American Journal of Criminal Justice

, Volume 39, Issue 1, pp 77–93

“Last Hired, First Fired”: The Effect of the Unemployment Rate on the Probability of Repeat Offending

  • Stewart J. D’Alessio
  • Lisa Stolzenberg
  • David Eitle
Article

DOI: 10.1007/s12103-013-9199-1

Cite this article as:
D’Alessio, S.J., Stolzenberg, L. & Eitle, D. Am J Crim Just (2014) 39: 77. doi:10.1007/s12103-013-9199-1

Abstract

Research examining the connection between the unemployment rate and the aggregate crime is inconclusive. One explanation for the inconsistent findings is that the unemployment rate influences the criminal activity of repeat and first-time offenders in different ways. Results support this thesis by revealing an inverted U-shaped association between the unemployment rate and the probability of repeat offending. The curvilinear relationship likely results from repeat offenders and those lacking a criminal record entering and exiting the labor force at different levels of unemployment. Our findings highlight the role that the unemployment rate plays in affecting repeat offending and underscore the importance of distinguishing between repeat and first-time offending when analyzing the effect of the unemployment rate on crime.

Keywords

Unemployment rate Repeat offending First-time offending 

The logic that a downturn in the economy amplifies criminal activity is one of the earliest theses of criminological inquiry (Sellin, 1972). Yet despite the long standing assumption that a robust relationship exists between the unemployment rate and crime, the results of the voluminous number of inquiries delving into this proposed association can best be summarized as equivocal.1 The failure to evince a consistent positive linkage between the unemployment rate and crime has engendered protracted scholarly debate about the nature, strength, direction, and even the noteworthiness of this purported relationship. The literature highlights several potential reasons for the incongruous findings, including the length of the unemployment lag needed to model appropriately expected motivational effects and whether the unemployment rate is an adequate proxy measure of general economic conditions (Greenberg, 2001).

One line of inquiry not explored fully is the possibility that the unemployment rate has a differential effect on repeat and first-time offending patterns. More than 70 years ago, Radzinowicz (1939) argued that the relationship between unemployment and crime varied for different subgroups in the population. While no published research to our knowledge has investigated whether the unemployment rate has a differential effect on initial and repeat offending, some suggest that such a paradoxical relationship exists (Blumstein, Cohen, & Farrington, 1988). Yet despite theorizing that the unemployment rate may influence repeat and first-time offending patterns in divergent ways, the exact nature of this differential effect remains unresolved.

Unemployment and First-Time Offending

Two competing perspectives can be identified from a close examination of the relevant literature. One position influenced by the work of Cantor and Land (1985) argues that a rise in the unemployment rate induces more first-time offending than repeat offending. The expectation that a high unemployment rate inspires initial rather than repeat offending is theoretically grounded in that a person with a paying job who is laid off has a direct economic incentive to commit crime because of the abrupt and dramatic change in his or her living conditions (Becker, 1968). The desire to engage in illegal activities can also increase among the employed during periods of economic upheaval since a rise in the unemployment rate adversely influences the earnings of workers (Parker, 1981). In such a situation crime simply becomes an alternative means for the continuance of a preexisting lifestyle or possibly even a means for subsistence. An individual previously lacking a criminal record delves into crime after losing his or her job because the benefits of partaking in illegal activities now outweigh the costs associated with such behavior.

In addition to the monetary motivation to initiate criminal behavior generated from the loss of a paying job, the recently unemployed have less of a stake in obeying the law because they no longer have a job to lose if arrested. The finding of a connection between unemployment and young adult crime is often adduced to strengthen the claim that unemployment acts to escalate initial rather than repeat offending because one would naturally expect to see most first-time adult arrestees emerge among young adults (Greenberg, 1985). Further evidence supporting this position can be found in studies demonstrating that the unemployed have higher participation rates in crime than do the employed, even after controlling for prior criminal history (Viscusi, 1986).

It is also asserted that the disappointment and anger associated frequently with losing a job may lead to first-time offending, particularly when the individual perceives that the loss of his or her job is unjustified. Considering the linkage between unemployment and mental health (Catalano et al., 2011), there is little doubt that the loss of a job can be an extremely frustrating and psychologically traumatic event. Being terminated abruptly from a job is considered one of the top ten traumatic events experienced by an individual over his or her life-course (Spera, Buhrfeind, & Pennebaker, 1994). Losing a job is not only psychologically distressing, but it can also engender disillusionment and anger because being terminated is often the result of a diminished demand for labor rather than the consequence of a person’s job performance. Aggressive behavior transpires when an individual is blocked from actualizing his or her goals, particularly in instances where the blockage is perceived by the individual to be unjust (Agnew, 1992). For example, an individual is more apt to berate others verbally (Ebbesen, Duncan, & Konecni, 1975) and to exhibit threatening behavior following the loss of a job (Catalano, Dooley, Novaco, Hough, & Wilson, 1993; Catalano, Novaco, & McConnell, 2002).

While unemployment may lead to more first-time offending because of a need effect or because of the amplification of a person’s anger and frustration, a high unemployment rate is not expected to have a commensurate effect on repeat offending patterns because repeat offenders are chronically unemployed (Parker & Horwitz, 1986). Thus, a rise in unemployment produced by the deterioration in economic conditions should have little direct impact on the economic situation of the majority of repeat offenders in the population. There is also not much of a substantive difference between being unemployed and the menial type jobs typically available to individuals with a prior criminal record (Rasmusen, 1992). Ample evidence exists to suggest that stable, quality employment is beyond the reach of individuals who are stigmatized by a criminal record.2 Individuals with a criminal record are 50 % less likely than are people without an official criminal record to be re-contacted by an employer after the initial job interview (Pager, 2003). Surveys also find that a majority of employers are unwilling to hire individuals with a criminal record under any circumstance (Holzer, Raphael, & Stoll, 2007). State licensing agencies further restrict the employment opportunities of ex-offenders by prohibiting them from being licensed in professions such as home healthcare, nursing, education, plumbing, and barbering (Kethineni & Falcone, 2007; Wheelock, Uggen, & Hlavka, 2011). In short, the stigma of a criminal record, coupled with ex-offenders’ low levels of educational attainment and lack of job skills (Visher, Debus-Sherrill, & Yahner, 2011), severely limits employment opportunities independent of the overall unemployment rate such that ex-offenders have little chance of obtaining stable quality employment.3

One might also argue that repeat offenders become more prolific in their offending when economic conditions are favorable because targets for crime are more plentiful and guardianship levels are curtailed due to the rise in employment among the citizenry. Along these lines, Piliavin, Gartner, Thornton, and Matsueda (1986) report that individuals with a criminal record are much more likely than individuals without a criminal record to engage in illegal activities when they perceive the opportunities for such activities to be abundant. Among those committed to a criminal lifestyle, employment also does not necessarily prevent the individual from participating in illegal activities. Nor does it preclude one from using the opportunities available because of employment to commit crimes while at work.

Guardianship levels also probably decrease as the unemployment rate declines. Borrowing from routine activities theory, Cantor and Land (1985) argue that a decline in the unemployment rate reduces guardianship because individuals employed in the workforce are not at home during the workday protecting their possessions from criminals. There is empirical evidence to support the notion that households are attractive targets for burglars during the daytime when they are most likely to be unoccupied. For example, when law enforcement agencies were able to establish the time that a burglary was committed (approximately 75 % of the residential burglaries), about 62 % of the residential burglaries were committed between the hours of 6:00 a.m. to 6:00 p.m. (Department of Justice, 2006). In a study that differentiated between residential burglaries that occurred on weekdays between the hours of 6:00 a.m. to 6:00 p.m. from residential burglaries that transpired on weeknights/weekends, D’Alessio, Eitle, and Stolzenberg (2012) found a negative association between the unemployment rate and the weekday residential burglary rate. The unemployment rate had little influence on the weeknight/weekend residential burglary rate or on residential burglaries generally. Their results suggest that unemployment attenuates residential burglary because of an immediate increase in guardianship levels engendered by the unemployed being at home during the workday. Other studies also report that unoccupied households have an elevated risk of burglary victimization (Wright & Decker, 1996), that major activities away from the home amplify the risk of property victimization (Miethe, Stafford, & Long, 1987) and that burglars attempt to avoid occupied residences (Bennett & Wright, 1984).

Unemployment and Repeat Offending

However, while it is certainly plausible that a burgeoning unemployment rate inspires more initial than repeat offending, some scholars remain unconvinced. The reason for this doubt is that it is just as plausible that a rise in the unemployment rate acts to amplify repeat rather than first-time offending because criminality during periods of economic distress is probably more likely to transpire among individuals already exhibiting a propensity for such activity (Blumstein et al., 1988). Empirical research supports this position by documenting that during periods of economic distress the average rate of juvenile offending only increases among juveniles already prone to delinquency (Farrington, Gallagher, Morley, Ledger, & West, 1986). Other studies also find that unemployment has a substantive impact on the criminal activity of individuals with a prior criminal record. Sviridoff and Thompson (1983) found in a study of male misdemeanants released from New York City’s correctional facility at Rikers Island that the largest percentage of ex-offenders who recidivated (38 %) were rearrested within 6 months of losing their last job. Harer (1994) also reports that job stability lessens the chance of recidivism. In another study, Berk, Lenihan, and Rossi (1980) found that experimentally induced unemployment heightened arrest levels for property and non-property crimes even after work effort was controlled statistically. They also observed that ex-offenders who received special job placement and counseling services experienced significantly fewer arrests for all types of crime. Wang, Mears, and Bales (2010) found that the unemployment rate for black males influenced the likelihood of recidivism for black males released from prison in 67 Florida counties. Finally, there is considerable evidence that convicted offenders have a much greater chance of being employed when overall unemployment rates are low (Schmitt & Warner, 2011). Such a finding suggests that during economic upswings maximal opportunities exist for repeat offenders to discontinue or at a minimum attenuate their criminal activity.

For a variety of reasons, it is also somewhat difficult to imagine a situation in which the loss of a job would induce large numbers of previously law-abiding citizens to participate in criminal activity especially violent forms of criminal behavior. First, although self-report studies document that a large percentage of the population has committed minor law violations at some point in their lives, only a small fraction of the population is responsible for the vast amount of serious crime that we experience in society (Wright, Tibbetts, & Daigle, 2008). Second, most people are only unemployed for a relatively short period of time (Clark & Summers, 1979). The average time the unemployed were out of work in 2007 was only 16.8 weeks nationally, according to federal data.4 Thus, because the average time a person is unemployed is about 3 months and unemployment benefits typically last longer, most people lacking an official criminal record would have already found another job long before their savings had been depleted and the motivation for crime took hold. Research further demonstrates that many of the unemployed reject at least one job offer before finally accepting a job (Blau, 1992). It is also plausible that a high unemployment rate may act to strengthen the inhibition against partaking in criminal activity among the employed and among those individuals wishing to enter the labor market (Catalano et al., 1993). The strengthening of the inhibition against criminal activity is speculated to occur because during periods of economic hardship employers can more easily replace workers with undesirable characteristics, such as those having aggressive tendencies (Catalano & Kennedy, 1998).

In the current study, we use judicial processing information drawn from 40 large urban U.S. counties to assess the effect of the unemployment rate on the likelihood of repeat offending.5 We determine this effect using a hierarchical generalized linear model (HGLM). A multilevel model is advantageous for our purposes because it affords us the ability to consider the influence of both offender-level and contextual factors such as the aggregate unemployment rate in predicting the probability of repeat offending.

Data

The offender-level data used in this study were gathered by the Bureau of Justice Statistics (2007) to track felony cases processed though the criminal justice system in 40 of the nation’s 75 most populous counties. This dataset is archived at the Inter-university Consortium for Political and Social Science Research at the University of Michigan (ICPSR02038). The data analyzed in this study contain information on 10,222 adult offenders charged with a felony crime during 2000 in 39 of the counties.6 Juvenile felony offenders prosecuted as adults are also included in the data. This dataset not only provides information on the demographic characteristics of individual criminal offenders such as race, sex and prior criminal record, but also geocode information that identifies the county of prosecution. These geographic identifiers can be used to match each case with the unemployment rate and other aggregate county variables theorized to be associated with repeat and first-time offending.

Dependent Variable

Prior criminal record is the dependent variable and is coded as a dichotomy. Repeat offenders are coded as one and first-time offenders are coded as zero. Repeat offenders are defined as individuals with one or more prior misdemeanor and or felony arrests. All the offenders included in the dataset were prosecuted for a felony crime. In regards to the two theoretical positions articulated at the beginning this article, it is not germane whether individuals defined as first-time offenders were actually engaging in illegal activities prior to their first arrest. What is salient for the perspectives advanced here is whether the offender had a prior criminal record at the time of his or her arrest because it is this record rather than a person’s participation in illegal activities per se that acts to impede the individual in actualizing his or her ambitions in the labor market.

Offender and County Variables

Our analysis seeks to account for several factors related to the characteristics of the offender that might reasonably be posited to impact criminal activity. These offender characteristics include the age of the offender, the race of the offender, the sex of the offender, and whether the offender is Hispanic. Several variables measured at the county-level are also included in the multilevel analysis. These contextual variables, except for the crime rate (UCR), were obtained from the 2000 Census. The unemployment rate is the variable of theoretical interest and is defined as the percent of the civilian labor force unemployed in each county in 2000. Although the unemployment rate is an imperfect measure of the health of the economy because it excludes part-time workers seeking full-time jobs and unemployed workers no longer on unemployment rolls, we use the unemployment rate in our analysis rather than some other measure because indicators of economic activity are usually highly correlated with each other and because unemployment is frequently perceived to be the most theoretically relevant proxy of overall economic conditions (Parker & Horwitz, 1986). Our use of the unemployment rate also helps us to maintain comparability with much of the previous research that examined the relationship between unemployment and crime.

Prior research also indicates that several other contextual factors besides the unemployment rate may influence the likelihood of repeat offending, including annual income for employees in manufacturing occupations, population density, percentage of young people in the population, crime rate, and community disadvantage. Annual income for employees in manufacturing occupations has been used previously as an indicator of the extent of quality employment available to low educated workers (Weiss & Reid, 2005). Population density is employed to examine how urbanization influences repeat offending. Because research has established that young people are disproportionately involved in serious criminal activity (Stolzenberg & D’Alessio, 2008), the percent of the population between the ages of 15 to 24 was also included in the analysis as a contextual variable. The percentage of the black population and the crime rate were also incorporated in the analysis as statistical controls. Lastly, we include factor scores from a principal components analysis of four indicators of community disadvantage: percent of the population (ages 25+) that never graduated from high school; percent of households with public assistance income; percent of households headed by a single female (ages 15–64) with children; and percent of families below the poverty level in 1999 (see Table 1). A high score on this composite variable indicates a greater level of community disadvantage. The means, standard deviations, and definitions for all the variables are displayed in Table 2.
Table 1

Principal component analysis

 

% of variance

Community disadvantage component

Percent of the population (ages 25+) that never graduated from high school

88.273

.917

Percent of households with public assistance income

5.783

.925

Percent of households headed by a single female (ages 15–64) with children

3.777

.951

Percent of families below the poverty level in 1999

2.167

.965

Table 2

Means, standard deviations, and definitions for the variables used in the analysis

 

Mean

Std. dev.

Definition

Repeat offender

.75

.43

Coded 1 if the offender was previously arrested, 0 otherwise.

Offender’s age

31.48

10.46

Age of the offender in years.

Offender white

.42

.49

Coded 1 if the offender is white, 0 if the offender is black.

Offender male

.80

.40

Coded 1 if the offender is male, 0 otherwise.

Offender Hispanic

.01

.10

Coded 1 if the offender is Hispanic, 0 otherwise.

Unemployment rate (2000)

3.87

1.06

Percent of the civilian labor force that is unemployed in 2000.

Unemployment rate (1999)

3.96

1.67

Percent of the civilian labor force that is unemployed in 1999.

Unemployment rate (2000)2

16.06

9.21

Percent of the civilian labor force that is unemployed in 2000 (squared).

Unemployment rate (1999)2

18.37

17.25

Percent of the civilian labor force that is unemployed in 1999 (squared).

Manufacturing income

28,593.46

6,165.63

Average annual income for employees in manufacturing occupations.

Population density

2,875.52

5,627.19

Population per square mile of land area.

Percent 15–24

14.22

1.95

Percent of the population prone to criminal activity (ages 15–24).

Percent black population

16.58

13.87

Percent of the population that is black or African American.

Crime rate

4,988.49

2,177.32

Number of index crimes divided by the county population and multiplied by 100,000,

Community disadvantage

.00

1.00

Factor scores from principal component analysis of four variables: a) percent of households with public assistance income, b) percent of the population (ages 25+) that never graduated from high school, c) percent of households headed by a single female (ages 15–64) with children, and d) percent of families below the poverty level in 1999. Larger scores indicate greater disadvantage.

Analysis and Results

We employed a nonlinear hierarchical modeling procedure to generate parameter estimates because the data are multilevel and the dependent variable is a dichotomy (Raudenbush, Bryk, Cheong, & Congdon, 2001). A penalized quasi-likelihood (PQL) procedure was employed to generate parameter estimates for the binary outcome (Breslow & Clayton, 1993). With our number of aggregate units, there should be sufficient power to show a connection between the unemployment rate and the probability of repeat offending (Bassiri, 1988). We grand centered all the variables included in the analysis by subtracting their grand means, so that the mean of each variable was zero across all cases. The centering of a variable is useful because it helps to mitigate multicollinearity. The offender characteristic variables were also modeled as fixed, or constrained to be the same for all the counties, since between-county variation in these parameters are not of interest in this study. The one randomly varying parameter is the intercept because we are interested in determining whether variation in unemployment among the counties influences the probability of repeat offending. The intercept can be interpreted as the mean level of repeat offending among the counties, after adjusting for the independent variables included in the model. Furthermore, because we permit the intercept to vary across counties, we are able to model any observable differences in the likelihood of repeat offending between the counties with contextual variables such as the unemployment rate. This variability between counties is the outcome to be explained in the between-county model.

We began our analysis by estimating a within-county regression model using the offender variables to predict the probability of an offender being a repeat offender. Since the intercept is allowed to vary, we were able to examine the size and significance of its estimated variance (τ) to determine whether it was worthwhile to use the unemployment rate and the other contextual variables to predict between-county differences in its slope. A chi-squared test indicated that the likelihood of repeat offending varied considerably among the counties (τ= .66, p < .001). That is, the probability of repeat offending is higher in some counties than in others even after the offender characteristic variables are taken into account. The purpose of the between-county model is to use the contextual variables, including the unemployment rate, to explain why this unexplained variation exists.

The results for the county-level analysis are reported in Model 1 of Table 3. A visual inspection of this model reveals that there is not a consequential effect of the unemployment rate on the probability of repeat offending. After the offender characteristic variables and the contextual variables are controlled, labor market conditions are not salient in explaining variation among the counties regarding the likelihood of repeat offending. This finding seems to call into question the assertion that the unemployment rate has a differential effect on repeat and first-time offending patterns.
Table 3

Nonlinear hierarchical models estimating the probability of repeat offending

  

Model 1

Model 2

Model 3

Model 4

  

Unem. 2000

Unem. 1999

Unem. 2000 w/Curvilinear

Unem. 1999 w/Curvilinear

Intercept

1.060 (.075)

1.061 (.073)

1.063 (.066)

1.062 (.066)

 

Unemployment rate

−.156 (.186)

−.162 (.100)

1.044** (.338)

.310* (.148)

Unemployment rate2

−.156*** (.038)

−.046*** (.010)

Manufacturing income

−.017e-3 (.011e-3)

−.018e-3 (.011e-3)

−.021e-3 (.012e-3)

−.020e-3 (.011e-3)

Population density

−.154e-3*** (.026e-3)

−.155e-3*** (.024e-3)

−.137e-3*** (.023e-3)

−.146e-3*** (.023e-3)

Percent 15–24

.074 (.052)

.056 (.052)

.072 (.046)

.069 (.051)

Percent black population

−.002 (.010)

−.005 (.009)

−.010 (.009)

−.008 (.009)

Crime rate

−.151e-3** (.043e-3)

−.146e-3*** (.041e-3)

−.180e-3*** (.046e-3)

−.155e-3*** (.043e-3)

Community disadvantage

.488 (.268)

.635** (.212)

.706** (.246)

.569** (.200)

Offender’s age

.034*** (.003)

.035*** (.003)

.035*** (.003)

.035*** (.003)

Offender white

−.621*** (.101)

−.623*** (.101)

−.626*** (.101)

−.624*** (.101)

Offender male

.708*** (.084)

.709*** (.084)

.713*** (.085)

.713*** (.085)

Offender Hispanic

−.226 (.262)

−.230 (.263)

−.237 (.267)

−.239 (.270)

Results are estimated from population-average models with robust standard errors. Standard errors are in parentheses

*p ≤ .05; **p ≤ .01; ***p ≤ .001 (two-tailed tests)

The results displayed in Model 1 of Table 3 also reveal effects of annual income for employees in manufacturing occupations, population density, the percent of young people in the population, the crime rate, and community disadvantage. Two contextual variables are salient in this model. In counties that are more densely populated and that have a higher crime rate, the likelihood of repeat offending is reduced. None of the other county-level variables are statistically significant in this equation. The coefficients for the offender’s age, race and sex are also noteworthy in this equation. The likelihood of repeat offending is enhanced when the offender is older, when the offender is black and when the offender is male.

The analysis presented in Model 1 focused on the contemporaneous relationship between the unemployment rate and the likelihood of repeat offending, but one has to wonder whether a lagged relationship might exist between the unemployment rate and the likelihood of repeat offending since people are probably more inclined to partake in criminal activity once their monetary savings are depleted. Cantor and Land (1985) estimate that it might take as long as 1 year for someone’s financial savings to be exhausted. Based on this estimate, we conducted a second analysis using a one-year lag for the unemployment rate. The results for this analysis, which are presented in Model 2 of Table 3, fail to show any evidence of a lagged relationship between the unemployment rate and the probability of repeat offending. The results for the other variables remain unchanged in this model. Taken in their totality, the results presented in Models 1 and 2 of Table 3 provide little evidence that the unemployment rate influences repeat offending.

Although the results presented in Models 1 and 2 of Table 3 suggest that the unemployment rate affects repeat and first-time offending similarly, we felt it prudent to consider whether a curvilinear relationship exists between the unemployment rate and the probability of repeat offending because people with a prior criminal record may differ from individuals lacking an official criminal record in their tendency to enter and exit the labor force at different levels of unemployment. It is plausible that repeat offenders are less prolific in their criminal offending when the unemployment rate is low because it easier for them to find employment when jobs are plentiful (Schmitt & Warner, 2011). The enhanced competition for workers effectuated by a tight labor market also forces employers to pay higher wages to workers, especially to workers at the low end of the wage scale (Bartik, 2001).7 A higher wage is effective in attenuating crime among repeat offenders (Needels, 1996). But when economic conditions begin to deteriorate, individuals with a prior criminal record along with other less desirable workers are the first in line to be laid off by employers (Holzer, Raphael, & Stoll, 2004). The proclivity for crime then intensifies among individuals with a prior criminal record as they are terminated from their jobs. Thus, when the unemployment rate is at a moderately high level, one would expect the probability of repeat offending to be at its highest. However, when the unemployment rate is extremely high, there is a much lower probability of repeat offending because many of these individuals already lost their jobs when the economy initially began its downward spiral. The effect of unemployment on the criminal activity of repeat offenders should thus be much less substantive when the unemployment rate is high.8

To assess the possibility of a curvilinear relationship, we modeled the effect of the unemployment rate as a second order polynomial, including as independent variables the unemployment rate and the unemployment rate squared. Because a preliminary analysis revealed that the unemployment rate variable and its square were highly correlated, we expressed these two variables as deviations from their respective means to reduce this collinearity (Aiken & West, 1991). If the relationship between the unemployment rate and the probability of repeat offending conforms to an inverted U-shaped pattern, the sign of the coefficient for the unemployment rate variable should be positive and its square negative.

The results of the curvilinear analysis, which are reported in Model 3 of Table 3, demonstrate that the unemployment rate and its square are consequential in determining the probability of repeat offending. The sign of the coefficient for the unemployment rate variable is positive and its square is negative indicating an inverted U-shaped pattern. To facilitate interpretation of this curvilinear relationship, we constructed a figure that portrays graphically the inverted U-shaped relationship between the unemployment rate and the probability of repeat offending. The plotting of the expected probabilities in Fig. 1 reveals that when the county unemployment rate is at a low and very high level, the expected probability of repeat offending is also low. But when the unemployment rate is at a moderately high level in a county, the expected probability of repeat offending is at its highest. For a county that has an unemployment rate that is one standard deviation greater (approximately 2.3 to 3.96) the discrete change in the expected probability of being a repeat offender is .27 (from an expected probability of .47 to .74). However, the discrete change is less dramatic for counties with an unemployment rate above the mean. To illustrate, in a county that has an unemployment rate of 7.3 % the expected probability of repeat offending is .05 greater (from .88 to .93) than in a county with an unemployment rate that is one standard deviation lower (5.6 %). The discrete change in the expected probability of repeat offending approaches zero in counties with an unemployment rate within this range. If a county had an unemployment rate of 14.1 %, there would be a .07 decrease (from .92 to .85) in the expected probability of repeat offending as compared to a county with an unemployment rate one standard deviation lower (12.4 %).9
Fig. 1

Expected probability of repeat offending for selected values of unemployment

Figure 1 reinforces the finding that the strength and direction of the relationship between the unemployment rate and the probability of repeat offending is dependent on the relative level of unemployment. The results for the other variables included in Model 3 remain unchanged from those estimated in Models 1 and 2, except for the effect of community disadvantage. The likelihood of repeat offending is higher in counties that are plagued by community disadvantage.

We also conducted a second nonlinear analysis to determine whether the curvilinear relationship between the unemployment rate and the likelihood of repeat offending would hold if the lagged unemployment rate was used as the variable of theoretical interest. The results for this analysis, which are reported in Model 4 of Table 3, mirror the results reported in Model 3. Both the lagged unemployment rate variable and its square are statistically significant. The results for the other variables included in the model remain unchanged from those reported in Model 3.10

Conclusion

The effect of the unemployment rate on criminal activity continues to remain a topic of interest. Theory suggesting that downturns in the economy engender a rise in criminal activity has induced researchers to focus their energies on trying to establish a nexus between the unemployment rate and criminal activity. This line of inquiry is rooted in the logic that a rise in the unemployment rate elevates crime because the unemployed are more criminally motivated than the employed, notwithstanding the underlying reason(s) for this enhanced desire. What is surprising and somewhat disconcerting is that research appraising the relationship between the unemployment rate and crime has produced contradictory findings. One potential reason for the incongruous findings reported in the literature is the implicit assumption made in previous work that unemployment affects repeat and first-time offending similarly, despite theorizing that this supposition is probably incorrect. We took a different path than that taken in prior research by investigating whether the unemployment rate has a differential impact on repeat and first-time offending patterns.

The initial results generated in the multilevel analysis indicated that the unemployment rate had a congruent effect on both repeat and first-time offending. Had we stopped here we might have drawn the faulty conclusion that the linkage between the unemployment rate and crime did not vary among different subgroups in the population as originally theorized by Radzinowicz (1939). However, once we added the squared term for the unemployment rate variable to the model, there was a strong curvilinear relationship between the unemployment rate and the probability of repeat offending. The often heard adage “Last hired, first fired” aptly describes the essence of this curvilinear relationship. Research shows that employers are much more hesitant to hire ex-offenders than other low-skilled, stigmatized workers such as welfare recipients or workers with inconsistent employment histories (Schmitt & Warner, 2011). The mark of a criminal record signals employers that the person is unskilled, untrustworthy, has a propensity to steal, and that he or she might harm customers (Holzer et al., 2007). While employers’ generally have a negative attitude toward people with a criminal record, their willingness to hire these individuals is dependent on labor market conditions. When economic conditions are favorable and the unemployment rate is low, the transition of ex-offenders into the workforce increases markedly. Employers are receptive to hiring less desirable workers when the labor market is tight and additional workers are needed, otherwise employers would invariably face economic costs related to their unwillingness to hire these less attractive workers (Becker, 1971).

But when economic conditions deteriorate and the unemployment rate is high, individuals with a criminal record and other less desirable workers are quick to be laid off by employers (Holzer et al., 2004). Crime activity then increases among these individuals. However, when the unemployment rate reaches a very high level, many previously employed individuals with a prior criminal record are no longer in the workforce. A very high unemployment rate should then have little direct impact on the likelihood of repeat offending because many repeat offenders, along with other less desirable workers, lost their jobs when the economy initially began to deteriorate.

The results generated in this study highlight the importance of distinguishing between repeat and first-time offending when analyzing the effect of the unemployment rate on crime. By creating such an analytic distinction, we were able to ascertain one probable reason for why previous studies investigating the relationship between the unemployment rate and crime produced disparate findings. We are by no means claiming that we have identified the sole reason for the inconsistent findings reported in the literature. What we can say is that the failure of researchers to consider explicitly the differential effect that the unemployment rate has on repeat and initial offending may lead one to reach erroneous conclusions.

While our results suggest that concentrating on subgroups such as repeat offenders when analyzing the effect of the unemployment rate on crime is a potentially fruitful analytic approach, the current study is preliminary in scope. The following issues need to be contemplated when evaluating the import of our findings. First, because the results generated in this study pertain to urban contexts, they may not hold true for sparsely populated rural areas. It is also of import to recognize that in addition to the effect of the unemployment rate, the influence of other contextual variables on initial and repeat offending may hinge on whether an urban or rural context is analyzed. Take the effect of community disadvantage for example. Social disorganization theory proffers that factors such as residential instability, ethnic diversity, family disruption, and concentrated poverty are prominent in influencing a community’s ability to develop and maintain an enduring system of social relationships. The urban/rural nature of the social environment is thus theorized to play a role in enhancing or impeding a community’s ability to control its residents (Osgood & Chambers, 2000). However, while it is certainly warranted to more fully explore whether the degree of urbanity of a community matters in influencing initial and repeat offending patterns, researchers need to remember that the unemployment rate remains a problematic measure in rural contexts because self-employment and farming, which both tend to be major occupations in rural settings, do not often qualify for unemployment compensation.

Our findings also need to be replicated with longitudinal data before they can be accepted without question. Because this study was cross-sectional in design, it is somewhat difficult to draw inferences about the robustness of our findings over lengthy periods. Longitudinal analyses also need to be undertaken to weigh the possibility of a feedback effect. The causal relationship between the unemployment rate and crime may not be unidirectional, but rather reciprocal because incarcerated individuals are removed from employment roles. Survey data should also be analyzed to help determine the specific reasons for why repeat and first-time offenders decided to perpetrate their criminal offenses, although a potential problem with this line of inquiry is that repeat offenders tend to be vastly unrepresented in self-report surveys (Cernkovich, Giordano, & Pugh, 1985).

Notwithstanding these caveats, our findings suggest that any attempt to reintegrate individuals with a criminal record into the society without either sealing or expunging their official criminal records is unlikely to attenuate repeat offending to any substantial degree. The stigma associated with a criminal record has a number of adverse consequences, including the hindering of an individual’s ability to secure and maintain worthwhile employment (Pager, 2003). While people with an official criminal record are able to enjoy the material benefits derived from a low unemployment rate, these same individuals suffer when economic conditions deteriorate because their criminal past makes them more apt to be terminated from their jobs than other types of workers. However, while the sealing or expunging of criminal records might reduce crime among repeat offenders, there still remains the possibility that first-time offending might escalate with such a policy because “… the opportunity cost of crime would fall for workers who had not been criminal in the past since they would receive the same pooled wage as the criminal workers” (Rasmusen, 1996:539).

The employment prospects of citizens lacking an official criminal record may also be impacted adversely because “[i]f accessibility to criminal history information is limited (because of cost or legal prohibitions), employers may infer the likelihood of past criminal activity from such traits as gender, race, or age” (Holzer, Raphael, & Stoll, 2006:452). Such a situation is especially problematic for black citizens lacking an official criminal record because of the widespread perception that African-Americans are more criminally inclined than whites (Hurwitz & Peffley, 1997). This perception among the public would invariably lead employers to view all black employees as potential criminals, notwithstanding whether or not they actually had a criminal record. Any concerted effort directed at making it more difficult for private companies to access criminal records would likely encounter stiff resistance (Gebo & Norton-Hawk, 2009). One way to resolve this paradox is to implement a policy whereby less serious offenders have their official criminal records sealed or expunged, while the most serious criminal offenders still maintain their official criminal records. Such a policy helps to attenuate the prospect for intensification in first-time offending while simultaneously giving less serious repeat offenders, who represent the bulk of the offending population, some hope that they will eventually be fully reintegrated into the civilian workforce (Berk, Sherman, Barnes, Kurtz, & Ahlman, 2009).

Footnotes
1

See Allen (1996) and Vieraitis (2000) for comprehensive reviews of this literature.

 
2

Besides having an economic impact, the stigma of a criminal record can also have social ramifications by magnifying an individual’s difficulty in finding a spouse or by attenuating the probability that a person will be admitted to a university (Rasmusen, 1996).

 
3

Another potential reason for the chronic unemployment among repeat offenders is that these individuals have characteristics, traits, and experiences prior to a conviction that greatly diminish their likelihood of obtaining and retaining quality employment and that serve to shape criminal propensity independent of fluctuations in employment rates and related economic conditions (Gottfredson & Hirschi, 1990).

 
4

However, it is important to note that 2007 marked the beginning of an economic recession in the U.S. (Mian & Sufi, 2010).

 
5

Although an individual is not typically classified as a criminal offender until he or she is convicted in a criminal court, we use offender throughout the manuscript to describe the defendant in order to remain consistent.

 
6

One county was dropped from the analysis because it was missing data on the crime rate variable.

 
7

A low unemployment rate has other positive consequences such as reducing welfare dependency and decreasing poverty levels (Layard, Nickell & Jackman, 2005).

 
8

Our decision to conduct a nonlinear analysis was further supported by a visual examination of the scatterplots for the variables included in the analysis, which suggested the possibility of an inverted U-shaped relationship between the unemployment rate and the likelihood of repeat offending.

 
9

However, it is important to note that while the expected probabilities of some of the unemployment rates were higher than the mean rate during the time period that these data were collected (i.e., ten of the counties had unemployment rates in 1999 that exceeded 5 %), only a few of the counties had extremely high unemployment rates. Because of the small number of counties with extremely high unemployment rates during the observation period, one should view the evidence supporting an inverted U-shaped relationship between the unemployment rate and the probability of repeat offending with some healthy skepticism until further research is conducted.

 
10

Although juveniles prosecuted as adults comprise only 2.6 % of the total number of prosecuted offenders, we excluded these juvenile offenders and re-estimated our equations to help ensure that our results remained robust across different specifications. The results produced from this analysis were nearly identical to the findings reported in Table 3.

 

Copyright information

© Southern Criminal Justice Association 2013

Authors and Affiliations

  • Stewart J. D’Alessio
    • 1
  • Lisa Stolzenberg
    • 2
  • David Eitle
    • 3
  1. 1.Department of Criminal JusticeFlorida International UniversityMiamiUSA
  2. 2.Department of Criminal JusticeFlorida International UniversityMiamiUSA
  3. 3.Montana State UniversityBozemanUSA

Personalised recommendations