Upload
j-h
View
212
Download
0
Embed Size (px)
Citation preview
Jealous of the Joneses: conspicuousconsumption, inequality, and crime
By Daniel L. Hicks* and Joan Hamory Hicksy
*Department of Economics, University of Oklahoma, 308 Cate Center Drive,
Norman, OK 73019, USA; e-mail: [email protected], University of California, Berkeley; e-mail: [email protected]
Empirical research on the relationship between economic inequality and crime has
focussed on income inequality, despite the fact that income is not easily observed by
potential criminals. We extend this literature by shifting the focus from income to its
visible manifestation—conspicuous consumption. Using variation within US states
over time, we document a robust association between the distribution of conspicuous
consumption and violent crime. Our results link violent crime to inequality in visible
expenditure, but not to inequality in total expenditure, suggesting that information
plays a key role in the determination of crime. Furthermore, focussing on conspicuous
expenditure allows for new tests of competing theories of crime. Our findings are
consistent with social theories that link crime with relative deprivation, but provide
little support for traditional economic theory.
JEL classifications: K42, E21, E25, D31, D8.
1. IntroductionThe distribution of resources and its social consequences are of fundamental
interest to economists. Recent research has documented a relationship between
economic inequality and crime both within and across countries (see, for
example, Kelly, 2000; Fajnzylber et al., 2002a; Demombynes and Ozler, 2005).
Across the board, studies have focussed on income inequality and income
poverty as factors motivating individuals to engage in crime. However, this
approach fails to recognize that income is actually an opaque measure to
potential criminals, particularly at the individual level. Instead, we argue that the
act of consumption and the display of opulence drive home the reality of social and
economic inequality within a community.1
..........................................................................................................................................................................1 In particular, we argue that individuals are typically unaware of their neighbors’ income and instead
form estimates of income based on other observable characteristics of their neighbors, such as their
occupation or their assets, which serve as a record of previous consumption activities.
! Oxford University Press 2014All rights reserved
Oxford Economic Papers 66 (2014), 1090–1120 1090doi:10.1093/oep/gpu019
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
We examine the relationship between crime and inequality in a framework
that incorporates observable information on household consumption signalling
using new measures of the visibility of consumption constructed by Charles et al.
(2009) and Heffetz (2011). Our analysis focusses on the distribution of visible
consumption and criminal behavior within US states over a period of nearly two
decades. First, we document a positive and significant association for components
of violent crime, but not for components of property crime, a finding that is
consistent with the existing literature, which employs income rather than expend-
iture (Kelly, 2000). Second, we show that the observed relationship between
the distribution of expenditure and crime holds only for inequality in visible ex-
penditure and not for inequality in total expenditure. This evidence suggests that
the visibility of expenditure conveys information which criminals factor into the
decision to commit crime. Finally, we outline an empirical method for sorting
between competing explanations of the relationship between inequality and
crime by exploiting variation in the visibility of consumption activities and in
types of criminal activity. Our results are consistent with social theories
that connect violent crime with relative deprivation and provide mixed evi-
dence for economic theories that explain property crime using a cost-benefit
framework.
Justification for distinguishing between consumption and income inequality can
be found in an established macroeconomic literature. Blundell and Preston (1998)
make the case that because households smooth consumption, expenditure may
better reflect both lifetime resources and welfare than income, making consump-
tion a preferable statistic for studying inequality. Similarly, Dahlberg and
Gustavsson (2008) show that increases in permanent income are associated with
higher levels of crime, whilst transitory income shocks produce no effects. The
permanent income hypothesis suggests that expenditure measures should provide
more information on lifetime resources than disposable income (Charles et al.,
2009). Together these facts suggest that consumption instead of income may be
a more appropriate approach to examining the motivations for criminal behavior.
This distinction is also particularly important when the measures diverge empir-
ically, and an active research agenda seeks to contrast consumption and income
inequality, producing mixed evidence on the extent to which these measures move
together in the USA (Krueger and Perri, 2006; Aguiar and Bils, 2011; Attanasio
et al., 2012).
Current economic theory on criminal behavior is derived primarily from Becker
(1968), who suggests that individuals evaluate the pros and cons of engaging in
crime and then choose their optimal course of action. In this framework, the size of
the gap between the poor (who may be considering crime) and the rich (with
resources to be stolen) accounts for much of the expected net benefit of criminal
activity.2 Thus, when inequality increases, ceteris paribus, the incentive faced by
potential criminals to engage in crime also becomes stronger. A high degree of
..........................................................................................................................................................................2 Probability and costs of being caught also enter into this calculation.
d. l. hicks and j. h. hicks 1091
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
inequality or a relatively severe level of poverty could also be interpreted as a low
opportunity cost of engaging in illicit activity because choosing a life of crime
would mean forgoing a smaller amount of legally obtained income.
Two prominent theories within the sociological literature provide alternative
explanations for the relationship between inequality and crime. Beginning with
Merton (1938), strain theory argues that the inability to attain pecuniary success
through legal endeavors creates a sense of disenfranchisement with societal con-
ventions and leads individuals to view crime as a viable alternative. Under strain
theory, rising inequality causes feelings of relative deprivation to heighten,
furthering the motivation for criminal behavior. In a similar vein, social disorgan-
ization theory suggests that inequality directly weakens community ties by reducing
social capital. Geographic areas with high economic inequality exhibit above
average levels of poverty, residential mobility, racial heterogeneity, and family in-
stability—factors that have been associated with higher levels of criminal behavior
because (according to the theory) they serve to attenuate social cohesion (Kelly,
2000).
These sociological explanations are consistent with the early consumer demand
theory of Duesenberry (1949) and Leibenstein (1950), which asserts that individ-
uals obtain utility not only from their own consumption but also from the level of
their consumption relative to others. Recently economists have begun to search for
consumption externalities of this nature in empirical data, often referred to in the
literature as demonstration or Veblen effects. For instance, Luttmer (2005) finds
evidence in the USA that increased earnings amongst an individual’s neighbors
negatively affects self-reported happiness.3 Similarly, Maurer and Meier (2008)
document that individuals make an effort to smooth their consumption relative
to the consumption of their neighbors; in an attempt to ‘keep up with the Joneses’.4
Likewise, the prominent explanation for the Easterlin paradox—in which survey-
based measures within a country show individuals with higher income reporting
greater happiness, yet corresponding increases in national income show no effect
on happiness—is that individuals care more about their relative position in society
than their absolute position (Easterlin, 1995; Luttmer, 2005). The idea that indi-
viduals lack perfect information concerning peer income and that knowledge of
relative earnings can affect personal well-being and happiness has also been
confirmed in recent experimental research. Card et al. (2012) demonstrate that
when employees of the University of California system were informed of
colleague salaries, those individuals earning less than the mean salary reported
significantly reduced job satisfaction.
..........................................................................................................................................................................3 This effect appears to be stronger for individuals who report being more social with their neighbors,
indicating they are more likely to place their neighbors within their reference or peer group.4 Maurer and Meier find that whilst this effect is robust, it is moderate in magnitude. One concern is that
employing panel data from the US Panel Study of Income Dynamics restricts their observation of
consumer behavior to food purchases, which may not truly capture the intended behavior they hope
to evaluate.
1092 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Our work is also related to and borrows from the literature on consumption
visibility. This concept can be traced to early thinkers including Plato, Thomas
Hobbes, and Adam Smith, all of whom recognized that expenditures could be used
to display wealth or to gain honor (Heffetz, 2011). The present analysis invokes the
notion of conspicuous consumption associated most prominently with the writings
of Thorstein Veblen (1899), who coined the term when describing visible expend-
iture for the purpose of demonstrating wealth within society. Modern economic
research often approaches conspicuous consumption within the context of
signalling models, wherein individuals derive utility from the ability to showcase
wealth or status through their consumption behavior (Corneo and Jeanne, 1997;
Charles et al., 2009; Heffetz, 2011).
Finally, a related literature examines the possibility that individuals take into
account the risk of expropriation when they make economic decisions that
might convey signals of wealth, such as when purchasing a car. An undesired
externality of visible expenditure may be that it offers criminals improved infor-
mation about the potential benefits of crime. For instance, De Mello and Zilberman
(2008) find increased savings rates in cities of Sao Paulo, Brazil, which have higher
rates of property crime. Similarly, in the USA, Mejıa and Restrepo (2010) argue that
visible goods convey information about an individual’s wealth to potential
criminals. They subdivide visible goods by their potential for loss due to theft
and document a negative and significant impact of property crime on the con-
sumption of goods they classify as ‘visible but non-stealable’, which they focus on
to isolate the role of information in the determination of criminal behavior.
This article proceeds as follows. Section 2 reviews the existing literature, which
examines the relationship between economic inequality and crime. Section 3
describes the data used in the current analysis and presents summary statistics.
Section 4 explores the relationship between inequality and crime by focussing on
visible consumption over time within US states. Section 5 outlines the leading
theories explaining this relationship and presents an empirical strategy for distin-
guishing amongst them. Section 6 concludes.
2. Inequality and crimeA sizeable literature explores the empirical relationship between inequality and
crime in the USA. For instance, Blau and Blau (1982) find that income
inequality has a sizeable and statistically significant relationship with violent
crime using data from 125 US metropolitan statistical areas (MSAs). They argue
that inequality helps explain the relationship between poverty and crime, accounts
for the higher rates of crime in the US South, and drives much of the observed
relationship between racial composition and crime rates. More recently, using data
from 829 urban counties in the USA, Kelly (2000) finds that income inequality is
correlated with violent crime but not with property crime. Consistent with an
economic explanation for criminal behavior, increases in deterrence measures,
such as the size of the police force, have been shown to significantly reduce crime.
d. l. hicks and j. h. hicks 1093
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Estimates from a number of international studies have generally agreed with the
results from the US-based literature. Analysing changes in the wage structure in
the UK from 1975 to 1996, Machin and Meghir (2004) show that declines in
the relative earnings of low-wage individuals are associated with higher levels of
property and vehicle crime. Nilsson (2004), studying Sweden, finds that the
proportion of relatively poor individuals is associated with higher levels of
property crime but produces mixed results for violent offences. Demombynes
and Ozler (2005) examine the relationship between income inequality and crime
in South Africa at the level of the police precinct. They find that inequality within a
precinct is highly correlated with property crimes but not with violent crime. The
nature of their data further permit the examination of income disparities between
neighborhoods, and the authors find that precincts which are the wealthiest of
their geographic neighbors experience burglary rates 25–43% higher than less
affluent nearby precincts. They also show that violent crime is more frequent
when inequality exists across larger regions, such as over sets of bordering
neighborhoods.
Although most empirical studies examine local, regional, or national crime, there
is some research that examines inequality and crime (primarily violent) across
countries. Combining information on inequality from the Deininger and Squire
database with the UN World Crime Surveys, Fajnzylber et al. (2002a, 2002b)
document that inequality and violent crime rates are positively correlated across
a panel of nearly 40 countries. Furthermore, they find that this correlation is
stronger across countries than within. Similarly, Soares (2004) contends that
while the within-country evidence on inequality and crime is somewhat inconclu-
sive, research to date across countries provides support for a positive relationship.5
After correcting for the fact that crime reporting rates are correlated with economic
development, he documents a relationship between income inequality and crime,
concluding that ‘reducing inequality from the level of a country like Colombia to
levels comparable to Argentina, Australia, or United Kingdom, would reduce thefts
by 50%, and contact crimes by 85%’ (2004, p.178).
Hsieh and Pugh (1993) conduct a meta-analysis of the relationship between
poverty rates, income Gini coefficients, and violent crime, and find that 74 of 76
estimates (using levels of aggregation varying from the neighborhood level to cross-
country studies) document a positive (and for the most part statistically significant)
relationship. Although estimates of the size of this relationship vary widely across
studies, nearly 80% find a correlation of 0.25 or greater. There are a number of
possible reasons studies of inequality and crime have produced estimates of varying
magnitude and statistical significance. For instance, Brush (2007) documents that
cross-sectional and time-series analyses in the USA yield divergent results, poten-
tially because time-series estimates rely on relatively short changes in inequality
whilst inequality may only affect crime over the long run. As previously discussed,
..........................................................................................................................................................................5 One potential explanation for the wide range of findings is the diversity of techniques applied to
understanding the within-country relationship across economists, sociologists, and criminologists.
1094 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
income inequality may capture transitory and not permanent income, leading to
misleading results, as suggested by the findings of Dahlberg and Gustavsson (2008).
This article departs from the existing literature by examining the inequality–
crime relationship in a new and unique way that more appropriately matches
both theory and intuition. A neighbor’s income and bank account balance are by
no means perfectly observable for academic researchers or individuals considering
committing crimes.6 Nor are these factors likely to create strain within society or
inform potential criminals of the expected gains from crime in the way they are
invoked in the theories we described in the introduction. Intuition suggests that the
relevant factor should be how individuals spend and display their income. Thus, by
focussing on visible consumption activities, this analysis extends the existing
literature by studying the relationship between crime and the physical manifest-
ations of inequality. Furthermore, we show that these novel measures of inequality
can help distinguish between existing theories of criminal behavior.
3. Data3.1 Expenditure data
We use information on household consumption expenditure, income, and demo-
graphics in the US from the Consumer Expenditure Survey (CEX). The CEX is a
nationally representative survey that contains some of the longest-running,
consistent, and most detailed information on consumer expenditure available.
Our records come from the National Bureau of Economic Research’s family-level
extracts, which have been recompiled by Charles et al. (2009) for the period 1986–
2002. Although CEX data are available over a longer time period, the NBER extracts
constructed by Harris and Sabelhaus (2000) for the specific period 1986–2002
include a set of consumption categories that remain uniform in definition over
time, making these particular years ideally suited to our analysis.7
The CEX is conducted as a rotating panel. In any given year, there are approxi-
mately 20,000 interviews nationwide, and the typical year includes approximately
5,000 unique households. Households in the CEX are interviewed quarterly for up
to five consecutive quarters, though not all households complete all interviews. We
include in our analysis all households that have a state of residence identifier, did
not change state of residence during the interview period, report positive total
expenditure, and have a household head who is at least 18 years of age.8
Our analysis is run at the level of the state-year, and in aggregating the
..........................................................................................................................................................................6 In spite of their access to personal income data, we are aware of no evidence that CPAs are more likely
to be involved in crime.7 As we detail later, our analysis only includes the years 1986–2001 due to missing information for other
covariates. Blundell et al. (2006, p.1) suggest that the ‘Consumer Expenditure Survey (CEX) provides
comprehensive data set on the spending habits of US households’.8 The most common reason for dropping a household from our data set is that it lacks a state identi-
fier—nearly 20,000 households lack this information in the data, likely because they live in smaller or
more sparsely populated states. Only one household was dropped because it reported negative average
d. l. hicks and j. h. hicks 1095
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
household-level CEX data we further limit the sample to state-years that contain at
least 25 households. After these restrictions, we obtain 68,624 households in our
sample for a total of 544 state-year observations. Nine states—Maine, Mississippi,
Montana, New Mexico, North Dakota, Rhode Island, South Dakota, West Virginia,
and Wyoming—as well as Washington, DC, are not included in our analysis due to
these restrictions.
Before reviewing summary statistics for the CEX, we note two limitations of this
data that are relevant for our analysis. First, the survey is not specifically designed to
be representative at the state level. Second, the public release files exclude
household-level data for some states and years as discussed (generally this is
done for states that are smaller or more sparsely populated) to maintain
respondent anonymity. Although neither of these limitations are especially
damaging for our estimation, which occurs within states over time, this does
alter the generalizability of our findings. With the notable exception of the
District of Columbia, many of the excluded states are both less densely
populated and likely experience lower levels of crime. Therefore, a caveat of our
analysis is that it may not perfectly generalize to the population at large and to
rural, less populated states in particular.
Table 1 presents summary statistics on the CEX data for our analysis sample. The
average household over the 1986–2001 period has 2.8 individuals, two of whom are
adults. Nearly 74% of household heads classify themselves as Caucasian, and 13%
are African American. Thirty-seven percent of households are headed by a female.
Amongst household heads, 16% did not graduate from high school, 32% have a
high school education, 26% have attended some college, and 26% have a college
degree or higher. Nearly all households live in an urban area, which is unsurprising
given that the CEX defines a region as urban if it is part of an MSA or an urbanized
area of over 2,500 individuals. The distribution of respondents across regions
suggests that the sample is well partitioned geographically.
We calculate real income in the CEX data as the sum of wages, business, farm,
and rental income, dividends, interest, pension and social security receipts, supple-
mental security income, unemployment and worker’s compensation insurance, and
welfare payments. Twenty-four percent of households report zero income during
the survey period, although this shrinks to only 12% when we focus amongst those
households with heads who report being employed and working positive hours.
Amongst those who report positive income, average annual income is roughly
$57,000 (in 2005 dollars).9
Table 2 displays summary statistics on household consumption in our analysis
sample. To derive a measure of annual household consumption, we calculate
..........................................................................................................................................................................annual expenditure—no households report zero total annual expenditure. We drop households with
heads under age 18 because these households are atypical.9 Following Charles et al. (2009), we deflate all prices used in our analysis to 2005 dollars using the 2005
June Consumer Price Index. We employ the non–seasonally adjusted series for All Urban Consumers:
All Items (CPIAUCNS).
1096 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
average quarterly expenditure per household (by dividing total reported household
consumption by the number of quarters the household was interviewed) and then
multiply this figure by four. Reassuringly, no households report zero total annual
expenditure. Total consumption expenditure in our sample averages just over
$44,000 per year.
Identifying the visibility of household consumption expenditure is not a trivial
task. Two recent studies have made an effort to define conspicuous consumption in
the United States. Heffetz (2011) measures the visibility of expenditure through the
use of a random national phone survey in which individuals are asked how long a
period of time would be required for them to ascertain a neighbor’s consumption
habits with regard to each of 31 expenditure categories derived from the CEX. The
author then constructs a visibility index that ranks consumption activities
according to their degree of conspicuity. Similarly, Charles et al. (2009)
Table 1 Summary statistics on CEX household characteristics
Mean SD
Household head characteristicsFemale 36.52 48.15Age 43.11 13.53White 73.88 43.81Black 12.69 33.22Hispanic 9.56 29.32Asian 3.14 17.28Less than high school 16.20 33.00High school 32.01 39.75Some college 25.93 36.99College grad or higher 25.86 39.37
Household demographicsFamily size 2.81 1.52Adults in household 2.06 0.92
Household incomeZero income households 23.77 42.57Among employed hh heads who
work positive hours11.79 32.25
Income levels (Y> 0) 57,253 47,930
Regional characteristicsUrban 99.11 9.42Northeast 22.22 41.57Midwest 22.33 41.64South 31.86 46.59West 23.60 42.46
Notes: The sample includes all households that appear in state-years of the main
regression sample (see the notes to Table 4 for more detail), for a total of 68,624
households. All figures represent the percentage of households except age, family
size, adults in household, and income. Income is stated in 2005 dollars.
d. l. hicks and j. h. hicks 1097
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
construct multiple measures of visibility using an online survey of graduate
business school students, in which they ask respondents to detail how close their
interaction would have to be with others to notice another individual’s specific
consumption habits. Both of these efforts classify expenditures on goods and
services such as clothing, jewelry, personal care, and vehicles as highly conspicuous
relative to consumption of other goods. Heffetz also finds that expenditures on
food away from home, alcohol and cigarettes, furniture and recreation are highly
visible. The online Appendix Table 1 details the definitions of conspicuous con-
sumption derived from these two studies and the exact consumption categories
included in each definition.
Because of their different focusses, each of these measures may be of interest. The
Heffetz measure may more closely approximate what he calls ‘socio-cultural’
Table 2 Summary statistics on CEX annual household expenditure
Mean SD Shareof totalexpenditure
Share of HHwith positiveexpenditure
Quarterly household expenditureTotal consumption expenditure 44,162 29,466 — 1.00Visible expenditure, Charles et al.
(narrow def.)6,654 11,168 0.12 0.99
Visible expenditure, Charles et al.(broad def.)
9,786 13,121 0.18 1.00
Visible expenditure, Heffetz 12,812 14,844 0.25 1.00
Visible expenditures (inclusive)Clothing, jewelry, and related services 2,154 2,569 0.05 0.98Personal care goods and services 387 402 0.01 0.90Vehicle purchases (new and used) 4,114 10,403 0.06 0.26Vehicle maintenance and accessories 3,132 4,225 0.06 0.89Food out 1,907 2,366 0.04 0.96Tobacco and alcohol 833 1,100 0.02 0.81Recreation goods and services 2,339 3,998 0.05 0.96Furniture and household durables 1,289 2,873 0.03 0.83
Nonvisible expenditureFood (excluding food out) 4,942 2,774 0.14 1.00Utilities 3,117 1,787 0.08 0.99Other transportation 2,776 2,101 0.07 0.98Books, magazines, and toys 958 1,404 0.02 0.94Education 937 3,038 0.02 0.38Health 1,963 2,831 0.04 0.88Housing 10,386 6,861 0.26 0.99Other nondurables and services 2,930 5,421 0.06 0.93
Notes: The sample includes all households that appear in state-years of the main regression sample (see
the notes to Table 4 for more detail), for a total of 68,624 households. Expenditures are stated in 2005
dollars. Visible goods listed in the second panel are goods defined as visible under at least one of the
Charles et al. (2009) or the Heffetz (2010) definitions. Similarly, goods never classified as visible by any
of these definitions comprise nonvisible expenditure summarized in the third panel.
1098 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
inequality, whilst Charles et al.’s measures (which we will refer to as C1 and C2
measures, with C2 employing a more broad definition of visible expenditure)
provide a more direct physical or material approximation of inequality.
Throughout the analysis we display results for all measures, though for brevity in
the figures we focus primarily on the Heffetz measure (the graphical results look
similar regardless of measure).
As the top panel of Table 2 shows, 12–25% of total expenditure can be classified
as visible depending on the definition of conspicuous consumption employed.
Importantly, nearly all households report expenditure on visible consumption,
which is driven primarily by the inclusion of clothing expenditure as a conspicuous
component of consumption. The second panel of Table 2 examines individual
components of visible consumption in further detail.10 Vehicle purchases and
vehicle maintenance make up a sizeable fraction of visible expenditure and
account for the differences between the narrow C1 and broad C2 definitions.
Figure 1 depicts kernel densities of visible consumption expenditures, nonvisible
consumption expenditures, and income for households with positive values less
than $100,000 per year of each measure respectively, and amongst households with
positive income.11 Two things in particular stand out. First, all distributions are
skewed right. Second, the distributions of nonvisible consumption and income are
much more disperse when compared to visible consumption.12
We calculate inequality by state-year for income, total expenditure, and visible
expenditure. Our analysis initially focusses on the Gini coefficient because of its
desirable properties such as Lorenz-consistency, but also because it is a familiar and
easily interpretable measure.13 Figure 2 presents Lorenz curves for the USA for
income, total expenditure, and visible expenditure. Income is more unequally
distributed than total expenditure. Furthermore, the Lorenz curve for visible ex-
penditure is everywhere to the right of that for total expenditure, suggesting that
the distribution of visible consumption is much more unequal than that for total
..........................................................................................................................................................................10 Visible goods listed in the second panel of Table 2 are goods defined as visible under at least one of the
Charles et al. (2009) or the Heffetz (2011) definitions. Similarly, goods never classified as visible by any
of these definitions make up nonvisible expenditure summarized in the third panel. For a complete
classification of goods under the various definitions of Charles et al. and Heffetz, see Appendix Table 1.11 We show the Heffetz definition of visible consumption here, but the results are similar regardless of
visibility measure used.12 The finding that a large number of households have a very small amount of visible consumption
expenditure is consistent with Heffetz (2011), who argues that many visible goods are luxuries with high
income elasticities.13 For our primary analysis, we construct the Gini coefficient by state-year, excluding households with
zero income from the inequality measure (and likewise for the expenditure inequality measures).
Households reporting zero income are often regarded as providing unreliable data, and including
them may decrease the signal-to-noise ratio of the resulting Gini (Szekely and Hilgert, 1999). As
discussed in the results for Table 5, we provide results using a Gini that includes these households
and find that these results are similar.
d. l. hicks and j. h. hicks 1099
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
expenditure, consistent with the notion suggested by Heffetz (2011) that visible
goods behave like luxury goods. The Lorenz curves for income and visible con-
sumption cross, making a direct comparison difficult.
Figure 3 plots the Gini coefficients for income, total expenditure, and visible
expenditure over time in the USA for the period 1986–2001. The first thing to
notice is that all three series have different levels and appear to fluctuate independ-
ently. Income inequality is higher than inequality in expenditures for the entire
period of focus for this study. In addition, income inequality is generally increasing
during this period, whilst inequality in total expenditure remains fairly flat.
Furthermore, inequality in visible consumption is greater than income inequality,
but also fairly steadily increasing over time.14 These overall trends mask hetero-
geneity both across states and over time within states.
0.0
0002
.000
04.0
0006
.000
08D
ensi
ty
0 20000 40000 60000 80000 1000002005 U.S. Dollars
Non-Visible Expenditure Visible ExpenditureIncome
Per Household Per Year
Fig. 1 The distribution of income and expenditure per household per yearNotes: Data for this figure were drawn from the NBER CEX family-level extracts compiled
by Charles et al. (2009) for the period 1986–2001. The sample includes all households
with nonmissing data that appear in state-years of the main regression sample (see the
notes in Table 4 for more detail), that have positive income, and where the income or
expenditure measure falls between zero and 100,000 (exclusive). Visible expenditure is
defined here using the Heffetz (2011) measure, but the figure looks very similar regardless
of definition used.
..........................................................................................................................................................................14 The correlation between income inequality and visible expenditure inequality ranges from 0.22 to 0.38
depending on the measure of visible consumption used.
1100 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
3.2 Crime data
Our estimates for criminal activity in the USA come from the Uniform Crime
Reporting (UCR) Program of the Federal Bureau of Investigation. Figure 4
presents property and violent crime rates in the USA from 1960 to 2008, the
total range of years currently available. Property crime shows relatively steady
growth from 1960 to 1980, and then declines somewhat thereafter. Violent
crime, in contrast, shows relatively steady growth through the early 1990s and
declines slightly after. Property crime rates are an order of magnitude larger than
those for violent crimes.
We focus on the period from 1986 to 2001, for which we have the consistent
Harris and Sabelhaus (2000) CEX expenditure data and a uniform set of covariates
described in the next section. We employ annual crime statistics on a state-by-state
basis for all categories of violent (assault, murder, rape, and robbery) and property
(burglary, larceny, and motor vehicle theft) crime. Figures 5a and 5b document
violent and property crime rates respectively for states in our analysis sample by
category over this period. Downwards trends are apparent for some property
crimes such as burglary and larceny, whereas violent crimes exhibit more
variation over the period.
0.2
5.5
.75
1
Cum
ulat
ive
% o
f Exp
endi
ture
0 .25 .5 .75 1Cumulative % of Population
Total Expenditure Visible ExpenditureytilauqEtcefrePemocnI
Fig. 2 Income and expenditure Lorenz curves.Notes: Data for this figure was drawn from the NBER CEX family-level extracts
compiled by Charles et al. (2009) for the period 1986–2001. The sample includes all
households with nonmissing data that appear in state-years of the main regression
sample (see the notes in Table 4 for more detail). Visible expenditure is defined here
using the Heffetz (2011) measure, but the figure looks very similar regardless of definition
used.
d. l. hicks and j. h. hicks 1101
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
There are a number of data quality concerns to consider when working with
crime statistics, perhaps the two most important being under-reporting of events
and differential reporting across geographic areas (Blau and Blau, 1982). Under-
reporting likely varies across types of crime (for example, murder is considered
more likely to be reported than larceny or rape). Nonetheless, for our analysis such
concerns are second order, because our identification is achieved using variation
within states over time. As long as reporting rates do not change differentially
within these geographic areas over time (although plausible, we have not seen
strong evidence that this would be a systematic concern), then neither under-
reporting nor spatial variation in reporting rates will affect our analysis.
Table 3 presents summary statistics on property and violent crime rates across
state-year observations in our analysis sample. Property crimes are substantially
more common than violent offences. As can be seen from the table, there is a high
degree of variation across state-years, with observations around the 75th percentile
experiencing crime rates more than two times those in the 25th percentile. As we
mentioned earlier, some smaller or more sparsely populated states have been
excluded from our analysis, having too few household observations with which
to calculate state-level inequality. The last column of Table 3 presents property and
violent crime rates for all states including those omitted from our analysis. The 50th
.3.3
5.4
.45
.5.5
5
Gin
i Coe
ffic
ient
1986 1990 1994 1998 2002Year
erutidnepxElatoTemocnIVisible Expenditure
Fig. 3 Inequality in the USANotes: Data for this figure was drawn from the NBER CEX family-level extracts compiled by
Charles et al. (2009) for the period 1986–2001. The sample includes all households with
nonmissing data that appear in state-years of the main regression sample (see the notes in
Table 4 for more detail). Visible expenditure is defined here using the Heffetz (2011)
measure, but the figure looks very similar regardless of definition used.
1102 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
percentile of state-year crime rates in our sample is larger, but not dramatically so.
Nevertheless, our sample comprises 40 of the 50 states, including those that contain
the largest populations, suggesting that our results are likely applicable to the vast
majority of the US population.
3.3 Additional covariates
We draw from the existing literature to construct a wide range of controls to
include in our regressions, in an attempt to encompass the other traditionally
cited determinants of crime. Some of these require additional data sources, but
several can be directly calculated from the CEX sample, which has the added benefit
of consistency with our expenditure data.
Following Kelly (2000) and Demombynes and Ozler (2005), we construct a
measure of family instability, defined as the percentage of households in our
CEX data with a single female head of household. We include percentages of
household heads identifying with different racial groups in the CEX data to
capture the effect of racial heterogeneity.15 We employ annual census estimates
of state population. We control for the fraction of the population aged 15–29,
and the fraction of the population with varying levels of education in a given
010
0020
0030
0040
0050
00
Rat
e pe
r 10
0k I
ndiv
idua
ls
1960 1970 1980 1990 2000 2010Year
Violent OffensesProperty Offenses
Fig. 4 Crime rates in the USANotes: Data for this figure were drawn from the UCR Program of the Federal Bureau of
Investigation (FBI) for the period 1960–2008. All states and all years are included in this
figure.
..........................................................................................................................................................................15 Similar measures are used in Blau and Blau (1982), Kelly (2000), and Demombynes and Ozler (2005).
d. l. hicks and j. h. hicks 1103
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
020
040
060
080
0
Rat
e pe
r 10
0k I
ndiv
idua
ls
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Year
Robbery AssaultRape Murder
01,
000
2,00
03,
000
4,00
05,
000
Rat
e pe
r 10
0k I
ndiv
idua
ls
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Year
ynecraLyralgruBMotor Vehicle Theft
(a)
(b)
Fig. 5 Crime rates by category. Panel a: violent crime rates. Panel b: property
crime rates.Notes: Data for these figures was drawn from the Uniform Crime Reporting (UCR) Program
of the Federal Bureau of Investigation for the period 1986-2001. The sample includes all
state-year observations that appear in the main regression sample (see the notes in Table 4
for more detail).
1104 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
year, both constructed from the CEX data. We also control for average per capita
expenditure in the CEX data, and the percentage of the population living below the
poverty line.16 This is consistent with Nilsson (2004), who documents a relation-
ship between the share of individuals living with less than 10% of median income
and rates of crime.
Because deterrence likely varies across states and over time, we control as well for
available law enforcement resources. Data on state and local police expenditures are
taken from the Bureau of Justice Statistics’s extracts of the Census Bureau’s Annual
Government Finance Survey and Annual Survey of Public Employment, and we lag
this measure by one year to alleviate concerns of endogeneity. Data on local police
expenditures is missing from the sample in 2001, and thus we drop 2002 (the year
in which the lagged measure would be included) from our analysis. Finally, we
obtain annual estimates of state unemployment rates from the Bureau of Labor
Statistics’s Regional and State Employment and Unemployment Series by averaging
over the 12 monthly values.
4. ResultsIn this section, we explore the relationship between inequality in expenditure and
crime within US states over time. We confirm that this relationship is robust to
controlling for standard determinants of crime cited in the existing literature.
Table 3 Reported UCR crimes by state
State-level crime rates per 100k population...............................................................................................
Incl. missingdataa
Min 25thPercentile
50thPercentile
75thPercentile
Max 50thPercentile
Violent offencesTotal 97 373 538 707 1,244 481Murder 1 4 7 9 20 6Rape 12 29 36 44 84 35Robbery 10 103 150 211 624 125Assault 45 219 334 449 786 292Property offencesTotal 2,185 3,896 4,464 5,195 7,820 4,218Burglary 308 788 1,009 1,222 2,294 935Larceny 1,654 2,627 2,937 3,412 5,106 2,865Motor vehicle theft 113 338 442 598 1,158 400
Notes: The sample includes all state-year observations that appear in the main regression sample (see the
notes to Table 4 for more detail). aDenotes data using the full sample from the UCR.
..........................................................................................................................................................................16 To estimate the fraction of the population living below the poverty line, we take the 2005 estimated
poverty lines from the US Department of Health and Human Services for each household size, and then
calculate the share of the population in poverty using the CEX data.
d. l. hicks and j. h. hicks 1105
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Finally, we explore the robustness of our results to using alternate regression spe-
cifications, measures of inequality, and controls.
Figures 6a and 6b depict the (unconditional) lowess plots of violent and property
crime on inequality. Figure 6a reveals a clear positive relationship between violent
crime and inequality for inequality in income, total expenditure, and all three
measures of visible expenditure. The evidence provided in Fig. 6b suggests a less
clear relationship between property crime and inequality without accounting for
other determinants of crime. Total expenditure and the Heffetz measure of visible
consumption appear to have positive slopes, but income and the Charles et al.
measures are relatively flat. We explore these relationships more rigorously using
regression analysis later.
Before settling on the appropriate regression model and specification for
investigating this relationship more rigorously, it is important to explore the char-
acteristics of our data. Figure 7 presents the distribution of counts of violent and
property crimes per state-year in our sample. Both samples exhibit low mass points
with extremely long right tails. A non-normal distribution of this nature suggests
that we should employ a count model in our analysis of crime, the Poisson being
the standard choice. However, simple summary statistics of the crime measures
indicate that over-dispersion is likely in our data—in all cases the variance is orders
of magnitude larger than the mean (see Appendix Table 2). Furthermore, upon
running our primary specification (described in eq. (1)) using a Poisson regression,
both the deviance and Pearson goodness-of-fit statistics indicate that the Poisson is
not an appropriate model (p< 0.0001 across all types of crime and all measures of
income and expenditure inequality). Together these results suggest that criminal
offences are in fact too disperse for the traditional Poisson count model, which
assumes equality of mean and variance. Hilbe (2007, p.30) describes the use of
negative binomial regression models as appropriate when the event counts are
‘intrinsically heteroskedastic, right skewed, and have a variance that increases with
the mean of the distribution’, which characterizes our data well.17 Thus, we estimate
the crime–inequality relationship using a negative binomial regression model.18
Because we employ panel data in our analysis, it is important to consider
potential econometric issues that are common with this type of data. In
particular, we test for panel unit roots and serial correlation (Tauchen, 2010).
We perform panel unit root tests on the crime and inequality series, using the
Fisher-type test, which is appropriate for data sets containing unbalanced panel
data.19 These tests suggest that we cannot reject the null hypothesis of stationarity
..........................................................................................................................................................................17 Because our estimation strategy involves aggregation to the level of US states, there is no particular
evidence to suggest that a zero-inflated model would be more appropriate.18 As a final check, a likelihood ratio test comparing the negative binomial to the Poisson model using
the specification we outline in eq. (1) indicates a strong preference for the former (p< 0.001).19 We employ the Phillips-Perron test using a Newey-West lag length of two. The Fisher-type tests
produce four distinct p-values—we consider a series stationary when at least two of these four tests
indicate it.
1106 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
060
012
00
Cri
mes
per
100
k
.3 .35 .4 .45 .5 .55
Gini Coefficient
Income
060
012
00
Cri
mes
per
100
k
.2 .25 .3 .35 .4
Gini Coefficient
Expenditure0
600
1200
Cri
mes
per
100
k
.55 .6 .65 .7 .75
Gini Coefficient
Visible Exp. (C1)0
600
1200
Cri
mes
per
100
k
.45 .5 .55 .6 .65 .7
Gini Coefficient
Visible Exp. (C2)
060
012
00
Cri
mes
per
100
k
.4 .45 .5 .55 .6 .65
Gini Coefficient
Visible Exp. (Heffetz)
040
0080
00
Cri
mes
per
100
k
.3 .35 .4 .45 .5 .55
Gini Coefficient
Income
040
0080
00
Cri
mes
per
100
k
.2 .25 .3 .35 .4
Gini Coefficient
Expenditure
040
0080
00
Cri
mes
per
100
k
.55 .6 .65 .7 .75
Gini Coefficient
Visible Exp. (C1)
040
0080
00
Cri
mes
per
100
k
.45 .5 .55 .6 .65 .7
Gini Coefficient
Visible Exp. (C2)
040
0080
00
Cri
mes
per
100
k
.4 .45 .5 .55 .6 .65
Gini Coefficient
Visible Exp. (Heffetz)
(a)
(b)
Fig. 6 Lowess plots of crime on inequality. Panel a: violent crime. Panel b:
property crime.Notes: Data for these figures were drawn from the UCR Program of the Federal Bureau of
Investigation and from the NBER CEX family-level extracts compiled by Charles et al. (2009)
for the period 1986–2001. The sample includes all state-year observations that appear in the
main regression sample (see the notes in Table 4 for more detail).
d. l. hicks and j. h. hicks 1107
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
of any of the income and consumption inequality measures. For the crime series,
we cannot reject that murder, rape, assaults, burglary, and motor vehicle theft are
stationary. Only robbery and larceny, and also total violent crime and total
property crime, are likely not stationary. Thus, our primary analysis will be
performed in levels, but for robustness we also present results in first differences
in Appendix Table 3 for the non-stationary series.20
Serial correlation is also a pertinent concern for estimation using panel data. To
examine this formally, we perform a Wooldridge test for autocorrelation in panel
data. The test statistics suggest that we should reject the null hypothesis of no first-
order autocorrelation for all series. To address this issue, we include the first lag of
the crime rate as a control in our regressions.21
Our regression equation takes the following form:
Yit ¼ �þ �lnXit þ �Yit�1 þ �lnZit þ �i þ �t þ "it ð1Þ
where i indexes states and t indexes time. Y is the count of offences. X denotes a
measure of income or expenditure inequality (which we vary across specifications).
0.0
0001
.000
02
Den
sity
0 100000 200000 300000 400000
kernel = epanechnikov, bandwidth = 6.3e+03
Violent Offenses0
2.00
0e-0
64.
000e
-06
Den
sity
0 500000 1000000 1500000 2000000
kernel = epanechnikov, bandwidth = 3.8e+04
Property Offenses
Fig. 7 Distribution of crimes across state-yearsNotes: Data were drawn from the UCR Program of the Federal Bureau of Investigation for
the period 1986–2001. The sample includes all state-year observations that appear in the
main regression sample (see the notes in Table 4 for more detail).
..........................................................................................................................................................................20 After we first-difference, we cannot reject the stationarity of any series, but the interpretation of the
results is different as identification comes from only short-run changes in inequality (Spelman, 2008).21 The exception is Appendix Table 3, which contains the first differenced regressions. The Wooldridge
test cannot reject the null hypothesis of no first-order autocorrelation in our first-differenced
specification.
1108 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Z is a vector of time-varying state-level controls, including the percent of
households that are female headed; percents of households that are African
American, Hispanic, and Asian American; population; percent of the population
aged 15–29; percent of population at different levels of education; average per
capita household expenditure; the unemployment rate; the poverty rate; lagged
state and local police expenditures; and indicators for geographic region.22 Wi
represents state fixed effects and Ft time fixed effects. The measures of inequality
and all non-dummy controls are logged so that coefficients in our negative
binomial regressions can be interpreted simply as elasticities.
Results from negative binomial regressions taking the form specified in eq. (1)
are presented in Table 4. Each cell represents a distinct regression, including the full
set of controls as already described. We suppress all results except for the coeffi-
cients on the inequality terms as the reporting of controls would be cumbersome
given the number of regressions presented. We report goodness-of-fit statistics for
the regressions at the bottom of each panel.23
Panel A of Table 4 presents results for violent criminal offences. Several findings
are immediately apparent. First, we do not see the association between income
inequality and violent crime that is found for the USA in other papers (see, for
example, Danzinger and Wheeler, 1975; Blau and Blau, 1982; Kelly, 2000). We
cannot rule out the possibility that this may be because income information is
not well captured in the CEX data as discussed in Section 3.1, or because the
relationship is attenuated in our state-level analysis (the aforementioned studies
examined MSAs).
At the same time, total expenditure may provide a reasonable proxy for
permanent income, as argued by Charles et al. (2009). In the estimates produced
in Table 4, inequality in total expenditure is not statistically significant, whilst
inequality in visible expenditure is positive and statistically significant (or very
nearly statistically significant at standard levels of confidence) for total violent
crime, murder, and assaults. Specifically, for murder we find a positive and signifi-
cant relationship with all three definitions of visible consumption. For assaults we
find a relationship at traditional levels of confidence using the C2 and Heffetz
measures, and near significance on the C1 measure (the p-value is 0.12). For
total violent crime, we see the relationship is statistically significant at traditional
levels of confidence using the Heffetz measure, and very nearly significant using the
other two measures of conspicuous consumption (with p-values of 0.14 for the C1
measure and 0.11 for the C2 measure).
These results suggest that there is a relationship between inequality and con-
spicuous consumption for violent crime, which is driven by murder and assaults.
These findings are also consistent with a situation in which visible consumption
inequality provides a stronger signal to criminals than income inequality and total
..........................................................................................................................................................................22 These controls are described in further detail in Section 3.23 Furthermore, a likelihood ratio test comparing the negative binomial to the Poisson model indicates a
strong preference for the former (p< 0.001 for all regressions).
d. l. hicks and j. h. hicks 1109
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Table 4 Negative binomial regression estimates of the relationship between
inequality and crime
Panel A: violent offences................................................................................................................................................................
Total Murder Rapes Robbery Assault
Income Gini 0.005 0.026 0.003 �0.028 0.020(0.027) (0.054) (0.028) (0.033) (0.038)
Total exp. Gini 0.025 0.110 0.019 �0.027 0.054(0.033) (0.070) (0.045) (0.038) (0.046)
Vis. exp. Gini (C1) 0.062 0.127** 0.012 0.036 0.088(0.042) (0.064) (0.047) (0.052) (0.057)
Vis. exp. Gini (C2) 0.058 0.133** �0.019 0.012 0.090*(0.036) (0.052) (0.045) (0.040) (0.050)
Vis. exp. Gini (H) 0.050* 0.124** 0.027 0.019 0.072*(0.030) (0.057) (0.038) (0.038) (0.044)
Prob> LR 0.000 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.253 0.313 0.281 0.263 0.241
Panel B: property offences................................................................................................................................................................
Total Burglary Larceny MV theft
Income Gini 0.019 0.011 0.019 0.060(0.019) (0.024) (0.022) (0.037)
Total exp. Gini �0.035* �0.020 �0.051** 0.017(0.020) (0.025) (0.025) (0.046)
Vis. exp. Gini (C1) 0.020 0.030 0.017 �0.019(0.023) (0.030) (0.024) (0.050)
Vis. exp. Gini (C2) 0.022 0.000 0.032 �0.015(0.024) (0.029) (0.026) (0.044)
Vis. exp. Gini (H) 0.001 0.015 �0.009 0.025(0.021) (0.025) (0.023) (0.040)
Prob> LR 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.224 0.233 0.227 0.224
Notes: Data for this table were drawn from a number of sources, as described in the text. The sample of
households includes those with positive real average annual expenditures, where the household has a
state identifier and did not change state of residence during interview period, and where the household
head is at least 18 years old. State-year observations are further limited to those composed of at least 25
household-level observations. Each cell represents a separate regression, employing state-level data for
the period 1986–2001. The dependent variable is a count of offences. Controls include the lagged
count of offences; percent of households that are female headed; percents of households that are
African American, Hispanic, and Asian American; population; percent of the population aged 15–29;
percent of population at different levels of education; average household expenditure; the unemploy-
ment rate; the poverty rate; lagged state and local police expenditures; indicators for geographic region;
and state and year fixed effects. All nondummy controls are logged. All income and expenditure
measures are per capita, in 2005 dollars. Each regression has 544 observations. Robust standard
errors, clustered at the state level, are presented in parentheses. *** p< 0.01, ** p< 0.05, and
*p< 0.10. The likelihood ratio test p-value is from a test of whether the model with covariates
predicts better than a model with an intercept only.
1110 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
consumption expenditure (which may be a reasonable proxy for permanent
income). The lack of results on total consumption expenditure is particularly
telling, because although income may be less accurately measured, total consump-
tion and visible consumption are likely equally well measured in the CEX.
Panel B of Table 4 includes the results for property crime and its components. It
is not uncommon in the existing literature to find no robust relationship between
income inequality and property crime, with Demombynes and Ozler (2005) being a
notable exception. Our results for property crime are generally close to zero and
occasionally negative. With the exception of inequality in total expenditure, the
estimated coefficients are rarely significant.24
One potential explanation for the lack of an observed relationship between
inequality and property crime is the possibility of reverse causality affecting our
estimates. Mejıa and Restrepo (2010) document a negative and significant effect of
property crime on the consumption of visible goods. If individuals in high-
property-crime areas limit their purchases of visible goods, this may drive down
observed inequality in regions with high crime, pushing down our estimated co-
efficients. It is plausible that one would observe this reverse connection for property
crime and not violent crime. In fact, Mejıa and Restrepo find no relationship
between consumption of visible goods and violent crime. We perform a Granger
causality test to check for bidirectional causality in our model, using the procedure
outlined in Spelman (2008). Thornton and Batten (1985) suggest that results of this
test can be sensitive to lag selection, so we run the test using everything from one to
five lags. The results suggest that bidirectional causality is not a major concern.
We next present several robustness checks on these results. The first check relates
to our regression methodology. Count models are regularly employed in the
analysis of crime and we believe they are appropriate in the context of our
analysis. We presented evidence that the distribution of crimes across states is
bounded by zero, has a long right tail, and contains excessive levels of
dispersion, factors suggesting that a negative binomial regression is more appro-
priate than both ordinary least squares (OLS) and Poisson models (Hilbe, 2007).
Nonetheless, if the distribution of criminal offences is normal in the log of the
crime rate, then log-linear OLS estimates should yield similar results. Table 5
reproduces the results of Table 4, this time using log crime rates as the
dependent variable. The OLS results are similar to those of the negative binomial
regressions—there is a strong and positive relationship between murder and
inequality across all three measures of visible expenditure inequality, and near
statistical significance for total violent crime and assaults across all three
measures of conspicuous consumption inequality as well. The OLS results also
suggest a relationship between inequality in total expenditure and violent crime.
For property crimes, the OLS results indicate a relationship between inequality in
conspicuous consumption and crime, but on closer inspection, this result is driven
..........................................................................................................................................................................24 We also ran our regressions for violent and property crimes including the inequality term squared, but
did not find a nonlinear relationship with inequality.
d. l. hicks and j. h. hicks 1111
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
primarily by larceny, which we previously found to be a non-stationary series. Our
first-differenced results, presented in Appendix Table 3, do not support this finding
either.
Next we explore the possibility that some of our sample restrictions drive the
results. As a check, we rethink how the Gini coefficient is constructed in the
primary analysis. In the Gini coefficient used in the preceding analysis, we
dropped individuals with zero income or expenditure out of concern that the
data are unreliable and their inclusion would decrease the signal-to-noise ratio
of the resulting Gini (Szekely and Hilgert, 1999). There are a large number of
households with zero and negative income in the CEX (approximately 25% of
Table 5 OLS estimates of the relationship between inequality and crime
Panel A: violent offences................................................................................................................................................................
Total Murder Rapes Robbery Assault
Income Gini �0.028 0.083 0.030 �0.095** �0.016(0.026) (0.069) (0.036) (0.045) (0.038)
Total exp. Gini 0.050 0.206** 0.054 �0.046 0.089*(0.035) (0.099) (0.051) (0.058) (0.053)
Vis. exp. Gini (C1) 0.071 0.245*** 0.017 0.075 0.097(0.046) (0.094) (0.050) (0.069) (0.062)
Vis. exp. Gini (C2) 0.056 0.228** �0.020 0.036 0.080(0.040) (0.092) (0.057) (0.053) (0.056)
Vis. exp. Gini (H) 0.054 0.189** 0.044 0.031 0.075(0.039) (0.077) (0.045) (0.047) (0.052)
R2 0.988 0.951 0.957 0.985 0.984
Panel B: property offences................................................................................................................................................................
Total Burglary Larceny MV theft
Income Gini 0.008 �0.015 0.013 0.029(0.024) (0.029) (0.026) (0.035)
Total exp. Gini �0.028 �0.000 �0.048 0.044(0.025) (0.028) (0.032) (0.045)
Vis. exp. Gini (C1) 0.042* 0.060* 0.037 0.018(0.023) (0.033) (0.027) (0.055)
Vis. exp. Gini (C2) 0.051** 0.033 0.058** 0.022(0.025) (0.032) (0.027) (0.052)
Vis. exp. Gini (H) 0.015 0.039 0.003 0.042(0.023) (0.029) (0.026) (0.040)
R2 0.976 0.979 0.975 0.970
Notes: Data for this table were drawn from a number of sources, as described in the text. For information
on the sample, see the notes for Table 4. Each cell represents a separate regression, employing state-level
data for the period 1986–2001. The dependent variable is the log crime rate. Controls are the same as
listed in Table 4. All nondummy controls are logged. All income and expenditure measures are per
capita, in 2005 dollars. Each regression has 498 observations. Robust standard errors, clustered at the
state level, are presented in parentheses. *** p< 0.01, ** p< 0.05, and * p< 0.10.
1112 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
our sample). Although there are no individuals with zero reported total expend-
iture, there are a very small number who report zero visible expenditure. In Table 6,
we recalculate the Gini coefficient, this time including zeros for income and ex-
penditure. Reassuringly, the results are effectively unchanged from our primary
specifications in Table 4.
A third robustness check entails exploring alternative regression controls. In
particular, it is more typical in the existing crime determinants literature to
control for average income per capita rather than average total expenditure per
capita. Table 7 presents the results using state-level average income per capita
(calculated using the CEX data) in lieu of average expenditure per capita.
Although the resulting coefficient estimates for violent crime are smaller and
the statistical significance is weaker, the results remain substantively the same.
Table 6 Negative binomial regressions, with a Gini including zeros
Panel A: violent offences................................................................................................................................................................
Total Murder Rapes Robbery Assault
Income Gini 0.015 0.061 �0.041 0.027 0.021(0.029) (0.062) (0.027) (0.033) (0.039)
Total exp. Gini 0.025 0.110 0.019 �0.027 0.054(0.033) (0.070) (0.045) (0.038) (0.046)
Vis. exp. Gini (C1) 0.065 0.144** 0.014 0.033 0.093(0.044) (0.065) (0.049) (0.051) (0.058)
Vis. exp. Gini (C2) 0.060 0.139** �0.018 0.012 0.093*(0.037) (0.053) (0.046) (0.040) (0.050)
Vis. exp. Gini (H) 0.054* 0.130** 0.024 0.024 0.076*(0.030) (0.057) (0.039) (0.038) (0.045)
Prob> LR 0.000 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.253 0.312 0.281 0.263 0.241
Panel B: property offences................................................................................................................................................................
Total Burglary Larceny MV theft
Income Gini 0.021 0.008 0.022 0.068(0.019) (0.022) (0.023) (0.047)
Total exp. Gini �0.035* �0.020 �0.051** 0.017(0.020) (0.025) (0.025) (0.046)
Vis. exp. Gini (C1) 0.016 0.025 0.014 �0.025(0.024) (0.030) (0.026) (0.053)
Vis. exp. Gini (C2) 0.017 �0.006 0.027 �0.021(0.024) (0.029) (0.026) (0.045)
Vis. exp. Gini (H) 0.000 0.012 �0.010 0.025(0.021) (0.026) (0.023) (0.040)
Prob> LR 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.224 0.233 0.227 0.224
Notes: Refer to the notes for Table 4.
d. l. hicks and j. h. hicks 1113
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
The Current Population Survey (CPS) is typically used to estimate income in the
USA and may provide a more reliable measure. As a further check, we control for
average state-level income drawn from CPS data instead and obtain similar results
(not shown).
One potential concern with running a state-level analysis using the CEX data is
the construction of state-level estimates from household data. The analysis we have
presented so far included state-year observations that were produced from at least
25 household observations. This could be an issue if having too few households in
some state-years induces too much measurement error. Table 8 presents the results
from Table 4, this time including state-year observations that were produced from
Table 7 Negative binomial regressions, controlling for income instead of total
expenditure
Panel A: violent offences................................................................................................................................................................
Total Murder Rapes Robbery Assault
Income Gini 0.001 0.030 0.001 �0.028 0.014(0.027) (0.057) (0.028) (0.033) (0.037)
Total exp. Gini �0.005 0.071 0.003 �0.022 0.005(0.031) (0.064) (0.041) (0.036) (0.041)
Vis. exp. Gini (C1) 0.042 0.108* 0.005 0.032 0.057(0.040) (0.059) (0.047) (0.048) (0.051)
Vis. exp. Gini (C2) 0.043 0.114** �0.021 0.011 0.067(0.035) (0.054) (0.047) (0.039) (0.046)
Vis. exp. Gini (H) 0.030 0.099* 0.019 0.016 0.041(0.030) (0.053) (0.038) (0.037) (0.040)
Prob> LR 0.000 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.252 0.313 0.281 0.263 0.241
Panel B: property offences................................................................................................................................................................
Total Burglary Larceny MV theft
Income Gini 0.020 0.010 0.021 0.062*(0.019) (0.023) (0.021) (0.037)
Total exp. Gini �0.017 �0.022 �0.025 0.032(0.017) (0.024) (0.019) (0.038)
Vis. exp. Gini (C1) 0.024 0.022 0.024 �0.007(0.023) (0.031) (0.024) (0.049)
Vis. exp. Gini (C2) 0.022 �0.009 0.035 �0.008(0.023) (0.030) (0.026) (0.044)
Vis. exp. Gini (H) 0.005 0.007 �0.001 0.033(0.021) (0.026) (0.022) (0.039)
Prob> LR 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.224 0.233 0.227 0.224
Notes: Refer to the notes for Table 4. The only departure here is that instead of controlling for average
household expenditure, we control for average household income.
1114 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
at least 50 household observations. This restriction drops our sample size from 544
observations to 460. Again, the violent crime results remain broadly similar, whilst
the property crime specifications produce mixed results that differ from our
previous specifications.
In this section, we documented a robust relationship between inequality and
conspicuous consumption for violent crime, which is driven by murder and
assaults. We also found a larger and more robust relationship for inequality in
visible consumption and crime than for inequality in either income or total ex-
penditure (which should proxy for permanent income) and crime. This result is
consistent with consumption providing a stronger signal to criminals than income.
Table 8 Negative binomial regressions, adjusting the number of households by
state-year
Panel A: violent offences................................................................................................................................................................
Total Murder Rapes Robbery Assault
Income Gini �0.046* 0.050 �0.022 �0.061* �0.043(0.025) (0.053) (0.035) (0.036) (0.037)
Total exp. Gini 0.014 0.105 0.024 �0.049 0.037(0.036) (0.074) (0.047) (0.040) (0.056)
Vis. exp. Gini (C1) 0.088 0.226** 0.017 0.081 0.112(0.063) (0.094) (0.073) (0.071) (0.087)
Vis. exp. Gini (C2) 0.051 0.180** �0.031 0.034 0.076(0.057) (0.072) (0.064) (0.053) (0.080)
Vis. exp. Gini (H) 0.079* 0.160** 0.008 0.075* 0.093(0.046) (0.081) (0.050) (0.043) (0.069)
Prob> LR 0.000 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.254 0.311 0.278 0.26 0.240
Panel B: property offences................................................................................................................................................................
Total Burglary Larceny MV theft
Income Gini �0.019 �0.026 �0.022 0.005(0.019) (0.027) (0.023) (0.037)
Total exp. Gini �0.041* �0.007 �0.070** 0.061(0.024) (0.034) (0.030) (0.043)
Vis. exp. Gini (C1) 0.045 0.081** 0.031 0.029(0.027) (0.032) (0.029) (0.076)
Vis. exp. Gini (C2) 0.035 0.046 0.035 0.014(0.028) (0.031) (0.031) (0.066)
Vis. exp. Gini (H) 0.019 0.055** �0.005 0.077*(0.020) (0.025) (0.024) (0.043)
Prob> LR 0.000 0.000 0.000 0.000McFadden’s adj. R2 0.223 0.230 0.226 0.221
Notes: Refer to the notes for Table 4. The only departure here is that instead of limiting state-year
observations to those composed of at least 25 households, we limit state-year observations to those
composed of at least 50 households. Each regression has 460 observations.
d. l. hicks and j. h. hicks 1115
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
The estimates produced for property crime were mixed across Tables 4–8 and failed
to provide evidence of a clear and robust relationship.
5. On the determinants of crimeIn this section we address existing theory, which posits a connection between
inequality and criminal behavior. Understanding the channel through which
socioeconomic inequality may be linked to crime is important for designing
crime prevention policy and formulating research. To date, studies have
struggled to explain the mechanism underlying this association. In this section,
we present an exploratory exercise in which we exploit one plausible channel
through which inequality may directly influence crime—the display of visible
material wealth—to sort between competing explanations for why one should
expect a relationship to exist in the first place.
Distinguishing between economic and sociological theories of inequality and
crime is a challenging endeavor for a number of reasons. First, the prominent
explanations discussed in Section 1 are similar in nature and yield some
overlapping predictions. Second, determinants of criminal behavior are multi-
faceted. For example, even evidence suggesting that the majority of crimes are
inflicted by the poor on other poor individuals does not necessarily conflict with
the economic explanation for crime because potential victims differ both in their
resources and in their accessibility to criminals.25 We attempt to sidestep these
issues by exploiting the visibility of consumption expenditures to provide new
tests to examine the fundamental connection between economic inequality and
crime. In particular, we present two propositions.
Proposition 1 Strain theory and social disorganization theory suggest that indi-
viduals may become disillusioned after observing others with greater economic
success. If this is correct, then the association between inequality in visible expend-
iture and crime should be stronger than the relationship between inequality in total
expenditure and crime.
Consistent with social disorganization and strain theories, the results presented
in Panel A of Tables 4, 5, 6, 7, and 8 suggest that for total violent crimes, and for
murder and assaults in particular, the estimated relationship with visible expend-
iture inequality is robust across a range of inequality measures and estimation
techniques, and this is not true for total expenditure inequality (which only
achieves statistical significance in Table 5). Panel B of these tables suggests that
the relationship between inequality in expenditure and property crime is not robust
..........................................................................................................................................................................25 For instance, research has shown that the rich have better access to protective services, and thus the
probability of being caught faced by potential burglars is higher (Bourguignon, 2000; Fajnzylber et al.,
2002a).
1116 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
to a range of samples, controls, and methodologies, so we consider that relationship
to be effectively zero.
We consider a second proposition based on Becker’s standard theory of crime:
Proposition 2 Because criminals face uncertainty concerning the benefits of crime,
visible consumption by potential victims may serve to reduce informational
asymmetries that would otherwise discourage criminal efforts. This reduction in
information costs should apply primarily to the pecuniary benefits of property
crime and not the psychic (or psychotic) gains attributable to violent crime. If
Becker’s standard economic theory is correct, then the relationship between
inequality in visible consumption and crime should be stronger than the relation-
ship between inequality in total consumption and crime for property crimes, but
not necessarily for violent crimes.
Table 4 suggests that we do not observe this predicted association in our data,
and the relationship between property crime and inequality in visible consumption
appears far from robust across our set of robustness checks, with property crime
estimates vacillating between positive, negative, and most frequently, insignificance.
Our results thus provide little support for the classical cost-benefit explanation for
criminal behavior provided in Proposition 2.
6. ConclusionThis article extends the existing literature examining the relationship between
inequality and crime by suggesting that the signalling nature of visible consumption
behavior may be the driving factor behind such a connection. Previous research on
this subject has focussed on inequality in income, despite the fact that income is an
opaque measure of relative well-being.
Examining variation within US states over nearly two decades, we document a
strong association between the distribution of visible consumption and violent
criminal offences, particularly for murder and assaults. This relationship proves
robust to the inclusion of zero income households, to changes in our sample
and control variables, and to alternative estimation strategies. We find no such
robust relationship between inequality in visible expenditure and property crime,
although we cannot definitively rule out estimates that are biased downwards due
to reverse causality. Furthermore, we find no evidence linking crime and total
expenditure, suggesting that information plays an important role in crime
determination.
Importantly, we document a larger and more robust relationship for inequality
in visible consumption and crime than for inequality in income and crime. This
result is consistent with consumption providing a stronger signal to criminals than
income. A caveat of this analysis is that our results likely suffer from aggregation of
the analysis at the state level. As such, a promising avenues for future research
would be to tackle the connection between conspicuous consumption and crime in
a more disaggregated setting, such as amongst MSAs, cities, or communities.
d. l. hicks and j. h. hicks 1117
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Finally, exploiting the nature of conspicuous consumption and its influence
across types of crime, we shed light on the existing causal stories explaining
crime. Our results on violent crime are consistent with social disorganization
and strain theories, which suggest that higher inequality should be related to
violent crime because it leads to greater relative deprivation. At the same time,
our findings provide little support for Becker-style explanations of criminal
behavior if the visibility of expenditure carries information about the potential
returns to illicit activities.
Supplementary materialSupplementary material is available online at the OUP website.
AcknowledgementsThe authors acknowledge comments provided during presentations at the University of
Oklahoma and the Royal Economic Society Annual Conference 2011, and the suggestions
of two anonymous referees. Jia Wang provided valuable research assistance. Any errors are
our own.
ReferencesAguiar, M. and Bils, M. (2011) Has consumption inequality mirrored income inequality?
NBER Working Paper Series No. 16807, Cambridge, MA.
Attanasio, O., Hurst, E., and Pistaferri, L. (2012) The evolution of income, consumption
and leisure inequality in the US, 1980–2010, NBER Working Paper Series No. 17982,
Cambridge, MA.
Becker, G. (1968) Crime and punishment: an economic approach, Journal of Political
Economy, 76, 169–217.
Blau, J. and Blau, P. (1982) The cost of inequality: metropolitan structure and violent crime,
American Sociological Review, 47, 114–29.
Blundell, R., Pistaferri, L., and Preston, I. (2006) Imputing consumption in the PSID using
food demand estimates from the CEX, available at http://www.stanford.edu/�pista/impute.
pdf.
Blundell, R. and Preston, I. (1998) Consumption inequality and income uncertainty,
Quarterly Journal of Economics, 113, 603–40.
Brush, J. (2007) Does income inequality lead to more crime? A comparison of cross-
sectional and time-series analysis of United States counties, Economic Letters, 96, 264–68.
Bourguignon, F. (2000) Crime, violence, and inequitable development, available at http://
siteresources.worldbank.org/DEC/Resources/84797-1251813753820/6415739-12518140107
99/bourg.pdf.
Card, D., Mas, A., Moretti, E., and Saez, E. (2012) Inequality at work: the effect of peer
salaries on job satisfaction, American Economic Review, 102, 2981–3003.
Charles, K., Hurst, E., and Roussanov, N. (2009) Conspicuous Consumption and Race, The
Quarterly Journal of Economics, 124, 425–68.
1118 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Corneo, G. and Jeanne, O. (1997) Conspicuous consumption, snobbism, and conformism,
Journal of Public Economics, 66, 55–71.
Dahlberg, M. and Gustavsson, M. (2008) Inequality and crime: separating the effects of
permanent and transitory income, Oxford Bulletin of Economics and Statistics, 70, 129–53.
Danzinger, S. and Wheeler, D. (1975) The economics of crime: punishment or income
redistribution, Review of Social Economy, 33, 113–31.
De Mello, J. and Zilberman, E. (2008) Does crime affect economic decisions? An empirical
investigation of savings in a high-crime environment, B.E. Journal of Economic Analysis &
Policy, 8, 1–26.
Demombynes, G. and Ozler, B. (2005) Crime and local inequality in South Africa, Journal of
Development Economics, 76, 265–92.
Duesenberry, J. (1949) Income, Saving and the Theory of Consumer Behavior, Harvard
University Press, Cambridge, MA.
Easterlin, R. (1995) Will raising the incomes of all increase the happiness of all? Journal of
Economic Behavior & Organization, 27, 35–47.
Fajnzylber, P., Lederman, D., and Loayza, N. (2002a) Inequality and violent crime, Journal
of Law & Economics, 45, 1–40.
Fajnzylber, P., Lederman, D., and Loayza, N. (2002b) What causes violent crime? European
Economic Review, 46, 1323–57.
Harris, E. and Sabelhaus, J. (2000) Consumer Expenditure Survey, family level extracts,
1980:1–1998:2, available at http://www.nber.org/ces_cbo/Cexfam.pdf.
Heffetz, O. (2011) A test of conspicuous consumption: visibility and income elasticities,
Review of Economics and Statistics, 93, 1101–17.
Hilbe, J. (2007) Negative Binomial Regression, Cambridge University Press, New York.
Hsieh, C. and Pugh, M. (1993) Poverty, income inequality, and violent crime: a meta-
analysis of recent aggregate data studies, Criminal Justice Review, 18, 182–202.
Kelly, M. (2000) Inequality and crime, Review of Economics and Statistics, 82, 530–39.
Krueger, D. and Perri, F. (2006) Does income inequality lead to consumption inequality?
evidence and theory, Review of Economic Studies, 73, 163–93.
Leibenstein, H. (1950) Bandwagon, snob, and Veblen effects in the theory of consumers’
demand, Quarterly Journal of Economics, 64, 183–207.
Luttmer, E. (2005) Neighbors as negatives: relative earnings and well-being, Quarterly
Journal of Economics, 120, 963–1002.
Machin, S. and Meghir, C. (2004) Crime and economic incentives, Journal of Human
Resources, 39, 958–79.
Maurer, J. and Meier, A. (2008) Smooth it like the ‘Joneses’? Estimating peer-group effects in
intertemporal consumption choice, Economic Journal, 118, 454–76.
Mejıa, D. and Restrepo, P. (2010) Crime and conspicuous consumption, Documento CEDE
No. 2010-32, Universidad de los Andes Department of Economics Research Paper Series,
Bogota.
Merton, R. (1938) Social structure and anomie, American Sociological Review, 3, 157–66.
Nilsson, A. (2004) Income inequality and crime: the case of Sweden, IFAU Working Paper
No. 2004-6, Uppsala.
d. l. hicks and j. h. hicks 1119
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from
Soares, R. (2004) Development, crime and punishment: accounting for the internationaldifferences in crime rates, Journal of Development Economics, 73, 155–84.
Spelman, W. (2008) Specifying the relationship between crime and prisons, Journal ofQuantitative Criminology, 24, 149–78.
Szekely, M. and Hilgert, M. (1999) What’s behind the inequality we measure: anunvestigation using Latin American data, Inter-American Development Bank WorkingPaper #409, Washington, DC.
Tauchen, H. (2010) Estimating the supply of crime: recent advances, in B. Benson andP. Zimmerman (eds) Handbook on the Economics of Crime, Edward Elgar, Cheltenham.
Thornton, D. and Batten, D. (1985) Lag-length selection and tests of Granger causalitybetween money and income, Journal of Money, Credit and Banking, 17, 164–78.
Veblen, T. (1899) The Theory of the Leisure Class, Penguin Classics, New York,(reprint 1994).
1120 jealous of the joneses
at Nova Southeastern U
niversity on Novem
ber 28, 2014http://oep.oxfordjournals.org/
Dow
nloaded from