clearance has made clearing crimes especially important

for police leadership’s evaluation of police rank-and-file and for city leadership’s evalu-

ation of police leadership (Rayman 2013). Police leadership care about department (and

thus officer) clearance rates because bureaucrats accrue influence and power within and

on behalf of their agencies by maintaining a reputation as highly competent and impar-

tially technocratic – evaluations that are typically also made using observable performance

measures (Carpenter 2014).

For these reasons, research on incentives in bureaucracies would predict police to fo-

cus on cases with the highest probability of clearance. With respect to burglaries, crimi-

nology research distinguishes between forced entry burglaries (such as those involving a

picked lock or broken window) and unforced entry burglaries (such as those involving en-

try through a door or window left open). Forced entry burglaries provide more analyzable

evidence are thus more solvable as compared to unforced entries (Coupe 2016; Shannon

and Coonan 2016; Killmier, Mueller-Johnson, and Coupe 2019). The incentive structure

that police face in investigating burglaries, then, can be expected to encourage them to

focus scarce investigatory resources on forced entry burglaries.

Several other areas of literature suggest the contrary conclusion: that there would be

inequality in police service provision. But existing work is not conclusive on the ques-

tion, and existing lines of work do not necessary translate to inequalities across service

provision activities.

First, and most directly, the little existing work on police service provision provides

evidence of inequalities in outcomes related to service provision. That research shows

that police are less likely to clear homicides when the victim is Black or Hispanic than

when the victim is white (Roberts and Lyons 2011; Fagan and Geller 2018). While some

scholars and journalists believe this is due to police’s deliberate lack of effort in investi-

gating homicide cases with Black and Hispanic victims (Leovy (2015)), others argue that

6

this disparity is more due to a lack of cooperation with homicide investigators in commu-

nities where police-community relations are strained (Roberts 2015; Mancik, Parker, and

Williams 2018). These studies illustrate the distinction between investigative thorough-

ness and clearance: if factors beyond investigators’ control largely determine clearance

rates, clearance is not an accurate reflection of investigative thoroughness. One advan-

tage of the present study is that I measure investigative thoroughness directly, without the

complications inherent in using clearance as a proxy for thoroughness.

Second, the literature on inequalities in other areas of policing show significant dispar-

ity in police treatment of suspects of different races. Nonwhites are stopped (in vehicles

and on foot) more than whites, nonwhite suspects more likely than white suspects to be ar-

rested for similar crimes, and nonwhite suspects are more likely to report experiencing ag-

gressive or unfair treatment by police (Carr, Napolitano, and Keating 2007; Walker, Spohn,

and DeLone 2012; Epp, Maynard-Moody, and Haider-Markel 2014). At least some of these

inequalities may be rooted in well-documented, unconscious associations between Black-

ness and criminality (Eberhardt et al. 2004).

But this literature is focused solely on the police’s roles stopping suspicious pedes-

trians and vehicles or arresting criminal suspects, rather than on their roles as service

providers. Inequalities in one context might not translate to the other. A key feature of

pedestrian and vehicle stops is that those contexts require police to make nearly instant

judgments about an individual’s likelihood of possessing drugs or weapons. A wealth of

psychological evidence shows that unintentional biases are magnified under “snap judg-

ment” decision conditions (Payne 2006; Freeman and Johnson 2016). Another situational

characteristic that can magnify the reliance on heuristics such as racial stereotypes is anx-

iety. When police are stopping persons and vehicles in search of contraband, they tend to

be alert to potentially dangerous situational developments (Woods 2018). Psychological

studies show that anxiety inhibits typical information processing and can increase the use

of stereotypes in decision making (Wilder 1993; Hilton and Von Hippel 1996; Hamilton

and Sherman 2014).

Third, and most broadly, other public services feature extreme racial and socioeco-

nomic inequalities. If police investigative services are understood as a public service, it

7

is reasonable to think that it – like many other public services – might be distributed un-

equally on the basis of race and/or socioeconomic class, although the absence of a private

equivalent to many police services (including burglary investigations) provides reason to

think that inequalities of the sort present in the context of education, housing, and other

public goods will not apply to the service provision context.

In summary, literature on public goods provision would lead us not to expect racial or

class inequalities in the provision of non-substitutable public goods, including residential

burglary investigation. Literature on incentives in bureaucracies would lead us to expect

police officers to direct resources in ways that promote their success on the performance

measures on which they are evaluated – here, attempts to clear as many crimes as possible

would lead police to focus on forced entry burglaries. But the literatures on racial disparity

in homicide clearance, racial discrimination by police officers, and inequalities in urban

public service provision would lead us to expect that race and class inequalities will exist

in police service provision as well.

3.

Data: Policing in Tucson, Arizona

The primary data for this project are detailed records for every residential burglary that

took place in Tucson, Arizona in calendar year 2016. Burglary is defined by Arizona statute

as “entering or remaining unlawfully in or on a residential structure with the intent to

commit any theft or any felony therein” (A.R.S. §13-1507). The data includes the street

address of the residence, the division (one of four geographic areas) in which it took place,

the priority level that the police dispatcher placed on the call,1 the time the call was placed,

the time that officers arrived, the time that officers left, the number and ages, races, and

genders of everyone at the scene, and detailed information on the nature of the incident

and the activities undertaken by the officers at the scene. Officers also record how many

officers were present at the scene and their badge identification numbers, along with how

many victims were present at the scene, and their races and genders. The TPD provided

employee records information which allowed me to merge in information on the ages,

1The priority levels range from 1, meaning an emergency call for which dispatched officers would use

lights and sirens in arriving, to 4, meaning lowest priority.

8

races, and genders of the officers present at each scene. The TPD system later adds infor-

mation on whether a detective was eventually assigned to the case, which is a decision

made by the burglary sergeant in the division in which the burglary took place.

For purposes of comparability of incidents, I exclude any residential burglary which

was initially reported as a larceny (a theft which does not involve an unlawful entrance)

and is later re-coded as a burglary based on further investigation, since the initial phase

of a larceny investigation is much less thorough than that of a burglary investigation. I

also exclude incidents reported using the TPD web interface rather than a phone call or an

alarm system, since they reflect a much lower level of urgency on the part of the individual

in need of police service. These exclusions leave 2,771 burglaries.2 All these burglaries

needed to be addressed at least partially in the year’s

36

5 days by Tucson’s 870 sworn

officers, reflecting a very high level of capacity constraint. These burglaries are shown at

the locations they occurred in Figure 1.

The criteria I use to define a thorough burglary investigation are (1) the amount of time

police spent at the scene, (2) whether the police dusted for fingerprints, and (3) whether a

detective was eventually assigned to the case. Conversations with TPD researchers, patrol

officers, burglary detectives, and burglary sergeants revealed that the standard practice at

all burglary scenes is to conduct a thorough interview with the victim or victims and to

dust for fingerprints in the location where fingerprints are most likely to have been left

by the perpetrator (which depends on the victim’s description of missing items and the

perpetrator’s likely point of entry). Other potential measures of thoroughness, such as

whether or not DNA was collected, whether full lists of stolen items were recorded, or

whether the police recorded stolen items’ serial numbers, were not widely applicable to

all cases: DNA is only collected if the perpetrator appears to have left a sample of bodily

fluid, serial numbers are only recorded for items that have serial numbers (such as high-

value electronics and firearms), and full lists of stolen items are sometimes not possible to

record at a scene because victims are often too distressed to know exactly what has been

taken. Sergeants assign detectives based primarily on their perception of the solvability

of the case. In practice, sergeants view cases as solvable when there is potential physical

2The FBI’s Uniform Crime Reporting data for Tucson in 2016 records 4,1

38

burglaries; the 2,771 in the final

data are only residential (rather than commercial) burglaries that meet the above criteria.

9

evidence which prosecutors could later use to definitively link a suspect to a burglary.

In classifying burglaries as forced or unforced entries, I follow both TPD and the FBI’s

Uniform Crime Reporting (UCR) Program, both of which distinguish between the two

types of burglaries (DOJ 2014). TPD classify as forced entry any burglaries in which the

perpetrator entered the residence through a locked door or window, whether by breaking

the window or door, by using any tool to force or pry open the window or door, or by

using a tool to disable the lock on a window or door. In many cases, a forced entry will

be obvious, as it has caused damage to the window or door and/or the surrounding area

of the residence. In some cases, if there are no signs of forced entry, patrol officers need to

decide whether the victim is being truthful in their account that they locked all doors and

windows, and the entry is classified as forced if officers believe the victim’s account and

unforced if they do not.

In addition to the police-provided information about each burglary, I linked the ad-

dresses of the burglaries to American Community Survey (ACS) data (2011-2015 esti-

mates) on Census block group demographics. Census block groups are the smallest level

of aggregation at which the Census Bureau reports demographic data; in dense urban ar-

eas, they typically correspond to individual square city blocks. The average population

of a Census block group in Tucson is 1,540. Summary statistics on the 2016 residential

burglaries and the Census block groups in which they took place is presented in Table 1.

Tucson is an ideal setting for this study because of the very high level of racial diversity

(47% white non-Hispanic, 42% Hispanic, 5% Black, 3% Native American, and 3% Asian,

as of the 2010 Census) as well as socioeconomic diversity (

25

% of persons in Tucson live

in poverty as of 2015, and the median household income was below the national median

at $

37

,149) means that there is a great deal of variation in types of neighborhoods and

individuals that the police serve.

As Figure 1 and Figure 2 show, burglaries are more prevalent in denser and poorer

areas of the city. The retirement communities of the west part of Tucson and the wealthy

neighborhoods of the most northern part of Tucson experienced few residential burglaries

in 2016. Burglaries were most concentrated in the poor neighborhoods immediately east

of Interstate 10, which divides the city’s eastern and western parts.

10

Table 1 displays summary statistics for key variables of interest. Where the number of

observations is less than 2,771, the relevant data is missing for some of the burglaries. Of-

ten this is due to officers not recording that information in their narrative reports. Missing

data on time spent at the scene is a result of patrol officers editing their narrative reports

after submitting them, which causes the time to default to zero, and failing to manually

input the time stamp again.

4.

Results and Discussion

4.1. Race and investigative thoroughness

Table 2 presents the results of OLS regressions predicting the three indicators of inves-

tigative thoroughness. Columns (1), (3), and (5) display the coefficients on the share of

white officers and victims at a scene when predicting the log of time spent at the scene

(in minutes), and the binary probability of the police collecting fingerprints and assign-

ing a detective to the case, without any additional controls. Columns (2), (4), and (6) add

fixed effects for TPD’s four (geographic) police divisions, to control for differences in ca-

pacity between different divisions, and fixed effects for month of the year, to control for

the basic seasonality of crime patterns. This minimum set of controls, especially in the

odd-numbered columns, illustrates that the comparatively large standard errors on the

coefficients for share of white officers and share of white victims are not an artifact of

collinearity in the control variables (Chatterjee and Simonoff 2013). Other incident- and

victim-level variables included in Table 3, such as 911 call priority level, the indicator that

the residence is an apartment, the number of victims, the number of officers, and the age

of victims, are omitted here because they are plausibly post-treatment with respect to the

race of the victims.

Table 2 shows that the relationships between the share of white victims or officers and

investigative thoroughness is substantively small, and for the most part not statistically

significant at conventional levels, even when only a minimum set of control variables are

included in the regression. Furthermore, the signs on the significant coefficients reflect a

result contrary to the conventional wisdom about race and policing: that police officers

spend slightly less time with groups of victims that are more heavily white (specifically,

11

that they spend about 7% less time, or 4 fewer minutes, with all white groups of victims

compared to all nonwhite groups of victims).

These findings on race and investigative thoroughness – despite being “null” in the

frequentist sense – are enormously informative in the Bayesian sense. Abadie (2018) ar-

gues that readers of scientific journals are agents in a limited-information Bayesian setting.

Readers have a prior belief about the value of an estimand which the author seeks to esti-

mate, and an author’s result is informative if “it has the potential to substantially change

the [prior] beliefs of the agents” – that is, the result is informative if the prior and pos-

terior distributions of the estimator substantially differ. Abadie shows that, even when

prior distributions are diffuse, the prior and posterior distributions of the estimator very

often differ more when the result is not statistically significant than when the result is

statistically significant, especially in large samples.

In the present case, even using a diffuse prior distribution (the standard normal prior),

the prior probability of rejection of the null hypothesis of zero mean difference is extremely

high (0.97). This fact further emphasizes how informative it is to fail to reject the null hy-

pothesis of zero difference in investigation thoroughness between burglaries with different

shares of white victims at the scene. Details of this calculation are included in Appendix

B.

This result is also extremely surprising in the context of existing research on racial bias

in policing. The bulk of existing research on this topic3 uses data on either pedestrian stops

or traffic stops, or on police-involved fatalities, to assess the extent of racial bias. Most

of the studies of pedestrian stops find that racial minorities are more likely than whites

to be stopped (Gelman, Fagan, and Kiss 2007) and that officers appear to use a lower

threshold of suspicion when deciding to stop nonwhites compared to whites (Goel et al.

2016), although some find limited evidence of bias in arrests conditional on stops (Coviello

and Persico 2015). Previous studies of traffic stops also mostly find significant bias against

minority drivers; for example, Baumgartner et al. (2017) uses data on over 18 million traffic

stops made in North Carolina between 2000 and 2016 and finds that, conditional on all

3The literature on racial bias in policing is vast and spans many disciplines in the social sciences and

humanities, but here I am referring to quantitative, empirical research using police administrative records,

rather than qualitative, game theoretic, or lab experimental research.

12

available background variables, Black drivers were significantly more likely to be stopped

and significantly less likely to be found with contraband compared to white drivers. Most

studies of police-involved fatalities also find that African-Americans are more likely than

whites to be killed by the police (Ross 2015; Knox, Lowe, and Mummolo 2019), although

agreement on this score is not universal (Fryer Jr 2019).

4.2. Forced entry and investigative thoroughness

Table 3 presents the results of fixed effects regressions predicting the three indicators of

investigative thoroughness. Columns (1), (3), and (5) show the coefficient of the binary

variable indicating a forced entry on the log of minutes spent at the scene, on the probabil-

ity of prints having been taken, and on the probability of a detective having been assigned,

with no additional controls. Columns (2), (4), and (6) make full use of the rich microdata

by employing controls for all available victim, officer, and incident characteristics, along

with three sets of fixed effects: block group, calendar date, and hour of the day.

Comparing the odd-numbered and even-numbered Columns reveals that, while some

of the variation that is explained by forced entry is explained by the battery of victim-

, incident-, and officer-level controls, and by the three sets of fixed effects, forced entry

nevertheless has a statistically significant and substantively large effect on investigative

thoroughness even when all of these variables are taken into account. The estimates in

Columns (2), (4), and (6) reflect that officers spend

21

% more time at the scenes of forced

entry burglaries, are 18 percentage points more likely to dust for fingerprints, and are 12

percentage points more likely to assign a detective to the case when the burglary featured

a forced entry compared to an unforced entry, conditional on all of the available situational

and demographic variables, and on block group, date, and hour fixed effects. The means

of these three dependent variables for unforced entry cases are 55 minutes,

27

% likelihood

of collecting fingerprints, and 25% likelihood of assigning a detective, and so these coeffi-

cients represent officers spending 11 more minutes at the scene, along with a 66% increase

in the likelihood of collecting fingerprints and a 47% increase in the likelihood of assigning

a detective.

The key identifying assumption for a causal interpretation of the relationship between

13

forced entry and investigative thoroughness is that within Census block group, calendar

date, and hour of the day, forced and unforced entry are randomly assigned to residences.

This assumption is plausible for three reasons, corresponding to the three sets of fixed ef-

fects. First, Census block groups are small: in Tucson, on average, they are composed of

just 1,540 individuals, or 616 households. The assumption that these fixed effects capture

neighborhood-level variation which would be related to the incidence of forced versus

unforced entry (such as the residences’ level of physical security or any other socioeco-

nomic characteristics) is thus not a strong one. Second, crime and police capacity are

time-varying in many ways: according to the seasons, according to whether it is a week-

day, a weekend day, or a holiday, and even, some have argued, according to the timing

of public benefits distribution (Foley 2011). Fixed effects for the individual calendar day,

then, account for all of that variation and any other unobservable time variation at the

level of the day or longer. Finally, crime and police capacity vary greatly according to the

time of day – because of light and darkness, because of when individuals are likely to be in

their homes versus in their workplaces, and because of changing police shifts. Allowing

effects to be estimated within the specific hour of the incident, then, accounts for all of this

variation.

These results are consistent with the large body of literature on bureaucratic decision-

making argues bureaucrats are focused on maximizing their performance on observable,

quantifiable measures which influence their reputations, and in consequence, their future

professional prospects (Carpenter (2014)). These results conform to a model of police of-

ficers as bureaucrats maximizing their main performance measure – clearance rates – just

as other types of bureaucrats seek to maximize their own performance measures Balla and

Gormley Jr (2017); Brodkin (2008). An officer, in other words, reasons that a forced en-

try is more likely to result in a cleared case than is an unforced entry, regardless of the

demographics of the victim.

4.3. Unconditional inequality in service provision

Although Tucson residents receive similar quality burglary investigations conditional on

whether the burglary featured a forced or unforced entry, they do not receive similar qual-

14

ity burglary investigations overall. Service provision is unconditionally unequal.

Because the regression models in Table 3 account for many types of variation that could

affect investigative thoroughness, the results mask neighborhood- and residence-level dif-

ferences in the incidence of forced and unforced entry. In particular, there are marked

unconditional differences in the rates of forced and unforced entry in different types of

residences and different types of neighborhoods; Table 4 shows that apartments, high

poverty neighborhoods, and high renter share (more than two-thirds rented residences)

neighborhoods are over-represented among unforced entry cases and under-represented

among forced entry cases (forced entry cases are 65% of overall cases).

Although it is impossible to tell with certainty why this is the case, the clearest expla-

nation is that landlords on the private rental market have much less incentive to invest in

security measures for their properties than owners of their own residences do (Hamilton-

Smith and Kent 2005). In addition, Tucson burglary officials speculated that in multi-

family residences it is easier to case many residences quickly for unlocked doors or win-

dows, and there is an economy of scale for a burglar in learning a way in to a housing

complex with a large number of units. This mechanism would be consistent with crimino-

logical literature which finds that event dependence for repeat and near-repeat burglaries

(that is, burglaries which occur more than once within a year at the same residence or

within the same small cluster of residences) is greater where the distance between homes

is smaller (Short et al. (2009)). Taken together, this evidence demonstrates that officers’

seemingly neutral decisions to attempt to maximize clearance rates in fact has distribu-

tional consequences.

15

5.

Conclusion: Bureaucratic Incentives, Resource Allocation, and Inequality

Policing, like other public services, involves the provision of services to a diverse set of

residents with differing interests and needs. Local police departments provide a public

service to residents by responding to calls for service. Scholars of policing in many disci-

plines have focused extensively on racial inequalities in the police’s punitive functions –

pursuing and arresting criminal suspects (Knowles, Persico, and Todd 2001; Persico and

Todd 2006; Grogger and Ridgeway 2006; Golub, Johnson, and Dunlap 2007; Persico and

Todd 2008; Beckett, Nyrop, and Pfingst 2006; Antonovics and Knight 2009; Gelman, Fagan,

and Kiss 2007; Mitchell and Caudy 2015; Legewie 2016; Horrace and Rohlin 2016; Baum-

gartner et al. 2017). And research that leverages incident-level data to measure inequalities

focuses on either traffic ticketing patterns (DeAngelo and Owens 2017; West 2018, 2019)

or racial patterns in criminal victimization (DeAngelo, Gittings, and Pena 2018). These

methods have not, however, been previously used to explore inequalities in police ser-

vices rendered to victims and witnesses of crimes or disturbances, despite the crucial role

of victim service provision in the police function (Goldstein 1977; Friedman 2020).

In this paper, I show that police officers in Tucson, Arizona primarily use a simple

rule – whether or not the burglary featured a forced entry into the residence – in decid-

ing how thoroughly to investigate a residential burglary. Conditional on basic seasonal

controls, demographic characteristics of victims and officers are not significantly associ-

ated with additional investigative thoroughness, but officers do devote greater investiga-

tive resources to forced entry burglaries. Officers spend 11 more minutes at the scenes of

forced entry burglaries, they are 18 percentage points more likely to dust for fingerprints,

and they are 12 percentage points more likely to eventually assign a detective to the case

when the burglary featured a forced entry, conditional on all of the available situational

and demographic variables, and on block group, date, and hour fixed effects.

Are these results likely generalizable to cities other than Tucson? While it is impossible

to say for certain, residential burglaries take place in every city and investigating them is

a core component of police training across the United States (Weisel 2002; Antrobus and

Pilotto 2016). Clearance rates, too, are a key metric of police performance in every depart-

ment (Mas 2006; Sonnichsen 2007; Pare 2014). For these reasons, it is not unreasonable

16

to expect the police incentive to clear residential burglaries to shape police behavior in

many cities. Further research is necessary to determine how the clearance incentive in-

teracts with other incentives – including, perhaps, the incentive to provide more racially

or socioeconomically advantaged citizens with higher-quality public services (Keefer and

Khemani 2005).

The findings in this study beg the question: why do street-level bureaucrats exhibit

different levels of bias in different contexts? As applied to police officers: why they might

they be less likely to discriminate against racial minorities in the residential burglary in-

vestigation context than they are in the vehicle stop, pedestrian stop, and use-of-force

contexts? Three areas of literature provide plausible answers.

A first hypothesis requiring further research, for which this paper provides prelim-

inary support, is that police behave differently when providing substitutable and non-

substitutable public goods. The article’s result is consistent with the famous Besley and

Coate (1991) prediction that inequalities in public service provision will be greater for non-

substitutable public goods. Not all police-provided services are non-substitutable: private

patrol is directly available to citizens from private firms (Trounstine 2015), and wealthy

citizens routinely call on civil lawyers or therapists for mediation and arbitration services

for which poorer citizens routinely call the police (Friedman 2020). Residential burglary

investigation, by contrast, is non-substitutable, and so this article’s finding that racial and

socioeconomic inequalities are not present, conditional on contextual variables, is consis-

tent with Besley and Coate (1991).

Second, police are incentivized to maximize performance on measurable performance

metrics, in this case the crime clearance rate. This incentive is described by with the classic

characterization of police as street-level bureaucrats (Lipsky 1980; Hill 2003), the relatively

low-level officials who exercise enormous discretion in allocating public resources. Lipsky

(1980) observed that street-level bureaucrats tend to focus on cases or clients that are the

easiest, the most straightforward, or are the most likely to lead to a positive outcome for

which they can take credit – a phenomenon he calls “creaming.” “Confronted with more

clients than can readily be accommodated,” Lipsky writes, “bureaucrats often choose (or

skim off the top) those who seem most likely to succeed in terms of bureaucratic success

17

criteria.” This insight raises the possibility that discrimination by street-level bureaucrats

may be more prevalent in contexts where the cases which are administratively simplest to

handle are also the cases where clients are more racially or socioeconomically advantaged.

Third, different contexts implicate different psychological responses. Burglary investi-

gations can be done slowly and methodically, and victims of residential burglaries are not

perceived threats to officer safety. This contrasts with the pedestrian and vehicle stop con-

texts, in which many officers are trained to be alert to the possibility that civilians might

be armed or otherwise dangerous (Woods 2018). We might therefore expect officers to

rely less on racial stereotypes when carrying out a residential burglary investigation than

when stopping persons or vehicles.

Each of these hypotheses open the door for further research on the determinants of

discrimination – or the lack thereof – by police in different context. A focus on the nature

of the goods that police provide, the incentives they face, and the psychological pressures

they encounter can provide us with a more textured view of police discrimination.

Finally, the distributional consequences illustrated here represent an example of biased

outcomes emerging from neutral rules (Jolls (2001)). Even if and when public officials do

not intend to discriminate, the incentive structure that they face can lead to disparate im-

pacts on disadvantaged communities. Outside the policing context, public officials are

incentivized to locate hazardous waste or other undesirable sites on cheap land or in com-

munities unable to mobilize opposition, and these incentives can lead to undesirable sites

being located in minority communities (Foreman 2011; Sherman 2012). Public officials

are similarly incentivized to hire government contractors with relevant experience or who

promise to deliver a project at low costs, but even seemingly neutral criteria such as these

can have disparate impacts on minority contractors (Sonn 1992). And public universi-

ties are incentivized to admit students with high standardized test scores, since having

students with higher test scores increases their ranking and prestige, but reliance on test

scores can have a disparate impact on minority and female applicants (Jencks and Phillips

2011). In each of these instances, a legitimate goal, and a legally unobjectionable incentive

structure, can lead government officials or entities to act in ways that replicate existing

racial, social, and economic inequalities.

18

6. Appendix

6.1. Appendix A: addressing selection

One reason to believe that racial minorities may be calling about more serious burglaries

would be if they place a disproportionate share of high priority burglary calls. The priority

level fixed effects which feature in the main analysis should eliminate the possibility that

any differences of this sort would bias the results, but Appendix Figure A1 below confirms

that whites do not place a disproportionate share of low priority calls relative to their share

of the Tucson population.

Figure A1: Share white among victims and in the Block group at each priority level

Appendix Table A1 shows that, even when call priority level is not taken into account,

the demographics of the victims and the neighborhood are not significant predictors of

investigative thoroughness. This suggests that call priority level, even though it bears a

statistically significant relationship to investigative thoroughness in the full model, is not

explaining any of the same variation in thoroughness outcomes as victim or neighborhood

demographics.

19

Table A1: Predicting investigative thoroughness without priority level

Dependent variable:

Time (mins) (log) Prints collected Detective assigned

(1) (2) (3)

Forced entry 0.186∗∗∗ 0.175∗∗∗ 0.116∗∗∗

(0.045) (0.0

26

) (0.025)
Share white officers 0.019 0.020 0.0

28

(0.051) (0.0

29

) (0.0

30

)
Share white victims −0.049 −0.036 −0.010

(0.044) (0.027) (0.026)
Apartment −0.059 −0.044 −0.098∗∗∗

(0.054) (0.028) (0.027)
Victim ages (mean) −0.0001 −0.001 −0.001

(0.001) (0.001) (0.001)
Share male officers 0.042 0.046 0.061∗

(0.063) (0.0

32

) (0.0

33

)
Share male victims −0.005 −0.015 0.0

24

(0.047) (0.026) (0.025)
Total victims −0.006 0.030∗ 0.028

(0.025) (0.016) (0.018)
Total officers 0.120∗∗∗ 0.010 0.0

35

∗∗∗

(0.011) (0.007) (0.005)
East 0.

31

9 0.151∗ 0.036

(0.198) (0.085) (0.099)
West −0.027 0.033 −0.041

(0.094) (0.056) (0.060)
Midtown 0.

34

1∗ 0.042 0.020

(0.193) (0.080) (0.089)
Constant 4.080∗∗∗ −0.748∗∗∗ −0.273

(0.459) (0.176) (0.193)

Block group fixed effects? Yes Yes Yes
Date fixed effects? Yes Yes Yes
Hour fixed effects? Yes Yes Yes
Observations 2,508 2,641 2,641
R2 0.424 0.354 0.355

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Robust standard errors are clustered at the Block

group level and are included in parentheses.
Omitted division category: South.

20

An ordered probit regression predicting priority level, presented in Appendix Table A2,

confirms that there is no statistically significant relationship between victim or neighbor-

hood demographics and assigned priority at conventional levels. Block group fixed effects

are omitted from this regression in order to see what, if any, relationship there is between

Block group demographics and priority level. Police division fixed effects are included to

account for some geographic variation.

Table A2: Predicting priority level (ordered probit regression)

Dependent variable:

Priority level (1-4)

Forced entry −0.494∗∗∗
(0.128)

Share Hispanic victims −0.214
(0.154)

Share Black victims −0.268
(0.272)

Hispanic share (Block group) −0.532
(0.342)

Black share (Block group) −1.815
(1.251)

Poverty share (Block group) −0.437
(0.479)

Renter share (Block group) 0.714∗∗

(0.332)
Apartment 0.491∗∗∗

(0.142)
Victim ages (mean) 0.003

(0.004)
Share male victims 0.542∗∗∗

(0.130)
Total victims −0.197∗∗∗

(0.069)
East −0.525∗∗

(0.243)
West −0.621∗∗∗

(0.212)
Midtown −0.703∗∗∗

(0.

22

7)

Date fixed effects? Yes
Hour fixed effects? Yes

Observations 2,556

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Omitted division category: South.

21

Another reason to suspect that nonwhites are being discriminated against relative to

whites in cases of the same seriousness would be if nonwhites and whites received dif-

ferent treatment in the subset of forced entry cases. Table A3, though, shows that victim

and officer demographics are not predictive of thoroughness in the subset of forced entry

cases.

Table A3: Predicting investigative thoroughness among forced entry cases

Dependent variable:
Time (mins) (log) Prints collected Detective assigned
(1) (2) (3)

Share white officers 0.005 0.012 0.021
(0.070) (0.045) (0.048)

Share white victims −0.072 −0.037 −0.055
(0.061) (0.041) (0.042)

Apartment −0.045 −0.024 −0.124∗∗∗
(0.072) (0.048) (0.044)

Victim ages (mean) 0.002 −0.0004 −0.001
(0.002) (0.001) (0.001)

Share male officers 0.066 0.046 0.054
(0.080) (0.048) (0.049)

Share male victims 0.044 0.001 0.060
(0.069) (0.0

39

) (0.040)

Total victims −0.009 0.020 0.039
(0.042) (0.026) (0.025)

Total officers 0.110∗∗∗ 0.003 0.031∗∗∗

(0.017) (0.009) (0.007)
East 0.520 0.220 0.142

(0.371) (0.150) (0.152)
West −0.027 0.071 0.011

(0.154) (0.125) (0.115)
Midtown 0.526 0.087 0.153

(0.359) (0.144) (0.164)
Constant 5.707∗∗∗ −0.830∗ −0.765∗

(0.775) (0.430) (0.459)

Block group fixed effects? Yes Yes Yes
Date fixed effects? Yes Yes Yes
Hour fixed effects? Yes Yes Yes
Observations 1,618 1,705 1,705
R2 0.554 0.469 0.467

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Robust standard errors are clustered at the Block group level and are included in parentheses. Omitted division category: South.

Finally, if residents of heavily minority Tucson neighborhoods were calling the police

less than residents of heavily white neighborhoods, that would provide suggestive evi-

dence that racial minorities were under-reporting crimes (or under-reporting some types

of crimes) to the police. To test this, I downloaded publicly available data on every 911

22

call made in Tucson in 2016. These calls are geo-located to the level of the 100-block. I

used GIS to match these locations to Census Block groups and merged in demographic

information from the American Community Survey. I included only calls that came from

citizen telephones (landlines and cellphones) and excluded officer-originated 911 calls. In

the regression table that follows, the unit of analysis is the Block group, the dependent

variables are the logged total number of 911 calls received in 2016, the logged number of

calls regarding burglaries,4 and the logged number of calls regarding suspicious persons.

I chose these latter two types of calls because they reflect quite serious and quite minor

crime concerns, respectively.

Table A4: Predicting the number of 911 calls in 2016 (Block group level)

Dependent variable:

Total 911 (log) Burglary (log) Suspicious person (log)

(1) (2) (3)

Share Hispanic −0.112 −0.224 −0.306
(0.283) (0.212) (0.

23

8)

Share Black 2.638∗∗ 1.291 1.795∗

(1.129) (0.845) (0.950)
Share other 0.189 0.286 0.263

(1.072) (0.802) (0.902)
Share 18 to 24 0.230 1.006∗∗ 0.151

(0.577) (0.432) (0.486)
Share 65 or over −1.946∗∗∗ −1.550∗∗∗ −1.510∗∗∗

(0.585) (0.438) (0.492)
Median HH income (log) −0.658∗∗∗ −0.527∗∗∗ −0.551∗∗∗

(0.153) (0.114) (0.129)
Total population (log) 0.110 0.155∗ 0.166∗

(0.115) (0.086) (0.097)
Constant 12.155∗∗∗ 6.996∗∗∗ 8.166∗∗∗

(1.755) (1.313) (1.477)

Observations 410 410 410
R2 0.161 0.209 0.143

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Appendix Table A4 shows that the strongest negative determinants of 911 call volume

are the share of the Block group over 65 years old and the median household income in

the Block group. The racial makeup of the Block group is mostly not associated with

911 call volume at conventional significance levels, although the share Black in the Block

4In principle 911 calls regarding burglaries could be compared to the TPD’s burglary records to calculate

the 911 reporting rates for burglaries. Unfortunately, because the 911 data does not distinguish between

residential and commercial burglaries, such a comparison is not possible with the data to which I was granted

access.

23

group is positively associated with total 911 call volume at the 0.05 level. Of course, this

could be because of a greater baseline amount of crime in mostly Black neighborhoods

that is being reported at the same rate as crime elsewhere. To the extent that it is because

of over-reporting of crime by residents of mostly Black neighborhoods, that phenomenon

is documented in some of the literature on sociological determinants of 911 use (Davis

and Henderson 2003; Bosick et al. 2012; Baumer and Lauritsen 2010; Desmond and Valdez

2013).

24

6.2. Appendix B: the significance of “null” results

Abadie (2018) argues that, in a limited-information Bayesian framework, where readers

have a prior over the value of an estimand Θ, “a statistical result [is] informative when

it has the potential to substantially change the beliefs of the agents over a large range

of values for θ.” It follows that rejection of the point null is only informative when the

prior probability of rejection is low (that is, below 0.5), and that failure to reject the null

is more informative than rejection of the null whenever the prior probability of rejection

is greater than 0.5. In the simple case where an agent has a prior θ ∼ N(µ, σ2) on θ,

with σ2 > 0, there are n independent observations of θ, x1, …, xn distributed N(θ, 1), and

θ̂n =
1
n Σ

n
i=1 xi ∼ N(θ,

1
n ), with statistical significance conventionally defined as

√
n|θ̂n|

greater than some threshold c > 0 (usually c is 1.96), the prior probability of rejection is

given by:

Pr(
√

n|θ̂n| > c) = Φ
√

nµ − c
√

1 + nσ2

+ Φ
−
√

nµ − c
√
1 + nσ2
.

This quantity, importantly, does not depend on θ, but rather on depends heavily on

the sample size (n). In the present case, where n = 2, 771, non-significance is extremely

informative. Using a standard normal prior (µ = 0 and σ2 = 1) on any regression coeffi-

cients θ and a standard significance threshold of c = 1.96 reveals that the prior probability

of rejection in this case is 0.97.

25

References

Abadie, Alberto. 2018. Statistical non-significance in empirical economics. Technical report

National Bureau of Economic Research.

Antonovics, Kate, and Brian Knight. 2009. “A New Look at Racial Profiling: Evidence from

the Boston Police Department.” Review of Economics and Statistics 91 (1): 163-177.

Antrobus, Emma, and Andrew Pilotto. 2016. “Improving forensic responses to residen-

tial burglaries: Results of a randomized controlled field trial.” Journal of Experimental

Criminology 12 (3): 319–345.

Balla, Steven J, and William T Gormley Jr. 2017. Bureaucracy and democracy: Accountability

and performance. CQ Press.

Baumer, Eric P, and Janet L Lauritsen. 2010. “Reporting crime to the police, 1973–2005:

A multivariate analysis of long-term trends in the National Crime Survey (NCS) and

National Crime Victimization Survey (NCVS).” Criminology 48 (1): 131–185.

Baumgartner, Frank R, Derek A Epp, Kelsey Shoub, and Bayard Love. 2017. “Targeting

young men of color for search and arrest during traffic stops: Evidence from North

Carolina, 2002–2013.” Politics, Groups, and Identities 5 (1): 107–131.

Becker, Gary S. 1968. “Crime and punishment: An economic approach.” In The economic

dimensions of crime. Springer.

Beckett, Katherine, Kris Nyrop, and Lori Pfingst. 2006. “Race, drugs, and policing: Under-

standing disparities in drug delivery arrests.” Criminology 44 (1): 105–137.

Benson, Bruce L, David W Rasmussen, and David L Sollars. 1995. “Police bureaucracies,

their incentives, and the war on drugs.” Public Choice 83 (1-2): 21–45.

Besley, Timothy, and Stephen Coate. 1991. “Public provision of private goods and the re-

distribution of income.” The American Economic Review 81 (4): 979–984.

26

Bosick, Stacey J, Callie Marie Rennison, Angela R Gover, and Mary Dodge. 2012. “Report-

ing violence to the police: Predictors through the life course.” Journal of Criminal Justice

40 (6): 441–451.

Braun, Michael, Jeremy Rosenthal, and Kyle Therrian. 2018. “Police discretion and racial

disparity in organized retail theft arrests: evidence from Texas.” Journal of empirical legal

studies 15 (4): 916–950.

Brodkin, Evelyn Z. 2008. “Accountability in street-level organizations.” Intl Journal of Pub-

lic Administration 31 (3): 317–336.

Carpenter, Daniel. 2014. Reputation and Power: Organizational Image and Pharmaceutical Reg-

ulation at the FDA. Princeton University Press.

Carpenter, Daniel P, and George A Krause. 2012. “Reputation and public administration.”

Public administration review 72 (1): 26–32.

Carr, Patrick J, Laura Napolitano, and Jessica Keating. 2007. “We never call the cops and

here is why: A qualitative examination of legal cynicism in three Philadelphia neigh-

borhoods.” Criminology 45 (2): 445–480.

Chalfin, Aaron, and Justin McCrary. 2018. “Are US cities underpoliced? Theory and evi-

dence.” Review of Economics and Statistics 100 (1): 167–186.

Chatterjee, Samprit, and Jeffrey S Simonoff. 2013. Handbook of regression analysis. Vol. 5 John

Wiley & Sons.

Cihan, Abdullah, Yan Zhang, and Larry Hoover. 2012. “Police response time to in-progress

burglary: A multilevel analysis.” Police Quarterly 15 (3): 308–327.

Coupe, Richard Timothy. 2016. “Evaluating the effects of resources and solvability on bur-

glary detection.” Policing and Society 26 (5): 563–587.

Coviello, Decio, and Nicola Persico. 2015. “An economic analysis of Black-White dispari-

ties in the New York Police Department’s stop-and-frisk program.” The Journal of Legal

Studies 44 (2): 315–360.

27

Davis, Robert C, and Nicole J Henderson. 2003. “Willingness to report crimes: The role of

ethnic group membership and community efficacy.” Crime & Delinquency 49 (4): 564–

580.

DeAngelo, Gregory, and Emily G Owens. 2017. “Learning the ropes: General experience,

task-specific experience, and the output of police officers.” Journal of Economic Behavior

& Organization 142: 368–377.

DeAngelo, Gregory J., R. Kaj Gittings, and Amanda Ross. 2018. “Police Incentives, Policy

Spillovers, and the Enforcement of Drug Crimes.” Review of Law & Economics 14 (March).

DeAngelo, Gregory, R Kaj Gittings, and Anita Alves Pena. 2018. “Interracial face-to-face

crimes and the socioeconomics of neighborhoods: Evidence from policing records.” In-

ternational review of law and economics 56: 1–13.

Desmond, Matthew, and Nicol Valdez. 2013. “Unpolicing the urban poor: Consequences

of third-party policing for inner-city women.” American sociological review 78 (1): 117–

141.

Di Tella, Rafael, and Ernesto Schargrodsky. 2004. “Do police reduce crime? Estimates using

the allocation of police forces after a terrorist attack.” American Economic Review 94 (1):

115–133.

DOJ. 2014. “Burglary.” Crime in the United States, 2013, Department of Justice, Federal Bureau

of Investigation. https://ucr.fbi.gov/crime-in-the-u.s/2013/crime-in-the-u.s.

-2013/property-crime/burglary-topic-page/burglarymain_final .

DOJ. 2016. “Crime in the United States, 2015.” Department of Justice, Federal Bureau

of Investigation. https://ucr.fbi.gov/crime-in-the-u.s/2015/crime-in-the-u.s.

-2015/resource-pages/2015-cius-summary_final .

Duncan, Greg J, and Richard J Murnane. 2011. Whither opportunity?: Rising inequality,

schools, and children’s life chances. Russell Sage Foundation.

Eberhardt, Jennifer L, Phillip Atiba Goff, Valerie J Purdie, and Paul G Davies. 2004. “Seeing

28

https://ucr.fbi.gov/crime-in-the-u.s/2013/crime-in-the-u.s.-2013/property-crime/burglary-topic-page/burglarymain_final

https://ucr.fbi.gov/crime-in-the-u.s/2013/crime-in-the-u.s.-2013/property-crime/burglary-topic-page/burglarymain_final

https://ucr.fbi.gov/crime-in-the-u.s/2015/crime-in-the-u.s.-2015/resource-pages/2015-cius-summary_final

https://ucr.fbi.gov/crime-in-the-u.s/2015/crime-in-the-u.s.-2015/resource-pages/2015-cius-summary_final

black: Race, crime, and visual processing.” Journal of Personality and Social Psychology

87 (6): 876.

Ehrlich, Isaac. 1973. “Participation in Illegitimate Activities: A Theoretical and Empirical

Investigation.” Journal of Political Economy 81 (May): 521–565.

Epp, Charles R, Steven Maynard-Moody, and Donald P Haider-Markel. 2014. Pulled over:

How Police Stops Define Race and Citizenship. Univ. of Chicago Press.

Fagan, Jeffrey, and Amanda Geller. 2018. “Police, Race, and the Production of Capital

Homicides.” Berkeley J. Crim. L. 23: 261.

Foley, C Fritz. 2011. “Welfare payments and crime.” The review of Economics and Statistics

93 (1): 97–112.

Foreman, Christopher H. 2011. The promise and peril of environmental justice. Brookings In-

stitution Press.

Freeman, Jonathan B, and Kerri L Johnson. 2016. “More than meets the eye: Split-second

social perception.” Trends in cognitive sciences 20 (5): 362–374.

Freeman, Richard B. 1999. “Chapter 52 The economics of

crime.” In Handbook of Labor Economics. Vol. 3. Elsevier.

https://linkinghub.elsevier.com/retrieve/pii/S1573446399300432.

Friedman, Barry. 2020. “Disaggregating the Police Function.” U. Pa. L. Rev.(forthcoming

2020-21).

Fryer Jr, Roland G. 2019. “An empirical analysis of racial differences in police use of force.”

Journal of Political Economy 127 (3): 1210–1261.

Gelman, Andrew, Jeffrey Fagan, and Alex Kiss. 2007. “An analysis of the New York City

police department’s “stop-and-frisk” policy in the context of claims of racial bias.” Jour-

nal of the American Statistical Association 102 (479): 813–823.

Goel, Sharad, Justin M Rao, Ravi Shroff et al. 2016. “Precinct or prejudice? Understand-

ing racial disparities in New York City’s stop-and-frisk policy.” The Annals of Applied

Statistics 10 (1): 365–394.

29

Goldstein, Herman. 1977. “Policing a free society.” Policing a Free Society Cambridge, Mass:

Ballinger Pub. Co.

Golub, Andrew, Bruce D Johnson, and Eloise Dunlap. 2007. “The race/ethnicity disparity

in misdemeanor marijuana arrests in New York City.” Criminology & public policy 6 (1):

131–164.

Goncalves, Felipe, and Steven Mello. 2020. “A few bad apples? Racial bias in policing.”

Working Paper available at https://papers.ssrn.com/sol3/papers.cfm?abstract_

id=3627809.

Grogger, Jeffrey, and Greg Ridgeway. 2006. “Testing for racial profiling in traffic stops from

behind a veil of darkness.” Journal of the American Statistical Association 101 (475): 878–

887.

Hamilton, David L, and Jeffrey W Sherman. 2014. “Stereotypes.” In Handbook of social cog-

nition. Psychology Press.

Hamilton-Smith, Niall, and Andrew Kent. 2005. “The prevention of domestic burglary.”

Handbook of crime prevention and community safety: 417–457.

Hill, Heather C. 2003. “Understanding implementation: Street-level bureaucrats’ re-

sources for reform.” Journal of Public Administration Research and Theory 13 (3): 265–282.

Hilton, James L, and William Von Hippel. 1996. “Stereotypes.” Annual review of psychology

47 (1): 237–271.

Holmström, Bengt. 1999. “Managerial incentive problems: A dynamic perspective.” The

Review of Economic Studies 66 (1): 169–182.

Horrace, William C, and Shawn M Rohlin. 2016. “How dark is dark? Bright lights, big city,

racial profiling.” Review of Economics and Statistics 98 (2): 226–232.

Jencks, Christopher, and Meredith Phillips. 2011. The Black-White test score gap. Brookings

Institution Press.

Jolls, Christine. 2001. “Antidiscrimination and accommodation.” Harvard Law Review 115:

642.

30

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3627809

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3627809

Keefer, Philip, and Stuti Khemani. 2005. “Democracy, public expenditures, and the poor:

understanding political incentives for providing public services.” The World Bank Re-

search Observer 20 (1): 1–27.

Kelly, Morgan. 2000. “Inequality and crime.” Review of Economics and Statistics 82 (4): 530–

539.

Killmier, Bronwyn, Katrin Mueller-Johnson, and Richard Timothy Coupe. 2019.

“Offender–Offence Profiling: Improving Burglary Solvability and Detection.” In Crime

Solvability Factors. Springer.

Knowles, John, Nicola Persico, and Petra Todd. 2001. “Racial bias in motor vehicle

searches: Theory and evidence.” Journal of Political Economy 109 (1): 203–229.

Knox, Dean, Will Lowe, and Jonathan Mummolo. 2019. “Administrative records mask

racially biased policing.” American Political Science Review: 1–19.

Krivo, Lauren J, and Robert L Kaufman. 2004. “Housing and wealth inequality: Racial-

ethnic differences in home equity in the United States.” Demography 41 (3): 585–605.

Legewie, Joscha. 2016. “Racial profiling and use of force in police stops: How local events

trigger periods of increased discrimination.” American Journal of Sociology 122 (2): 379–

424.

Lemos, Margaret H, and Max Minzner. 2014. “For-Profit Public Enforcement.” Harvard Law

Review 127: 853.

Leovy, Jill. 2015. Ghettoside: A True Story of Murder in America. New York: PenguinRandom-

House.

Lipsky, Michael. 1980. Street-Level Bureaucracy: Dilemmas of the Individual in Public Services.

New York: Russel Sage Foundation.

Mancik, Ashley M, Karen F Parker, and Kirk R Williams. 2018. “Neighborhood context

and homicide clearance: Estimating the effects of collective efficacy.” Homicide studies

22 (2): 188–213.

31

Mas, Alexandre. 2006. “Pay, reference points, and police performance.” The Quarterly Jour-

nal of Economics 121 (3): 783–821.

Mast, Brent D, Bruce L Benson, and David W Rasmussen. 2000. “Entrepreneurial Police

and Drug Enforcement Policy.” : 24.

McCrary, Justin, and Deepak Premkumar. 2019. “Why We Need Po-

lice.” In The Cambridge Handbook of Policing in the United States, ed.

Tamara Rice Lave and Eric J. Miller. 1 ed. Cambridge University Press.

https://www.cambridge.org/core/product/identifier/9781108354721

Mitchell, Ojmarrh, and Michael S Caudy. 2015. “Examining racial disparities in drug ar-

rests.” Justice Quarterly 32 (2): 288–313.

Mitchell, Ojmarrh, and Michael S Caudy. 2017. “Race differences in drug offending and

drug distribution arrests.” Crime & Delinquency 63 (2): 91–112.

Natapoff, Alexandra. 2006. “Underenforcement.” Fordham L. Rev. 75: 1715.

Niskanen, William A. 1971. Bureaucracy & Representative Government. New York: Rout-

ledge.

Pare, Paul-Philippe. 2014. “Indicators of police performance and their relationships with

homicide rates across 77 nations.” International Criminal Justice Review 24 (3): 254–270.

Payne, B Keith. 2006. “Weapon bias: Split-second decisions and unintended stereotyping.”

Current Directions in Psychological Science 15 (6): 287–291.

Persico, Nicola. 2002. “Racial profiling, fairness, and effectiveness of policing.” American

Economic Review 92 (5): 1472–1497.

Persico, Nicola, and Petra E Todd. 2008. “The hit rates test for racial bias in motor-vehicle

searches.” Justice Quarterly 25 (1): 37–53.

Persico, Nicola, and Petra Todd. 2006. “Generalising the hit rates test for racial bias in law

enforcement, with an application to vehicle searches in Wichita.” The Economic Journal

116 (515): F351–F367.

32

Pinotti, Paolo. 2017. “Clicking on Heaven’s Door: The Effect of Immigrant Legalization on

Crime.” American Economic Review 107 (January): 138–168.

Rayman, Graham A. 2013. The NYPD Tapes: A Shocking Story of Cops, Cover-Ups, and

Courage. St. Martin’s Press.

Ritter, Joseph A. 2017. “How do police use race in traffic stops and searches? Tests based

on observability of race.” Journal of Economic Behavior & Organization 135: 82–98.

Roberts, Aki. 2015. “Adjusting rates of homicide clearance by arrest for investigation diffi-

culty: Modeling incident-and jurisdiction-level obstacles.” Homicide studies 19 (3): 273–

300.

Roberts, Aki, and Christopher J Lyons. 2011. “Hispanic victims and homicide clearance by

arrest.” Homicide Studies 15 (1): 48–73.

Ross, Cody T. 2015. “A multi-level Bayesian analysis of racial bias in police shootings at

the county-level in the United States, 2011–2014.” PloS one 10 (11): e0141854.

Shannon, Stephen, and Barry Coonan. 2016. “A solvability-based case screening checklist

for burglaries in Ireland.” European Law Enforcement Research Bulletin (15): 31–41.

Sherman, Daniel J. 2012. Not here, not there, not anywhere: politics, social movements, and the

disposal of low-level radioactive waste. Routledge.

Short, Martin B, Maria R D’orsogna, Patricia J Brantingham, and George E Tita. 2009.

“Measuring and modeling repeat and near-repeat burglary effects.” Journal of Quanti-

tative Criminology 25 (3): 325–339.

Sigelman, Lee. 1986. “The Bureaucrat as Budget Maximizer: An Assumption Examined.”

Public Budgeting & Finance 6 (1): 50–59.

Sonn, Paul K. 1992. “Fighting minority underrepresentation in publicly funded construc-

tion projects after Croson: A Title VI litigation strategy.” The Yale Law Journal 101 (7):

1577–1606.

Sonnichsen, Richard C. 2007. “Measuring police performance.” Monitoring Performance in

the Public Sector: Future Directions from International Experience: 219–235.

33

Trounstine, Jessica. 2015. “The privatization of public services in American cities.” Social

Science History 39 (3): 371–385.

Truman, Jennifer L., and Rachel E. Morgan. 2016. “Criminal Victimization, 2015.” United

States Bureau of Justice Statistics. http://www.bjs.gov/index.cfm?ty=pbdetail&iid=

5804.

Walker, Samuel, Cassia Spohn, and Miriam DeLone. 2012. The color of justice: Race, ethnicity,

and crime in America. Cengage Learning.

Weisel, Deborah Lamm. 2002. Burglary of single-family houses. Vol. 18 US Department of

Justice, Office of Community Oriented Policing Services . . . .

West, Jeremy. 2018. “Racial Bias in Police Investigations.” Working Paper available at

https://people.ucsc.edu/~jwest1/articles/West_RacialBiasPolice .

West, Jeremy. 2019. “Learning by Doing in Law Enforcement.” Working Paper available at

https://people.ucsc.edu/~jwest1/articles/West_LBDPolice .

Wilder, David A. 1993. “The role of anxiety in facilitating stereotypic judgments of out-

group behavior.” In Affect, cognition and stereotyping. Elsevier.

Woods, Jordan Blair. 2018. “Policing, Danger Narratives, and Routine Traffic Stops.” Mich.

L. Rev. 117: 635.

34

http://www.bjs.gov/index.cfm?ty=pbdetail&iid=5804

http://www.bjs.gov/index.cfm?ty=pbdetail&iid=5804

https://people.ucsc.edu/~jwest1/articles/West_RacialBiasPolice

https://people.ucsc.edu/~jwest1/articles/West_LBDPolice

7.

Tables

Table 1: Summary statistics

Statistic N Mean St. Dev. Min Max

Forced entry 2,771 0.646 0.478 0 1
Poverty share (Block group) 2,771 0.297 0.173 0 0.749
Hispanic share (Block group) 2,771 0.423 0.258 0.023 0.960
White share (Block group) 2,771 0.458 0.247 0.012 0.964
Black share (Block group) 2,771 0.042 0.050 0 0.339
Victim ages (mean) 2,647 43.270 17.721 0 99
Share white officers 2,771 0.573 0.435 0 1
Share Black officers 2,771 0.039 0.171 0 1
Share Hispanic officers 2,771 0.315 0.406 0 1
Share male officers 2,771 0.783 0.370 0 1
Share white victims 2,771 0.509 0.486 0 1
Share Black victims 2,771 0.051 0.214 0 1
Share Hispanic victims 2,771 0.271 0.434 0 1
Share male victims 2,771 0.495 0.462 0 1
Total victims 2,771 1.355 0.797 1 15
Total officers 2,771 2.168 2.145 1 24
Apartment? 2,771 0.311 0.463 0 1
Priority level (1-4) 2,686 3.603 0.781 1 4
East 2,771 0.212 0.409 0 1
West 2,771 0.275 0.447 0 1
Midtown 2,771 0.254 0.436 0 1
South 2,771 0.202 0.402 0 1
Incident on weekend or holiday? 2,771 0.248 0.432 0 1
Incident hour 2,765 14.689 6.293 1 24
Time spent (mins) 2,626 86.928 78.015 1 910
Prints collected? 2,771 0.399 0.490 0 1
Detective assigned? 2,771 0.344 0.475 0 1

35

Table 2: Race of officers, race of victims, and investigative thoroughness

Dependent variable:
Time (mins) (log) Prints collected Detective assigned

(1) (2) (3) (4) (5) (6)

Share white officers 0.039 0.044 0.032 0.041∗ 0.038∗ 0.025
(0.037) (0.040) (0.021) (0.022) (0.021) (0.022)

Share white victims −0.082∗∗ −0.066∗ −0.014 −0.009 −0.026 −0.027
(0.033) (0.035) (0.019) (0.020) (0.019) (0.019)

Constant 4.190∗∗∗ 4.119∗∗∗ 0.388∗∗∗ 0.314∗∗∗ 0.336∗∗∗ 0.421∗∗∗

(0.031) (0.067) (0.018) (0.038) (0.017) (0.037)

Month fixed effects? No Yes No Yes No Yes
Division fixed effects? No Yes No Yes No Yes
Observations 2,626 2,626 2,771 2,771 2,771 2,771
R2 0.003 0.010 0.001 0.026 0.002 0.020

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 36

Table 3: Forced entry and investigative thoroughness

Dependent variable:
Time (mins) (log) Prints collected Detective assigned
(1) (2) (3) (4) (5) (6)

Forced entry 0.254∗∗∗ 0.184∗∗∗ 0.202∗∗∗ 0.176∗∗∗ 0.139∗∗∗ 0.120∗∗∗

(0.046) (0.046) (0.027) (0.025) (0.019) (0.024)
Share white officers 0.025 0.011 0.024

(0.051) (0.028) (0.027)
Share white victims −0.014 −0.035 −0.009

(0.046) (0.026) (0.025)
Apartment −0.044 −0.039 −0.089∗∗∗

(0.055) (0.029) (0.028)
Victim ages (mean) −0.0004 −0.001 −0.001

(0.001) (0.001) (0.001)
Share male officers 0.014 0.032 0.056∗

(0.066) (0.033) (0.032)
Share male victims −0.015 −0.010 0.031

(0.048) (0.025) (0.024)
Total victims −0.007 0.027∗ 0.028∗

(0.027) (0.015) (0.015)
Total officers 0.115∗∗∗ 0.014∗ 0.032∗∗∗

(0.013) (0.007) (0.007)
East 0.383 0.063 −0.014

(0.274) (0.120) (0.116)
West −0.030 −0.030 −0.047

(0.107) (0.069) (0.067)
Midtown 0.292 −0.077 0.030

(0.334) (0.122) (0.118)
Constant 4.007∗∗∗ 3.893∗∗∗ 0.269 −0.850∗∗ 0.254∗∗∗ −0.299

(0.501) (0.501) (0.222) (0.429) (0.015) (0.413)

Priority level fixed effects? No Yes No Yes No Yes
Block group fixed effects? No Yes No Yes No Yes
Date fixed effects? No Yes No Yes No Yes
Hour fixed effects? No Yes No Yes No Yes
Observations 2,626 2,434 2,771 2,559 2,771 2,559
R2 0.021 0.432 0.039 0.355 0.020 0.359

Notes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Robust standard errors are clustered at the Block

group level and are included in parentheses.
Omitted division category: South.
37

Table 4: Representation of economic disadvantage in forced and unforced entry burglaries

Overall mean Share among Share among p-value
forced entry unforced entry

Apartment 0.31 0.28 0.37 < 0.01 High poverty neighborhood 0.26 0.24 0.29 < 0.01 High renter share neighborhood 0.39 0.36 0.44 < 0.01

38

8.

Figures

Figure 1: Map of burglaries in Tucson in 2016, with Census block group population density
information (persons per square mile)

Notes: Graduated colors and legend refer to population density of the Census block group. Shaded areas
are the independent municipality of South Tucson, which has its own police department (left), and Davis-
Monthan Air Force Base (right).

39

Figure 2: Map of burglaries in Tucson in 2016, with Census block group income informa-
tion (median annual income)

Notes: Graduated colors and legend refer to the median annual household income in the Census block group.
Shaded areas are the independent municipality of South Tucson, which has its own police department (left),
and Davis-Monthan Air Force Base (right).

40

Introduction
Policing: Public goods, bureaucratic behavior, and inequality
Data: Policing in Tucson, Arizona
Results and Discussion
Race and investigative thoroughness
Forced entry and investigative thoroughness
Unconditional inequality in service provision
Conclusion: Bureaucratic Incentives, Resource Allocation, and Inequality
Appendix
Appendix A: addressing selection
Appendix B: the significance of “null” results
Tables
Figures

How High-Income Areas Receive More Service From

Municipal Government: Evidence From City

Administrative Data∗

James J. Feigenbaum†

Department of Economics
Princeton University and Boston University

Andrew B. Hall‡

Department of Political Science
Stanford University

August

2

0,

20

1

6

Abstra

ct

Do economic inequalities translate into political inequalities, and if so, how? Combining
data on over

5

00,000 requests for services from the 2

4

-Hour Constituent Service Hotline in
Boston, Massachusetts with fine-grained census data on income, we show that higher-income
census tracts request and receive more services from the city’s government than do lower-income
census tracts located in the same neighborhood. To ensure that these results are not driven by
differences in the service needs of higher- vs. lower-income areas, we first estimate them using
only requests for snow removal, because snowfall affects the entire city. We report that a

10

%
increase in the per-capita income of a census tract predicts roughly a

3

% increase in the number
of requests for snow removal (p < 10−5). We then show that higher-income areas are more likely to place requests using the city’s smartphone app, and patterns in the timing of requests suggest, as expected, that the smartphone app is more convenient than the alternative methods for placing requests available to those who do not own smartphones. Higher-income citizens are thus advantaged in the non-electoral components of local politics because it is easier for them to participate.

∗Authors are listed in alphabetical order and contributed equally. For comments and suggestions the authors
thank Justin Grimmer, Melissa Sands, Brandon Stewart, and Jessica Trounstine.
†James J. Feigenbaum is a Post-Doc in the Department of Economics at Princeton University. In 201

7

, he will be

an Assistant Professor in the Department of Economics at Boston University.
‡Corresponding Author. Andrew B. Hall is an Assistant Professor in the Department of Political Science at

Stanford University (andrewbenjaminhall@gmail.com, http://www.andrewbenjaminhall.com).

http://www.andrewbenjaminhall.com).

1

Introduction

More than 50% of the world’s population,1 and more than

8

0% of the United States’ population,2

now reside in cities, yet we still have much to learn about the political workings of city governments.

Given the sharp rise in income inequality in the U.S. and globally (Piketty 20

14

), and the apparent

links between economic and political inequalities in the national political process (Gilens 2005,

20

12

; Gilens and Page 2014), it is especially important now to revisit age-old questions about the

economics and politics of our cities. To ask a question first posed in Dahl’s seminal study of New

Haven, Connecticut, “How does a ‘democratic’ system work amid inequality of resources?” (Dahl

1

9

61).

We are newly able to answer questions of this sort thanks to the recent explosion in the avail-

ability of quantitative data about U.S. cities. Taking advantage of this development, we combine

administrative data on over 500,000 constituent requests for a variety of municipal services—like

snow removal, tree removal, graffiti removal, and pothole repairs—with fine-grained census data on

incomes for the city of Boston, Massachusetts. We follow Dahl (

19

61: 3) in using this new data to

answer basic questions about municipal politics, such as: “To what extent do various citizens use

their political resources? Are there important differences that in turn result in differences in influ-

ence?” We find that higher-income areas devote more effort to requesting government services—such

as snow plowing, pothole repairs, and traffic light repairs—and, as a result, receive more service

from local government.

Our answer to these basic questions is thus somewhat different from Dahl’s. Though city

politics may be more pluralistic in other ways, when it comes to the provision of basic city services,

it is the wealthy who participate in the process, and gain from it, at a greater rate. Why is

this so? After presenting our main results, we examine one important mechanism. For a variety

of reasons, it is simply easier for higher-income individuals to use the requesting process. In

particular, we show that requests that come from higher-income census tracts are more likely to

be filed via the city’s smartphone app, while requests from lower-income census tracts are more

likely to be placed by telephone. We also document that non-smartphone requests are heavily

1
http://www.un.org/en/development/desa/news/population/world-urbanization-prospects-2014.html, Ac-
cessed 20 July 20

15

.

2
http://www.reuters.com/article/2012/03/26/usa-cities-population-idUSL2E8EQ5AJ20120326, Accessed 2

0

July 2015.

2

http://www.un.org/en/development/desa/news/population/world-urbanization-prospects-2014.html

http://www.reuters.com/article/2012/03/26/usa-cities-population-idUSL2E8EQ5AJ20120326

clustered during working hours on weekdays—when it is perhaps easier for individuals to place

phone calls—while smartphone app requests are more evenly distributed across weekdays and

weekends. This timing pattern strongly suggests that the smartphone app makes requesting easier

and more convenient, as we might expect. Because smartphones are expensive,3 higher-income

areas have disproportionate access to easier means of lodging requests. This result helps explain

the inequalities we document and suggests that, while the introduction of new technologies for

fielding constituent complaints is an overall boon to the transparency of local government, it can

exacerbate inequalities in participation when the underlying technology is costly for citizens to

obtain.

In general, statistical associations between income and service requests are uninformative be-

cause the underlying need for services may vary across individuals and space in ways the researcher

cannot observe. Suppose we find that one part of a city tends to have residents with higher in-

comes and also originates more requests for tree removals, relative to other parts of the city. Is

this because higher-income residents are more proactive in requesting tree removals, or could it

be that higher-income people simply live in places with more trees, and thus have more need to

request tree removals? Biases like these confound simple comparisons. We attack this problem

in two ways. First, we make all of our comparisons within-neighborhood, comparing census tracts

located within the same neighborhood but differing in their income levels. This removes any bias

from underlying service needs that vary across neighborhoods but not within neighborhoods and,

because neighborhoods are relatively small, makes the resulting comparisons more trustworthy.

Second, going further, we perform our primary analysis only on requests for snow removal. Unlike

many other services, the need for snow removal is relatively constant across places within a snowy

city, which makes Boston the ideal context for our study (Levine and Gershenson 2014; O’Brien,

Sampson, and Winship 20

13

). Across a wide variety of request types, we continue to find a robust

association between the per-capita income of census tracts and the frequency of service requests,

just as we do for snow removal requests.

3A recent Pew survey estimates that 84% of Americans earning $75,000 or more own a smartphone while only 50% of
those earning less than $30,000 do (http://www.pewinternet.org/files/2015/03/PI_Smartphones_040

11

51 ).
The survey also reports that the lower-income individuals of own smartphones are more “smartphone dependent”—
meaning it is their only main access to the internet—than higher-income owners. As such, even for lower-income
smartphone owners, filing requests may remain difficult since their phone use will often be tied to other more pressing
purposes.

3

http://www.pewinternet.org/files/2015/03/PI_Smartphones_0401151

The remainder of the paper is organized as follows. In the next section, we situate our study

in the literatures on local politics and economic and political inequality, and we lay out the impli-

cations of our analyses. Following that, we describe the datasets we use to answer these questions

empirically. Next, we explain our empirical approach and describe why it is helpful in removing

unobserved variables that could confound the relationship between income and service request. Sub-

sequently, we present our results, showing that higher-income areas request more services and that

smartphone use is an important mechanism that helps explain the findings. Finally, we conclude.

2

Local Politics and Economic Inequality: An Overview

The study of municipal politics has a long tradition in the social sciences both because of the

central role the city plays in the lives of so many Americans, and also because of the general lessons

about the nature of democratic society that the scrutiny of cities can reveal. Reviewing the local

politics literature, for example, Jessica Trounstine writes: “Knowing how benefits are distributed

and who wins and who loses in American politics requires understanding the functioning of local

representative democracy” (Trounstine 2010).4

Perhaps no single study exemplifies this goal more than Dahl (1961), who explores local poli-

tics in New Haven, Connecticut, but, in doing so, illuminates broader themes about representative

democracy. Dahl explores how political resources are distributed across New Haven society, high-

lighting the ways in which these resources are non-accumulating, i.e., not distributed in the same

manner as economic resources. As a result, he concludes that local democratic politics can be sur-

prisingly pluralistic. Though the wealthy possess many economic resources that the less wealthy do

not, Dahl argues that other sources of political power, like popularity, produce a complex political

system in which the wealthy—at least in historical New Haven—do not possess undue political

influence.

4Recent work takes up this goal and applies modern empirical techniques to municipal politics, though this burgeoning
literature mainly on the exclusively electoral links between city dwellers and their representatives (Tausanovitch and
Warshaw 2013, 2014; Ferreira and Gyourko 2009; Gerber and Hopkins 2011; Gamm and Kousser 2013; Hajnal
and Trounstine 2005). Several studies also investigate disparate aspects of the non-electoral arena in municipal
politics, examining, for example, the share of community security and education that is provided by private entities
(Trounstine N.d.), the role that economic constraints and political and institutional factors play in determining how
cities allocate their funds (Hajnal and Trounstine 2010) or the provision of public goods across cities with different
characteristics (Rugh and Trounstine 2011). We offer a different angle, focusing instead on how citizens request and
receive local services. In so doing, we aim to understand the links between economic and political inequality.

4

There are good reasons to revisit the fundamental questions that Dahl poses, given the economic

and political changes that have taken place between the writing of his book and the present day.

Income inequality in the U.S. has risen sharply since the publication of Who Governs? (e.g., Piketty

2014), meaning that the economic resources that Dahl studies are distributed more unevenly than

in previous eras. If this inequality affects political inequalities, then its effects will be larger today

than at the time that Dahl was writing.

Indeed, studying a modern context, we find evidence that economic and political resources—at

least in the sphere of local government services—are cumulative, with those earning more money

also wielding more political influence simply because they are better able to participate in the

local political process. Though focused on local service requests, our findings suggest deeper links

between economic and political inequality, too. Many other important political activities beyond

requesting services require similar amounts of time, effort, and information. It is likely that wealth-

ier individuals and places create similar political advantages by engaging in other such activities at

a disproportionate rate as well.

The results also speak directly to the literature on civic engagement. Previous research, for

example, reports that “Citizen activists tend to be drawn disproportionately from more advantaged

groups” (Verba, Schlozman, and Brady 1995). One type of activist, as the authors explain, is

a person who contacts local government with particularized requests.5 We confirm that more

advantaged groups place more requests for services, and we highlight the consequences that this

behavior has at the local level.

Finally, because our analyses establish a clear link between the income of city areas and their

propensity to request services from local government, they build on work that documents apparent

political inequalities that result from economic inequalities at the national level (Gilens 2005, 2012),

too. In local politics, as in national politics, the political system appears to place disproportionate

attention on the needs and requests of those with greater economic resources. This link exists not

just in the ideological arena—like in the roll-call votes that the aforementioned papers investigate—

but in the non-ideological arena, too, as we show.

5For other work on contacting local officials, see also Butler and Broockman (2011) and White, Nathan, and Faller
(2014).

5

We are far from the first to exploit municipal services data to study political questions. Minkoff

(2015) studies New York City request data, examining which of a number of important demographic

and political variables predict the filing of requests and offering a general theoretical framework

for thinking about the request of services. Though not focused on income levels, the article does

report only a small association between census-tract median income and the frequency of requests,

in contrast to the results we will present below. We suspect there are two reasons for this difference.

First, of course, the contexts in which we estimate these associations differ. Second, unlike the

main variables that Minkoff (2015) focuses on, the income of census tracts may be especially

tied to the underlying, unobserved service needs the tracts have. Not observing these needs may

produce a downward bias if higher-income areas have a higher propensity to file requests but a

lower underlying need for services. In studies of Boston request data, both Levine and Gershenson

(2014) and O’Brien, Sampson, and Winship (2013) outline the snow-based control strategy which

forms one of the two empirical approaches we employ in this paper to circumvent problems of

this kind. In establishing that demographic characteristics map to service request behavior, the

article also highlights why accounting for these variables is important for our study. Also studying

Boston, O’Brien (2015) provides a number of interesting analyses concerning the ways that citizens

interact with the request system, showing that they tend to report hyper-local issues and do not

specialize by issue type. Finally, Christensen and Ejdemyr (2015) studies San Francisco request

data and shows how electoral incentives can induce representatives to become more responsive to

service requests. Although we do not study response times in this paper—focusing instead on

non-electoral participation—this latter study thus suggests fruitful areas for follow-up research.

3

Data on Constituent Requests and Income

The analysis draws on three main datasets. Data on service requests to the Mayor’s

24

-hour

hotline comes directly from the city of Boston, which makes these and other datasets available at

https://data.cityofboston.gov/. This dataset covers the years 2011–2015. Citizens are able

to make requests via the phone, a mobile app, the city website, or in person. In addition, city

employees can add requests directly. Typically we include these worker-generated requests because

they likely come from constituents (e.g., citizens who call a particular department instead of the

6

https://data.cityofboston.gov/

24-hour hotline). However, we also re-estimate the results without the worker-generated requests

to make sure they are not driving the results; see Table A.2.

All requests are categorized according to type by the city. Table A.1 lists the universe of request

types. As the table shows, there are several categories of snow services. For the purposes of the

snow-only analyses in the paper, we include all categories that use the word

“snow.”

For every request, we are able to observe the time and date of complaint and of resolution, as

well as the request type, the assigned department, and the source of the complaint (phone call,

app, etc). Crucially for our purposes, we also observe the exact latitude and longitude location

of the problem, enabling us to map complaints to census tracts and neighborhoods. We exclude

two census tracts from our analyses: the tract containing city hall, which we suspect contains

a large number of erroneous geo-locations, and the tract containing Logan airport, which has no

population. These omissions do not drive the estimated results. For more geographical information,

see Figure S.3 and following in the Supporting Information.

We merge the service request data with census tract demographic and economic data from the

2009-2013 American Community Survey (ACS). We use the 2013 TIGER/Line shapefiles to ma

p

census tracts and neighborhoods in Boston. We locate each request point, by latitude and longitude,

within a census tract and a neighborhood layer. Both the ACS and shapefiles are available from

NHGIS, https://www.nhgis.org/.

4

Empirical Approach: Within-Neighborhood Design

As discussed in the Introduction, the main empirical concern is that places that very in income

may also vary in their underlying need to request services, confounding simple comparisons of

request behavior and income. To circumvent this issue, we do two things. First, as in Levine and

Gershenson (2014), we focus initially on requests for snow removal, because, unlike other service

types, we know that all areas of Boston have relatively similar needs in this area since snowfall

affects the whole city.6 As Levine and Gershenson (2014: 614) writes, “Because snowstorms are

6In conversations with a Boston-area meteorologist, we have become aware that snowfall is not truly constant across
the city; certain swaths of the city receive more snowfall than others. However, it is our impression that these areas
are quite a bit larger, geographically, than the neighborhoods within which we estimate our effects.

7

https://www.nhgis.org/

Figure 1 – Broader Neighborhoods of Boston, As Defined by the City.

East Boston

Charlestown

South BostonSouth End
Fenway/Kenmore

Allston/Brighton

Jamaica Plain
Roxbury

Dorchester

Mattapan
Roslindale

West Roxbury

Hyde Park

Back Bay
Financial District/Downtown

North EndWest End

Chinatown

Beacon Hill

Mission Hill

entirely exogenous to urban politics and neighborhood demographics, they generate equal levels of

objective need across the city.”7

Second, we also make all of our statistical comparisons within neighborhood, as well as within-

time, using census tracts-by-month as our unit of analysis. Census tracts are small geographical

units used for census-taking purposes. In our data, a tract has an average population of just over

4,000 individuals. These tracts are contained within larger city neighborhoods which are defined

by the city; on average in our data, a Boston neighborhood contains just over

16

census tracts.

Figure 1 shows these neighborhood definitions.

We only calculate the association between income levels and service requests by comparing

the incomes and requests of census tracts that are both located within the same neighborhood.

This further helps us remove confounding from unobserved underlying service needs, since these

census tracts are likely to be similar in many respects, especially since the divisions of census tracts

7Similarly, O’Brien, Sampson, and Winship (2013: 9) explains: “When it snows, it typically snows throughout the city,
and, controlling for certain infrastructural characteristics (e.g., the total road length, dead ends), all neighborhoods
should have a roughly equal need for snow plows.” We do not directly control for road characteristics but the
neighborhood fixed effects likely absorb much of this variation.

8

are arbitrary and driven by the statistical needs of the Census Bureau rather than political or

socioeconomic considerations.

Formally, we estimate equations of the form

log Num Requestsijst = β1 log Incomeit + β2GINI it + γj + δt +

2015∑
z=2011

1{Y ear = z} + Xitη + �ijst,

where the outcome variable log Num Requestsijst measures the logged number of service requests of

type s placed during month t in census tract i located in neighborhood j. The main explanatory

variable of interest is log Incomeit, which measures the logged per-capita income in census tract i

in month t. Thus, β1 represents the main quantity of interest, the estimated association between

per-capita income and request behavior across census tracts. We control for neighborhood fixed

effects (γj), month fixed effects (δt), year fixed effects (the summation term), and an optional

vector of additional control variables measured at the census tract and month level (Xit). When

included, this vector contains the following variables: GINI; log census tract population; a set of

share variables to indicate the racial composition of the tract as measured in the Census; and

population density of the census tract.

5

Higher-Income Areas Request More Service

5.1 Main Analysis

First, we compute the per-capita income and the number of service requests submitted in each

census tract in the city of Boston, and we use this to make our within-neighborhood comparisons.

Figure 2 presents binned averages of log income and log total requests by census tract, covering the

full time period of our data, where both variables are residualized using neighborhood fixed effects

as well as month and year fixed effects—thus ensuring that the comparisons are only made using

income and requests as measured within the same month and year and among census tracts within

the same neighborhood. As the plot shows, there is a noticeable, positive relationship between the

per-capita income of census tracts and the number of requests made.

9

Figure 2 – Relationship Between Income and Service Requests at the Census-
Tract Level, Boston, 2011-2015. Higher-income census tracts tend to request more
services from local government. Points are averages in equal-sample-sized bins of
census tract log income. Data residualized by neighborhood, month, and year fixed
effects. Each point represents roughly

25

,000 raw data points.

●

●
●
●
●

●
●

●

●
●
●

●
●
●
●
●
●
●
●
●

L
o

g
#

o

f
R

e
q

u
e

st
s

in
C

e
n

su
s

Tr
a

ct

9.5 10.0 10.5 11.0

0.4

0.5

0.6

0.7

0.8

Log Per Capita Income of Census Tract

Next, Figure 3 presents the overall geographical correlation between census-tract per-capita

income and requests. Darker colored areas are census tracts where the per-capita income is higher

above the neighborhood’s average, in the left panel, and where the number of complaints are higher

above the neighborhood’s average, in the right panel.

Table 1 presents the formal regression results. The first column shows the simple results consid-

ering only snow-removal requests and including neighborhood, month, and year fixed effects, but

no additional control variables. As the first row shows, a one percent increase in per-capita income

at the census-tract level predicts roughly a 0.3 percent increase in the number of snow-removal

requests (p < 10−5). Because of the neighborhood fixed effects, we can think of this estimate as

a within-neighborhood estimate. So, for example, for two census tracts both located within the

neighborhood of Back Bay, we would expect those where the per-capita income is higher to make

more snow-removal requests—even though underlying snow conditions are likely to be very similar

across the whole of Back Bay.

In the second column, we replicate the snow-removal analysis but with the addition of the

following control variables: GINI (a measure of income inequality within each census tract); log

census tract population; a set of share variables to indicate the racial composition of the tract as

10

Figure 3 – Income and Snow-Removal Requests Per Capita by Census Tract,
Boston.

(a) Income per Capita, 2013 ACS

Income pc
(1000$)

Relative to
Neighborhood Mean

under −12.0
−12.0 to −4.3

−4.3 to −0.6
−0.6 to 3.4
3.4 to 9.5

over 9.5

(b) Snow-removal requests per capita in 2015

Complaints
per 1000

Relative to
Neighborhood Mean

under −20.0
−20.0 to −7.0

−7.0 to 0.3
0.3 to 7.3

7.3 to 20.0
over 20.0

measured in the Census; and population density of the census tract. The addition of these controls

has little impact on the main quantity of interest, as shown in the first row of the second column.

The latter two columns replicate these same specifications but using all request types. This

alternate approach has advantages and disadvantages. By bringing in more request types, the

results are obviously more general as well as more powerful from a statistical perspective. On the

other hand, by including request types where the underlying need surely varies more by area—

graffiti clean-up, for example, does not occur evenly across tracts even within neighborhoods—the

associations we measure here may be less informative. As we see in the third and fourth columns,

we continue to find a strong and positive association between census-tract income and the number

of requests.

11

Table 1 – Association Between Income, Income Inequality, and Requests for Service
at the Census Tract Level. Wealthier census tracts file more requests, both for snow
removal and across all included request types.

DV: Log Number of Requests

Snow Removal All Requests

Log Income 0.32 0.

21

0.14

0.12

(0.06) (0.08) (0.02) (0.02

)

# Individual Obs

# Census Tract-Month Obs 4,878 4,849 157,691 157,245

Controls No Yes No Yes
Neighborhood Fixed Effects Yes Yes Yes Yes
Year Fixed Effects Yes Yes Yes Yes
Month Fixed Effects Yes Yes Yes Yes

Standard errors clustered by census tract in parentheses. When
included, controls are: GINI, log population, racial composition
of tract, and population density.

12

5.2 Effects Across Service Types

Figure 4 decomposes overall estimates by service type. The figure plots the estimated coefficient

on log income at the census-tract level where equation 1 is estimated repeatedly, restricting the

sample in turn to each of the service request types listed in Table A.1. As we see, there is a positive

association between census-tract per-capita income and requests for essentially all types of requests,

with just a few categories displaying negative (but never statistically significant) coefficients, most

likely due to sampling variability.

The association for snow removal requests is among the largest. We suspect that this evidence

in favor of the snow-based design; our estimates for other request types are likely downward biased

because higher-income areas have lower unobserved needs for those services (e.g., graffiti removal).

Naturally, the opposite could also be true; the estimated association for snow-removal requests could

instead be upward biased if, say, wealthier areas have greater needs for snow removal. We think

this is less likely since snowfall affects all neighborhoods, and more importantly, all areas within a

neighborhood, but we cannot rule it out completely. We should point out that, if anything, one

might expect the snow results to be downward biased by road quality, too. If higher-income areas

have better roads, and thus a lower underlying need to request snow removal, then our estimate

of the link between income and snow removal requests likely understates the full difference in the

propensity for higher-income vs. lower-income areas to file requests.

5.3 Technology Disproportionately Helps Higher-Income Areas

We now turn to considering why higher-income areas submit more requests, even when underlying

need is held constant. There are no doubt many explanations. Citizens in higher-income areas may

be more educated, may have more information about local government—including even the simple

fact of being aware that the request system exists—and may have more reason to believe that the

government will heed their requests than do citizens in lower-income areas. While we cannot test

for all of these mechanisms in our data, we can investigate one possibility: namely, that it is simply

easier for people with more money to find the time and opportunity to file requests.

To test this idea, we first investigate the means by which each complaint in the dataset is

submitted. We calculate the proportion of requests within each census tract that were submitted

13

Figure 4 – Estimated Association Between Census-Tract Per-Capita Income and
Requests for Services Across Service Types, Boston, 2011-2015. The independent
and dependent variables are both logged, so resulting coefficients can be interpreted
as percentage effects. Higher-income areas request all services except traffic signal
repairs at higher rates; this relationship is especially pronounced for snow removal.

●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●

Schedule a Bulk Item Pickup
Snow Removal
Transfer Not Completed
Pothole Repair (Internal)
Missed Trash/Recycling/Yard Waste/Bulk Item
Schedule a Bulk Item Pickup SS
Request for Snow Plowing
Request for Recycling Cart
Graffiti Removal
Street Light Outages
Sign Repair
Sidewalk Repair (Make Safe)
Sticker Request
Requests for Street Cleaning
Improper Storage of Trash (Barrels)
Empty Litter Basket
Highway Maintenance
PWD Graffiti
Abandoned Vehicles
Request for Pothole Repair
Poor Conditions of Property
Tree Emergencies
Space Savers
Tree Maintenance Requests
Parking Enforcement
Building Inspection Request
Recycling Cart Return
Sidewalk Repair
Catchbasin
General Lighting Request
Unsafe Dangerous Conditions
Utility Call−In
New Tree Requests
Illegal Dumping
Heat − Excessive Insufficient
Sidewalk Repair (Internal)
Notification
Unsatisfactory Living Conditions
Electrical
Public Works General Request
Pick up Dead Animal
Work w/out Permit
Rodent Activity
Major System Failure
New Sign Crosswalk or Pavement Marking
Missing Sign
Snow Plow (Emergency)
Street Light Knock Downs
Overflowing or Un−kept Dumpster
Contractor Complaints
Traffic Signal Repair
Street Light Outages (Internal)
Misc. Snow Complaint

0 0.5
Coefficient on Log Income

Outcome: Log # of Requests

14

Figure 5 – Estimated association between census-tract per-capita income and the
proportion of requests made via the smartphone app. Higher-income areas are more
likely to use the smartphone app to submit requests. Points are averages in equal-
sample-sized bins of census tract log income. Data residualized by neighborhood,
month, and year fixed effects. Each point represents roughly 25,000 raw data points.

●
●

●●

● ●

●
●
●
●
●
●
●
●
●
●
●
●
●
●

P
ro

p
o

rt
io

n
o

f
R
e
q
u
e
st
s

M
a

d
e

V
ia

A
p

p
9.5 10.0 10.5 11.0

0.08

0.10

0.12

0.14

0.16

0.

18

0.20

0.

22

Log Per Capita Income of Census Tract

via the city’s smartphone app, and we repeat the simple analysis from above with this new outcome

variable. As Figure 5 shows, higher-income census tracts are more likely than lower-income ones

to submit requests via smartphone.

This relationship is important because, as discussed in the Introduction, smartphones are ex-

pensive, and ownership patterns vary strongly with income. Higher-income individuals are more

likely to own smartphones, and thus more likely to have access to the city’s smartphone app, which

allows citizens to submit requests on-the-fly. There is good reason to suspect this makes filing

requests easier. Although both the app and the hotline phone are available 24 hours a day, 7 days

a week, calling requires effort and planning in ways the app does not.

We test for evidence that users find the smartphone app more convenient by investigating the

timing patterns of smartphone requests vs. non-smartphone requests. Figure 6 shows the frequency

of requests across days of the week for non-smartphone requests (left panel) and smartphone re-

quests (right panel). Non-smartphone requests, which are largely personal phone calls to the city’s

office, are clustered on the work days. Smartphone requests, in contrast, are almost evenly dis-

tributed across all seven days of the week, with Saturday and Sunday requests much more frequent

that they are for non-smartphone requests. We confirm this same difference in patterns when we

look at the times of day when requests are made, too, in Figures S.2 and S.3. Using the smartphone

15

Figure 6 – Frequency of requests across days of the week. Requests made through
means other than the smartphone app (mainly phone calls) are much more common
on weekdays; requests made through the smartphone app are nearly constant across
the entire 7-day week.

ta

b
le

(w
d

a
y(

d
a
ys

.n
o

n
.a

p
p

))
/s

u
m

(t
a

b
le
(w
d
a
y(
d
a
ys
.n
o
n
.a
p
p

))
)

S M T W Th F Sat

0

0.1

0.2

Non−

Smartphone Requests

t

a
b
le

(w
d
a
y(
d
a
ys

.a
p

p
))

/s
u

m
(t

a
b
le
(w
d
a
y(
d
a
ys
.a
p
p
))
)
S M T W Th F Sat
Smartphone Requests

app is much more convenient, allowing individuals to submit requests at any time without having

to place a call to an office.

We also investigate the times of day at which citizens place requests. Figures 7 and 8 present

these time distributions by month for all requests made through all non-smartphone channels (first

figure) and all requests made by smartphone (second figure). The hours 9:00am–5:00pm (9:00–

17

:00) are highlighted in each panel. As we see in the first plot, regular requests (those not made

via the smartphone app) peak during normal business hours.

In the second figure, on the other hand, we see that business hours are barely prioritized over

other hours by smartphone users. Like the days of the week analysis in the body of the paper, this

suggests the convenience that the app offers. Smartphone users do not need to worry about whether

they can reach a person or find a time when they can place a phone call. Because higher-income

individuals are more likely to own smartphones, they are more likely to enjoy this convenience,

thus helping to explain why higher-income areas submit more requests in general.

16

Figure 7 – Frequency of requests across the hours of the day, by month; all non-
smartphone requests. Most requests are submitted during daytime working hours.
The month of February shows the most requests, largely driven by the massive
February snowfall of 2015.

Work Hours

0

4000

8000

January February March April

0
4000
8000

May June July August

0 5 10 15 20

0

4000
8000

September

0 5 10 15 20

October

0 5 10 15 20

November

0 5 10 15 20

December

Hour of Day

#
o

f
R
e
q
u
e

st
s

17

Figure 8 – Frequency of requests across the hours of the day, by month; smartphone
requests only. Requests via smartphone display less difference between work/non-
work hours. The month of February shows the most requests, largely driven by the
massive February snowfall of 2015.

Work Hours
0

500

1000

1500
January February March April

0
500
1000

1500
May June July August

0 5 10 15 20
0

500
1000

1500
September

0 5 10 15 20
October
0 5 10 15 20
November
0 5 10 15 20
December
Hour of Day
#
o
f
R
e
q
u
e
st
s
18

6

Discussion and Conclusion

In this paper, we have documented a link between economic resources and political resources.

Controlling for underlying need by studying requests for snow removal, as well as by making

comparisons only within neighborhoods of Boston, we have demonstrated that higher-income census

tracts place more requests for service, and thus receive more service, than do lower-income tracts.

Why does this inequality exist, and what could be done about it? While we certainly do not

have a full answer, we have offered evidence for one explanation. It is easier—more convenient, less

time-consuming, and requiring less effort—for higher-income individuals to request services from

local government. In particular, higher-income areas are more likely than lower-income areas to

use smartphones to request services, probably because smartphones are expensive and therefore

more prevalent among wealthier people. Usage timing patterns suggest, as one might expect,

that this smartphone app in turns makes submitting requests easier and more convenient, thereby

advantaging those who own smartphones. The rolling out of other request channels that cater to

lower-income areas might be one way to address this inequality.

More generally, new efforts may be necessary in order to persuade individuals in lower-income

areas to file requests. A lack of faith in the government’s willingness to respond to their requests

and/or a lack of information about even the possibility of submitting these requests are likely to be

important reasons why these areas submit fewer of them. Future work should investigate important

questions like these.

Local government is arguably the most important layer of government in Americans’ daily lives.

Like any democracy, our municipal governments draw much of their power from, and partially

deploy their resources in response to, citizen activity. When this activity is uneven across economic

strata, it is likely that the government’s response will also be uneven. In our case, studying the city

of Boston, it is clear that those of greater economic means participate more in the non-electoral

components of local government, asking for and receiving more service from their city’s government

than do those of lesser financial means.

19

References

Butler, Daniel M., and David E. Broockman. 2011. “Do Politicians Racially Discriminate Against
Constituents? A Field Experiment on State Legislators.” American Journal of Political Science
55(3): 463–477.

Christensen, Darin, and Simon Ejdemyr. 2015. “The Effects of Election Eligibility and Timing on
Shirking: An Empirical Reappraisal.” Working Paper.

Dahl, Robert A. 1961. Who Governs?: Democracy and Power in an American City. Yale University
Press.

Ferreira, Fernando, and Joseph Gyourko. 2009. “Do Political Parties Matter? Evidence from U.S.
Cities.” The Quarterly Journal of Economics 124(1): 399–422.

Gamm, Gerald, and Thad Kousser. 2013. “No Strength in Numbers: The Failure of Big-City Bills
in American State Legislatures, 1880–2000.” American Political Science Review 107(4): 663–678.

Gerber, Elisabeth R., and Daniel J. Hopkins. 2011. “When Mayors Matter: Estimating the Impact
of Mayoral Partisanship on City Policy.” American Journal of Political Science 55(2): 326–339.

Gilens, Martin. 2005. “Inequality and Democratic Responsiveness.” Public Opinion Quarterly 69(5):
778–796.

Gilens, Martin. 2012. Affluence and Influence: Economic Inequality and Political Power in America.
Princeton University Press.

Gilens, Martin, and Benjamin I. Page. 2014. “Testing Theories of American Politics: Elites, Interest
Groups, and Average Citizens.” Perspectives on Politics 12(3): 564–581.

Hajnal, Zoltan L., and Jessica Trounstine. 2005. “Where Turnout Matters: The Consequences of
Uneven Turnout in City Politics.” Journal of Politics 67(2): 515–535.

Hajnal, Zoltan L., and Jessica Trounstine. 2010. “Who or What Governs?: The Effects of Eco-
nomics, Politics, Institutions, and Needs on Local Spending.” American Politics Research 38(6):
1130–1163.

Levine, Jeremy R., and Carl Gershenson. 2014. “From Political to Material Inequality: Race,
Immigration, and Requests for Public Goods.” Sociological Forum 29(3): 607–627.

Minkoff, Scott L. 2015. “NYC 311: A Tract-Level Analysis of Citizen-Government Contacting in
New York City.” Urban Affairs Review pp. 1–36.

O’Brien, Daniel Tumminelli. 2015. “Custodians and Custodianship in Urban Neighborhoods: A
Methodology Using Reports of Public Issues Received by a City’s 311 Hotline.” Environment
and Behavior 47(3): 304–327.

O’Brien, Daniel Tumminelli, Robert J. Sampson, and Christopher Winship. 2013. “Ecometrics in
the Age of Big Data: Measuring and Assessing ‘Broken Windows’ Using Administrative Records.”
Boston Area Research Initiative Working Paper No. 3.

Piketty, Thomas. 2014. Capital in the 21st Century. Harvard University Press.

20

Rugh, Jacob S., and Jessica Trounstine. 2011. “The Provision of Local Public Goods in Diverse
Communities: Analyzing Municipal Bond Elections.” The Journal of Politics 73(4): 1038–1050.

Tausanovitch, Chris, and Christopher Warshaw. 2013. “Measuring Constituent Policy Preferences
in Congress, State Legislatures, and Cities.” Journal of Politics 75(2): 330–342.

Tausanovitch, Chris, and Christopher Warshaw. 2014. “Representation in Municipal Government.”
American Political Science Review 108(03): 605–641.

Trounstine, Jessica. 2010. “Representation and Accountability in Cities.” Annual Review of Political
Science 13(1): 407–4

23

.

Trounstine, Jessica. N.d. “The Privatization of Public Services in American Cities.” Forthcoming,
Social Science History .

Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady. 1995. Voice and Equality: Civic
Voluntarism in American Politics. Harvard University Press.

White, Ariel, Noah Nathan, and Julie Faller. 2014. “What Do I Need to Vote? Bureaucratic
Discretion and Information Provision by Local Election Officials.” American Political Science
Review 109(1): 129–142.

21

A.1

Online Appendix

In total, we observe 202 distinct service request types, as categorized by the city of Boston. Table

A.1 lists those request types for which we observe at least 1,000 distinct requests in the data. Some

request types are similar and could be combined; to depict the raw data we leave these categories

separate here. In the analyses on snowfall, however, we combine all types that contain the word

“snow.”
22

Table A.1 – Service Request Types and Frequencies. All request types with at
least 1,000 appearances in the data are displayed.

Type Total

Schedule a Bulk Item Pickup 60,414
Request for Snow Plowing 47,079
Requests for Street Cleaning 32,383
Pothole Repair (Internal) 29,731
Missed Trash/Recycling/Yard Waste/Bulk Item 28,614
Street Light Outages 24,823
Request for Pothole Repair 23,620
Snow Removal 20,281
Sidewalk Repair (Make Safe) 12,406
Graffiti Removal 12,334
Request for Recycling Cart 11,896
Schedule a Bulk Item Pickup SS 11,575
Improper Storage of Trash (Barrels) 11,501
Tree Maintenance Requests 10,491
Traffic Signal Repair 10,329
Unsatisfactory Living Conditions 9,263
Sticker Request 7,837
Pick up Dead Animal 7,231
Rodent Activity 6,851
Abandoned Vehicles 6,687
Sign Repair 6,528
Poor Conditions of Property 6,035
Tree Emergencies 5,055
Highway Maintenance 4,844
Parking Enforcement 4,632
Sidewalk Repair 4,358
New Tree Requests 4,100
Illegal Dumping 4,073
New Sign Crosswalk or Pavement Marking 3,649
General Lighting Request 3,558
Building Inspection Request 3,485
Empty Litter Basket 3,425
Notification 3,413
Recycling Cart Return 2,999
PWD Graffiti 2,988
Transfer Not Completed 2,899
Sidewalk Repair (Internal) 2,710
Street Light Outages (Internal) 2,491
Utility Call-In 2,447
Major System Failure 2,436
Work w/out Permit 2,302
Heat – Excessive Insufficient 2,276
Unsafe Dangerous Conditions 2,227
Street Light Knock Downs 2,184
Missing Sign 2,139
Contractor Complaints 1,744
Public Works General Request 1,674
Space Savers 1,526
Request for Snow Plowing (Emergency Responder) 1,415
Misc. Snow Complaint 1,364
Catchbasin 1,213
Electrical 1,193
Overflowing or Un-kept Dumpster 1,095

23

A.2

Removing Worker Submitted Requests

As discussed in the Materials section, our main analyses include all requests submitted from all

sources. Some requests are filed directly by city employees. We have good reason to suspect most of

these requests still originate from citizens; on its own website, for example, the city of Boston refers

to these categories as “direct department contacts” (http://www.cityofboston.gov/mayor/24/

requests.asp, Accessed 22 July 2015). Nevertheless, we want to be sure these alternate sources do

not drive our results. Accordingly, Table A.2 re-estimates the main results using only the following

two city-identified sources: “Citizens Connect App” and “Constituent Call.” As the table shows,

the results are nearly identical to those reported in the paper.

24

http://www.cityofboston.gov/mayor/24/requests.asp

http://www.cityofboston.gov/mayor/24/requests.asp

Table A.2 – Association Between Income, Income Inequality, and Requests for
Service at the Census Tract Level; Removing Worker-Reported Requests. Wealthier
census tracts file more requests, both for snow removal and across all included
request types.

DV: Log Number of Requests
Snow Removal All Requests

Log Income 0.34 0.24 0.11 0.09
(0.06) (0.08) (0.02) (0.02)

# Individual Obs 58,496 58,496 311,098 311,098

# Census Tract-Month Obs 4,625 4,599 121,706 121,405

Controls No Yes No Yes
Neighborhood Fixed Effects Yes Yes Yes Yes
Year Fixed Effects Yes Yes Yes Yes
Month Fixed Effects Yes Yes Yes Yes

Standard errors clustered by census tract in paren-
theses. When included, controls are: GINI, log pop-
ulation, racial composition of tract, and population
density.

25

Introduction
Local Politics and Economic Inequality: An Overview
Data on Constituent Requests and Income
Empirical Approach: Within-Neighborhood Design
Higher-Income Areas Request More Service
Main Analysis
Effects Across Service Types
Technology Disproportionately Helps Higher-Income Areas
Discussion and Conclusion
Online Appendix
Removing Worker Submitted Requests

RACIAL DISCRIMINATION IN LOCAL PUBLIC
SERVICES: A FIELD EXPERIMENT IN THE
UNITED STATES

Corrado Giulietti
University of Southampton

Mirco Tonin
Free University of Bozen-Bolzano

Michael Vlassopoulo

s

University of Southampton

Abstract
We examine whether racial discrimination exists in access to public services in the United States. We
carry out an email correspondence study in which we pose simple queries to more than 19,000 loca

l

public service providers. We find that emails from putatively black senders are almost 4 percentage
points less likely to receive an answer compared to emails signed with a white-sounding name

.

Moreover, responses to queries coming from black names are less likely to have a cordial tone.
Further tests suggest that the differential in the likelihood of answering is due to animus toward
blacks rather than inferring socioeconomic status from race. Finally, we show that attitudes toward
the government among blacks are more negative in states with higher discrimination. (JEL: D73,
H41, J15

)

1. Introduction

African Americans have a disadvantaged position in American society in terms
of economic outcomes, educational achievements, incarceration, health and life

The editor in charge of this paper was M. Daniele Paserman.

Acknowledgments: We would like to thank the Editor Daniele Paserman and three anonymous referees
for very constructive comments. We also thank Matthew Blackwell, Raj Chetty, Rajeev Dehejia, Stefano
DellaVigna, Uri Gneezy, Jonathan Guryan, Nathaniel Hilger, Guillermina Jasso, Gary King, Peter Kuhn,
Uwe Sunde and Adam Szeidl for useful discussions. We also thank participants of seminars at IZA,
Central European University, Temple, Köln, Newcastle, Surrey, Innsbruck, Verona, Rotterdam, and
participants of the 2015 Society for Government Economists Conference, the Workshop on the Economics
of Discrimination in Naples, the 2016 Royal Economic Society Conference, the 39th NBER Summer
Institute, the 33rd European Association of Law and Economics for their comments. We thank Kerwin
Kofi Charles and Jonathan Guryan for kindly providing us with their data. This project has received ethical
approval from the Institute for the Study of Labor (IZA) in Bonn, Germany, and from the University of
Southampton.

E-mail: C.Giulietti@soton.ac.uk (Giulietti); Mirco.Tonin@unibz.it (Tonin);
M.Vlassopoulos@soton.ac.uk (Vlassopoulos)

Journal of the European Economic Association 2019 17(1):165–204 DOI: 10.1093/jeea/jvx045
c� The Author(s) 2017. Published by Oxford University Press on behalf of European Economic Association.

All rights reserved. For permissions, please e-mail: journals.permissions@oup.co

m

D
ow

nloaded from
https://academ

ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U

niversity of C
alifornia, S

an D
iego user on 05 June 2020

mailto:C.Giulietti@soton.ac.uk

mailto:Mirco.Tonin@unibz.

it

mailto:M.Vlassopoulos@soton.ac.uk

mailto:journals.permissions@oup.com

166 Journal of the European Economic Association

expectancy.1 Discrimination is commonly proposed as one of the possible causes
of this predicament and has been documented in several realms, including the labor
market, the judicial system, housing and product markets.2 In his review of racial
inequality, Fryer (2011) underlines that “the new challenge is to understand the
obstacles undermining the achievement of black and Hispanic children in primary and
secondary school” (p. 857). Local public service providers like school districts and
libraries have a major role to play in this regard; thus, discrimination in access to these
services represents an important obstacle toward addressing racial inequality. More
generally, public institutions at the local level—so-called street-level bureaucracies
(Lipsky 1980)—are at the front-line of service delivery and thus play a key role in the
policy-implementation process, exerting great influence on how policies are actually
carried out and experienced by citizens. It is hence important to examine their attitudes
and behavior vis-à-vis discrimination.

A central tenet of the U.S. law is the prohibition of racial discrimination by
the government, with racial discrimination by public authorities prohibited and
the principle of nondiscrimination central to governmental policy throughout the
country (U.S. Government 2013). For instance, the Civil Rights Act of 1964 bans
discrimination based on race by government agencies that receive federal funding. This
is supplemented by several other provisions in federal and state law. For instance, under
the Minnesota Human Rights Act (363A.12), “It is an unfair discriminatory practice
to discriminate against any person in the access to, admission to, full utilization of or
benefit from any public service because of race […]”. Thus, discrimination by providers
of public services not only has a potentially detrimental impact on the economic and
social lives of those affected, but is also illegal. Furthermore, taste-based discrimination
à la Becker (1957) is predicted to fade with intensified market competition and
lower barriers to entry. Although deregulation and globalization may have increased
competition in the U.S. economy, thus placing pressure on discriminatory attitudes
in the private sector (see Levine, Rubinstein, and Levkov 2014, for relevant evidence
from the financial industry), this has certainly been much less the case for the public
sector.

In this study, we investigate racial discrimination across a wide range of public
services. Targeted services include school districts, local libraries, sheriff offices,
county clerks, county treasurers and job center veteran representatives. In particular, we
collect all available emails of the targeted local public service providers, which gives us
more than 19,000 cases, corresponding to roughly half of the total number of providers.

1. Altonji and Blank (1999, Chap. 48) provide an overview of race differential in the labor market.
Fryer (2011, Chap. 10) focuses on the racial achievement gap in education. Sabol, West, and Cooper
(2009) provide figures about incarceration by race. CDC (2011) reports on race disparities in mortality and
morbidity.

2. Charles and Guryan (2011) discuss research on discrimination against blacks in labor market outcomes.
Alesina and La Ferrara (2014) show evidence of racial bias in capital sentencing. Ewens, Tomlin, and Wa

ng

(2014) is a recent contribution on discrimination in housing. Product market discrimination is studied, for
instance, by Doleac and Stein (2013).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 167

Some of these services play important roles in relevant economic domains as they
directly relate to employment (job centers) or education (school districts). Libraries
also perform an important role by both promoting literacy and providing access to
information and computer technology, thus facilitating activities like job-searching.3

The other services that we study involve typical government functions, like law
enforcement (sheriffs), taxation (county treasurers), and general public administration
(county clerks).4

To identify in a credible way whether there is racial discrimination in access
to local public services, we conduct an email correspondence study. In particular,
we solicit information relevant to access a public service (the office opening hours
or some specific information, for example, the documentation needed for school
enrollment) from 19,079 local public offices and observe whether or not we receive a
reply depending on whether the request was signed with a distinctively white or black
name. Correspondence and audit studies have been used to investigate discrimination
in a variety of settings, including employment (Bertrand and Mullainathan 2004),
housing (Ewens et al. 2014), product markets (Gneezy, List, and Price 2012; Doleac
and Stein 2013), financial markets (Bayer, Ferreira, and Ross forthcoming), and
along different dimensions, including race, ethnicity, gender, age, disability, sexual
orientation, obesity, caste, and religion.5

Failing to provide information about a service is not equivalent to denying access to
a service. However, there is growing evidence showing that the provision of information
has an important impact on decisions and take-up rates. For instance, regarding the
Earned Income Tax Credit, Bhargava and Manoli (2015) show that the mere receipt of
a second notice just months after the receipt of an initial IRS notice led to substantial
additional claiming. Hoxby and Turner (2013) show that providing information on
the application process and colleges’ net costs has an effect on college enrollment.

3. A nationally representative survey by the Pew Research Center (2013) finds that over half of Americans
aged 16 and older have used a public library in some way in the previous 12 months, with many using
facilities provided by libraries for purposes related to education (42%, e.g., taking online classes or working
on assignments and schoolwork), employment (40%, e.g., searching for job opportunities, submitting online
job applications or working on resumes), and health (37%, e.g., learning about medical conditions, finding
health care providers, and assessing health insurance options). Many also report using a library computer
to download government forms or find out about a government program or service. Interestingly, the study
shows that library services are particularly important to “[w]omen, African Americans and Hispanics,
adults who live in lower-income households, and adults with lower levels of educational attainment”.

4. County clerks are generally responsible for keeping records of deeds and marriage licenses and
most other public records. They also issue many licenses and often have responsibilities for elections,
including the preparation of ballots and the recruitment and training of poll workers. County treasurer
offices’ responsibilities include, for instance, the issuance of vehicle titles and registrations, the collection
of vehicle fees, and the collection of property taxes for local governments. The responsibilities of school
districts may include the selection of textbooks and other curriculum materials, the hiring and dismissal of
staff, the monitoring of finances, ensuring compliance with relevant laws and the maintenance of school
buildings. Sheriff offices have a range of duties that include criminal investigations, traffic enforcement,
and operation of the jail. Their responsibilities may include serving warrants and evictions.

5. See Riach and Rich (2002), Guryan and Charles (2013), Rich (2014), Neumark (2016), and Bertrand
and Duflo (2017), for earlier and more recent reviews of the literature.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

168 Journal of the European Economic Association

Hastings and Weinstein (2008) find that providing low-income families with direct
information about school-level academic performances has an impact on parents’
school choice. Duflo and Saez (2003) show the effects of information on the decision
to enroll in a tax deferred account retirement plan. Daponte, Sanders, and Taylor (1999)
document that ignorance about the food stamp program contributes to nonparticipation.
Thus, making it more difficult for a citizen to obtain information about a service
is not merely a nuisance, but can also have an important impact on whether and
how the citizen engages with the service. Moreover, experiencing even relatively
small episodes of discrimination in a specific domain may erode the trust that
an individual has in government institutions in general, potentially leading to the
development of an “oppositional culture” with negative consequences, for instance,
in terms of educational achievement (Fryer and Torelli 2010). Furthermore, the
medical and psychological literature provides ample evidence of the negative effect
of discrimination on physical and mental health (Harrell, Hall, and Taliaferro 2003),
including so-called racial microaggressions, that is, subtle everyday experiences of
racism (Wong et al. 2014). Finally, it seems implausible that a discriminatory attitude
would only express itself in a very specific element of the service delivery, without
having a more general impact. In other words, a librarian not replying to requests for
information coming from blacks may also treat blacks differently in other aspects of
the service, for example, by being less forthcoming when asked about a certain library
resource. Thus, our measure of discrimination is informative about the general attitude
permeating local public services.

Our results show that emails signed with a distinctively African-American name
are less likely to receive a reply than identical emails signed with a distinctively whi

te

name, thus indicating the presence of discrimination in access to public services. In
terms of magnitude, given a response rate of almost 72% for white senders, emai

ls

from putatively black senders are almost 4 percentage points less likely to receive an
answer. This differential response is particularly strong among sheriff offices, but is also
present in libraries and school districts. We also find that responses to inquiries coming
from African-American names are less likely to have a cordial tone. Thus, despite a
rising sentiment among whites that the so-called reverse discrimination is on the rise
in the United States (Norton and Sommers 2011) due, for instance, to affirmative
actions, what we find is evidence of discrimination against African Americans by
public service providers. Note that email is an important channel of communication
between the public and government agencies in the United States. A recent survey
(Smith 2010) finds that contacting a government agency via email or a website is the
second most preferred channel (28%)—telephone being the most preferred (35%)—
and is a more preferred option than in-person visits (20%). Among internet users,
online contact is the preferred option.

In the context of our study, the case for statistical discrimination is much weaker
than in a labor market setting. Nevertheless, we also investigate whether the differential
response rate is due to animus or to a form of statistical discrimination arising from
assigning low socioeconomic status to a sender with a distinctively African-American
name. In particular, we deploy two approaches, whereby the first entails predicting the

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 169

race of the recipient and checking whether black recipients are more likely to respond
to emails signed by black senders, as we would expect if taste-based discrimination
by white recipients were at play. In the second approach, we attempt to directly fi

x

the socioeconomic status of the sender—in a second wave of emails that we sent—
by signaling his occupation. A change in the racial difference in responses across
waves due to this additional information would indicate the importance of statistical
discrimination. Both approaches are suggestive of taste-based discrimination being an
important driver of the race gap in the response rate.

Our results are consistent with other studies that uncover evidence of racial/ethnic
discrimination in public services, mostly involving various aspects of law enforcement.
For instance, Alesina and La Ferrara (2014) find evidence consistent with the
presence of racial prejudice in capital sentencing, driven exclusively by Southern
states. Glaeser and Sacerdote (2003) look at vehicular homicides and find that
drivers who kill blacks receive significantly shorter sentences. Abrams, Bertrand, and
Mullainathan (2012) find support for the hypothesis that some judges treat defendants
differently based upon their race. A recent study by political scientists regarding
discrimination in the electoral process (White, Nathan, and Faller 2015) finds that
emails about voting requirements sent to over 7,000 local U.S. election officials from
Latino aliases are significantly less likely to receive a response and, if granted, to
receive responses of lower quality than those sent from non-Latino white aliases.
A related study by Butler and Broockman (2011)—albeit focusing on lawmakers
rather than bureaucrats—involved sending emails asking for help with registering to
vote to almost 5,000 U.S. state legislators. They find that putatively black requests
receive fewer replies than requests coming from white aliases, even when the email
signaled the sender’s partisan preference. Furthermore, Distelhorst and Hou (2014)
find discriminatory behavior against ethnic Muslims by unelected public officials in
China.

To the best of our knowledge, our study is the first to explore racial discrimination
in a variety of local public services that perform important functions and constitute the
majority of interactions between government institutions and citizens. The fact that
we find evidence of discrimination has important implications for public policy, which
we will discuss in the conclusions, after presenting the experimental set up in the next
section and the results in Sections 3 and 4. In Section 5, we explore the association
between our measure of discrimination and the racial gap in wages and attitudes toward
government.

2. The Field Experiment

The field experiment—conducted in March/April 2015—entailed us sending email
queries to over 19,000 local public offices, signing the emails with names that strongly
evoke the race of the sender (white or black). In what follows, we describe the
procedures surrounding the selection of public services, email queries, and names
of sender, as well as the experimental design.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

170 Journal of the European Economic Association

TABLE 1. Emails by type of service.

Recipient Number Percentage

School district 9,873 51.75
Library 4,894 25.65
Sheriff 1,836 9.62
Treasurer 1,129 5.92
Job center 731 3.83
County clerk 616 3.23

Total 19,079 100

Notes: Figures refer to the number of emails sent to each type of service.

2.1. Type of Public Services, Emails and Names of Senders

The first step is to select which public services to target. We chose public services
according to two criteria: (i) the provision of the service is at the county or subcoun

ty

level, to ensure a large number of observations and broad geographic coverage; and
(ii) email addresses are publicly available or a directory of email addresses could be
obtained. We came up with six types of public services that span a variety of local
public services: school districts, local libraries, sheriff offices, county treasurers, job
center veteran representatives, and county clerks. We were able to obtain over 21,000
email addresses and finally use 19,079 valid ones (the sources of email addresses are
listed in Table A.1 in Appendix A).6,7 This constitutes our target population. Table 1
presents the breakdown of numbers and shares of emails by type of public service. The
three most numerous public services are school districts, libraries and sheriff offices,
which jointly account for almost 90% of the emails sent. The emails used in the field
experiment account for nearly 50% of all existing potential recipients (Table B.1 in
Appendix B reports the detailed number of existing recipients and of emails in the
sample).

Figure 1 illustrates the geographic coverage and dispersion of our field experiment.
It is evident that more populated counties and regions—which hence have a larger
number of available recipients (such as the North-East)—receive a relatively large
number of emails.

Figure 2 plots the share of emails against the share of recipients across states. As
can be seen, most observations are clustered around the 45-degree line, suggesting that

6. A small random sample was used for testing; about 2,000 emails were eliminated because they were
undelivered. We checked and corroborated that the probability that an email is eliminated does not correlate
with our key variables.

7. Most of our email addresses are of the “personal” type rather than “generic”, that is, they include the
name or surname of the receiver and are not, for instance, of the type “info@…” or “office@…”. We do not
know, however, who the actual reader is. It could well be the case that an email addressed to the sheriff is
actually read by a deputy or a personal assistant.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 171

Number of recipients per county

1/2 3/4 5/8 9 or more

FIGURE 1. Location of recipients. White fill indicates counties with no recipients.

AK

AL

AR

AZ

CA

CO

CT

DC

DE

FL

GA

HI

IA

ID

IL

IN

KS

KY

LA

MA

MD

ME

MI

MN

MO

MS

MT

NC

ND

NE

NH

NJ

NM

NV

NY

OH

OK

OR

PA

RI

SC

SD TN

TX

UT

VA

VT

WA

WI

WV

WY

0
.0

2
.0

4
.0

6
.0

8
S

h
a

re
o

f
e

m
a

ils
a

cr
o

ss
s

ta
te

s

0 .02 .04 .06 .08
Share of potential recipients across states

FIGURE 2. Sample representativeness. Share of emails corresponds to the number of emails in a
state over the total number of emails in the United States; share of potential recipients corresponds to
the number of potential recipients in the state over all potential recipients in the entire United States.

the number of email addresses is proportional to the number of potential recipients
across states. We will account for any discrepancies in one of our robustness checks.

Our emails contain simple queries that were chosen not to impose significant
effort/investment on the recipients’ part. Specifically, we use two types of email:
simple and complex. Simple emails ask about the opening hours of the office, whereas
complex emails ask for some additional yet basic information that an ordinary citizen
might need to know to carry out a transaction with the office. For example, in the case

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

172 Journal of the European Economic Association

of school districts, the simple email asks the following question: “I would like to enroll
my son in a school in this district and I have some questions. Could you please tell
me what your opening hours are?”. The complex email asks: “I would like to enroll
my son in a school in this district. Could you please let me know the documents I
would need to do this? Do I also need an immunization record?”. The idea of having
both types of emails is to check whether the degree of discrimination changes with the
complexity (i.e., amount of effort) required by the task. Every email has the following
format:

����������������������������������
From: [Black/White Name]
Subject: [Opening Hours] or [Inquiry]

Hi,
My name is [Black/White Name] and I live in [City Name].
[Simple Query/Complex Query]
Thank you,
[Black/White Name]
����������������������������������
To make the name of the sender as noticeable and salient as possible, we show it

three times: in the sender field, the main body and the signature. The complete set of
questions are presented in Table A.2 in Appendix A.

Names of senders were chosen to evoke race as much as possible. We use
two distinctively white names (Jake Mueller and Greg Walsh) and two distinctively
African-American names (DeShawn Jackson and Tyrone Washington). Both the first
names and surnames of our chosen names are among the most recognizable black and
white names and have been previously used in correspondence studies (Bertrand and
Mullainathan 2004; Butler and Broockman 2011; Broockman 2013; White et al. 2015).

We created four email addresses, with the local part comprising two letters and six
numbers and the domain part being “gmail.com”, corresponding with the four chosen
names. In each case, the display name of the email sender was the sender’s full name.
The field “City Name” contains the name of the city where the public service provider
is located.

2.2. Experimental Design

We sent the emails over a period of two weeks due to limits in the number of emails
that can be sent daily. Emails are differentiated by the race of the person who signs
it (white or black) and the type of query that it contains (simple or complex). This
gives rise to a 2 � 2 research design with four treatments that correspond to the four
possible pairs of race/email complexity. In most of the analysis, we pool the two black
and two white names. We randomized the treatments at the state/public service type.
This means, for example, that we randomized across school districts in California, and
then across sheriff offices in California, and so forth. We also randomized the order in

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 173

TABLE 2. County characteristics of Email recipients.

Black White t-test (p-value)

% of blacks in employment 6.10 6.07 0.83
(10.01) (10.25)

Unemployment rate (%) 6.13 6.12 0.82
(2.14) (2.11)

% of Hispanic 10.05 9.83 0.27
(13.91) (13.81)

Average labor income (USD) 794 789 0.17
(239) (237)

Crime rate (%) 2.45 2.44 0.47
(1.31) (1.29)

% of Dem votes 43.05 42.96 0.67
(15.13) (14.99)

Urbanization 71.80 70.68 0.09
(45) (45.53)

N 9,472 9,607

Notes: Standard deviations in parentheses. p-value refers to the t-test for the difference between two means.

which emails were sent across the treatments, as well as across types of public services
and states.

Table 2 shows summary statistics of various county characteristics of recipients
broken down by whether the email that they received was signed by a distinctively
white or black name. As can be seen, our sample is balanced across all of these
characteristics (the data sources of these characteristics are presented in Table A.3 in
Appendix A). As evident from Figure 1, the vast majority of counties (98.37%) are
present in our sample with at least one recipient. Nevertheless, this is not the case
for each type of recipient. Hence, in Table B.2 in Appendix B, we present summary
statistics of the local characteristics of both counties that are covered in our sample
and those that are not, by type of recipient.

Six weeks after the first set of emails were sent to all recipients, we sent a second
wave of emails. The structure of the email was the same as in the first wave, aside from
modifying the signature as illustrated in what follows:

����������������������������������
[Black/White Name]
Real Estate Agent
Buy – Sell – Rent
����������������������������������
The purpose of this is to fix the emails recipients’ perceptions about the

socioeconomic background of the sender. In the second wave, we randomized the
race of the sender and changed the email type, whereby those recipients who received
a simple email in the first wave were sent a complex one in the second wave and vice
versa. To avoid any suspicion that may arise from receiving two emails from the same

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

174 Journal of the European Economic Association

person, we used a different name within the same race for those cases—about half of
the sample—where a recipient was randomized to receive an email signed by the same
race in both waves.

Our main outcome is whether the email is answered. Furthermore, we also
investigate measures of the quality of response by focusing on the number of responses
received, the length of the response (number of words), the delay in the response and
how cordial the response is.

3. Results

3.1. Descriptive Statistics

Overall, about 70% of the 19,079 emails that we sent received a response (see Table B.3
for detailed statistics). This indicates that public service providers are generally quite
responsive to queries coming from the public, despite a non-negligible share of them
going unanswered. The response rate was 68.88% for simple emails and 70.84% for
complex emails, which is surprising given that complex emails seemingly require more
effort from the recipient. A possible explanation could be that responders may consider
the information solicited by simple emails (i.e., opening times) to be easily available
from various sources and thus they feel less compelled to provide an answer to such a
query.

Emails signed by white-sounding names (we will refer to them as “white emails”
hereafter) receive a response in 71.66% of the cases, whereas those signed by black-
sounding names (henceforth: “black emails”) in only 67.96% of the cases, with
the difference of 3.7 percentage points being strongly statistically significant (z-stat,
p-value < 0.000). The response rate to emails coming from the two white-sounding names is almost identical (71.74% for Jake Mueller and 71.57% for Greg Walsh; z-stat, p-value 0.85), whereas there is a difference between the two black-sounding names (69.05% for DeShawn Jackson and 66.91% for Tyrone Washington; z-stat, p-value 0.03). Given that both first names are among the most recognizable African-American names according to Fryer and Levitt (2004), it is possible that this difference emerges because one of the last names has a stronger association with black people than the other.8 Indeed, the 2000 Census shows that among the persons who are called Jackson, 53.02% are black and 41.93% are white, whereas for Washington the figures are 89.87% and 5.16%, respectively. In both cases, the response rate is significantly lower than that for white emails. Hence, hereafter we will consider the difference between white and black emails without distinguishing between the two names within each category.

8. In their analysis of the distribution of first names in California, Fryer and Levitt (2004) find that the
first name Tyrone is slightly more represented among whites relative to DeShawn. In particular, they report
that “[t]here are 463 children named DeShawn, 458 of whom are Black. The name Tyrone is given to 502
Black boys and only 17 Whites.” (Fryer and Levitt 2004, p. 770).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 175

To summarize, the descriptive evidence indicates considerable racial differences
in the response rate to the emails.

3.2. Main Results

Next, we examine whether there are racial differences in response rates in a regression
framework, which allows us to control for various factors such as the type of public
service, state fixed effects and several county characteristics. Specifically, we estimate
linear probability regressions of the form

Responsei D ˇ0
X

Service Typei C � Complex Emaili
C ıBlacki C X 0i � C s C d C “i ; (1)

where Response is a binary variable indicating whether a response to the email was
provided. This variable is coded as 1 if the response was received within 40 days and 0
otherwise. The variable Service Type indicates the type of public service to which the
email was sent, Complex Email is a binary variable indicating whether the email was
simple or complex, Black is a binary variable indicating whether the email was signed
by a distinctively black name, X is a vector of county level characteristics that we use as
controls, s represents state fixed effects and d are indicators for the calendar days when
emails were sent out. Standard errors are clustered at the state/public service type level.
Estimating regressions with state/public service fixed effects produces numerically
similar results across all our specifications. The main coefficient of interest in these
regressions is ı, which tells us whether there is a differential response according to the
racial identity of the sender.

Table 3 summarizes the main regression results. Column (1) includes state, public
service type and sending day fixed effects. The estimated racial gap in response rates—
at 3.84 percentage points—does not substantially differ from that emerging from the
raw comparison reported in the previous section (3.7 percentage points). Column (2)
adds a dummy variable that takes the value of one if the email question is complex.
In line with the raw comparison, complex emails are more likely (1.84 percentage
points) to receive a response than simple emails. In column (3), we examine whether
the differential response rate between white and black emails varies according to
complexity by adding an interaction term between the black name and the complex
email dummies. The estimate of the interaction term proves to be small and statistically
insignificant, indicating that the differential in the response rate is not specific to the
nature of the query. In column (4), which represents our baseline specification, we
include various county level characteristics (unemployment rate, average wage, share
of Hispanic population, crime rate, share of democratic votes, rural/urban dummy).
Unsurprisingly, since the emails are randomly assigned, we find that the inclusion
of these controls does not change the racial difference in response rate estimated in
column (1). Finally, in column (5) we exploit the second wave of emails and particularly
the fact that half of the recipients receive emails from senders with different races across

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

176 Journal of the European Economic Association

TABLE 3. Difference in response rates.

1 2 3 4 5

Black � 0.038��� � 0.038��� � 0.037��� � 0.038��� � 0.032���
(0.006) (0.006) (0.009) (0.006) (0.005)

Complex 0.018��� 0.019�� 0.018��� 0.029���
(0.007) (0.009) (0.007) (0.004)

Black � Complex � 0.002
(0.013)

xY 0.698 0.698 0.698 0.698 0.665
R2 0.045 0.045 0.045 0.049 0.023
N 19,079 19,079 19,079 19,079 38,168

State/service/date F.E. Y Y Y Y Y
County controls N N N Y N
Office F.E. N N N N Y

Notes: Robust standard errors in parentheses clustered at the state/public service type level. ��p < 0.05; ���

p

< 0.01. Dependent variable is a binary variable indicating whether a response to the email was provided (linear probability model). County controls are: unemployment rate, average wage, share of Hispanic in the population, crime rate, share of votes to democrats in presidential elections, and a dummy for rural/urban counties. Office fixed-effects refer to a regression that uses data from the two waves. In the model in column (5), R2 represents the within R2 .

the two waves. We hence estimate a model with office (recipient) fixed effects. The
within-recipient variation in the responsiveness to white and black emails is similar to
that estimated in column (1) (3.22 percentage points).

As mentioned in the experimental design Section, the share of emails for each
state does not perfectly match the share of potential recipients in each state. Despite
generally not being very large, this discrepancy makes some states under-represented
and others over-represented. To correct for this, in unreported regressions, we have
reestimated the model in column (4) of Table 3 by weighting observations by the ratio
of the number of recipients in each state to the number of emails sent in the state. The
estimate (�0.036, s.e. 0.007) is remarkably close to that of the baseline specification.
Next, we checked the sensitivity of our results to the clustering of the standard errors.
Clustering at a level other than state/public service type does not affect the precision of
our estimates. In particular, if we were to cluster at the state/public service type/sender
name level, the standard error of the black dummy coefficient in the baseline model
would be 0.008, whereas if we were to cluster at the county/public service type level,
the standard error would be 0.007.9

9. Two further checks that we perform concern the functional form and the email structure. First, we
estimated the baseline specification using a probit model. The marginal effect (for the baseline �0.040, s.e.
0.006) is remarkably similar to the estimate of the linear probability model, reassuring us that results are
not sensitive to the chosen functional form. Second, we excluded from the sample emails with a generic
structure (i.e., “office@…”, “admin@…”). When performing a regression in the resulting sample (15,851
observations), we find a coefficient estimate of �0.033 (s.e. 0.007).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 177

TABLE 4. Type of public service.

School district Library Sheriff Treasurer Job center County clerk

Black � 0.035��� � 0.041��� � 0.074��� � 0.040 0.008 � 0.016
(0.009) (0.010) (0.021) (0.029) (0.031) (0.048)

xY 0.748 0.670 0.498 0.718 0.725 0.649
R2 0.040 0.047 0.108 0.082 0.164 0.092
N 9,873 4,894 1,836 1,129 731 616

Notes: Robust standard errors in parentheses clustered at the state level. ���p < 0.01. Dependent variable is a binary variable indicating whether a response to the email was provided (linear probability model). xY refers to the average response rate. All regressions include controls of column (4), Table 3.

3.3. Type of Public Service

It is important to recall that our sample comprises six different public services with
different sizes in the sample due to a combination of differences in how many of them
are present in the country and in email availability. One might be interested in knowing
whether our results might be driven by one particular type of public service. Hence, we
analyze the results by type of service. When considering the response rates, the pattern
of higher response rates for white emails holds in all cases, except for job centers. The
raw differences in response rates are presented in Table B.4 in Appendix B.

In Table 4, we estimate the econometric model in equation (1) for each type of
public service. The results essentially confirm the patterns of the descriptive statistics,
with estimates only being statistically significant for school districts, libraries and
sheriff offices and the largest racial difference found in the latter group. This is in line
with the literature mentioned in the Introduction finding evidence of discrimination in
law enforcement. The finding of a lack of discrimination for more administrative-type
jobs, like county clerks or job centers, goes against what found for the U.S. election
officials (White et al. 2015), but the roles are of course very different.10

3.4. Geographic Heterogeneity

Racial disparities might not be equally distributed across the United States. For
example, recent evidence from Stephens-Davidowitz (2014) shows that Google search
queries with racially charged language are particularly intense in Southern states. We
therefore explore whether there is geographic heterogeneity in the racial difference in
the response rate. For this purpose, we split our sample into the four regions defined
by the Census Bureau (North-East, Mid-West, South, and West) and estimate our

10. In unreported analysis, we estimated a regression model that attributes equal importance to each
service. We achieve this by weighting observations by the ratio of the total number of emails sent to the
number of emails of each type of public service. The estimated coefficient (�0.032, s.e. 0.012) is not too
dissimilar from that of the unweighted regressions, showing that the differential treatment between black
and white emails is robust to giving equal weight to each of the six services.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

178 Journal of the European Economic Association

TABLE 5. Heterogeneity by geographical areas.

Regions Counties

North-East Mid-West South West Rural Urban

Black � 0.026�� � 0.049��� � 0.031�� � 0.039�� � 0.058��� � 0.030���
(0.010) (0.010) (0.013) (0.017) (0.011) (0.007)

xY 0.728 0.722 0.635 0.708 0.660 0.714
R2 0.036 0.049 0.075 0.036 0.087 0.038
N 3,666 7,346 4,975 3,092 5,488 13,591

Notes: Robust standard errors in parentheses clustered at the state/public service type level. ��p < 0.05; ���p < 0.01. Dependent variable is a binary variable indicating whether a response to the email was provided (linear probability model). xY refers to the average response rate. All regressions include controls of column (4), Table 3.

baseline specification on each subsample.11 Table 5 summarizes the results. We find a
significant racial gap in all four regions, with the estimate ranging from �0.026 in the
North-East to �0.049 in the Mid-West.12

To further document this pattern of geographical variation, in columns (5) and (6),
we classify counties into urban and rural and split the sample along this dimension.13

This gives rise to 1,312 rural counties and 1,780 urban counties. The results indicate
that the racial gap in response rate is substantially larger in rural areas, namely almost
double than urban areas. This is consistent with the fact that we find a larger racial gap
in the Mid-West, where the incidence of rural counties is highest.

Evidence that the differential treatment in response vis-à-vis African Americans
is not worse in Southern states might appear as striking given the relatively higher
density of black population. However, an important consideration is that the percentage
of blacks employed in public services is also higher in such regions. As we will
document in Section 4, the race of the recipient plays an important role in determining
the magnitude of the racial disparity.

11. The state composition of each region is the following: North-East includes Connecticut, Maine,
Massachusetts, New Hampshire, Rhode Island, and Vermont; Mid-West includes Illinois, Indiana,
Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South
Dakota; South includes Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia,
Washington, D.C., West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana,
Oklahoma, and Texas; West includes Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah,
Wyoming, Alaska, California, Hawaii, Oregon, and Washington.

12. When we split the sample according to the 9 Census Divisions (see Figure C.1), all point estimates
for the black dummy are negative albeit some of them are not statistically significant.

13. We apply the six-level classification developed by the National Center for Health Statistics: Large
central metro, Large fringe metro, Medium metro, Small metro, Micropolitan, and Noncore. We designate
the last category as rural.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 179

TABLE 6. Other outcomes.

Number Length Cordial Delay

Black � 0.001 � 0.114 � 0.062��� � 0.918
(0.003) (1.847) (0.007) (1.147)

xY 1.029 171.041 0.708 24.684
R2 0.015 0.055 0.114 0.031
N 13,321 13,321 13,321 13,321

Notes: Robust standard errors in parentheses clustered at the state/public service type level. ���p < 0.01. Dependent variables are, respectively: number of replies obtained, length of replies (number of words), whether the sender was addressed by name or with salutations, and delay in obtaining a reply (number of hours). xY refers to the average of the outcome variable reported in each column header. All regressions include controls of column (4), Table 3.

3.5. Additional Results: Other Outcomes

The outcome analyzed so far is whether an inquiry receives a reply. In this section we
investigate whether there are differences in the quality of the reply, as measured by the
number of replies sent by the receiver and the length of the email (number of words).
We also use a measure of cordiality of the response: a binary variable concerning
whether the respondent addresses the sender by name or with a salutation.14 Finally,
we consider the intensive margin of replies, measuring the number of hours it takes
for the recipient to reply.

Table B.5 in Appendix B shows descriptive statistics related to these outcomes.
Most respondents sent just one reply, although a few also send some follow-up emails.
The average length of emails is just above 171 words, and it takes on average just over
a day (25 h) to receive a reply. For these three outcomes, a raw comparison suggests
no difference between black and white senders. There appears to be a difference in
the measure of cordiality, with 73.52% of responses to white emails being classified
as cordial as opposed to 67.9% of responses to black emails. This is confirmed by
the regression analysis in Table 6, whereby cordiality represents the only significant
difference between black and whites. Furthermore, Table C.1 in Appendix C shows
that this difference is statistically significant in 5 out of the 6 types of public services
(all but the job centers). Therefore, it appears that black emails are not only less likely
to receive a response but also that—conditional on receiving a response—it is less
likely to have a cordial tone.15

This result seems consistent with evidence of prejudice rather than statistical
discrimination. Even if, for instance, dealing with citizens of low socioeconomic

14. For salutations, we search the text for the following keywords “Hi”, “Mr”, “Dear”, “Hello”, “Good”,
“Thank”.

15. For robustness, we also constructed an alternative definition of cordial reply, defined as 0:5 �
.# cordial words/ C 0:5 � .# characters of all cordial words/, that takes more into account the “intensive” margin
of cordiality. This variable produces a coefficient estimate of �0.110 (s.e. 0.030).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

180 Journal of the European Economic Association

background is more costly in terms of time or effort and recipients use race to infer the
socioeconomic background of the sender, once a response is sent it seems unjustified to
use a less cordial tone toward African-American senders. We explore the interpretation
of our overall results in depth in the next section.

4. Interpretation

Thus far, our results indicate a statistically and economically significant difference in
the response to emails signed by white and African-American names. One possible
interpretation is that this represents taste or prejudice-based discrimination, whereby
responders may have an aversion to interacting with citizens with black-sounding
names due to racially prejudicial attitudes or they may consider such citizens less
worthy of their effort and attention. Another possibility is that the lower response to
black emails represents a form of statistical discrimination, whereby the distinctively
African-American names might signal some other personal trait, besides race, such as a
certain socioeconomic background (Fryer and Levitt 2004). In the labor market context,
it has been argued that employers may use race to infer unobserved characteristics that
are relevant for productivity. Thus, profit maximizing employers may statistically
discriminate against some groups even if they are unprejudiced. In our setting,
characterized by a one-shot interaction between the sender and a public officer, the case
for statistical discrimination is weaker and less likely to be of first order importance.
Nevertheless, it is conceivable to think that public officers may attempt to minimize
effort and perceive senders with a lower socioeconomic status (SES) as requiring more
help. In absence of direct information about SES, public officers may use race to infer
it and, thus, be less likely to respond to queries coming from black-sounding names.
Alternatively, officers may try to minimize the likelihood of being reprimanded, may
consider senders with a low SES as less assertive and, thus, infer that senders with a
black-sounding name are less likely to complain if their query is left unanswered. In
what follows, we explore the relevance of taste-based vis-à-vis statistical discrimination
adopting two approaches: the first uses the inferred race of the respondent and the
second uses the socioeconomic background signaled by the occupation of the sender.
Moreover, we also explore the correlation of a discrimination index constructed using
our data to measures of racial prejudice used in the literature.

4.1. Race of the Recipient

In the first approach, we consider that if statistical discrimination were the primary
driver of the difference, we would expect the recipient’s race not to be an important
predictor of a response to a black email. Accordingly, white and black recipients
should have a similar propensity to respond to names conveying low socioeconomic
background, that is, African-American names.

As a first attempt at assessing this view, consider Figure 3, which plots
a discrimination index from our data against the share of black population in

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 181

AK
AL
AR
AZ
CA

CO

CT
DE
FL
GA
HI
IA

ID IL
IN

KS
KY
LA
MA
MD
ME
MI

MN MO

MS
MT
NC
ND
NE
NH
NJ
NM

NV
NY

OH

OKOR
PA

RI SC

SD

TN
TX

UT

VA
VT

WA

WIWV

WY

Slope: -0.271
Std. Err.: 0.166
R2: 0.042

-.
1

-.
0

5
0

.0
5

.1
.1

5
D

is
cr

im
in

a
tio

n
in

d
e

x

0 .05 .1 .15
% of blacks in employment

FIGURE 3. Difference in response rates and % of blacks in employment. The discrimination index is
obtained by pooling the data of the two waves and aggregating the data at the state level. Observations
are weighted by the number of emails sent in each state. N D 50 (Washington D.C. is excluded).

employment, both at the state level. The discrimination index is obtained by aggregating
responses at the state level and then calculating the racial gap in response rates
(with larger values indicating higher discrimination). To increase the precision of our
measure, we pooled data from both the first and second wave. The relationship between
these two variables—weighted by the number of emails sent in each state—appears to
be negative.16

Since we do not have exact information about the race of the recipient, we try to
proxy for it by using two methodologies. In the first approach, for which we report the
results in Table 7, we proxy for the probability of the person who receives the email
being black using four measures: the share of blacks among employed individuals in
the county (columns (1) and (2)); the share of blacks among employed individuals in
the public sector in the county (columns (3) and (4)); the share of blacks working as
education administrators, sheriffs and librarians in the county (columns (5) and (6));
the share of blacks among school principals in the state (columns (7) and (8)).17

16. To improve readability, we excluded from the graph Washington D.C., where only 44 emails where
sent (23 in wave I and 21 in wave II). In Washington D.C. the share of blacks in employment is 38%
(vis-à-vis 8% at the national level) and the share of blacks in public employment is 34% (whereas only
11% for the whole the United States). In any case, including Washington D.C. does not change the slope
of the regression line in the figure.

17. The correlation between the first two measures is 0.91. Note that, with the exception of the share of
blacks in employment, data are available only for a subset of counties and states. The fact that the estimate
in column (3) is lower than our baseline is not surprising, given that the available counties are all classified
as urban areas, where we know that the racial gap is relatively less pronounced.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

182 Journal of the European Economic Association

T
A

B
L

E
7

.
M

ec
ha

ni
sm

s:
R

ac
e

of
re

ci
pi

en
t.

%
of

bl
ac

ks
In

e

m
pl

oy
m

en
t

In
pu

bl
ic

se
ct

or
A

m
on

g
ed

uc
at

or
s,

A
m

on
g

sc
ho

ol
li

br
ar

ia
ns

an
d

sh
er

if
fs

p

r
in

ci
pa

ls

B
la

ck
�

0.
03

8�
��

�
0.

05
1�

�

�
�

0.
01

9�
�

�
0.

04
8�

��
�

0.
02

3�
�

�
0.
05
1�
��
�
0.
03

2�
��

�
0.

07
3�

��
(0

.0
06

)
(0

.0
08

)
(0

.0
09

)
(0

.0
13

)
(0

.0
10

)
(0
.0
13
)
(0
.0
09
)
(0

.0
15

)
%

of
bl

ac
ks

�
0.

24
3�

��
�

0.
34

4�
��

�
0.

18
1

�
0.

32
3�

��
�

0.
16

5
�

0.
31

6�
�

�
0.

09
0

�
0.

26
5�

(0
.0

72
)

(0
.0

78
)

(

0
.1

15
)

(0
.1

24
)

(0
.1

32
)

(0
.1

40
)

(0
.1

58
)

(0
.1
32
)
B
la
ck
�
%
of
bl
ac

ks
0.

20
8�

��
0.

27
2�

��
0.

28
8�

��
0.

36
6�

��
(0

.0
70

)
(0

.0
90

)
(0

.1
01

)
(0

.1
22

)

x Y
0.

69
8

0.
69

8
0.

72
5

0.
72

5
0.

72
4

0.
72

4
0.

74
4

0.
74

4
R

2
0.

05
0

0.
05

1
0.

02
8

0.
02

9
0.

02
8
0.
02
9
0.

00
9

0.
01

0
N

19
,0

79
19

,0
79

7,
40

6
7,

40
6

6,
04

7
6,

04
7

6,
79

3
6,

79
3

N
ot

es
:R

ob
us

ts
ta

nd
ar

d
er

ro
rs

in
pa

re
nt

h

e
se

s
cl

us
te

re
d

at
th

e
st

at
e/

pu
bl

ic
se

rv
ic

e
ty

pe
le

ve
l.

�
p

< 0.

10
;�

�
p
< 0.

05
;�

�
�

p
<

0.
01

.D
ep

en
de

nt
va

ri
ab

le
is

a
bi

na
ry

va
ri

ab
le

in
di

ca
ti

ng
w

he
th

er
a

re

sp

o

n
se

to
th

e
em

ai
l

w
as

pr
ov

id
ed

(l
in

ea
r

pr
ob

ab
il

it
y

m
od

el
).

x Y
re

fe
rs

to
th

e
av

er
ag

e
re

sp
on

se
ra

te
.

%
of
bl
ac

k

s
in

em
pl

oy
m
en
t

re
fe

r

s
to

th
e

co
un

ty
sh

ar
e

of
bl
ac
ks

am
on

g
th

e
em

pl
oy

ed
po

pu
la

ti
on

.
D

at
a

ha
ve

be
en

ob
ta

in
ed

fr
om

th
e

20
06

–2
01

0
A

m
er

ic
an

C
om

m
un

it
y

S
ur

ve
y

(

A
C

S
).

%
of
bl
ac

ks
in

em
pl
oy
m
en
t
re
fe
rs
to
th
e
co
un
ty
sh
ar
e
of
bl
ac
ks
am
on
g
th

e
po

pu
la
ti
on
em
pl

oy
ed

in
th

e
pu

bl
ic
se
ct

or
.D

at
a
ha
ve
be
en
ob
ta
in
ed
fr
om
th
e
20
06
–2
01
0
A
m
er
ic
an
C
om
m
un
it
y
S
ur
ve
y
(A
C

S
)

an
d
ar
e

av
ai

la
b

l

e
fo

r
57

3
co

un
ti

es
.

%
of
bl
ac

ks
am

on
g

ed
uc

at
or

s,
li

br
ar
ia
ns
an
d
sh
er
if
fs
re
fe
rs
to
th
e
co
un
ty
sh
ar
e
of
bl
ac
ks

w
or

ki
ng

in
th
e
fo

ll
ow

in
g

oc
cu

pa
ti

on
s:

E
du

ca
ti

o

n
A

dm
in

is
tr

at
or

s

,
L

ib
ra

ri
an

s,
L
ib
ra

ry
T

ec
hn

ic
ia

ns
,L

ib
ra

ry
A

ss
is

ta
nt

s,
C

le
ri

ca
l

an
d

S
he

ri
ff

s,
B

ai
li

ff
s,

C
or

re
ct

io
na

l
O

ffi
ce

rs
,a

nd
Ja

il
er

s.
D

at
a
ha
ve
ob
ta
in
ed
fr
om

C
en

su
s

20
00

an
d
A
C

S
2

00
1–

20
15

an
d
ar
e
av
ai

la
bl

e
fo

r
46

9
co

un
ti
es
.
%
of
bl
ac
ks
am
on
g
sc
ho

ol
pr

in
ci

pa
ls

re
fe
rs
to
th
e

st
at

e
sh

ar
e
of
bl
ac
ks
w
or
ki
ng

as
sc

ho
ol

pr
in
ci
pa

ls
.

D
at

a
ha

ve
ob

ta
in

e

d
fr

om
th

e
20

11
–2

01
2

S
ch

oo
ls

an
d

S
ta

ffi
ng

S
ur
ve
y
an
d
ar
e
av
ai
la
bl
e
fo

r
28

st
at

es
.A

ll
re

gr
es

si
on

s
in

cl
ud

e
co

nt
ro

l

s
of

co
lu

m
n

(4
),

T
ab

le
3.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 183

.6
.6

5
.7

.7
5

P
re

d
ic

te
d

p
ro

b
a

b
ili

ty

1 2 3 4 5 6 7 8 9 10

% of blacks in employment, county – deciles

White sender Black sender

FIGURE 4. Race of recipient. Predictions from column (2) of Table 7. The x-axis represents deciles
of the share of blacks in employment. The estimates are calculated at the values of the 5%, 15%, …,
95% percentiles. These values are: 0, 0.002, 0.004, 0.008, 0.013, 0.023, 0.042, 0.072, 0.123, 0.272.

We observe that the share of blacks in the local area (columns (1), (3), (5), and (7))
is associated with a significant reduction in the probability of receiving a response.18

When we interact the share of blacks with the black dummy variable (columns
(2), (4), (6), and (8)), we obtain a positive and statistically significant coefficient,
indicating that the higher the probability of the recipient being African American, the
higher the probability that a black email receives a response. This result is consistent
across the various measures considered. In additional tests, we estimated models
interacting the quartiles of the share of blacks with the black dummy variable, finding
similar results.19

To facilitate an interpretation of these estimates, Figure 4 shows the predicted
probability of response using the estimates in column (2) of Table 7 by deciles of the
distribution of the share of blacks in employment. Predictions are calculated by varying
the values of the share of blacks in employment (represented by the midpoint of each
decile) and averaging over the remaining covariates. The figure shows that there is a
statistically significant difference in the predicted probability of response across races
where the likelihood of the recipient to be black is less than 10% (bottom eight deciles).
In the top two deciles, where the probability that a recipient is black becomes more
substantial, the predicted response rates for the two races become indistinguishable.

18. For example, moving from the 1st to the 3rd quartile of the share of blacks in employment (column
(1)) implies a reduction in the response rate of nearly 2 percentage points (from 71.19% to 69.55%).

19. In particular, in the case of the model with the share of blacks in employment, the estimate for the
baseline category (1st quartile) is �0.073 (s.e. 0.014) and the estimates for the interactions between the
2nd, 3rd, and 4th quartiles and the black dummy are: 0.028 (s.e. 0.017), 0.045 (s.e. 0.020), and 0.063 (s.e.
0.020).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

184 Journal of the European Economic Association

The evidence presented thus far suggests that African-American recipients are
less likely to ignore black emails, supporting the interpretation that the estimated gap
reflects taste-based discrimination. In the second approach, we attempt to identify
the race of the recipient more directly by inferring it from the surname associated
with each email address. Given that each email address in our data (except for Job
Centers) is provided with a contact person, we have name and surname information
for both respondents and nonrespondents. For each surname, we compute two indices
for the “probable race” of recipient, one for black names and one for white names,
corresponding to the frequencies of surnames by race and ethnicity as reported in the
2000 Census.20 The idea is to proxy for the probability that a certain name is white
or black. We can then order surnames in our database according to the confidence by
which we can associate them with a certain race, thereby obtaining a distribution for
the “probable race” index. Subsequently, we set several thresholds corresponding to
fixed percentiles of this distribution. For example, a threshold of 1% means that we
select the top 1% values of the distribution. In the case of African Americans, this
threshold includes values of the probable race index that range from 48.38% (e.g.,
the surname Mack, with the census showing that nearly half of the people holding
this surname are black) to 94.39% (e.g., the surname Ravenell, for which blacks
represent the great majority). In the case of whites, the 1% threshold includes values
ranging from 99% (e.g., the surname Kobylski) to 99.82% (e.g., the surname Sickle).
Lower thresholds include surnames that are less characteristic, for example, for blacks
Nicholson (with a value of 18.74%) and for whites Kline (with a value of 95.38%). We
can subsequently select samples of recipients for which we are increasingly confident
about their association with a specific race and estimate the corresponding racial gap in
response. We present these estimates in Figure 5, which shows that: (i) samples where
the name of the recipient is identified as being black are associated with a smaller race
gap in response than those identified as being white, and (ii) the more accurately (e.g.,
a threshold of 5% or 1%) we can designate the race of the recipient as being black
(white), the smaller (larger) the estimated adverse treatment experienced by blacks,
although estimates become more imprecise. Again, these results are consistent with
the taste-based discrimination interpretation of the differential response rate that we
find.

We also explore whether there exists a negative association between racial
segregation and discrimination, as the contact hypothesis (Allport 1954) would suggest.
We estimate models similar to Table 7 using indices for residential and school
segregation instead of share of blacks. In the case of residential segregation, we
found a negative, statistically significant coefficient for the interaction term (�0.018,
s.e. 0.006). In the case of school segregation, we found a positive but weaker and

20. Since we could not match all surnames, we use both waves for this exercise. We were able to match
surnames for 33,372 observations (87.43% of the sample). Out of the unmatched 4,796 observations, 1,452
relate to Job Centers (for which we did not have a contact name) and 3,344 pertain to surnames that were
present in our data but not in the Census (which reports only surnames with a frequency higher than 100).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 185

-0.03
-0.03

-0.04

-0.02

-0.03
-0.03
-0.04

0.00

-0.06

0.060.06

-0.06
-0

.2
0

-0
.1

5
-0

.1
0

-0
.0

5
0

.0
0

0
.0
5
0
.1
0
0
.1
5
0
.2
0

C
o

e
ffi

ci
e

n
t
o

f
B

la
ck

White 25% 15% 10% 5% 1%
Black 25% 15% 10% 5% 1%

FIGURE 5. Discrimination conditional on probable race of recipient. Points represent regression
coefficients estimated on subsamples of data. Subsamples are defined first by constructing an index
of the probability of belonging to the white or black race (matching surname from our data to the
2000 Census). Thresholds indicate the upper portion of the frequency distribution of the index and
are used to define the various subsamples.

statistically insignificant coefficient for the interaction term (0.015, s.e. 0.014). We
report these results in Table C.2 of Appendix C.

4.2. Fixing the Socioeconomic Background

We next turn attention to the second wave of emails that, as mentioned in the description
of the experiment, includes a signature indicating the sender’s occupation (real
estate agent). Note that, according to data from the Bureau of Labor Statistics for
2014, the annual mean wage of real estate agents ($55,530) is above the annual mean
wage for all occupations ($47,230).21 The literature generally finds no significant
income differences by race.22 Hence, this occupation should act as a signal to the
recipient that the sender is a middle-class person or at least that he does not belong
to a particularly low socioeconomic group. The idea behind this test is that, similar to
Oreopoulos (2011), if providing additional information regarding the sender changes
the racial gap in terms of responsiveness by public officials, then this would be an
indication of statistical discrimination playing a role. We may have an increase or a
decrease in the gap, depending on how black and white real estate agents compare
to generic citizens with black and white sounding names in terms of—for instance—

21. http://www.bls.gov/oes/current/oes_nat.htm.

22. For instance, in their study on almost 7,000 real estate licensees in the United States for 1999,
Benjamin et al. (2007) find no significant differences in hourly wages by race. The same is true for Sirmans
and Swicegood (2000), who focuses on Texas, and Sirmans and Swicegood (1997), who look at Florida.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://www.bls.gov/oes/current/oes_nat.htm

186 Journal of the European Economic Association

TABLE 8. Fixed socioeconomic background.

Reply Cordial reply
Wave II Pooled waves Wave II Pooled waves

Black � 0.037��� � 0.038��� � 0.039��� � 0.061��� � 0.061��� � 0.062���
(0.007) (0.005) (0.006) (0.008) (0.006) (0.007)

Black � Wave II 0.002 0.002
(0.008) (0.011)

xY 0.632 0.665 0.665 0.702 0.705 0.705
R2 0.051 0.052 0.052 0.105 0.107 0.107
N 19,089 38,168 38,168 12,072 25,393 25,393

Notes: Robust standard errors in parentheses clustered at the state/public service type level. ���p < 0.01. Dependent variable in columns (1)–(3) is a binary variable indicating whether a response to the email was provided (linear probability model) and in columns (4)–(6) is a binary variable indicating whether the sender was addressed by name or with salutations (linear probability model). xY refers to the average response rate (columns (1)–(3)) and to the average of the cordiality outcome (columns (4)–(6)). Wave II refers to the follow-up wave where occupation is signaled in the email. Pooled waves refers to the pooling of Wave I and Wave II. All regressions include controls of column (6), Table 3.

the perceived effort that a public officer expects to exert when dealing with a query
from them. Also, the overall response rate may increase or decrease. The critical
aspect behind this test is that information about socioeconomic background may
affect only statistical discrimination, as this depends on information other than race,
whereas animus should depend only on information about race. However, it is possible
that providing information about socioeconomic status shifts the expected cost of
dealing with blacks and whites in a parallel way or that the signal of socioeconomic
status is uninformative or goes completely unnoticed, thus leaving the gap unchanged
even if statistical discrimination plays a role. Thus, although finding a change in
the gap would be an indication of the importance of statistical discrimination,
not detecting a change is not necessarily an indication of a lack of importance.
In Appendix D, we provide a simple analytical framework to illustrate these
points.

Looking at the results, the overall response rate to the second wave’s emails is
slightly lower than the first wave (63.24%). In column (1) of Table 8, we report
the racial difference in response rate estimated from our baseline linear probability
regression using only data from the second wave. We find the racial gap to be 3.68
percentage points, almost identical to what we estimated in the first wave where there
was no occupation signal. Under the assumption that recipients notice the profession
indicated in the signature, this provides evidence that the differential response to white
versus African-American names is not specific to black names that are associated
with low socioeconomic background, but is also present when we compare emails
sent by individuals who belong to a middle income occupational group. In column
(2), we pool observations of the two waves, finding remarkably similar results. In
column (3), we test whether there is a difference in the treatment between black

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 187

and white emails across the two waves, finding that the interaction term between the
black dummy and the dummy for the second wave is very small and not statistically
significant.23 In columns (4)–(6), we estimate the same regressions using cordial reply
as the outcome variable. The estimates are remarkably similar to those obtained in
Table 6, indicating that the differential in the likelihood of receiving a reply with a
cordial tone is not attributable to black names that might evoke low socioeconomic
background.

To summarize, through this exercise we fail to detect the presence of statistical
discrimination.24

4.3. Relationship with Existing Racial Prejudice Measures

We next explore the correlation of our measure of discrimination with racial prejudice
measures used in the literature. In particular, we considered three distinct measures:
the average prejudice index used by Charles and Guryan (2008), the racially charged
search rate used by Stephens-Davidowitz (2014) and an index capturing the race
implicit association test created within the Harvard Project Implicit and described in
Xu, Nosek, and Greenwald (2014). In Table 9 we report the results of models where
we regress our discrimination index (the same that appears in Figure 3 with opposite
sign) on racial prejudice indices used in the above-mentioned studies, and which
are available at the state level. Despite that existing measures capture dimensions of
discrimination different than ours, and that the samples and time periods are different,
we find a moderate positive correlation with our index. This correlation becomes
stronger when controlling for geographic area by introducing dummies for the four
U.S. regions. Together with the results presented in the previous section, this evidence
suggests that prejudice-based discrimination is an important explanation behind our
finding.

23. In addition, exploiting the fact that half of our recipients in the second wave have already seen an
email signed by a white name and the other half have received an email signed with an African-American
name, we checked whether the racial difference in response rates in the second wave depends on whether
the email in the first wave was signed by a black or a white name. To this end, we estimate the baseline
model using only wave II, including an indicator for whether the recipient has received an email signed
by a white name in wave I and its interaction with our key indicator Black. The estimate of the interaction
term is very small and insignificant (0.001, s.e. 0.013), suggesting that the racial gap we estimate in
wave II does not depend upon the race of sender in wave I.

24. To further probe the potential sources of statistical discrimination in our setting, we looked into the
possibility that in areas with high electoral competition public officials may be more likely to worry about
a “vocal voter” (as signaled by the sender’s occupation in Wave II) who is more prone to complain or
communicate with politicians. Using county-level data from the 2012 presidential elections, we constructed
the “effective number of parties”—an index measuring how tight election competitiveness is and that is
widely employed in political sciences (see Laakso and Taagepera 1979). Adding this index to our baseline
regression and interacting it with a wave II dummy, we find that the interaction is statistically insignificant,
suggesting that the above-mentioned mechanism may not be a relevant driver of statistical discrimination
in our case.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

188 Journal of the European Economic Association

TABLE 9. Relationship with racial animus indices.

Average prejudice Racially charged Race implicit
index search rate association test

Discrimination index 0.257� 0.401��� 0.187 0.260�� 0.282�� 0.292���
(0.143) (0.097) (0.140) (0.109) (0.137) (0.085)

R2 0.070 0.643 0.035 0.519 0.080 0.708
N 45 45 51 51 51 51

Region dummies N Y N Y N Y

Notes: Standard errors in parentheses. �p < 0.10; ��p < 0.05; ���p < 0.01. The dependent variable is indicated in the column headers. The independent variable is the index of discrimination in local public services at the state level constructed using Waves I and II. See text for details. Average prejudice index refers to the measure used in Charles and Guryan (2008); Racially charged search rate refers to the measure used in Stephens-Davidowitz (2014); Race implicit association test refers to the index created within Project Implicit and described in Xu et al. (2014).

5. Discrimination and Racial Gaps in Attitudes toward the Government
and Wages

To provide an initial exploration into whether the discrimination that we found has real
consequences, in this section, we explore if our measure of discrimination is associated
with racial gaps in attitudes toward government and wages.

5.1. Attitudes toward Government

To analyze attitudes toward government, we use data from the American National
Election Studies (ANES), an opinion survey on voting attitudes and behavior. To
increase sample size and since some of the questions of interest are asked only in
some years, we pooled all available waves from 1948 to 2012. We constructed several
indicators ranging from trust in government to interest in elections. The details on
how the variables have been constructed and the years for which they are available
are reported in Table A.4 in Appendix A. We then match information on the state of
residence of respondents in the ANES with our discrimination index and estimate the
following regression:

Yi D ˛ C ˇ1Discrimination Indexs C ˇ2Blacki C ˇ3Blacki
� Discrimination Indexs C X 0i � C “i ; (2)

where Y is the outcome of interest (e.g., faith in government) of individual i,
Discrimination Index is the discrimination index at the state level derived from our
data (as described in Section 4.3) and Black is a variable equal to 1 if the respondent’s
reported race is black and 0 if it is white. X is a vector that includes the following
control variables: sex, age, education, marital status, social class, occupation, work
status, race-specific year dummies and four indicators for the region of residence. The

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 189

TABLE 10. Black-white gaps in attitudes toward government.

Faith in local Government Govt poor Not
government performance rating attention interested in

Most Least Local State Federal to people elections

Black � 0.101�� 0.154�� � 0.081� � 0.031 � 0.058 0.135�� � 0.070
(0.044) (0.059) (0.046) (0.042) (0.042) (0.055) (0.044)

Discrimination index � 0.299 0.161 � 0.024 0.043 � 0.204 0.048 � 0.216��
(0.260) (0.271) (0.251) (0.263) (0.235) (0.137) (0.087)

Black � Discrimination � 0.548� 0.057 � 0.470� � 0.426 � 0.958�� 0.441 0.433���
index (0.286) (0.294) (0.253) (0.430) (0.454) (0.313) (0.147)

xY 0.353 0.393 0.771 0.788 0.700 0.294 0.256
R2 0.024 0.025 0.024 0.032 0.042 0.055 0.091
N 4,547 4,156 4,314 4,330 4,412 19,872 27,832

Notes: Robust standard errors clustered at the state level in parentheses. �p < 0.10; ��p < 0.05; ���p < 0.01. Data are from the 1948–2012 American National Election Studies. The sample includes individuals aged 16 or above. The dependent variable is indicated in the column headers. The key independent variable is the index of discrimination in local public services at the state level constructed using Wave I and II. See text for details. Controls include a dummy for sex, age, race-specific year dummies, region dummies, and dummies for: education, marital status social class, occupation, and work status. Observations are weighted by the weights provided in the ANES.

coefficient of interest is ˇ3, which represents how the white-black gap in the outcome
of interest varies with our measure of discrimination in local public services. The
results of this analysis are reported in Table 10.

The outcome in column (1) is an indicator that is equal to one if the individual has
most faith in local government and zero if the individual has most faith in state/federal
government. The second column considers a similar outcome, but individuals are asked
about the level of government in which they have least trust. The estimates of ˇ3 for
both regressions indicate that the white-black gap in terms of faith in local government
(compared to other levels of government) is larger in states where discrimination in
public services is higher, albeit results are statistically significant at the 10% level only
in column (1).

The outcomes in columns (3)–(5) capture individuals’ subjective rating of
government’s performance. We consider all levels of government (local/state/federal)
and construct an indicator that is equal to 1 if the rating is above the median and 0
otherwise. The point estimates of ˇ3 suggest that African Americans have an even
lower rating of government performance if they reside in states where the value of
the discrimination index is higher. However, coefficient estimates are statistically
significant for local and federal governments only.

Next, we consider individuals’ opinions about how much the government pays
attention to what people think. The outcome variable is defined as 1 if the individual
thinks that the government is not paying sufficient attention and 0 otherwise. Estimates
suggest that while African Americans living in states where discrimination in public
services is low are less likely to report a negative opinion, the opposite happens for
those living in states with higher discrimination. However, the estimate of ˇ3 is not
statistically significant.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

190 Journal of the European Economic Association

TABLE 11. Black-white wage gaps.

All Males Females High school Above high
degree or below school degree

Black � 0.136��� � 0.206��� � 0.066��� � 0.125��� � 0.142���
(0.013) (0.017) (0.019) (0.022) (0.014)

Discrimination index � 0.242 � 0.268 � 0.200 � 0.006 � 0.338
(0.224) (0.228) (0.276) (0.241) (0.249)

Black � Discrimination � 0.379 � 0.326 � 0.459 � 0.788� � 0.199
index (0.241) (0.321) (0.283) (0.394) (0.222)

R2 0.334 0.324 0.325 0.205 0.283
N 27,799 15,102 12,697 9,717 18,082

Notes: Robust standard errors clustered at the state level in parentheses. �p < 0.10; ���p < 0.01. Data are from the 2013–2015 May CPS Merged Outgoing Rotation Group. The sample includes individuals aged 16–64 who are in full-time employment. The dependent variable is the log weekly wage. Top and bottom 1% of the wage distribution are trimmed. Controls include a dummy for sex, dummies for education categories, a quadratic on experience, year dummies and region dummies. Observations are weighted by the CPS basic final weight.

Finally, we consider the link between discrimination in local public services and
voting attitudes and behavior by using an indicator that equals 1 if individuals are not
interested in elections and 0 otherwise. The estimate of ˇ3 is positive and statistically
significant at conventional values, suggesting that in states with higher discrimination,
African Americans are more likely to show lower political interest.25

5.2. Racial Wage Gap

To look at racial wage gap, we use data from the Current Population Survey (CPS)
for individuals aged 16–64 in full-time employment. In Table 11 we relate log
weekly wages to race and the discrimination index. What emerges is that African
Americans earn less, both overall and looking at subsamples characterized by gender
or education (above/below high school). The interactions between the black dummy
and our index are all negative, indicating that the gap is wider in states displaying
higher discrimination. The coefficient of the interaction is however significant only for
the low-education subsample.

Overall, in this Section we find that our measure of discrimination has some
correlation with attitudes toward government and racial wage gaps, with states
characterized by higher discrimination displaying more negative attitudes toward
government by African Americans, as well as wider wage gaps.

25. We also explored the correlation between our discrimination index and racial gaps in voting turnout.
We accessed voting data by race related to the 2012 presidential elections—which are available for 34
states—and constructed a voting gap variable defined as the difference in the shares of white and black
people who voted. We then regressed this variable on the discrimination index. We found a positive, albeit
small and statistically insignificant correlation (0.088, s.e. 0.118). Results are nearly unchanged when
adding four indicators for the U.S. regions in the regression (the estimate for the voting gap is 0.024, s.e.
0.085).

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 191

6. Conclusions

We carry out an email correspondence study that aims to identify whether racial
discrimination exists in the provision of information regarding public services offered
by local offices in the United States (school districts, libraries, sheriff offices, treasurers,
job centers and county clerks). Overall, we find that requests of information coming
from a person with a distinctively black name are less likely to receive a reply than those
from a person with a distinctively white name. In our context, statistical discrimination
is unlikely to be of first order importance and a series of tests indeed points toward the
role of animus in explaining our findings.

Besides being illegal, discrimination by public service providers is particularly
startling, since governments could be major players in the effort to eradicate
discrimination in American society. For instance, school districts and libraries can play
an important role in closing the educational achievement gap of black children. Indeed,
our interest in the local level government relates to the fact that low-level bureaucrats
are responsible for the implementation of policy enacted at both the federal and state
level.

One criticism of correspondence studies in the labor market is that these analyses
may not measure labor market discrimination that blacks experience in equilibrium.
The explanation is that blacks may respond to the presence of discrimination by sorting
themselves across firms (e.g., minimizing their contact with the most discriminatory
ones) or adopting different job-search strategies than whites (e.g., sending more
resumés, see Charles and Guryan 2011). This is less of an issue in the case of local
public services, since providers are local monopolies in many cases. Thus, residents
of a given locality cannot usually choose with which school district or sheriff office to
interact. It is indeed true that black citizens may respond to the differential treatment
that we have uncovered by becoming more vocal in asking public officials to fulfill their
duties (for instance, by sending more “reminders” to unresponsive offices). However,
this entails a cost, both psychologically and in terms of time. Moreover, besides “voice”,
there is the alternative option of “exit” (Hirschman 1970), whereby black citizens who
feel discriminated by public offices may reduce their interaction with them as much
as possible, with potentially high costs in terms of the foregone consumption of
public services. In our settings, we cannot investigate which type of reaction prevails.
However, the analysis in the previous Section showed that discrimination is associated
with less engagement with government (as measured by interest in elections).

Overcoming discriminatory practices in local public services is a complex issue.
The persistence of such practices despite their illegality suggests that they will not be
eradicated through a quick legislative fix. Possible interventions include hiring policies
aimed at increasing diversity among the workforce or promoting racial matching
between employees and the communities they serve (Lang 2015). At the practical
level, instituting modes of communication with citizens that do not disclose the name
of the sender (e.g., through online forms) rather than through emails could also be
useful. What this paper shows is that discriminatory practices are present in terms of
access to public services and that policy makers should consider such interventions.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

192 Journal of the European Economic Association

Appendix A: Data Sources and Variable Definitions

TABLE A.1. Data Sources of Email addresses.

Recipient Source of email addresses Accessed/obtained

School district http://schoolinformation.com/ November 3, 2014
Libraries http://www.americanlibrarydirectory.com October 7, 2014
Sheriffs http://www.sheriffs.org October 7, 2014
Treasurers http://www.uscounties.org October 8, 2014
Job centers http://www.servicelocator.org November 18, 2014
County clerks http://www.uscounties.org October 8, 2014

TABLE A.2. Email queries by recipient.

Recipient Simple Query Complex Query

School district I would like to enroll my son in a
school in this district and I have some
questions. Could you please tell me
what your opening hours are?

I would like to enroll my son in
a school in this district. Could you
please let me know the documents I
would need to do this? Do I also need
an immunization record?

Library I would like to become a member of
the library. Could you please tell me
what your opening hours are?

I would like to become a member of
the library. Could you please explain
what I need to do for this? Do I need
proof of address?

Sheriffa I am performing a background check
on a local individual. Could you
please tell me what your opening
hours are?

I am performing a background check
on a local individual. Could you
please tell me what the procedure is
for a criminal record search and how
much it would cost?

Treasurer I am about to purchase a house and I
have some questions about property
taxes. Could you please tell me what
your opening hours are?

I am about to purchase a house. Could
you please explain how I can check
whether there are unpaid taxes on the
house? If there are unpaid taxes, who
would be liable for them?

Job center I am recently unemployed and have
some questions about benefits. Could
you please tell me what your opening
hours are?

I am recently unemployed. Could you
tell me what conditions I need to meet
to be eligible for benefits and how
would I apply to receive them?

County clerk My partner and I would like a
marriage license. Could you please
tell me what your opening hours are?

My partner and I would like a
marriage license. Could you please
let me know the procedure for
applying for one? Also would such
a license only be valid in this county,
or would it be recognized elsewhere?

a Background checks are used, for instance, by prospective employers and landlords.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://schoolinformation.com/

http://www.americanlibrarydirectory.com

http://www.sheriffs.org

http://www.uscounties.org

http://www.servicelocator.org

http://www.uscounties.org

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 193

T
A
B
L

E
A

.3
.

D
at

a
so

ur
ce

s
of
co
un

t

y
va

ri
ab

le
s.

V
ar

ia
bl

e/
da

ta
D

es
cr

ip
ti

on
O

bs
er

va
ti

on
al

S
ou

rc
e

Y
ea

r

W

eb
li

nk
da

ta
un

it
%
of
bl
ac
ks
in
em
pl
oy
m
en
t

N
um

be
r

of
bl
ac
ks
em
pl
oy
ed

ov
er

to
ta

l
em

pl
oy
ed
po
pu
la
ti
on

C
ou

nt
y

A
m

er
ic

an
C

o

m
m

un
it

y
S

ur
ve

y
20

10
ht

tp
:/

/f
ac

tfi
nd

er
.c

en
su

s.
go

v/

%
of
bl
ac
ks
in
pu
bl
ic
se

ct
or

N
um
be
r
of
bl
ac
ks
em
pl
oy
ed

in
pu

bl
ic
se
ct

or
ov

er
to

ta
l

po
pu

la
ti

on
em

pl
oy

ed
in

pu
bl
ic
se
ct
or
C
ou
nt
y
A
m
er
ic
an
C
om
m
un
it
y
S
ur
ve
y
20

06
–2

01
0

ht
tp

:/
/w

w
w

.c
en

su
s.

go
v/

pe
op

le
/e

eo
ta

bu
la

ti
on

/d
at

a/

ee

ot
ab

le
s2

00
62

01
0.

ht
m

l
%
of
bl
ac
ks
am
on
g
ed
uc
at
or
s,
li
br
ar
ia
ns
an
d
sh
er
if
fs
N
um
be
r
of
bl
ac
ks
em
pl
oy
ed

a

s
E

du
ca

ti
on

A
dm

in
is

tr
at

or
s,

L
ib

ra
ri

an
s,

L
ib

ra
ry

T
ec

hn
ic

ia
ns
,
L
ib
ra
ry
A
ss
is
ta
nt
s,
C
le
ri
ca
l
an
d
S
he
ri
ff
s,
B
ai
li
ff
s,
C
or
re
ct
io
na
l
O
ffi
ce

rs
,

an
d

Ja
il

er
s

ov
er
to
ta

l
po

pu
la
ti
on
em
pl
oy
ed
in
th

e
sa

m
e

oc
cu
pa
ti

on
s

C
ou
nt
y
C
en
su
s
an
d
A
m
er
ic
an
C
om
m
un
it
y
S
ur
ve
y
20

00
–2

01
5

ht
tp

s:
//

us
a.

ip
um

s.
or

g/
us

a/
%
of
bl
ac
ks
am
on
g
sc
ho
ol
pr
in
ci
pa
ls
N
um
be
r
of
bl

ac
k

sc
ho
ol
pr
in
ci
pa
ls
ov
er
to
ta

l
nu

m
be

r
of

sc
ho
ol
pr
in
ci
pa
ls
S
ta

te
N

at
io

na
l

C
en

te
r

fo
r

E
du
ca
ti

o

n
S

ta
ti

st
ic

s
20

11
/2

01
2
ht
tp
s:
//

nc
es

.e
d.

go
v/

su
rv

ey

s/

sa
ss

/
ta

bl
es

/s
as

s1
11

2_
20

13
31

3_
p1

s_
00

1.
as

p
%
of

H
is

pa
ni

c
N

um
be

r
of

hi
sp

an
ic

ov
er
to
ta
l
po
pu
la
ti
on
C
ou
nt
y
C
en

su
s:

P
ro

fi
le

of
ge

n-
er

al
po

pu
la
ti
on

20
10

ht
tp

:/
/f

ac
tfi

nd
er

.c
en
su
s.
go
v/

U
ne

m
pl
oy
m
en
t

ra
te

N
um
be
r

of
un

em
pl
oy
ed
pe
op

l

e
ov

er
po

pu
la
ti
on

in
la

bo
r

fo
rc

e
C

ou
nt

y
L

oc
al

A
re

a
U

ne
m

pl
oy

–
m

en
t
S
ta

ti
st

ic
s

20
14

ht
tp
:/
/w
w
w

.b
ls

.g
ov

/l
au

/#
cn

ty
aa

A
ve

ra
ge

la
bo

r
in

co
m

e
A

ve
ra

g

e
w

ee
kl

y
w

ag
e

in
th

e
U

.S
.

do
ll

ar
s

ba
se

d
on

th
e

12
–

m
on

th
ly

em
pl
oy
m
en
t

le
ve

ls
C

ou
nt

y
Q

ua
rt

er
ly

C
en
su
s

of
E

m
pl
oy
m

en
ta

nd
W

ag
es

20
14
ht
tp
:/
/w
w
w
.b
ls
.g
ov

/c
ew

/d
at

at
oc

.
ht

m

C
ri

m
e
ra
te
N
um
be
r

of
vi

ol
en

t
an

d
pr

op
ri

et
y

cr
im

e
ov
er
to
ta
l
po
pu
la
ti

on
C

ou
nt

y
U

ni
fo

rm
C

ri
m

e
R

ep
or

t-
in

g
P

ro
gr

am
D

at
a

20
12

ht
tp

:/
/e

nc
el

ad
us

.i
sr

.u
m

ic
h.

ed
u/

ra
ce

/r
ac

es
ta

rt
.a

sp
%
of

D
em

vo
te

s
N

um
be
r
of
vo
te

s
gi

ve
n

to
B

ar
ac

k
O

ba
m

a
ov

er
to
ta
l
vo
te

s
du

ri
ng

20
12
P
re

si
de

nt
ia

l
el

ec
ti

on
s.

C
ou
nt
y

T
he

G
ua

rd
ia

n
an

d
A

la
sk

a.
go

v
20

1

2
ht

tp
:/

/w
w

w
.t

he
gu

ar
di

an
.c

om
/n

ew
s/

da
ta

bl
og

/2
01

2/
no

v/
07

/u
s-

20
12

–
el

ec
ti

on
-c

ou
nt

y-
re

su
lt

s-
do

w
nl

oa
d

ht
tp
:/
/w
w
w

.e
le

ct
io

ns
.a

la
sk
a.
go

v/
re

su
lt

s/
12

G
E

N
R

/d
at

a/
re

su
lt

s.
ht

m

U
rb

an
iz

at
io

n
R

ur
al

co
rr

es
po

nd
s

to
th

e
ca

te
go

ry
“N

on
co

re
”,

U
rb

a

n
to

“L
ar

ge
ce

nt
ra

l
m

et
ro

,
L

ar
ge

fr
in

ge
m

et
ro

,
M

ed
iu

m
m
et
ro

,
S

m
al

l
m
et
ro
an
d

M
ic

ro
po

li
ta

n”

C
ou
nt
y

N
at

io
na

l
C

e

n
te

r
fo

r
H

ea
lt

h
S

ta
ti
st
ic
s
20

13
ht

tp
:/
/w
w

w
.c

dc
.g

ov
/n

ch
s/

da
ta

_
ac

ce
ss

/u
rb

an
_r

ur
al

.h
tm

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://factfinder.census.gov/

http://www.census.gov/people/eeotabulation/data/eeotables20062010.html

http://www.census.gov/people/eeotabulation/data/eeotables20062010.html

http://www.census.gov/people/eeotabulation/data/eeotables20062010.html

https://usa.ipums.org/usa/

https://nces.ed.gov/surveys/sass/tables/sass1112_2013313_p1s_001.asp

https://nces.ed.gov/surveys/sass/tables/sass1112_2013313_p1s_001.asp

https://nces.ed.gov/surveys/sass/tables/sass1112_2013313_p1s_001.asp

http://factfinder.census.gov/

http://www.bls.gov/lau/#cntyaa

http://www.bls.gov/cew/datatoc.htm

http://www.bls.gov/cew/datatoc.htm

http://enceladus.isr.umich.edu/race/racestart.asp

http://enceladus.isr.umich.edu/race/racestart.asp

http://www.theguardian.com/news/datablog/2012/nov/07/us-2012-election-county-results-download

http://www.theguardian.com/news/datablog/2012/nov/07/us-2012-election-county-results-download

http://www.theguardian.com/news/datablog/2012/nov/07/us-2012-election-county-results-download

http://www.elections.alaska.gov/results/12GENR/data/results.htm

http://www.elections.alaska.gov/results/12GENR/data/results.htm

http://www.cdc.gov/nchs/data_access/urban_rural.htm

http://www.cdc.gov/nchs/data_access/urban_rural.htm

194 Journal of the European Economic Association

T
A
B
L
E
A

.4
.

D
at
a
so
ur
ce
s
of

an
ci

ll
ar

y
va
ri
ab
le
s.
V
ar
ia
bl
e/
da
ta
D
es
cr
ip
ti
on
O
bs
er
va
ti
on
al
S
ou
rc
e
Y
ea
r
W
eb
li

nk
un

it

D
is

si
m

il
ar

it
y

in
de

x
(r

es
id

en
ti

a

l
se

gr
eg

at
io

n)

C
al

cu
la

te
d

as

1 2

P
i

ˇ̌ ˇ̌b
i

B
�

w
i

W

ˇ̌ ˇ̌ ,
w

he
re

b
i
is

th
e
bl
ac

k
po

pu
la
ti
on
of
bl

oc
k

i,
B

is
th

e
to

ta
l
bl
ac
k
po
pu
la
ti
on
in
th

e
tr

ac
t,

w
i
is
th
e
w

hi
te

po
pu
la
ti

o

n
of

bl
oc

k
i,

W
is

th
e
to
ta

l

w
hi

te
po

pu
la
ti
on
in
th
e
tr

ac
t.

T
he
co
un

ty
-l

ev
el

in
de

x
is

ob
ta
in
ed

as
a

w
ei

gh
te

d
av

er
ag

e
of

th
e

tr
ac

t-
le

ve
l

di
ss

im
il

ar
it

y
in

di
ce

s
C
ou
nt
y

R
ac

ia
l

R
es

id
en

ti
al

S
eg

re
ga

–
ti

on
M

ea
su

re
m

en
t
P
ro

je
ct

20
00
ht
tp
:/
/e
nc
el
ad
us
.i
sr
.u
m
ic
h.
ed
u/
ra
ce
/r
ac
es
ta
rt
.a
sp
D
is
si
m
il
ar
it
y
in
de

x
(s

ch
oo

l
se
gr
eg
at
io

n)
C

al
cu

la
te

d
as

1 2
P
i
ˇ̌ ˇ̌b
i
B
�
w
i
W
ˇ̌ ˇ̌ ,
w
he
re
b
i
is
th
e

nu
m

be
r
of
bl
ac
k

st
ud

en
ts

of
sc

ho
ol
i,
B
is
th

e
nu

m
be
r
of
bl
ac

k
st

ud
en

ts
in

th
e
sc
ho

ol
di

st
ri

ct
,

w
i
is
th
e
nu
m
be
r
of
w
hi

te
st

ud
en

ts
of

sc
ho

ol
i,

W
is
th
e
nu
m
be
r

of
w

hi
te
st
ud
en
ts
in
th

e
sc

ho
ol

di
st

ri
ct

.
S
ch

oo
l

di
st
ri
ct
N
at
io
na
l
C
en
te
r
fo

r
E

du
ca
ti
on
S
ta
ti
st
ic
s
20
14
/2
01

5
ht

tp
s:

//
nc

es
.e

d.
go

v/
cc

d/
pu

bs
ch

un
iv

.a
sp

A
ve
ra
ge

pr
ej

ud
ic

e
in

de
x

B
as

ed
on

m
ea

n
of

re
sp

on
se
s
to
pr
ej
ud
ic

e
qu

es
ti

on
s

(s
ee

C
ha

rl
es

an
d

G
ur

ya
n

20
08

fo
r

de
ta

il
s)

S
ta

te
G

en
er

a

l
S

oc
ia

l
S
ur
ve

y
19

72
–

20
04

O
bt

ai
ne

d
fr
om
th

e
au

th
or

s
R
ac

ia
ll

y
ch

ar
ge

d
se

ar
ch

ra
te

G
oo

gl
e

se
ar

ch
es

t

h
at

in
cl

ud
e

a
ra

ci
al

ep
it

he
t

ov
er
to
ta

l
G

oo
gl

e
se
ar
ch

es
S

ta
te
G
oo
gl
e

T
re

nd
s
20
14
/2
01
5
ht
tp
:/

/s
et

hs
d.

co
m

/r
es

ea
rc

h/

R
ac

e
im

pl
ic

it
as

so
ci

at
io
n
te

st
In

de
x
th
at
m
ea

su
re

s
ho

w
lo

ng
it

ta
ke

s
to

a
pe

r

s
on

to
so

rt
as

so
ci
at
io

ns
be

tw
ee

n
A

fr
ic

an
-A

m
er
ic
an
an
d

E
ur

o-
A

m
er
ic
an

fa
ce

s
w

it
h

ev
al

ua
ti

on
s

(g
oo

d/
ba

d)
.

H
ig

he
r

le
ve

l
of

th
e
in
de

x
m

ea
n

st
ro

ng
er

im
pl

ic
it

pr
ef

er
en

ce
fo

r
E

ur
o-

A
m
er
ic

an
fa

ce
s.

O
bs

er
va

ti
on

s
ag

gr
eg

at
ed

by
st

at
e.

A
fr

ic
an
A
m
er
ic

an
s

ar
e

ex
cl

ud
ed

.
S
ta

te
H

ar
va

rd
P

ro
je

ct
Im

pl
ic

it
20

1

4
ht

tp
s:

//
os

f.
io

/q
hp

q3
/

M
os

t
(l

ea
st

)
fa

it
h

in
lo

ca
l

go
ve

rn
m

en
t

E
qu

al
s

on
e

if
th

e
in

di
vi

du
al

ha
s

m
os

t(
le

as
t)

fa
it

h
in

lo
ca

lg
ov

er
nm

en
ta

n

d
ze

ro
if

th
e
in
di

vi
du

al
ha

s
m

os
t

fa
it
h
in
st
at

e/
fe

de
ra

l
go

ve
rn

m
en

t
an
d
ze

ro
ot

he
rw

is
e

In
di

vi
du

al
A

m
er
ic
an
N
at
io
na

l
E

le
ct

io
n

S
tu

di
es

19
68

/1
99

6
ht

tp
:/
/w
w

w
.e

le
ct

io
ns

tu
di

es
.o

rg
/

st
ud

yp
ag

es
/a

ne
s_

ti
m

es
er

ie
s_

cd
f/

an
es

_t
im

es
er
ie
s_

cd
f.

ht
m

L
oc

al
/S

ta
te

/F
ed

er
al

G
ov

er
nm
en
t

pe
rf

or
m

an
ce

ra
ti

ng

In
de

x
ra

ng
in

g
fr

om
0

(l
ow

es
t

pe
rf
or
m
an
ce

)
to

on
e

(h
ig

he
st

pe
rf
or
m
an
ce

)
In

di
vi
du
al
A
m
er
ic

an
N

at
io
na
l

E
le

ct
io
n
S
tu
di

es
19

74
/1

98
0

ht
tp
:/
/w
w
w
.e
le
ct
io

ns
tu

di
es

.o
rg

/
st

ud
yp

ag
es

/a
ne

s_
ti

m
es

er
ie

s_
cd

f/
an

es
_t

im
es

er
ie
s_
cd

f.
ht

m
G
ov
er
nm
en
t

po
or

at
te

nt
io

n
to
pe
op

le
E

qu
al

s
on

e
if

th
e
in
di
vi
du

al
fe

el
s

th
at
go
ve
rn
m
en
t

is
pa

yi
ng

no
t

m
uc

h
at

te
nt

io
n

to
w

ha
t

pe
op

le
th

in
k

an
d

0
ot

he
rw
is
e
In
di
vi
du
al
A
m
er
ic
an
N
at
io
na
l
E
le
ct
io
n
S
tu
di
es

19
64

/2
00

4
ht
tp
:/
/w
w
w
.e
le
ct
io
ns
tu
di
es
.o
rg
/
st
ud
yp
ag
es
/a
ne
s_
ti
m
es
er
ie
s_
cd
f/
an
es
_t
im
es
er
ie
s_
cd
f.
ht
m
N
ot

in
te

re
st

ed
in

el
ec

ti
on
s
E
qu
al
s
on
e
if
th
e
in
di
vi
du

al
is

no
t
m
uc
h
in

te
re

st
ed

in
th

e
el

ec
ti
on
s
an
d
0
ot
he
rw
is
e
In
di
vi
du
al
A
m
er
ic
an
N
at
io
na
l
E
le
ct
io
n
S
tu
di
es

19
52

/2
01
2
ht
tp
:/
/w
w
w
.e
le
ct
io
ns
tu
di
es
.o
rg
/
st
ud
yp
ag
es
/a
ne
s_
ti
m
es
er
ie
s_
cd
f/
an
es
_t
im
es
er
ie
s_
cd
f.
ht
m

W
ag

es
L

og
w

ee
kl
y
w
ag
es
of
w

or
ki

ng
-a

ge
po

pu
la
ti
on
In
di
vi
du

al
C

ur
re

nt
P

op
ul

at
io
n
S
ur
ve
y
20

11
–

20
15
ht
tp
s:
//

cp
s.

ip
um
s.
or

g/
cp

s/
D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://enceladus.isr.umich.edu/race/racestart.asp

http://enceladus.isr.umich.edu/race/racestart.asp

https://nces.ed.gov/ccd/pubschuniv.asp

https://nces.ed.gov/ccd/pubschuniv.asp

http://sethsd.com/research/

https://osf.io/qhpq3/

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

http://www.electionstudies.org/studypages/anes_timeseries_cdf/anes_timeseries_cdf.htm

https://cps.ipums.org/cps/

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 195

Appendix B: Additional Tables and Figures

TABLE B.1. Details of emails.

Recipient N. recipients Sent Undelivered/ Final sample Emails in sample/
emails testing size N. recipients (%)

Wave I
School district 13,567 10,882 1,009 9,873 73
Library 14,638 5,350 456 4,894 33
Sheriff 3,080 2,087 251 1,836 60
Treasurer 3,143 1,252 123 1,129 36
Job center 3,146 890 159 731 23
veteran representative
County clerk 3,143 691 75 616 20
Total 40,717 21,152 2,073 19,079 47

Wave II
School district 13,567 10,882 1,029 9,853 73
Library 14,638 5,350 420 4,930 34
Sheriff 3,080 2,087 247 1,840 60
Treasurer 3,143 1,252 118 1,134 36
Job center 3,146 890 169 721 23
veteran representative
County clerk 3,143 691 80 611 19
Total 40,717 21,152 2,063 19,089 47

Notes: N. recipients refer to the existing number of recipients (i.e., potential recipients, including those with no
email address). Undelivered/testing refers to emails that were not part of the final sample.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

196 Journal of the European Economic Association

TABLE B.2. County characteristics of recipients in sample and not in sample.

School district Library Sheriff

In sample Not in sample In sample Not in sample In sample Not in sample

% of blacks in employment 7.63 9.84��� 5.02 11.28��� 7.77 7.89
(13.28) (14.06) (9.72) (16.15) (13.02) (13.83)

Unemployment rate (%) 6.23 6.68��� 5.95 6.66��� 6.11 6.50���

(2.29) (2.38) (2.15) (2.42) (2.19) (2.44)
% of Hispanic 8.26 8.53 8.64 7.85� 8.70 7.69��

(12.98) (15.29) (13.55) (12.72) (13.5) (12.72)
Average labor income (USD) 716 715 729 701��� 721 710

(190) (201) (199) (179) (186) (197)
Crime rate (%) 2.42 2.25� 2.27 2.56��� 2.48 2.29���

(1.48) (1.64) (1.37) (1.62) (1.52) (1.44)
% of Dem votes 38.75 36.90� 38.85 38.26 38.77 38.33

(14.76) (15.43) (14.49) (15.24) (14.35) (15.48)
Urbanization 58.43 47.97��� 58.40 56.44 61.11 52.49���

(49.29) (50.05) (49.3) (49.6) (48.76) (49.96)

N. counties 2,872 271 1,738 1,405 1,836 1,307

Treasurer Job center County clerk
veteran representative

In sample Not in sample In sample Not in sample In sample Not in sample

% of blacks in employment 3.97 9.97��� 10.04 7.24��� 3.84 8.79���

(8.52) (15) (13.63) (13.23) (6.41) (14.4)
Unemployment rate (%) 5.53 6.69��� 6.37 6.24 6.12 6.31�

(2.1) (2.31) (2.04) (2.37) (1.92) (2.39)
% of Hispanic 10.52 7.03��� 9.22 8.04�� 10.49 7.75���

(15.86) (11.24) (12.71) (13.31) (14.84) (12.7)
Average labor income (USD) 712 719 803 694��� 713 717

(175) (199) (213) (178) (188) (191)
Crime rate (%) 2.18 2.52��� 3.44 2.13��� 2.41 2.40

(1.42) (1.52) (1.49) (1.37) (1.33) (1.53)
% of Dem votes 36.55 39.73��� 44.75 36.98��� 34.96 39.47���

(14.44) (14.92) (14.36) (14.53) (13.6) (14.98)
Urbanization 53.59 59.73��� 86.88 49.90��� 56.66 57.74

(49.89) (49.06) (33.79) (50.01) (49.6) (49.41)

N. counties 1,129 2,014 648 2,495 616 2,527

Notes: Standard deviations in parentheses. �p < 0.10; ��p < 0.05; ���p < 0.01. p-values refer to the statistical significance of the two-sample t-statistic for the difference between the mean characteristics of “In sample” and “Not in sample” counties.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 197

TABLE B.3. Response rates—by sending name.

DeShawn Tyrone Total Greg Jake Total
Jackson Washington Black Walsh Mueller White

Response rate 69.05 66.91 67.96 71.57 71.74 71.66
(46.23) (47.06) (46.67) (45.11) (45.03) (45.07)

N 4,637 4,835 9,472 4,918 4,689 9,607

Difference within race (abs) 2.15 0.17
z-stat (p-value) 0.025 0.855

Difference B–W �2.60 �4.75
z-stat (p-value) 0.001 0.000

Notes: Figures refer to response rates multiplied by 100. Standard deviations in parentheses.

TABLE B.4. Response rates—by type of recipient.

School district Library Sheriff Treasurer Job center County clerk Total

White 76.51 69.08 53.23 73.90 71.93 65.72 71.66
(42.4) (46.23) (49.92) (43.96) (44.99) (47.54) (45.07)

Black 73.10 64.96 46.26 69.57 73.08 64.09 67.96
(44.35) (47.72) (49.89) (46.05) (44.42) (48.05) (46.67)

Difference B–W �3.41 �4.12 �6.98 �4.33 1.14 �1.63 �3.70
z-stat (p-value) 0.000 0.002 0.003 0.107 0.730 0.672 0.000

Notes: Figures refer to response rates multiplied by 100. Standard deviations in parentheses.

TABLE B.5. Other outcomes—summary statistics.

Black White t-test (p-value)

Number of replies 1.03 1.03 0.78
(0.18) (0.19)
[1–3] [1–6]

Length of reply (# words) 171.4 170.7 0.68
(99.2) (101.2)

[0–4213] [2–5324]
Cordial reply 0.679 0.735 0

(0.467) (0.441)
[0–1] [0–1]

Delay in reply (# hours) 24.2 25.2 0.41
(66.8) (73.9)

[0–960] [0–960]

Notes: N D 13,321. Cordial reply refers to whether the sender was addressed by name or with salutations. Standard
deviations in parentheses. Values in square brackets represent the min and max, respectively.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

198 Journal of the European Economic Association

Appendix C: Additional Analysis

TABLE C.1. Cordiality in response—by type of public service.

School district Library Sheriff Treasurer Job center County clerk

Black �0.060��� �0.047��� �0.063�� �0.135��� �0.037 �0.098�
(0.009) (0.014) (0.029) (0.031) (0.047) (0.051)

xY 0.771 0.722 0.462 0.503 0.643 0.502
R2 0.088 0.069 0.168 0.143 0.174 0.192
N 7,385 3,281 914 811 530 400

Notes: Robust standard errors in parentheses clustered at the state level. �p < 0.10; ��p < 0.05; ���p < 0.01. Dependent variable is a binary variable indicating whether the sender was addressed by name or with salutations (linear probability model). NY refers to the average of the cordiality outcome. All regressions include controls of column (4), Table 3.

TABLE C.2. Residential and school segregation.

Residential School

Black �0.038��� �0.038��� �0.031��� �0.031���
(0.006) (0.006) (0.010) (0.011)

Segregation index �0.005 0.004 �0.001 �0.008
(0.005) (0.006) (0.006) (0.007)

Black � Segregation index �0.018��� 0.015
(0.006) (0.014)

R2 0.049 0.050 0.040 0.041
N 19,079 19,079 5,216 5,216

Notes: Robust standard errors in parentheses clustered at the state/recipient level. ���p < 0.01. Dependent variable is a binary variable indicating whether a response to the email was provided (linear probability model).

All regressions include controls of column (4), Table 3. Residential segregation is measured using the dissimilarity
index for all counties. The index has been obtained from the Racial Residential Segregation Measurement

Project website and is defined as 1
2

P
i

ˇ̌
ˇ̌ bi

B
� wi

W

ˇ̌
ˇ̌,where b

i
is the black population of block i, B is the total

black population in the tract, w
i

is the white population of block i, W is the total white population in the
tract. The county-level index is obtained as a weighted average of the tract-level dissimilarity indices (see
http://enceladus.isr.umich.edu/race/racestart.asp for details).

School segregation is measured using the dissimilarity index for school districts obtained from the 2014–15
Common Core of Data from the National Center for Education Statistics. Districts with less than 10 black
students have been dropped. The index of dissimilarity is defined as previously, where b

i
is the number of black

students of school i, B is the number of black students in the school district, w
i

is the number of white students
of school i, W is the number of white students in the school district. Both indices of segregation have been
standardized.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://enceladus.isr.umich.edu/race/racestart.asp

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 199

-.
1
-.
0
5
0
.0
5
B
la

ck
/w

h
ite

g
a

p
in

r
e

sp
o

n
se

r
a

te

NE MA ENC WNC SA ESC WSC MO PA
Census Division

FIGURE C.1. Difference in response rates by census divisions. Black/white gap in response rate is
obtained by estimating the model of column (4) in Table 3 for each Census Division. Bars represent
confidence intervals. NE includes Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island
and Vermont; MA includes New Jersey, New York and Pennsylvania; ENC includes Illinois, Indiana,
Michigan, Ohio and Wisconsin; WNC includes Iowa, Kansas, Minnesota, Missouri, Nebraska, North
Dakota and South Dakota; SA includes Delaware, District of Columbia, Florida, Georgia, Maryland,
North Carolina, South Carolina, Virginia and West Virginia; ESC includes Alabama, Kentucky,
Mississippi and Tennessee; WSC includes Arkansas, Louisiana, Oklahoma and Texas; MO includes
Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah and Wyoming; PA includes Alaska,
California, Hawaii, Oregon and Washington.

Appendix D: Framework for Public Official’s Decision Problem

Here, we sketch a simple analytical framework, along the lines of Neumark (2012)
and Heckman (1998), to think about the decision-making of a public official who has
received an email and needs to decide whether to answer or not.

Suppose that the public official applies the following decision rule when it comes
to responding to email queries: a response will be provided if the perceived cost of
the response is below a certain threshold c. The cost here can be associated with the
effort that the official would need to exert if the citizen asking the initial question (e.g.,
opening hours of library), comes back with further questions or actually decides to
visit the office and interact with the official. In this perspective, the official does not
respond to citizens that are likely to be very demanding in terms of the official’s time
and effort in future interactions.26

Let us denote the fussiness of the citizen by F and suppose that fussiness depends
on the citizen’s characteristics, X. The treatment that the citizen receives, T, will depend

26. Alternatively, the cost may be interpreted as the penalty inflicted to the official by the supervisor in
case of no response and complain by the citizen. In that case, the decision rule is to respond to queries that
have a perceived cost that is above a certain threshold, that is, respond to citizens that are more likely to
complain.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

200 Journal of the European Economic Association

on the fussiness of the citizen and possibly also on the race, R (with R D B for blacks
and R D W for whites). Let us assume for simplicity that T(F(X), R) D X C � IR, with
IR D 1 if R D B and IR D 0 if R D W, and � > 0 capturing taste-based discrimination
against African Americans. The treatment in this example is continuous. However, in
our setting the treatment of senders is dichotomous (reply or not reply). As mentioned,
this can be modeled through a cut-off rule, so that a response is offered (T D 1) if the
perceived cost of the response is below a certain threshold:

T .F .XW /; W / D 1 if E.XW / < c; for white senders (D.1) and

T .F .XB /; B/ D 1 if E.XB / C � < c; for black senders; (D.2) where c is the cost of effort for the official, � is the taste-based discrimination factor, and E(XW), E(XB) are, respectively, the expected personal characteristics of citizens with white- and black-sounding names.

The difference in responses to whites and blacks in wave 1 (what we refer to as
the racial gap) is �Twave1 D T(F(XW), W) � T(F(XB), B) D E(XW) � E(XB) � � ,
and is a combination of taste-based discrimination, � , and statistical discrimination,
E(XW) � E(XB).

In wave 2, we attempt to affect the extent of statistical discrimination,
E(XW) � E(XB), present in wave 1. The logic of what we do in wave 2 is to try
to change the term E(XW) � E(XB) by adding information about the identity of the
sender. Then, if we see the racial gap change across the two waves, we can deduce
that this change can be attributed to a change in the extent of statistical discrimination
across the two waves and, thus, obtain a lower (upper) bound for statistical (taste-
based) discrimination.27 On the other hand, taste-based discrimination depends only
on the perceived race of the sender and, thus, is insensitive to additional information.
From now on, to make the argument sharper we assume a continuous treatment.28 The
intuition that information about socioeconomic background may affect only statistical
discrimination holds, however, also to the case with dichotomous treatment. This is

27. Oreopoulos (2011) applies this logic in the context of a résumé correspondence study by adding
information about the candidates’ unobserved productivity and observing whether it leads to a reduction of
the gap in call back rates across native and immigrant job candidates. He finds little evidence of statistical
discrimination, similarly to us.

28. The assumption of a continuous treatment is not innocuous. Heckman and Siegelman (1993),
Heckman (1998), and Neumark (2012) show that, when the treatment is nonlinear, as in our context,
then even if there are no mean differences in unobservables between the two racial groups, differences in
the variance of unobservables may give rise to differences in response due to statistical discrimination. To
see this, note that the probability that a white and a black sender receives a response, assuming that X is
normally distributed with mean zero, is: ˆ(c=�

w
) and ˆ((c � � )=�

b
), where ˆ denotes the standard normal

distribution and �
w
, �

b
are the standard deviations for whites and blacks respectively. One can readily see

that, if the standard deviations are unequal, then even in the absence of differences in mean characteristics,
one can observe different probabilities of response across the two racial groups also in the absence of
taste-based discrimination, that is, � D 0.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 201

because statistical discrimination depends on information other than race, whereas
taste-based discrimination should depend only on information about race. Indicating
with E.XW jReal Estate/ and with E.XB jReal Estate/ the expected fussiness when
the real estate signal is present of an agent with a white and black sounding name
respectively, then we should consider three cases.

Case (i):

E.XW jReal Estate/ � E.XB jReal Estate/ < E.XW / � E.XB /; (D.3) that is, the official expects real estate agents to be closer in terms of expected fussiness than wave 1 senders. In this case, we would expect to see a reduction in the racial gap comparing wave 1 to wave 2, and this reduction would be evidence of the presence of statistical discrimination.

Case (ii):

E.XW jReal Estate/ � E.XB jReal Estate/ > E.XW / � E.XB /; (D.4)
that is, the official expects real estate agents to be further apart in terms of expected
fussiness than wave 1 senders. In this case, we expect to find an increase in the racial
gap comparing wave 1 to wave 2, and this increase would, again, be evidence of the
presence of statistical discrimination.

Case (iii):

E.XW jReal Estate/ � E.XB jReal Estate/ D E.XW / � E.XB /; (D.5)
that is, there is a parallel shift in terms of expected fussiness. In this knife-edge case,
we would expect to find no change in the racial gap, regardless of the presence or not
of statistical discrimination, and comparing wave 1 to wave 2 would not provide any
evidence regarding the presence of statistical discrimination.

Notice that what has been discussed previously is true also if the signal is negative,
meaning that a real estate agent of a given race is expected to be more fussy
than a generic sender of the same race, that is, E.XW jReal Estate/ > E.XW / and
E.XB jReal Estate/ > E.XB /. This would reduce the probability of responses in the
second wave, which is something that we indeed find, but would still allow us to detect
the presence of statistical discrimination by comparing wave 1 to wave 2, as long as
we are not in case (iii), that is, as long as E.XW jReal Estate/ � E.XB jReal Estate/ ¤
E.XW / � E.XB /.

To summarize, as long as the change in the perceived fussiness of a black citizen
when the real estate identity is attached is different from that of his white counterpart,
we can still detect the presence and provide a lower bound for the importance of
statistical discrimination by comparing the racial gap in response rates across waves.

Finding, as we do, that the racial gap is constant across waves is consistent with
absence of statistical discrimination if E.XW jReal Estate/ � E.XB jReal Estate/
< E.XW / � E.XB / or E.XW jReal Estate/ � E.XB jReal Estate/ > E.XW / �
E.XB /. If, instead, E.XW jReal Estate/ � E.XB jReal Estate/ D E.XW / � E.XB /

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

202 Journal of the European Economic Association

or if the signal goes completely unnoticed, then comparing the two waves is not
informative about the presence or not of statistical discrimination.

References

Abrams, David, Marianne Bertrand, and Sendhil Mullainathan (2012). “Do Judges Vary in Their
Treatment of Race?” Journal of Legal Studies, 41, 347–383.

Alesina, Alberto and Eliana La Ferrara (2014). “A Test of Racial Bias in Capital Sentencing.”
American Economic Review, 104(11), 3397–3433.

Allport, Gordon W. (1954). The Nature of Prejudice. Addison, New York.
Altonji, Joseph G. and Rebecca M. Blank (1999). “Race and Gender in the Labor Market.” In

Handbook of Labor Economics 3C, edited by O. Ashenfelter and D. Card. Elsevier, New York,
pp. 3143–3259.

Bayer, Patrick, Fernando Ferreira, and Stephen L. Ross (forthcoming). “What Drives Racial and
Ethnic Differences in High-Cost Mortgages? The Role of High-Risk Lenders.” Review of Financial
Studies. https://doi.org/10.1093/rfs/hhx035.

Becker, Gary S. (1957). The Economics of Discrimination. University of Chicago Press, Chicago.
Benjamin, John D., Peter T. Chinloy, G. Donald Jud, and Daniel T. Winkler (2007). “Do Some People

Work Harder than Others? Evidence from Real Estate Brokerage.” The Journal of Real Estate
Finance and Economics, 35, 95–110.

Bertrand, Marianne and Esther Duflo (2017). “Field Experiments on Discrimination.” Handbook of
Economic Field Experiments, 1, 309–393.

Bertrand, Marianne and Sendhil Mullainathan (2004). “Are Emily and Greg More Employable Than
Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” American Economic
Review, 94(4), 991–1013.

Bhargava, Saurabh and Dayanand Manoli (2015). “Psychological Frictions and the Incomplete Take-
Up of Social Benefits: Evidence from an IRS Field Experiment.” American Economic Review,
105(11), 3489–3529.

Broockman, David E. (2013). “Black Politicians Are More Intrinsically Motivated to Advance
Blacks’ Interests: A Field Experiment Manipulating Political Incentives.” American Journal of
Political Science, 57, 521–536.

Butler, Daniel M. and David E. Broockman (2011). “Do Politicians Racially Discriminate against
Constituents? A Field Experiment on State Legislators.” American Journal of Political Science,
55, 463–477.

CDC (2011). “Health Disparities and Inequalities Report.” http://www.cdc.gov/mmwr/pdf/
other/su6001 , retrieved 27 July 2015.

Charles, Kerwin Kofi and Jonathan Guryan (2008). “Prejudice and Wages: An Empirical Assessment
of Becker’s The Economics of Discrimination.” Journal of Political Economy, 116, 773–809.

Charles, Kerwin Kofi and Jonathan Guryan (2011). “Studying Discrimination: Fundamental
Challenges and Recent Progress.” Annual Review of Economics, 3, 479–511.

Daponte, Beth Osborne, Seth Sanders, and Lowell Taylor (1999). “Why do Low-Income Households
not use Food Stamps? Evidence from an Experiment.” Journal of Human Resources, 34, 612–628.

Distelhorst, Greg and Yue Hou (2014). “Ingroup Bias in Official Behavior: A National Field
Experiment in China.” Quarterly Journal of Political Science, 9, 203–230.

Doleac, Jennifer L. and Luke C.D. Stein (2013). “The Visible Hand: Race and Online Market
Outcomes.” Economic Journal, 123, F469–F492.

Duflo, Esther and Emmanuel Saez (2003). “The Role of Information and Social Interactions in
Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of
Economics, 118, 815–842.

Ewens, Michael, Bryan Tomlin, and Liang Choon Wang (2014). “Statistical Discrimination or
Prejudice? A Large Sample Field Experiment.” Review of Economics and Statistics, 96,
119–134.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

https://doi.org/10.1093/rfs/hhx035

http://www.cdc.gov/mmwr/pdf/other/su6001

http://www.cdc.gov/mmwr/pdf/other/su6001

Giulietti, Tonin, and Vlassopoulos Racial Discrimination in Local Public Services 203

Fryer, Roland G. (2011). “Racial Inequality in the 21st Century: The Declining Significance of
Discrimination.” In Handbook of Labor Economics 4B, edited by O. Ashenfelter and D. Card.
Elsevier, New York, pp. 855–971.

Fryer, Roland G. and Steven D. Levitt (2004). “The Causes and Consequences of Distinctively Black
Names.” Quarterly Journal of Economics, 119, 767–805.

Fryer, Roland G and Paul Torelli (2010). “An Empirical Analysis of Acting White.” Journal of Public
Economics, 94, 380–396.

Glaeser, Edward L. and Bruce Sacerdote (2003). “Sentencing in Homicide Cases and the Role of
Vengeance.” Journal of Legal Studies, 32, 363–382.

Gneezy, Uri, John List, and Michael K. Price (2012). “Toward an Understanding of Why People
Discriminate: Evidence from a Series of Natural Field Experiments.” NBER Working Paper
17855. National Bureau of Economic Research, Cambridge, MA.

Guryan, Jonathan and Kerwin Kofi Charles (2013). “Taste-Based or Statistical Discrimination: The
Economics of Discrimination Returns to its Roots.” Economic Journal, 123, F417–F432.

Harrell, Jules P., Sadiki Hall, and James Taliaferro (2003). “Physiological Responses to Racism
and Discrimination: An Assessment of the Evidence.” American Journal of Public Health, 93,
243–248.

Hastings, Justine S. and Jeffrey M. Weinstein (2008). “Information, School Choice, and Academic
Achievement: Evidence from two Experiments.” Quarterly Journal of Economics, 123, 1373–
1414.

Heckman, James J. (1998). “Detecting Discrimination.” The Journal of Economic Perspectives, 12(2),
101–116.

Heckman, James J. and Peter Siegelman (1993). “The Urban Institute Audit Studies: Their Methods
and Findings.” In Clear and Convincing Evidence: Measurement of Discrimination in America,
edited by M. Fix and R. Struyk. Urban Institute Press.

Hirschman, Albert O. (1970). Exit, Voice and Loyalty. Harvard University Press,
Cambridge/Massachussets.

Hoxby, Caroline and Sarah Turner (2013). “Expanding College Opportunities for High-Achieving,
Low Income Students.” Stanford Institute for Economic Policy Research Discussion Paper 12-014,
Stanford Institute for Economic Policy Research, Stanford, CA.

Laakso, Markku and Rein Taagepera (1979). ““Effective” Number of Parties: A Measure with
Application to West Europe.” Comparative Political Studies, 12, 3–27.

Lang, Kevin (2015). “Racial Realism: A Review Essay on John Skrentny’s After Civil Rights.”
Journal of Economic Literature, 53, 351–359.

Levine, Ross, Yona Rubinstein, and Alexey Levkov (2014). “Bank Deregulation and Racial Inequality
in America.” Critical Finance Review, 3, 1–48.

Lipsky, Michael (1980). Street-Level Bureaucracy: Dilemmas of the Individual in Public Services.
Russel Sage Foundation, New York.

Neumark, David (2012). “Detecting Discrimination in Audit and Correspondence Studies.” Journal
of Human Resources, 47, 1128–1157.

Neumark, David (2016). “Experimental Research on Labor Market Discrimination.” NBER Working
Paper 22022. National Bureau of Economic Research, Cambridge, MA.

Norton, Michael I. and Samuel R. Sommers (2011). “Whites See Racism as a Zero-Sum Game That
They are now Losing.” Perspectives on Psychological Science, 6, 215–218.

Oreopoulos, Philip (2011). “Why do Skilled Immigrants Struggle in the Labor Market? A Field
Experiment with Thirteen Thousand Résumés.” American Economic Journal: Economic Policy,
3, 148–171.

Pew Research Center (2013). How Americans Value Public Libraries in Their Community.
Washington DC. http://libraries.pewinternet.org/2013/12/11/libraries-in-communities/, retrieved
12 December 2017.

Riach, Peter A. and Judith Rich (2002). “Field Experiments of Discrimination in the Market Place.”
Economic Journal, 112, F480–F518.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://libraries.pewinternet.org/2013/12/11/libraries-in-communities/

204 Journal of the European Economic Association

Rich, Judith (2014). “What do Field Experiments of Discrimination in Markets Tell us? A Meta
Analysis of Studies Conducted since 2000.” IZA Discussion Paper 8584. Institute for the Study
of Labor, Bonn, Germany.

Sabol, William J., Heather C. West, and Matthew Cooper (2009). “Prisoners in 2008.”
Bureau of Justice Statistics Bulletin, (NCJ 228417), 1–45. http://media.arkansasonline.com/
news/documents/2009/12/08/prisoners2008 , retrieved 12 December 2017.

Sirmans, G and Philip Swicegood (1997). “Determinants of Real Estate Licensee Income.” Journal
of Real Estate Research, 14, 137–153.

Sirmans, G. and Philip Swicegood (2000). “Determining Real Estate Licensee Income.” Journal of
Real Estate Research, 20, 189–204.

Smith, Aaron (2010). “The Internet Gives Citizens New Paths to Government Services and
Information.” Pew Research Center. http://www.pewinternet.org/2010/04/27/government-online/,
retrieved 12 December 2017.

Stephens-Davidowitz, Seth (2014). “The Cost of Racial Animus on a Black Candidate: Evidence
Using Google Search Data.” Journal of Public Economics, 118, 26–40.

U.S. Government (2013). “United States Periodic Report to the Committee on Elimination of
Racial Discrimination.” http://www.state.gov/documents/organization/210817 , retrieved 27
July 2015).

White, Ariel R., Noah L. Nathan, and Julie K. Faller (2015). “What Do I Need to Vote? Bureaucratic
Discretion and Discrimination by Local Election Officials.” American Political Science Review,
109, 1–14.

Wong, Gloria, Annie O. Derthick, E.J.R. David, Anne Saw, and Sumie Okazaki (2014). “The What,
the Why, and the How: A Review of Racial Microaggressions Research in Psychology.” Race and
Social Problems, 6, 181–200.

Xu, Kaiyuan, Brian Nosek, and Anthony Greenwald (2014). “Psychology Data from the Race Implicit
Association Test on the Project Implicit Demo Website.” Journal of Open Psychology Data, 2,
p.e3. DOI: http://doi.org/10.5334/jopd.ac.

Supplementary Data

Supplementary data are available at JEEA online.

D
ow
nloaded from
https://academ
ic.oup.com
/jeea/article-abstract/17/1/165/4756072 by U
niversity of C
alifornia, S
an D
iego user on 05 June 2020

http://media.arkansasonline.com/news/documents/2009/12/08/prisoners2008

http://media.arkansasonline.com/news/documents/2009/12/08/prisoners2008

http://www.pewinternet.org/2010/04/27/government-online/

http://www.state.gov/documents/organization/210817

http://doi.org/10.5334/jopd.ac

https://academic.oup.com/jeea/article-lookup/doi/10.1093/jeea/jvx045#supplementary-data

Gomez 1

Are Local Governments Representative?: The Link Between Low Turnout at Local Elections and

Minority Representation

David Alexander Gomez

June 12, 2020

Abstract

Democracy is rule by the people, which is not to be confused with rule by the majority. If

the United States is governed by a democracy, then why are most American minorities

unsatisfied with the representation they have in their local governments? Some theories have

claimed that in the end, democracy will only represent the privileged, but is that where America

intends to remain? Current protests in support of the Black Lives Matter Movement have

solidified minority dissatisfaction with local governments. Politicians in office claim that

protestors are ignorant and do not understand their government. Low levels of turnout make it

plausible that minorities in the United States are detached from local governments. This comes

down to a simple question: why is that the case? Turnout at local elections is considered high if

half of a city’s eligible voters show up to vote. Low turnout has been associated with less

representative local governments. In this literature review, I seek to provide an analytical

meaning of a representative bureaucracy and use literature to assess why local governments in

the United States fail to be representative bureaucracies. The major obstacle to increasing

representation that I find through literature is the low level of turnout at local elections. Local

governments in the United States continue to endorse and maintain structural and institutional

factors that reduce minority participation in local governments.

Gomez 2

Introduction

The idea of a representative democracy was founded with a significantly different

meaning than what it implies today (Elias 2013). As political scientists have shifted their views

on diversity, a more valuable meaning of representative democracy has been created. It holds

that passive representation is the strongest factor that leads to active representation. In this paper

I use literature to answer the question of why minorities are more detached from local

governments than the majority and how that impacts minority representation in local

governments. My goal is to provide why local governments in the United States are not

representative bureaucracies, how local governments push away minority participation, and how

low participation or turnout affects minority representation in local governments. I begin with an

analysis of the qualities of a representative bureaucracy because I believe that most local

governments in the United States fail to meet the guidelines. “Representative bureaucracy” refers

to a local government made up of politicians who represent the local population through duties

such as policy enactment. There is debate that bureaucrats’ race is not the primary cause of

minority representation. Some theorist state that the race of a bureaucrat is not as important as

that bureaucrat’s adaptation of a minority representative role (Bradbury & Kellough 2007).

However, a study on active representation, which states that the beneficiaries are those who are

represented by those in office, implied otherwise. Passive representation, or the employment of

minorities to government jobs, is the best way to create active representation. This draws into my

next point of why representation is important. Then I examine the major reason that local

governments are not representative bureaucracies. Most of the literature I examine allows me to

draw links between low turnout and low representation in local governments. I provide an

analysis of each of these major factors that reduce turnout and tie them directly or indirectly to a

Gomez 3

reduction on minority representation. The factors I examine that reduce turnout are

accountability, election timing, contestation, incumbency effect, voting style, mobilization, and

stake holder interest. All of these factors reduce turnout and are obstacle to fair representation in

local governments. Several, if not all, of these components can be linked to why minorities are

becoming more detached from their local governments.

Representative Bureaucracy

The term representative bureaucracy responds to the question of what is required of a

local government to produce fair representation. Originally, the term was developed by Kingsley

in 1944 to serve White elites who owned land or property. However, since its origin, the term

representative bureaucracy implied that those who are elected into power should mirror the

characteristics of the people they serve (Elias 2013). The term evolved and eventually the factors

of passive and active representation were added by Mosher in 1968. Passive representation is

achieved when a local government hires or appoints bureaucrats with the same demographic

characteristics as the people they serve. If passive representation is achieved, then active

representation is more likely to take place. Active representation takes place when policies that

are enacted properly represent or benefit the citizens who they affect. In more recent times, the

question of fair representation has pushed for policies such as President Obama’s executive order

13583 which requires local governments to promote diversity in the workforce (Elias 2013).

It may seem obvious that passive representation will lead to active representation, but

literature opens the door to new questions on what is required for a local government to be

considered a representative bureaucracy. The literature on representative bureaucracies speaks to

the debate of whether it is essential for bureaucrats to be of the same demographic, primarily

Gomez 4

race, as the citizens they serve to effectively exercise active representation and make local

governments representative bureaucracies. In her work published in 2013, Elias analyzes the

discourse of representative bureaucracy to find why bureaucrats of similar race, or passive

representation, are essential for fair representation. Moreover, Bradbury and Kellough provide a

study that strengthens the essentiality of passive representation. Before examining whether or not

the race of bureaucrats impacts representation, it is important to acknowledge that passive

representation has significant benefits that are undeniable. Passive representation ensures that the

voice of differing citizens is heard during policy enactment, it promotes the legitimacy of

government, and it gives all groups including minorities a stake in their local government

(Bradbury & Kellough 2007).

In Elias’s study of the discourse of representative bureaucracy she examines the

production (text), distribution (discursive practice), and consumption (social practice) on the

subject. She finds that there is a good understanding of what a representative bureaucracy should

hold, but there are huge flaws with distribution and consumption. Although policies, such as

executive order 13583, were enacted to increase diversity, a disagreement on the definition of

diversity causes flaws when put into social practice. A study in 1997, mentioned by Bradbury

and Kellough, provided that bureaucrats did not have to be the same race as the citizens they

served as long as they adopted a “representative role.” A representative role implies that a

politician in office is capable of enacting policies that benefit minorities if they decide to. This

claim does not outweigh the necessity of passive representation, but it is important to analyze the

literature on the topic to draw a conclusion on a representative bureaucracy.

The literature responds to the question of whether or not bureaucrats must share the

demographics of their constituents, which is the problem in the consumption stage of Elias’s

Gomez 5

analysis. Put simply, a White politician in office can absolutely enact redistributive policies or

policies that enhance the social welfare of minorities, but some translate that to mean minorities

are just as “well off” with representatives who do not share their demographics. The study by

Bradbury and Kellough concluded that 76% of Black administrators agreed with Black citizens

while only 44% of White administrators agreed with Black citizens on matters regarding the

wellbeing of Blacks. The questions in their study included support for simple welfare policies to

help Blacks, and the results make it clear that White and Black administrators have different

opinions about the importance and urgency for care assistance for a specific racial category.

Finally, I conclude that a representative bureaucracy describes a local government that

practices passive representation which in return ensues active representation, the entire purpose

of democracy. Politicians better represent citizens when they share critical demographics,

primarily race (Bradbury & Kellough 2007). This leads to the section of why representation is

important and how local governments fail to promote representative bureaucracies.

Fair Representation

Representation is a quality of democracy and all citizens living under a democratic

government are entitled to fair representation. A 2017 study by Sances and You discovered that

local governments that do not have Black representatives in office use a tickets and fines system

that disproportionately affects Blacks. When a Black representative was introduced to the local

government the amount of revenue accumulated by fines disproportionately affecting Blacks

reduced significantly. Representation not only ensures the political motivation and welfare of all

citizens, but it also impacts more practical matters such as how local taxes are spent. A 2018

study by Beach, Jones, Twinam, and Walsh studied the changes of housing prices in minority

Gomez 6

and non-minority neighborhoods and concluded that similar racial demographics of

representatives to the citizens increases the benefits of policies enacted. However, their findings

imply that multiple members must share the demographic and ideologies to actually impact

policy. Policies have been created to push for a more diverse and representative government

workforce, but minorities have not gained the political representation their populations are

entitled to. A 2016 chapter by Hajnal and Troustine provides that nationwide, in regard to city

council seats, Blacks are 12% of the American population but only hold 5.2% of seats, Latinos

are 19% of Americans but only hold 2.7% of seats, and Asians are 5.4% and only hold 0.5%

seats. Local governments spend around 25% of American taxes but mostly represent the White

population (Warshaw 2005). Important to note, increases in immigration populations have

increased the number of White partisans to the republican party (Hajnal & Rivera 2014).

Republicans are more conservative, support a small size (less representative) local government,

and spend far less on redistributive policies. Democrats have become more liberal and are likely

to enact redistributive policies to enhance social welfare, so they are more likely to be the

candidates favored by minorities.

If literature is clear on what fair representation entails, then why do minorities not have

fair representation in local governments? The literature I analyzed covers several different

factors that reduce representation and they all tie to low turnout in local elections, which leads to

the next section.

Low turnout at local elections and its impact on representation

Voter turnout at local elections has been on the decline and is a major cause of poor

representation in local governments. A 2016 study by Hajnal and Troustine found that Whites

Gomez 7

turnout to vote at significantly higher rates than minorities. On average 63% of Whites vote in

presidential elections where only 39% of Latinos, 36% Asians, and 55% Blacks turnout to vote

(Hajnal & Troustine 2016). More importantly, their study used a simulation to check for results

in local elections if turnout was perfect and found that increased turnout by minorities would

absolutely increase representation. In a 2005 article, also by Hajnal and Troustine, they

acknowledge that presidential elections are likely to not be impacted by higher turnout since

plurality voting allows the majority to win, even if that is a 51% majority. This led the authors to

examine local government and the impact of increased turnout on representation. Here it is

important to note that data shows that Blacks and Latinos are likely to live in cities where each

race is around 30% of the population (Hajnal & Troustine 2005). Literature has made it evident

that minorities can promote representation by showing up to vote, but that is easier said than

done. Structural and institutional factors that are kept alive by incumbents push minorities away

from local government. Many factors that reduce turnout are not directly endorsed by

incumbents or the majority, but their unresponsiveness to promote inclusive change is added

contribution to the unchanging status quo of lacking minority representation. In the following

sections I cover several subpoints that decrease turnout and discuss how they directly or

indirectly decrease representation in local governments.

Accountability

The purpose of decentralization and the foundation of over 90,000 local governments in

the United States was to move power closer to those affected by the power (Ribot 1999;

Warshaw 2005). Decentralization was meant to foster a growth in participation by making

politicians in office accessible to the citizens. Since citizens and politicians are technically

Gomez 8

neighbors, it is more difficult for politicians to undermine marginalized citizens. The proximity

allows for accountability, or the ability of citizens to hold politicians responsible for certain

actions. Several factors including low turnout and partisan elections have made politicians less

accountable for their actions. Moreover, the lack of media attention to local politics makes it

difficult to know who is responsible for what (Warshaw 2005). This makes minorities, who are

already much less aware of their local government’s actions, to become more detached from

local government. Intense media scrutiny can increase incumbent politicians’ accountability, but

literature provides other ways for minorities to increase the accountability of elected politicians

and increase voter turnout at local elections that may or may not feature other contests.

In an article published in 2020 Cook, Kogan, Lavertu, and Peskowitz discovered that

increased enrollment in charter schools decreased participation in local elections. The authors

also tied their findings to the fact that the cities they studied had lower Black representation in

local governments in comparison to cities with less charter school enrollment. The key to their

findings is that outsourcing public services, such as education, decreases accountability for local

governments and reduces participation, primarily of minorities, in local government elections. In

other words, outsourcing public services has a negative effect on minority participation in local

government, which in turn reduces representation of minorities. Since politician’s accountability

is reduced, they can undermine minority interests even though citizens are not able to undermine

taxes or local policies that may work against them. With this conclusion, minorities should not

endorse the outsourcing of public services and instead become more involved with their local

government. The resulting increased level of turnout should also increase representation, making

the idea of outsourcing public services unnecessary. This is obviously easier said than done,

since the primary reason for outsourcing education was due to distrust in local government, but if

Gomez 9

local governments are to satisfy the requirements of a representative bureaucracy then

accountability must be scrutinized, and turnout must increase.

Election Timing

Election timing is a major proponent to low turnout in local elections. Countless studies

have found that simply moving an election to on-cycle, or concurrent with presidential elections,

can significantly increase turnout. On a study based in California the researchers found that

elections that were on-cycle had a 40% turnout while off-cycle elections had an average of 18%

turnout (Marschall & Lappie 2018). There are several reason why off-cycle elections receive

such a low level of turnout. Marschall and Lappie note that there is vague information available,

media attention is lacking, and the increased voting cost is more likely to negatively affect

minorities. The policy SB415 in California requires that local governments switch to on-cycle

elections if turnout falls under a certain threshold but around 35% of elections in California are

still off-cycle (Hajnal, Lewis, & Louch 2002). To exacerbate the fact that current elected

officials hold on to non-representative practices, it has been found that holding an election on-

cycle significantly reduces the cost. Researchers found that on-cycle elections expenses average

$25,000 while hosting an off-cycle election averages $58,000. Reform in this area can

significantly increase minority participation and representation in local governments.

Contestation

As shocking as it may sound, the lack of contestation is a factor that reduces turnout at

local elections (Bowler & Donavan 2013). When there is only one candidate running for office it

is easy to wonder why hosting an election has a purpose. In general, it makes the elected

Gomez 10

politicians feel legitimate, but in reality, no competition at local elections defies democracy. A

study by Marschall and Lappie found that more than half of the elections in six states they

studied only featured one candidate. This is an obvious threat to fair representation and

significantly reduces minority participation in local governments. The researchers also found that

only 71% of mayoral elections in California featured more than one candidate from 2011-2014

(Marschall & Lappie 2018). Several factors contribute to the lack of candidates, such as the

incumbency effect and a lack of information, are greatly appreciated by incumbents because it

keeps them in office.

Incumbency Success

Contestation is a major problem that reduces turnout and an important factor contributing

to low contestation is the incumbency effect. Incumbents, or politicians in office, are

significantly more likely to win re-election. Researchers provide that incumbents already have

better resources than opponents to promote themselves. The increased probability of incumbents

winning ranges from 30-80%. In 2002, researchers concluded that incumbents had an 80%

chance of winning at re-election, while a 2005 study concluded that incumbents had a 32%

higher chance than opponents of winning at re-election (Warshaw 2005; Hajnal, Lewis, & Louch

2002). The incumbency effect exists for countless reasons beginning with the fact that

incumbents have power in a locality while opponents do not. This makes it easier for incumbents

to reach and influence voters. Name recognition at the poles increase the probability that less

informed voters will vote for incumbents (Benedictis-Kessner 2017). Benedictis-Kessner

examines different reason why the incumbency effect remains strong. In many cases, competitive

opponents are strayed away due to their knowledge of a significantly lower chance of winning.

Gomez 11

This leaves incumbents to face less competitive opponents (Benedictis-Kessner 2017). With the

incumbent success rate so high, minorities are pushed away from competing in local elections.

Voting Style

According to literature, the most effective way to increase turnout and representation is

by moving away from plurality voting. In the United States, plurality voting has created a system

where the majority always wins. This is a major flaw in democracy and studies have found that

switching to a proportional voting system increases representation. Local governments that use

plurality voting eliminate fair representation since excellent turnout would still mean that

minorities can either agree with the majority or lose to it. Although some forms of proportional

voting, such as cumulative voting, sometimes require a strategy to achieve better representation,

proportional voting has the ability to create fair representation and increase voter turnout (Hajnal

& Troustine 2016). A study conducted in Amarillo, Texas illustrates the effects of proportional

voting systems versus plurality voting (Richie, Amy, McBride 2000). In this study, the

researchers examined the result of the shift from plurality voting to proportional voting in

Amarillo, a local government that severely lacked minority representation. Over 20% of citizens

were minorities but only White candidates had taken office for over twenty years. On the first

election that the local government used proportional voting, two of the four open seats were

taken by minorities. These researchers provide that only a little over two hundred local

governments in the United States have switched to proportional voting but reforming all

elections to proportional voting would open the door to the fair representation that plurality

voting eliminates.

Gomez 12

Mobilization

Studies have found that mobilization can have a huge impact on increasing voter turnout.

Mobilization is as simple as a phone call to inform citizens about their local elections. Minorities

are the least likely to turnout to vote and they are also the least likely to be mobilized.

Information for local elections is poor which makes this lack of mobilization to minorities a

problem with transparency from local governments (Marschall & Lappie 2018). Researchers

have found that mobilization efforts by incumbents are correlated with turnouts twice as high

than when there are no mobilization efforts (Hajnal and Troustine 2016). City councilors are

capable of promoting voter mobilization, but most efforts are directed to Whites (Newman

2014). A survey provided that Whites were more likely to receive calls regarding candidates up

for elections than minorities (Marschall & Lappie 2018). Simple efforts of mobilization can

increase turnout. If voters are well informed by mobilization efforts representation is likely to

increase.

Stake-holder Interest

The final factor I analyze that contributes to minorities’ reduced participation in local

elections is stake-holder interest. Citizens who are less informed and live in poverty are less

likely to vote; both characteristics are more likely to relate to minorities. Minorites are less likely

to attain a college education and suffer from poverty (Ryan and Bauman 2016). These factors

reduce the stake of poor and less-informed minorities in local elections. For example, a poor

citizen who rents an apartment has less stake than a wealthy landowner in an election where

property taxes or land-use reforms are taking place. Moreover, researchers found that while less

than half of citizens earning less than $15,000 a year turnout to vote in presidential elections,

Gomez 13

over 75% of citizens earning more than $75,000 annually turnout to vote (Richie, Amy, McBride

2000). This means that at elections, voters with higher social standings significantly outweigh

those of lower socio-economic status by more than double. The Tiebout model states that

citizens vote with their feet, which means that they can simply move cities if the local

government does not satisfy their needs. However, minorities with low education or in poverty

are constricted from simply packing up and leaving, so instead they become detached from their

local governments (Kelleher & Lowery 2004).

Background

It is important to acknowledge that these studies are estimates of reality. Many of the

studies were conducted in a small number of cities and there are many factors that are

unaccounted for. In particular, city size and demographics can skew results or make certain

conditions nonexistent in some localities. Furthermore, much of the literature that I analyzed is

over a decade old and conditions have likely changed. It is fair to note that these studies may

over-exaggerate or misrepresent some local governments, but the general truth is that minorities

are underrepresented in local governments. Minorities’ lack of participation in local government

affairs is a major factor contributing to the lack of representation but structural and institutional

factors that incumbent politicians fail to reform ensure that participation will not increase. All in

all, the numbers in the studies may be outdated or inaccurate but they are not misleading. The

lack of fair representation is a problem within local governments of the United States.

Gomez 14

Discussion

Studies have found that minorities turnout the least to vote and simulation results provide

that increased turnout can increase representation. There are several barriers that require reform

for minorities to be attracted to their local governments. Although citizens can be reformers, they

are poorly informed of their options and probably don’t know how. Minorities are the least likely

to acquire information about their local governments and therefore it is in the hands of current

politicians to push for reform. Bowler and Donavan provide that attempts to reform are more

likely to come from losing partisans who are self-interested, but they fail to realize that the

purpose is not who wins the election but how well the local government promotes fair

representation. They provided that reforms did not fix much in the countries they studied

(Australia, New Zealand, and Japan), but neither of these countries are as large and diverse as the

United States (Bowler & Donovan 2013). Researchers acknowledge that reform that increases

turnout will not entirely create fair representation, but increased participation from minorities is

more valuable than imaginable (Hajnal, Lewis, & Louch 2002). Less than half of the reforms to

local governments’ structural or institutional proponents are approved (Bowler & Donovan

2013). For reform to be effective it must come from the top, or politicians already in power. It is

obvious that incumbents will not push for change since it is likely to jeopardize their jobs, but

that is more corrupt than self-interested politicians pushing for reforms that will benefit the

citizens of the locality. Minorities must overcome countless obstacles to reach a state where local

governments are representative bureaucracies, but the best starting point is to simply begin to

participate and gain awareness.

Gomez 15

Conclusion

In conclusion, the majority of local governments in the United States are not

representative bureaucracies. Institutional and structural reforms are required for fair

representation to even become a possibility. Incumbents are less likely to push for these reforms

in comparison to minorities who are aware of the ways in which their local government fails

them; therefore it is more likely for change to come by the increase of minority participation in

their local governments, but this requires the mass spread of awareness. Minorities have become

detached from governments that do not serve them. Politicians have found ways to be less

accountable for their actions by being less transparent. The timing of elections not only cost local

governments more to host but increase the cost of voting for minorities in poverty. Low

contestation removes most, if not the entire, purpose of elections. Why would anyone cast a vote

if it is obvious who is going to win? Moreover, the rate of incumbent success sways away

competition and maintains the same structural and institutional factors that could promote fair

representation. Simple efforts from local governments could help mobilize voters and increase

turnout but it doesn’t benefit incumbents, so they have no reason to. In the same vein,

incumbents have no reason to push for reform to switch from plurality voting to proportional

voting because it can cause them to be voted out of office. Minorites stake in local government is

low and declining. Policies such as Obama’s executive order 13583 make diversity a requirement

from local governments but increased participation from minorities is essential for change to

happen. A local government cannot instill passive representation if minorities choose to not

participate. Moreover, resorting to segregation is not feasible and studies provide that

segregation does not increase participation in local government (Kelleher & Lowery 2004).

Local governments fail to represent their diverse population and incumbents have no reason to

Gomez 16

foster change, so the best way to increase minority representation is through minority

participation. Policies that are already in place will facilitate the growth of passive representation

to representative bureaucracies.

Gomez 17

References

– Beach, Brian, Daniel B. Jones, Tate Twinam, and Randall Walsh. 2018. “Minority
Representation in Local Government.” NBER Working Paper.

– Benedictis-Kessner, Justin De. 2017. “Off-Cycle and Out of Office: Election Timing and
the Incumbency Advantage.” Boston Area Research Initiative.

– Bowler, Shaun and Todd Donovan 2013. The Limits of Electoral Reform. Oxford
University Press. Chapter 1-2.

– Bradbury, Mark D. and J. Edward Kellough. 2007. “Representative Bureaucracy:
Exploring the Potential for Active Representation in Local Government.” Oxford

University Press. 697-714

– Cook, Jason B., Vladimir Kogan, Stephane Lavertu, and Zachary Peskowitz. 2020.
“Government Privatization and Political Participation: The Case of Charter Schools.” The

Journal of Politics.

– Elias, Nicole m. Rishel. 2013. “Shifting Diversity Perspectives and New Avenues for
Representative Bureaucracy.” Southern Public Administration Educational Foundation.

331-372

– Kelleher, Christine and David Lowery. 2004. “Political Participation and Metropolitan
Institutional Contexts.” Urban Affairs Review. 720-757

– Marschall, Melissa J. and John Lappie. 2018. “Turnout in Local Elections: Is Timing
Really Everything?” Election Law Journal. Vol.17

– Newman, Ines. 2014. Reclaiming Local Democracy: A Progressive Future for Local
Government. Bristol University Press, Policy Press.

– Ribot, J. C. 1999. “Accountable Representation and Power in Participatory and
Decentralized Environmental Management.” Research Gate.

– Richie, Robert, Douglas Amy, and Frederick McBride. 2000. “How Proportional
Representation Can Empower Minorities and the Poor.” Fair Vote.

– Ryan, Camille L. and Kurt Bauman. 2016. “Educational Attainment in The United States:
2015.” Current Population Reports. 1-11

– Sances, Michael W., and Hye Young You. 2017. “Who Pays for Government?
Descriptive Representation and Exploitative Revenue Sources.” Journal of Politics.

Gomez 18

– Warshaw, Christopher. 2005. “Local elections and representation in the United States.”
Annual Review of political Science. Vol. 22: 461-479

– Zoltan, Hajnal L., Paul G. Lewis, and Hugh Louch. 2002. Municipal Elections in
California: Turnout, Timing, and Competition. Public Policy Institute of California.

– Zoltan, Hajnal L. and Jessica L. Troustine. 2016. “Race and Inequality in Local Politics.”
American Political Science Association. 1-17

– Zoltan, Hajnal L. and Jessica L. Troustine. 2005. “Where Turnout Matters: The
Consequences of Uneven Turnout in City Politics.” The Journal of Politics. Vol.67: 515-

535