Poverty and Crime
Crime
exists everywhere in the United States - in rural and urban areas, in the East
and West, and among all types of people. This has led many government officials, especially those in urban
areas, to focus largely on the reduction of crime among their respective
constituencies and has led others to speculate on the factors that influence
the amount of crime and how those factors can be controlled
In
the US, urban crime is often perceived as a problem amid areas with high
poverty levels. This may be the case;
however, many other factors, such as unemployment, population density, minority
population, age distribution, and locality in the US, are correlated with crime
and affect poverty as well. When these
factors are controlled for, how much does poverty affect crime? Knowing more accurately how poverty affects
crime can help us know if focusing on the reduction of poverty can aid in crime
reduction or if money and effort should go to other areas.
Poverty's
effects on crime can be explained through a variety of reasons. There is a higher rate of mental illness in
the poor than in the rich (Brill 40). Poverty
can lead to high levels of stress that in turn may lead individuals to commit theft,
robbery, or other violent acts. Moreover,
poverty may lead to an actual or perceived inferior education, which would
cause youth to count on less access to quality schools, jobs, and role models,
decreasing the opportunity costs of crime and increasing the probability of
youth spending time on the street associating with gangs, etc (Ludwig 1).
Crime
offers a way in which impoverished people can obtain material goods that they
cannot attain through legitimate means. Often threat or force can help them acquire even more goods, this induces
them to commit violent acts such as robbery, which is the second most common
violent crime. For many impoverished
people, the prize that crime yields may outweigh the risk of being caught,
especially given that their opportunity cost is lower than that of a wealthier
person. Thus, poverty should increase
crime rates.
However, many
other factors influence crime and are correlated with poverty as well. Higher unemployment would certainly increase
poverty and at the same time lead to more crime due to depression associated
with being unemployed. Personal income
per capita, which is inversely correlated with the poverty level, still may
increase crime since greater wealth means greater benefits to thieves and
robbers. Furthermore, because of social
class gaps, personal income per capita rates may not affect poverty to a great
extent (the income may be concentrated in a small percentage of the
population). It might even accentuate
the difference between the upper and lower classes, thereby inducing more
crime.
Variations
in the composition of population can affect crime in different ways. First, adolescents are often responsible for
crimes committed. "The poor delinquent
child... is more apt to be expelled from school or have a police record than a
well-to-do delinquent..." (Brill 40). A
higher percentage of inhabitants under the age of twenty-five may lead to
higher crime rates. On the other hand,
the elderly, because of their possessions and vulnerability, are believed to be
the most frequent victims of crime.
The degree of
minority population in an area is also correlated with poverty due to the disproportional
amount of minorities living in impoverished urban areas. In addition, racism towards minorities can
lead to lower wages and fewer jobs, resulting in higher poverty rates. In 1995, all Metropolitan Areas with
unemployment rates over 12% also had a population composed of at least 30%
minorities.
Geographic
regions within the US have different characteristics and therefore lead to
differing levels of both crime and poverty. The 1999 UCR report, for example, indicates that law enforcement
personnel varied between 2.5 and 4.3 persons per 1000 population among
differing regions of the US. Climate,
associated with geographical location, is also believed to affect crime - more
temperate climates being positively correlated with crime. Cultural factors such as recreational
activities, religious characteristics, and family cohesiveness are all
associated with geographic regions of the US and influence crime.
Because
of the manner in which population density influences living conditions
(ie: houses vs. apartment complexes), it
is also likely to be correlated with both poverty and crime. Studies have found that "more densely
populated neighborhoods tend to be poorer, have higher percentages of persons
in the age range of 12 to 20, have larger concentrations of single-parent
households, and larger nonwhite populations" (Short 52).
This
study first examines how poverty affects crime in the simple regression model. Then, controlling for the aforementioned factors
- race, unemployment, personal income, population density, geographic location,
and age distribution - it again examines the relationship between crime and
poverty and how this relationship is influenced when these factors are held
constant.
All
the data used in this study come from the 1997-1998
State and Metropolitan Area Data Book from the Bureau of the Census. The data covers 322 Metropolitan Areas,
including 245 MSA's, 17 CMSA's, 15 PMSA's, and 12 NEMSA's (which are
essentially MA's of different sizes and characteristics). Each MA contains a population of at least
100,000 inhabitants of which fifty percent or more live in urban areas. Seventeen of the 322 MA's have been
eliminated due to a deficiency in the data.
Since
this study is based on crime in urban areas, these data provide an excellent
resource for determining the causes of crime. Data from larger areas (such as states) would be too general and too
many conflicting characteristics within the area would be incorporated into the
data. The Metropolitan Areas provide a
sample composed of similar makeup, but with sufficient variation to provide a
good model. Each area has different
poverty levels, population density, etc.
List and Description of variables
|
Dependent Variables
|
Total
Crime
|
total
number of the seven index crimes reported to police per 100,000 inhabitants
|
Violent
Crime
|
total
number of the four violent crimes reported to police per 100,000 inhabitants
|
Independent Variables
|
Population
Dens.
|
thousands
of inhabitants per square mile
|
Pop.
Under 25
|
percentage
of population under 25 years old
|
Pop. Over
65
|
percentage
of population over 65 years old
|
Poverty
|
percentage
of population below the poverty level
|
Unemployment
|
percentage
of population unemployed
|
Black
|
percentage
of black population
|
Asian
|
percentage
of Asian and Pacific Islander population
|
Hispanic
|
percentage
of Hispanic population; may be of any race
|
PI per
Capita
|
personal
income per capita measured in thousands of US (1995) dollars
|
Dummy Variables - Regions of the
US
|
NE
|
New
England: ME, NH, VT, MA, RI, CT
|
MA
|
Middle
Atlantic: NY, NJ, PA
|
SA
|
South Atlantic: DE, MD, DC, VA,
WV, NC, SC, GA, FL
|
ENC
|
East
North Central: OH, IN, IL, MI, WI
|
WNC
|
West
North Central: MN, IA, MO, ND, SD, NE, KS
|
ESC
|
East
South Central: KY, TN, AL, MS
|
WSC
|
West
South Central: AR, LA, OK, TX
|
PAC
|
Pacific:
WA, OR, CA, AK, HI
|
MNT
|
Mountain:
MT, ID, WY, CO, NM, AZ, UT, NV
|
The
population data from the Data Book are based on the assumption that "population
change can be represented by administrative data in a statistical model." The Bureau of the Census conducts a
nationwide census every ten years. It
then uses data from documents that somehow reflect the change in the population
- birth and death certificates for example - to estimate how the population is
changing and what its current level is. This paper includes independent variables based on population density,
race/ethnic group, and age which are taken from this data (State and Metropolitan Area Data Book A-1).
The Data
Book obtains crime data from the Uniform Crime Reporting Program, which
consists of data voluntarily submitted to either the FBI or state UCR Programs
by law enforcement agencies across the United States. The two dependent variables in this study are based on this
data. "Total Crime" represents the
seven index crimes: murder and nonnegligent
manslaughter, forcible rape, robbery, aggravated assault, burglary,
larceny-theft, and motor vehicle theft. These are known as "index crimes" because of their seriousness,
frequency of occurrence, and likelihood of being reported to police. "Violent Crime" represents the four violent
crimes as defined by the UCR program: murder and nonnegligent manslaughter, forcible rape, robbery, and
aggravated assault. The crime data are
reported as crime rates - the number of crimes committed per 100,000
inhabitants. Although Violent Crime is
incorporated into Total Crime, it is used in this study to determine if
poverty has effects specific to violent crimes; it also provides a second data
set to observe. (A-9)
Unemployment
data are based on the Current Population Survey, which are annual averages of
monthly figures. The unemployed "are
all civilians who did not work during the survey week, who were available for
work during the survey week (except for temporary illness), and who made specific
efforts to find a job in the prior 4 weeks. Persons waiting to be recalled to a job from which they had been laid
off also are counted as unemployed" (A-11).
Personal
Income per Capita figures, as found in the Data Book, were taken from the Survey of Current Business conducted by
the US Bureau of Economic Analysis. They
consist of the personal income received by, or on behalf of, all members of the
area less personal contributions for social insurance. That total is then divided by the resident
population. The figures provide a
picture of the overall wealth of the area.
Poverty
data is based on the Current Population Survey as well. "Families and persons were classified as
below poverty level if their total family income or unrelated individual income
was less than the poverty threshold specified for the applicable family size,
age of householder, and number of related children under 18 present" (A-12).
Descriptive Statistics
|
Variable
|
Mean
|
Std. Dev.
|
Minimum
|
Maximum
|
Total
Crime
|
5414.5
|
1788.7
|
1114
|
13148
|
Violent
Crime
|
607.28
|
326.75
|
33
|
1881
|
Pop.
Density
|
.42359
|
.87778
|
.0053
|
11.808
|
Under 25
|
36.236
|
4.064
|
21.8
|
55.9
|
Over 65
|
12.646
|
3.554
|
4.7
|
34.3
|
Poverty
|
14.96
|
4.9998
|
5.9
|
41.1
|
Unemployment
|
5.6997
|
2.8551
|
1.8
|
28.7
|
Black
|
11.238
|
10.588
|
0.1
|
48.1
|
Asian
|
2.8049
|
4.7823
|
0.3
|
64.3
|
Hispanic
|
9.2669
|
14.293
|
0.4
|
94.9
|
PI per
Capita
|
20.449
|
3.6567
|
10.347
|
34.282
|
NE
|
0.03934
|
0.19473
|
0
|
1
|
MA
|
0.11803
|
0.32318
|
0
|
1
|
ENC
|
0.14426
|
0.35193
|
0
|
1
|
WNC
|
0.05246
|
0.22332
|
0
|
1
|
SA
|
0.19672
|
0.39817
|
0
|
1
|
ESC
|
0.06885
|
0.25362
|
0
|
1
|
WSC
|
0.14754
|
0.35523
|
0
|
1
|
MNT
|
0.08525
|
0.27971
|
0
|
1
|
PAC
|
0.14754
|
0.35523
|
0
|
1
|
The first
analysis will examine two simple regressions of total crime and violent crime
on poverty. These reflect the affect that
poverty has on crime before controlling for other variables.
Total Crime = β0 + β1(poverty)
+ є
|
R-square = 0.1421, R-square Adjusted = 0.1392
|
Standard Error of the
Estimate-Sigma = 1659.5
|
Variable
|
Est.
Coefficient
|
Standard
Error
|
|
|
Poverty
|
134.85
|
19.04
|
|
|
Constant
|
3397.1
|
300.2
|
|
|
Violent Crime = α0 + α1(poverty)
+ є
|
R-square = 0.1439, R-square Adjusted = 0.1411
|
Standard Error of the
Estimate-Sigma = 302.83
|
Variable
|
Est.
Coefficient
|
Standard
Error
|
|
|
Poverty
|
24.789
|
3.474
|
|
|
Constant
|
236.42
|
54.79
|
|
|
In both regressions,
the estimated coefficient of poverty is very significant. These regressions show that a one percent
increase in the population below poverty level will lead to an increase of
about 135 total crimes and about 25 violent crimes.
Now the
regressions of total and violent crime will be expanded to include all of the previously
mentioned variables. The models have
been reported with heteroskedasticity-robust standard errors. The geographic region that has been left out
is the South Atlantic. In addition,
Caucasians are not included in the race/ethnic group list.
Total Crime
|
|
Violent Crime
|
R-square = 0.4363, R-square Adjusted =�
0.4029
|
|
R-square = 0.4727, R-square Adjusted = 0.4414
|
Standard Error of the
Estimate-Sigma = 1382.2
|
|
Standard Error of the
Estimate-Sigma = 244.21
|
Variable
|
Est. Coefficient
|
Standard Error
|
|
Variable
|
Est. Coefficient
|
Standard Error
|
Poverty*
|
134.08
|
40.11
|
|
Poverty*
|
16.237
|
7.146
|
Population
Dens.
|
4.6788
|
117.1
|
|
Population
Dens.*
|
30.294
|
10.1
|
Unemployment*
|
-119.07
|
49.11
|
|
Unemployment
|
-11.171
|
10.03
|
Under 25
|
58.77
|
46.69
|
|
Under 25
|
-0.69455
|
7.19
|
Over 65o
|
63.89
|
37.87
|
|
Over 65o
|
11.651
|
6.345
|
Black*
|
60.187
|
14.64
|
|
Black*
|
17.431
|
2.735
|
Asian
|
3.4373
|
24.58
|
|
Asian
|
-1.4666
|
3.798
|
Hispanic
|
17.223
|
12.81
|
|
Hispanic*
|
5.6293
|
2.173
|
PI per
Capita*
|
142.52
|
37.98
|
|
PI per
Capita*
|
18.19
|
6.67
|
MA*
|
-1547
|
312.7
|
|
MA*
|
-146.54
|
49.8
|
ENC
|
-342.37
|
305.6
|
|
ENC
|
43.305
|
57.54
|
WNC
|
531.09
|
743.8
|
|
WNC
|
73.512
|
102
|
NE*
|
-653.64
|
332.1
|
|
NE
|
-31.975
|
76.68
|
ESC
|
-554.77
|
350.9
|
|
ESC
|
-93.855
|
66.74
|
WSC
|
23.448
|
331.7
|
|
WSC
|
24.489
|
64.23
|
MNT*
|
845.23
|
412.8
|
|
MNT
|
38.401
|
72.89
|
PACo
|
844.6
|
430.2
|
|
PAC*
|
187.04
|
69.01
|
Constant
|
-2544.4
|
2451
|
|
Constant
|
-342.46
|
388.2
|
oSignificant at the 10% level
|
|
oSignificant at the 10% level
|
*Significant
at the 5% level
|
|
*Significant
at the 5% level
|
The
estimated coefficient of poverty shows that holding all the other variables
constant, a 1% increase in the population under the poverty level will increase
total crimes by 134 per 100,000 inhabitants. It is surprising that this coefficient is almost identical to the simple
model, even after adding and holding constant sixteen independent variables,
but the standard error has increased which yields slightly less significance -
although it still is significant at the 0.1% level. The coefficient in the multivariate Violent Crime model decreased
by about 33% compared with the simple model; its significance decreased as well. In any case, both of these amounts are
substantial.
In order
to obtain the elasticity of crime with respect to poverty, the regressions will
be performed using the natural log of the crime rate and dividing the poverty
data by 100 (so that the data reads as a number between 0 and 1 instead of a
percentage between 0 and 100).
ln(Total Crime)
|
|
ln(Violent Crime)
|
R-square
= 0.4764, R-square Adjusted = 0.4454
|
|
R-square
= 0.4946, R-square Adjusted = 0.4647
|
Standard
Error of the Estimate-Sigma = 0.27270
|
|
Standard
Error of the Estimate-Sigma = 0.46917
|
Variable
|
Est
Coefficient
|
Std
Error
|
|
Variable
|
Est
Coefficient
|
Std
Error
|
Poverty*
|
2.1587
|
0.775
|
|
Poverty o
|
2.5682
|
1.31
|
Population Dens.
|
0.014904
|
0.02266
|
|
Populatn. Dens. o
|
0.037154
|
0.02088
|
Unemployment*
|
-0.019119
|
0.009085
|
|
Unemployment
|
-0.01676
|
0.01854
|
Under 25*
|
0.021036
|
0.009248
|
|
Under 25
|
-0.000049444
|
0.01441
|
Over 65o
|
0.014082
|
0.007229
|
|
Over 65
|
0.018493
|
0.0127
|
Black*
|
0.012319
|
0.002719
|
|
Black*
|
0.033024
|
0.004836
|
Asian
|
-0.00075461
|
0.004671
|
|
Asian
|
-0.0060179
|
0.005357
|
Hispanic
|
0.0029442
|
0.00205
|
|
Hispanic*
|
0.010896
|
0.003262
|
PI per Capita*
|
0.034803
|
0.007276
|
|
PI per Capita*
|
0.046468
|
0.01199
|
MA*
|
-0.39484
|
0.07122
|
|
MA*
|
-0.37916
|
0.1042
|
ENC
|
-0.067366
|
0.06495
|
|
ENC
|
0.055824
|
0.1081
|
WNC
|
0.046439
|
0.1208
|
|
WNC
|
-0.096352
|
0.2354
|
NE*
|
-0.11344
|
0.06982
|
|
NE
|
-0.14647
|
0.185
|
ESC
|
-0.077228
|
0.06542
|
|
ESC
|
-0.14096
|
0.12
|
WSC
|
0.035607
|
0.05517
|
|
WSC
|
0.12404
|
0.09242
|
MNT*
|
0.18064
|
0.07099
|
|
MNT
|
0.097143
|
0.1261
|
PAC*
|
0.17643
|
0.08775
|
|
PAC*
|
0.40353
|
0.1247
|
Constant
|
6.5173
|
0.4742
|
|
Constant
|
4.2651
|
0.7393
|
oSignificant at the 10% level
|
|
oSignificant
at the 10% level
|
*Significant at the 5% level
|
|
*Significant at the 5% level
|
The estimated
coefficient of poverty now yields crime-poverty elasticity. The ln(Total Crime) model shows that holding
the other variables constant, a 1% increase in poverty leads to a 2.16%
increase in total crime. This
coefficient is significant at the 0.6% level. The ln(Violent Crime) model shows that a 1% increase in poverty leads to
a 2.57% increase in violent crime and is significant only at the 5.1% level.
The
results of this analysis are consistent with other studies that have been
conducted by economists, for example:
[Studies] found that
homicides were disproportionately concentrated in areas of poverty. Three of these (by Bullock [1955], Beasley
and Antunes [1974], and Mladenka and Hill [1976]) studied violent crime in
Houston. Like Shaw and McKay [1969],
each reported high correlations between violent crime rates and measures of
poverty. Areas in Houston with high
rates of violent crime were also characterized by high population density and a
high proportion of black residents. (Short 51)
These studies as well as the above analysis show that
poverty is correlated with violent crimes - increased poverty leads to
increased violent crime.
Another study by Jens Ludwig, Greg J. Duncan, and
Paul Hirschfield also that teens committed more violent crimes when they lived
in high poverty areas than when they were relocated to low poverty areas. "The offer to relocate families from high- to
very low-poverty neighborhoods (census tracts with poverty rates below 10
percent) reduces juvenile arrests for violent offenses on the order of 30 to 50
percent of the arrest rate for controls" (Ludwig 13). However, in their paper, Ludwig and his associates state that "high-poverty
areas have on average higher rates of violent crime but lower rates of property
crime than low-poverty areas" (13). This contradicts the findings in the above analysis. Nevertheless, Ludwig and associates offer
explanations as to why this may be the case: "teens are subject to heightened scrutiny in low-poverty areas, and
national victimization surveys suggest that victims are more likely to report
less-serious crimes to the police in suburbs than cities" (14). They also believe that the experimental
group used in their experiment may actually change their behavior when they are
moved to a new area.
Although this study did not include regressions of
nonviolent crime on poverty, the Total Crime models capture more nonviolent
than violent effects. This can be seen
by comparing the means of the Total Crime and Violent Crime - Total Crime includes
about 4800 nonviolent crimes on average. The crimes Ludwig referred to may be less serious than the three
nonviolent index crimes, thus resulting in a different correlation.
Conclusion: Evidence greatly suggests that increases in
poverty lead to increased violent crime. Evidence from this paper also suggests the same for all seven index
crimes and careful studies can yield more verification of the matter.
Works Cited
Short, James F., Jr. Poverty, Ethnicity, and Violent
Crime. Boulder, CO: Westview Press,
1997.
Brill, Norman Q. America's Psychic Malignancy. Springfield, IL: Charles C Thomas
Publisher, 1993.
Ludwig, Jens, Greg J. Duncan, and Paul Hirschfield. Urban
Poverty and Juvenile Crime:
Evidence From a
Randomized Housing-Mobility Experiment. 20 April 2000.
<
http://www.jcpr.org/wpfiles/duncan.ludwig.revise4-25.PDF>