The Relationship between Poverty and Recidivism in Indiana Matthew Channing PA 706 December 10, 2014 Abstract Recidivism is the instance of convict released from incarceration and returning to incarceration within three years (Petersilia, 1999, p. 512). The causes of this are not universally known nor agreed upon. This paper presents a literature to investigate the background research of poverty’s effect on reoffending. As a caveat, poverty and recidivism are both difficult to measure meaningfully, as the literature shows. Simple linear regression analyses were conducted on four measures of poverty against recidivism rates for counties in Indiana. What was found was that poverty as defined in this paper was not strongly associated with recidivism rates, but indicators of real and potential economic need, unemployment rate and rates of receiving federal benefits via SNAP, are. Channing 1 Introduction I. Incarceration is usurping a larger share of state budgets than previously, while at the same time, states have less money to dedicate to the process (Orrel & Blackwell, 2008, p. 1). Yet, recidivism is likely a symptom of greater societal conditions. If all a state does is seek to treat the symptom, then the trend of resource-drain demanded by the criminal justice system will continue. The purpose of this paper is to determine if living in an area of poverty is related to recidivism. Literature Review II. The review of existing literature is divided into two main categories: review of material on poverty and a review of material on recidivism. In regards to poverty, the literature spotlights the method through which it is calculated and the controversies surrounding it. Also, other methods for determining a poverty statistic were examined to create a comparison in their validity and applicability. For the literature on recidivism, the articles centered around demographic statistics and likely contributing factors. Additionally, this paper reviews literature about the reentry process for ex-offenders. Poverty Poverty is measured in multiple forms, each with their own purpose. The most common measurement is the Census Bureau’s Current Population Survey’s Annual Social and Economic Supplement (CPS ASEC), which is the official measurement of poverty. Channing 2 The Census Bureau produces its figures from “the Current Population Survey's Annual Social and Economic Supplement (CPS ASEC)” (Institute for Research on Poverty, 2013). As context, the poverty measurement was useful because in 1964, it coincided with President Johnson’s War on Poverty (Fisher, 1992). These numbers came from “comparing pre-tax cash income against a threshold that is set at three times the cost of a minimum food diet in 1963, updated annually for inflation using the Consumer Price Index, and adjusted for family size, composition, and age of householder” (Institute for Research on Poverty, 2013). In addition to the pre-tax cash, other forms of income included are “cash benefits from the government…Social Security and Unemployment Insurance…Supplemental Security Income (SSI), public assistance benefits, such as Temporary Assistance for Needy Families… and workers compensation benefits” (Short, 2013). Calculating poverty status is expressed in a simple equation, “if a family's total income is less than the family's threshold, then that family and every individual in it is considered in poverty” (United States Census Bureau, 2013). An interesting note about the CPS ASEC is its definition of family and its insistence on that unit for calculating the poverty measure. The poverty thresholds themselves are determined by the size of the family and the age of the family members (United States Census Bureau, Poverty, 2013). As defined by the CPS portion of the census, a family is the related people within a household and a family household is the total of all people within a (United States Census Bureau, Definitions, 2013). As stated earlier, the CPS ASEC is the barometer for poverty in America, but the measure is not without its criticisms. There are several areas wherein the CPS ASEC falls short and may Channing 3 not capture the true nature of poverty. The most striking shortcoming of the measure is that it does not take account of taxes or noncash benefits aimed at improving the economic situation of the poor” (Short, 2013). This means that medical and housing assistance do not factor into the census’s poverty statistic and may over inflate the actual need that an individual experiences. For example, if a family’s threshold is compared against the actual cost of the residence, they can be found in poverty even though the difference between need and income is actually made up by a housing assistance supplement from the government. While this may be considered good news for poverty, the CPS ASEC can also deflate the poverty number. One way the census’s account of poverty can misrepresent the number is by omitting certain factors. By its own admission, the CPS ASEC does not adjust for the diversity of the country, saying, “The official poverty thresholds do not vary geographically” (United States Census Bureau, Definitions, 2013). Problematically, this tactic drastically discounts the proportion of a person’s income that may go to housing costs across the nation. The effect of this is that the ratio spent should be “the same whether a family lives in rural Mississippi or Manhattan” (Renwick, 2009, p. 2). Self-evidently, there are obvious differences in what it costs for a family of four to provide for housing across America. What this does is reduce the number of households who should be considered in poverty, but who are unrepresented by the census’s official statistic for their family size Recidivism The literature on recidivism is as equally varied as poverty’s. However, many articles are consistent in their findings of the state of recidivism and incarceration in general. Broadly Channing 4 defined, recidivism is the measurement of a released offender reoffending. A recidivism rate would be calculated by the reentrants in a population against the number of them who have reoffended. Traditionally, recidivism rates are calculated in three year intervals. Precedent for this calculation was set by a survey done in Oregon which “found that, overall, 63 percent of inmates were arrested for a felony or serious misdemeanor offense within three years of release from prison” (Petersilia, 1999, p. 512). A question remains as to what recidivism actually measures and if it approximates what its users want it to. At its heart, recidivism gives a number that approximates the people who have been found guilty of violating a law or condition of release (although it is possible to calculate recidivism for people that have simply been arrested but not gone to court) against those who have previously been found guilty of violating a law or condition of release. One author characterizes the ambiguous situation thusly, “Recidivism is currently an unsophisticated, dichotomous, terminal variable, incapable of measuring incremental progress toward postrelease success” (Gehring, 2000, pp. 511-512). As a reentrant, the individual is not given many resources upon which to build their successful life as a non-offender. What is termed gate money is the sum a reentrant is given to start the process of being self-sufficient. The “‘gate money’” provided by the state is rarely sufficient for more than a few days of subsistence” (Berk, Lenihan, & Rossi, 1980, p. 768). More disparagingly, the Economist reported that “in Illinois, released prisoners receive $50, a set of clothes, and a bus ticket” (as cited in Visher & Travis, 2003, p. 96). And Travis et al reports “one third of all state departments of corrections report that they do not provide any funds upon Channing 5 release” (as cited in Visher & Travis, 2003, p. 96). Without the requisite resources to generate a non-offending lifestyle, the odds are stacked against the reentrant. When the reentrant returns to the community, they are faced with a lack of opportunities. The most notable problem for reentrants is their unemployment. “The more urban the area, the higher the unemployment rate” (Harer, 1995, p. 99). “Unemployment is closely correlated with drug and alcohol abuse. Losing a job has similar effects” (Petersilia, 2000, p. 4). The employment opportunities are lower for a reentrant who returns to a more disadvantaged neighborhood compared to one that returns to a more better off area (Morenoff & Harding, 2011, p. 1). As reported above, for blacks, Hispanics, and whites, employment status was one of the leading causes of recidivism, yet the literature habitually express the difficulties reentrants face when attempting to find work. What this demonstrates is that the environment a reentrant returns to has a large impact on whether or not that reentrant recidivates. About their living situations, one study reported that “most offenders (60%) lived in poverty earning less than $15,000 annually, with only 9% having an annual income between $25,000 and $34,999 and 7% having an annual income of $35,000 or more” (Stalans, Yarnold, Seng, Olsen, & Repp, 2004, p. 259). Echoing the above, Travis et al says, “other research has shown that returning prisoners are increasingly concentrated in our nation's central cities and within them, in a relatively small number of neighborhoods that often are characterized by severe poverty, social disorganization, and high crime rates” (as cited in Visher & Travis, 2003, p. 102). Additionally, criminologists argue “that poverty represents a relative lack of opportunities for licit employment, making illicit opportunities for economic gain…an attractive alternative” (Harer, 1995, p. 108). From these Channing 6 articles, it is clear that the socioeconomic factors of a reentrants community are going to have a significant impact on their ability to not reoffend. As distinguished by the literature, opportunities for success in these neighborhoods are rare to begin with and so reentrants returning who are already disadvantaged cognitively and have a criminal background, are going to have a more arduous task of providing form themselves. On the link between poverty itself and recidivism, the literature bore mixed conclusions. By one account, “a higher percentage of persons below the poverty level…decreases the chances of a repeat offense. This could be possibly due to the lower propensity to commit property offenses among the probationers” (Sirakaya, 2006, p. 872).This article was the only one to make that conclusion about the relationship between poverty and recidivism. Most other pieces either determined that there was a slight effect, or that poverty was not a statistically significant indicator for recidivism rates. In so far as to recidivate, a person must commit a crime and “strong individual and aggregate level correlations between poverty and official measures of crime are perhaps among the most firmly established of social science empirical generalization” (Berk, Lenihan, & Rossi, 1980, p. 766). Another study found that amongst its data, “the strongest observed correlation between the independent variables is between risk and poverty” (Holtfreter, Resig, & Morash, 2004, p. 196). As so many studies are composed of different data sets, across different time periods, variance in their conclusion is to be expected. Yet, the nature of the conclusions is still similar in some fashions. Finally, one study attempted to bridge the connection between the poverty measure and recidivism rates. What was found was that “modest amounts of financial aid can reduce Channing 7 recidivism among ex-felons. Experimentally induced unemployment can increase recidivism” (Berk, Lenihan, & Rossi, 1980, p. 784). Future Research Future research should be done on the causal link between poverty and recidivism. As the government is already taking steps to reduce poverty, it would benefit policy-makers to know what external effects their policies have on the criminal justice system. This would certainly factor into their decisions and empower them to take future opportunities to directly diminish the rate of recidivism. So few studies have been done on this link because it is problematic to meaningfully identify what poverty is in the environment. To this end, what would be ideal for a study is one involving a control group of similar individuals. However, ethical considerations should be stated due to the harmful effects of poverty and recidivism both at the individual and at the community level. Intentionally subjecting a group of people to the conditions which could lead to them committing crimes is in itself a crime, legally and morally. III. Methods This paper’s research question was to determine the effect of poverty on recidivism rates. The hypothesis was that they are positively related, that an increase in the poverty rate will cause an increase in the recidivism rate. To explore this relationship, this paper selected the available data at the county level for Indiana. Indiana provides a good example to be studied because its poverty rate was about equal to the national average, lending it external validity, and Channing 8 more practically, the data for recidivism and poverty at the county level were available to be examined. Variables The following table expresses the methodological framework for the research question. Included is the independent variable, the rate of poverty in a given county, and the dependent variable, the adult recidivism rate for the same county. To test spuriousness, the table presents variables which may be acting upon, or obscuring, the relationship between the independent and dependent variable. A conditional variable is presented that could create a link between the independent and dependent variable. Crime as a need to survive is a conditional variable because it would be caused by increasing poverty rates, which could cause reentrants to supplement their income with crime, thus recidivating. Table 1 displays the variables used Table 1: Variables to Test Independent variable Dependent variable Possible spurious variables Conditional variable Poverty Rate Under 125% of Federal Poverty Level Recidivism Rate of adults 18 and over Unemployment Rate Rate of SNAP reception Crime as a need to survive Data Sources The data for the Poverty Rate and Unemployment Rate in Indiana counties were from the US Census’s Small Area Income and Poverty Estimates. They are “single-year ACS direct survey estimates from all counties and states regardless of population size” (United States Census Channing 9 Burea, American Community Survey, 2013).The American Community Survey itself is a yearly mandatory survey of about three million households in counties and metropolitan areas of a set metropolitan size (United States Census Bureau, Description of Income, 2013). In this experiment, the data were collected from the 2012 data set found at (United States Census Bureau, State and County, 2013). To augment the ACS information, data sets about SNAP recipients and Federal Poverty Level information were taken from an altered set from the New York Times. Food Research and Action Center applied, to the New York Times data, “county poverty data, using Census Bureau estimates of the number of people in each county with incomes under 125% of the Federal Poverty Level,” (Cooper 2010). This was done to account for changes in poverty levels based on geography, as opposed to applying a blanket level to all regions. While these data may present a more accurate picture of real poverty levels, it must be noted that the data were from 20062008 and “poverty has increased in most places” by the time the ACS was completed in 2012. Therefore, mapping the FRAC data over the ACS data may still be unreliable. The data for the Recidivism Rate in Indiana were taken from the 2012 recidivism report of 2009 releases from the Indiana Department of Corrections (Garner, 2012). Indiana measures recidivism as offender’s return to incarceration within three (3) years of their release date from a state correctional institution. For the returning offense, there categories are “a new conviction or a technical violation of post-incarceration supervision” (Garner, 2012, p. 2). Reentrants fall into four categories “Community Transition Program (CTP), Probation, Parole, or Discharged” (Garner, 2012, p. 2) in terms of the nature of their release from incarceration. Finally, if a Channing 10 reentrant reoffends multiple times within the same calendar year, only the earliest release dates impact the recidivism classification so that the data only reflect unique releases, not total releases (Garner, 2012, p. 2). Operationalization of measures The level of analysis for this study will be the county level. County averages are more reflective of population figures than the state level, and with ninety-two counties, there is a sufficient sample size to capture the diversity of counties and make the experiment generalizable to the entire population of counties. Therefore, there are ninety-two cases, corresponding with the reported total recidivism rate for the ninety-two counties, measuring the entire population. Using the SNAP and FPL measures does not give accounts of all counties as data were not available. These two data sets only measured 72 of the 92 counties in Indiana There are two things to say about the internal and external validity of the experiment. The concept of internal validity is, for the most part, not an issue, as this study examines all counties within the state of Indiana. On external validity, however, it should be noted that due to the possible limitations in meaningfully measuring the poverty rate, it is possible that Indiana may show that there is no correlation between poverty and recidivism rate, when in fact, a different sample group with a different effective poverty rate may actually express correlation. Channing 11 Research Design Because the data sets for this research are all interval level, the appropriate test to run in order to determine a relationship is a simple linear regression analysis. To test the goodness of fit between the two variables, the regression analysis reveals two statistics that express how well the independent variable explains variation in the dependent variable: the Adjusted R Squared and the Significance Coefficient of the alpha level (Meier et al, 2010, pgs. 345-346). The simple linear regression tests to run will use the Recidivism Rate (Recidivism) per county as a dependent variable. Tests will be run with County Poverty Rate (Poverty), County Unemployment Rate (Unemployment), Rate of County Population Under 125% of the Federal Poverty Level (FPL Rate), and Rate of County Population Receiving SNAP benefits (SNAP). There are many confounding variables not taken into account that could obscure any relationship or the lack thereof. As identified from the literature, these could be demographic data such as racial makeup of the reentrant population, and perhaps more importantly, the type and intensity of social services they receive upon reentry into society that would mitigate recidivating. IV. Results After running a correlation test (see Appendix: Table 3), each independent variable appears to correlate positively somewhat strongly with another independent variable. Running the pairs together in a regression test, while they may have high explanatory power for the dependent variable may lead to overlapping explanations, or multicolinearity. Channing 12 Therefore, when running the regressions, there will only be one independent variable run at a time. The independent variables hypothesized to have the most significant relationship with Recidivism did not. Below, Table 2 compares the relationship between the two variables yielded by the regression analysis. Table 2: Results of Linear Regression Independent Dependent Adjusted R Significance Equation Variable Variable Squared Score Poverty Recidivism .01 .165 =25.03+.364X Unemployment Recidivism .025 .069 =16.216+1.324X FPL Recidivism .022 .111 =24.311+.362X SNAP Recidivism .031 .072 =24.668+.530X These data show that while the four variables, as shown by the Adjusted R Squared, explain between 1% and 3% of variation within Recidivism, only two of them have a significant relationship with the dependent variable, Unemployment and SNAP. The Significance scores, when compared to an alpha of 10%, indicate that with 90% confidence, we can reject the null hypothesis that Unemployment and SNAP are not unassociated with Recidivism, respectively. By itself, the regression tests do not prove causation, but by using the equations build by the regression analyses, the relationship shows that an increase in Unemployment and Snap will result in an increase in Recidivism. Channing 13 V. Conclusion The regression analyses do not support the original hypothesis that Poverty as such would have the strongest relationship with Recidivism. This is not to say that adverse economic conditions don’t have a relationship with Recidivism, as evidenced by Unemployment and SNAP. Interestingly, a difference between these variables and the other two is that they may more likely indicate economic need, as shown by receiving benefits in the case of SNAP. One conclusion to draw from is that people who are in positions of economic need may be more likely, as supported by the literature. Also supported by one article in the literature review is that poverty may lead to decreases in recidivism as potential recidivists may be less likely to commit crimes against people living in poverty. Implications of economic need contributing to increases in recidivism are mixed. It would be fallacious to say that by increasing benefits, as SNAP shows, would increase recidivism. More evidence needs to be gathered on the economic conditions of those who recidivate and the motivations behind their crimes. Channing 14 Appendix Figure 1: Histogram of Recidivism Rate Channing 15 Table 3: Correlations between all Variables Percent of Poverty Rate Poverty Rate Pearson Correlation Recidivism Rate 1 Sig. (2-tailed) N Recidivism Rate Population under population on 125% Federal Rate SNAP Poverty Limit .146 .408** .546** .704** .165 .000 .000 .000 92 92 92 74 74 Pearson Correlation .146 1 .190 .210 .187 Sig. (2-tailed) .165 .069 .072 .111 92 74 74 1 .538** .200 .000 .087 N Unemployment Rate Percent of Unemployment Pearson Correlation Sig. (2-tailed) N Percent of population on Pearson Correlation SNAP Sig. (2-tailed) N Percent of Population under Pearson Correlation 125% Federal Poverty Limit Sig. (2-tailed) N **. Correlation is significant at the 0.01 level (2-tailed). 92 92 .408** .190 .000 .069 92 92 92 74 74 .546** .210 .538** 1 .545** .000 .072 .000 .000 74 74 74 74 74 .704** .187 .200 .545** 1 .000 .111 .087 .000 74 74 74 74 74 Channing 16 Poverty Rate & Recidivism Figure 2: Scatter Plot - Poverty Rate and Recidivism Rate Channing 17 Table 4: Linear Regression Analysis - Poverty Rate and Recidivism Rate Variables Entered/Removeda Variables Model Variables Entered 1 Poverty Rateb Removed Method . Enter a. Dependent Variable: Recidivism Rate b. All requested variables entered. Model Summary Std. Error of the Model 1 R R Square .146a Adjusted R Square .021 Estimate .010 9.56198 a. Predictors: (Constant), Poverty Rate Coefficientsa Standardized Unstandardized Coefficients Model 1 B (Constant) Poverty Rate a. Dependent Variable: Recidivism Rate Std. Error 25.030 3.859 .364 .260 Coefficients Beta t .146 Sig. 6.486 .000 1.399 .165 Channing 18 Unemployment & Recidivism Figure 3: Scatter Plot - Unemployment Rate and Recidivism Rate Channing 19 Table 5: Linear Regression Analysis - Unemployment Rate and Recidivism Rate Variables Entered/Removeda Variables Model Variables Entered 1 Unemployment Removed Method . Enter Rateb a. Dependent Variable: Recidivism Rate b. All requested variables entered. Model Summary Std. Error of the Model 1 R R Square .190a Adjusted R Square .036 Estimate .025 9.48883 a. Predictors: (Constant), Unemployment Rate Coefficientsa Standardized Unstandardized Coefficients Model 1 B (Constant) Unemployment Rate a. Dependent Variable: Recidivism Rate Std. Error 18.216 6.617 1.324 .720 Coefficients Beta t .190 Sig. 2.753 .007 1.838 .069 Channing 20 Percent of Population on SNAP & Recidivism Figure 4: Scatter Plot – Percent of Population on SNAP and Recidivism Rate Channing 21 Table 6: Linear Regression Analysis - Percent of Population on SNAP and Recidivism Rate Variables Entered/Removeda Variables Model Variables Entered 1 Percent of Removed Method population on . Enter SNAPb a. Dependent Variable: Recidivism Rate b. All requested variables entered. Model Summary Std. Error of the Model 1 R R Square .210a Adjusted R Square .044 Estimate .031 8.14394 a. Predictors: (Constant), Percent of population on SNAP Coefficientsa Standardized Unstandardized Coefficients Model 1 B (Constant) Percent of population on SNAP a. Dependent Variable: Recidivism Rate Std. Error 24.668 3.182 .530 .291 Coefficients Beta t .210 Sig. 7.753 .000 1.824 .072 Channing 22 Percent of Population under 125% of Federal Poverty Level & Recidivism Figure 5: Scatter Plot - Percent of Population under 125% of Federal Poverty Level and Recidivism Channing 23 Table 7: Linear Regression Analysis - Percent of Population under 125% of Federal Poverty Level and Recidivism Rate Variables Entered/Removeda Variables Model Variables Entered 1 Percent of Removed Method Population under . Enter 125% Federal Poverty Levelb a. Dependent Variable: Recidivism Rate b. All requested variables entered. Model Summary Std. Error of the Model R R Square .187a 1 Adjusted R Square .035 .022 Estimate 8.18303 a. Predictors: (Constant), Percent of Population under 125% Federal Poverty Level Coefficientsa Standardized Unstandardized Coefficients Model 1 B (Constant) Percent of Population under 125% Federal Poverty Level a. Dependent Variable: Recidivism Rate Std. Error 24.311 3.774 .362 .224 Coefficients Beta t .187 Sig. 6.442 .000 1.615 .111 Channing 24 References Berk, R. A., Lenihan, K. J., & Rossi, P. H. (1980, October). Crime and Poverty: Some Experimental Evidence From Ex-Offenders. American Sociological Review, 45(5), 766-786. Retrieved from http://www.jstor.org/ Fisher, G. M. (1992, May). The Development of the Orshansky Poverty Thresholds and Their Subsequent History as the Official U.S. Poverty Measure. Retrieved from United States Census Bureau: https://www.census.gov/ Cooper, R (2010, January). County-by-County Review of SNAP/Food Stamp Participation. Retrieved from Food Research and Action Center: http://frac.org/pdf/ny_times_snap_poverty_formatted.pdf Garner, A. (2012). Indiana Department of Corrections 2012 Recidivism Report (2009 Releases). Retrieved from State of Indiana: http://www.in.gov/ Gehring, T. (2000, June). Recidivism as a Measure of Correctional Education Program Success. Journal of Correctional Education, 51(2), 197-205. Retrieved from http://www.jstor.org/ Harer, M. D. (1995, September). Recidivism Among Federal Prisoners Released in 1987. Journal of Correctional Education, 98-128. Retrieved from http://www.jstor.org/ Institute for Research on Poverty. (2013). How is poverty measured in the United States? Retrieved from Institute for Research on Poverty: http://www.irp.wisc.edu/index.htm Meier, K., Brudney, J., & Bohte, J. (2010). Applied statistics for public and nonprofit administration (8th ed., pp. 345-346). Belmont, CA: Wadsworth Morenoff, J. D., & Harding, D. J. (2011, November). Final Technical Report: Neighborhoods, Recidivism, and Employment Among Returning Prisoners. Retrieved from National Criminal Justice Reference Service: https://www.ncjrs.gov/ Orrel, B. R., & Blackwell, A. G. (2008). The Poverty Forum: Community Factors Reducing Recidivism. Retrieved from The Poverty Forum: http://www.thepovertyforum.org/ Petersilia, J. (1999). Parole and Prisoner Reentry in the United States. Crime and Justice, 26, 479-529. Retrieved from http://www.jstor.org/ Petersilia, J. (2000, November). When Prisoners Return to the Community: Political, Economic, and Social Consequences. Sentencing and Corrections, 1-8. Retrieved from https://www.ncjrs.gov/ Renwick, T. (2009, August). Alternative Geographic Adjustments of U.S. Poverty Thresholds: Impact on State Poverty Rates. Retrieved from United States Census Bureau: http://www.census.gov/ Short, K. (2013, November). The Research Supplemental Poverty Measure: 2012. Retrieved from United States Census Bureau: http://www.census.gov/ Sirakaya, S. (2006, September). Recidivism and Social Interactions. Journal of the American Statistical Association, 101(475), 863-877. Retrieved from http://www.jstor.org/ Channing 25 Stalans, L. J., Yarnold, P. R., Seng, M., Olsen, D. E., & Repp, M. (2004, June). Identifying Three Types of Violent Offenders and Predicting Violent Recidivism while on. Law and Human Behavior, 253271. Retrieved from http://www.jstor.org/ United States Census Burea. (2013). American Community Survey. Retrieved from United States Census Bureau: http://www.census.gov/ United States Census Bureau. (2013). Current Population Survey (CPS) - Definitions. Retrieved from United States Census Bureau: http://www.census.gov/ United States Census Bureau. (2013). Description of Income and Poverty Data Sources. Retrieved from United States Census Bureau: http://www.census.gov/ United States Census Bureau. (2013). How the Census Bureau Measures Poverty. Retrieved from United States Census Bureau: http://www.census.gov/ United States Census Bureau, Small Area Income and Poverty Estimate. (2013). State and County Data File 2012. Retrieved from http://www.census.gov/ Visher, C. A., & Travis, J. (2003). Transitions from Prison to Community: Understanding Individual Pathways. {Annual Review of Sociology, 29, 89-113. Retrieved from http://www.jstor.org/