Research Methods in Criminal Justice Assignment 4

The bluntness of incarceration: crime and punishment in Tallahassee neighborhoods, 1995 to 2002 Geert Dhondt Published online: 29 March 2012 #Springer Science+Business Media B.V. 2012 Abstract The relationship between crime and incarceration is vexed. Some scholars suggest that incarceration deters crime. Others suggest that incarceration can increase crime in certain situations by undermining the social fabric and making the environ- ment more criminogenic. Most of the empirical work on the question is undertaken at an aggregate level (county, state, or national data). Yet, criminologists have long argued that the complex intertwining of crime and punishment is best understood at the neighborhood level, where the impacts of incarceration on social relationships are most closely felt. This paper examines the question using a panel of neighborhoods in Tallahassee, Florida for the period 1995 to 2002. I find evidence to support the contention that the high levels of prison admissions and prison cycling (admissions plus releases) is associated with increasingcrime rates in disadvantaged neighbor- hoods. This effect is not found in other neighborhoods. Looking more closely at the issues of race and class, I find that while marginalized neighborhoods experience slightly higher crime rates, they are faced with much higher incarceration rates. In Black neighborhoods in particular, prison admissions are an order of magnitude higher in comparison with non-Black neighborhoods even though underlying crime rates are not very different.

By employing incarceration —the bluntest of instruments —as the primary re- sponse to social disorder, policy makers have significantly missed the mark.

The very laws intended to punish selfish behavior and to further common social interests have, in practice, strained and eroded the personal relationships vital to family and community life. Crime cannot go unpunished. But by draining the Crime Law Soc Change (2012) 57:521 –538 DOI 10.1007/s10611-012-9376-z This paper would not have been possible without the assistance, feedback and support of Arjun Jayadev, Kade Finnoff, Michael Ash, David Kotz, John Bracey, Michael Carr, Mathieu Dufour, Jay Hamilton, Todd Clear, Natasha Frost, and the two anonymous reviewers. This paper is much stronger because of their contributions; its faults remain my own.

G. Dhondt ( *) Department of Economics, John Jay College of Criminal Justice, The City University of New York, New York, NY, USA e-mail: [email protected] resources of families, by frustrating the norm of reciprocity that inheres in family life, and by stigmatizing poor and minority families, our current regime of criminal sanctions has created a set of second-order problems that furthers social detachment.—Donald Braman, 2004 1 Criminal justice policy has used incarceration as a primary tool to combat crime since 1980. This has led to a seven-fold increase in incarceration during this time period, and at the current juncture more than 1 in every 100 Americans are in the prison system. In the 1990s, however, crime rates began to decline dramatically. For supporters of the “lock ’em up and throw away the key ”approach, this fact has been welcomed as vindication that more punitive incarceration has worked. Several researchers have provided evidence for the benefits of incarceration in reducing crime [ 1 –6], while other scholars have argued that other factors —an improving economy, the decrease in the price of drugs, the ending of the crack-cocaine epidemic, the legalization of abortion, changing demographics and the measurement and reporting of crime data, among others —were equally or more relevant [ 7–9].

While the economics literature has for the most part supported the idea that incarceration reduces crime, researchers in the sociology and anthropology of crime [ 10 –17 ] have challenged this notion. Through a series of case studies, theoretical and empirical papers, they argue that incarceration may fail to reduce crime rates in socially dysfunctional, highly fractured neighborhoods. Their argument is straight- forward: Incarceration works to destroy the social fabric of already disadvantaged populations and limit the ability of informal means of social control to prevent crime.

As a result, after a certain point, incarceration is powerless to have an impact on crime prevention, so punitive incarceration becomes self-defeating and also leads to several socially damaging side effects. Given that informal modes of social control are only effective on a relatively small scale, such an approach suggests that econometric analyses of the relationship between incarceration and crime at aggregate levels (city, region, state or country) would gloss over this set of effects. Most importantly, aggregate empirical studies miss critical social processes that occur at the level of neighborhoods which might in fact be more relevant as a guide to social policies. This paper assesses these competing claims. I use a unique panel dataset which records crime and incarceration at the most disaggregated level yet— that of neigh- borhoods —in a city that experienced remarkable decreases in crime rates in the 1990s: Tallahassee, Florida. The data was collected by Clear et al. [ 13] and has not been used previously. 2It is the first panel data analysis of crime and incarceration at the neighborhood level of which I am aware. 3I find that although crime fell in Tallahassee between 1995 and 2002 while incarceration rates rose, there is no evidence at the level of the neighborhood to indicate that increased incarceration reduced crime. Instead, I find relatively strong evidence to support the hypothesis that increased incarceration in 1 year is associated with increased crime the following 1Anthropologist Donald Braman, 2004, page 221, based on studies of families of the incarcerated in Washington, D.C. Quoted in Clear 2007, 148.

2I would like to thank Todd Clear for generously sharing this data. This study is based and inspired by Clear et al. [ 13] and uses their construction of neighborhoods, but while their study uses prison data in 1996 and 1997, this current study is a panel-data analysis between 1995 to 2002.

3For this analysis, I used STATA/SE 10 and STATA/SE 12.

522 G. Dhondt year. Furthermore, I find little evidence that incarceration may be associated with decreased crime in other neighborhoods. In examining the correlates of incarceration in this paper, I find evidence that incarceration has as much or more to do with the social and demographic characteristics as with observed crime in particular neighborhoods. The outline of this paper is as follows. In Section “The relationship between crime and incarceration ”, I discuss the literature on the relationship between incarceration and crime, showing in particular that there is a considerable theoretical and empirical ambiguity, more than one might expect, in terms of direction of effect. Several approaches suggest and provide evidence that incarceration decreases crime, while other approaches suggest that incarceration can increase crime in certain social contexts. I put particular emphasis on the importance of the level of aggregation in the analysis. In Section “Data and methodology ”, I outline the data and methodology used in this paper, and in Section “Results ”, I present the results. Section “Conclusions and further research ”offers a brief conclusion and directions for further research.

The relationship between crime and incarceration Incarcerating offenders immediately and effectively reduces crime in the future by removing ‘bad apples’ from society and by deterring other would-be criminals through the demonstration effect that this entails. Such a ‘common sense ’perspective began governing penal policy in the 1970s, and since then the United States has increased the size of its prison population sevenfold [ 15,18 –22 ]. However, the rate of expansion of the penal system has not been accompanied by an equivalent decrease in crime. This and other facts have caused some scholars to argue that the popular view of the prison is simplistic because it fails to account for the unintended consequences of imprisonment [ 23]. This “critical ”perspective argues that unforeseen effects in terms of social destabilization are subtle and modest but can have major accumulated effects that undermine the efficacy of incarceration to reduce crime. The critical perspective suggests that the debate about incarceration policy has been wrongly dominated by an atomistic view of criminal behavior —a view of the world in which individuals who engage in crime are influenced by simple cost-benefit analyses independent of the contexts in which they live. In the “mainstream ”view, the threat of incarceration then becomes a significant cost to consider and prevents rational but criminally minded individuals from carrying out their illicit desires.

Decisions to engage in crime are seen as products of the likelihood and degree of punishment if caught, and little else. The critical perspective provides an alternative view of criminal behavior. These accounts are typically less parsimonious and attempt to paint a more holistic perception of the potential offender —as a person who lives in particular areas with more or less criminogenic potential, interacts with fellow citizens, and responds to various life circumstances with choices based on a grounded understanding of the consequences of those choices. Given this perspec- tive, one must question whether increasing incarceration is always likely to reduce crime, since the criminogenic environment is a key element. It is useful at the outset to briefly examine and classify the various sub-strands of theoretical accounts used to support the mainstream perspective and the critical Bluntness of incarceration: crime and punishment in Tallahassee 523 perspective, respectively. There are broadly four reasons that might lead a researcher to believe that incarceration reduces crime: retribution, rehabilitation, deterrence and incapacitation. Retribution suggest that the‘moral ’punishment involved in incarcer- ation decreases crime over the medium term through the shock of the loss of liberty experienced by the criminal. This shock keeps the convict from committing a crime again upon release. Retribution also establishes a moral order of what is right and what is wrong [ 22,24 ] that becomes prevalent in social mores. As a variant on this theme, some scholars suggest that the process of incarceration is primarily a way to promote rehabilitation toward more pro-social behavior. Thus theorists suggest that incarceration is used to ‘correct ’an offender ’s behavior [ 18,19 ,25 ] so that a new and more well-behaved citizen is remolded while behind bars. A third reason for incar- ceration is that it provides a demonstration effect, in that a potential offender calculating the expected utility of crime is faced with a potent and observable disutility from undertaking the action —in other words, a deterrence. Clearly, the actual workings of the criminal justice system contribute to the perceived threat of incarceration, and three factors —certainty, severity and swiftness of punishment — come into play [ 7, 26 ]. The more certain, severe and swift a punishment is, the more effectively it will deter potential offenders. Finally, incarceration is said to reduce crime through the process of incapacitation. In such an account, incarceration works to reduce crime by locking up criminally minded predators since such offenders are likely to commit crimes if freed [ 2, 27 ]; locking up the few but potent ‘super predators’ or criminal elements will therefore have large impacts on crime control by inoculating society from these people.

Throughout most of the 20th century, conversation about crime policy was dom- inated by the idea that individual offenders require reform or rehabilitation [ 18,19 ].

The belief now in vogue is that they require incapacitation. Of course, these approaches differ in several important respects, but they share a common analytic foundation: crime and its control are best understood in regard to the thoughts and emotions of specific individuals who commit crimes or want to commit crimes, with offenders seen as individual actors who behave largely in isolation of their environ- ments. The tendency to view crime as a phenomenon defined by wayward individuals and their desires is not only ingrained in penology but also reinforced by popular media. The critical perspective suggests that these accounts are seriously mistaken in that they focus on the individual as the locus of crime and almost definitionally ignore the social causes of crime [ 8]. As a result, the critical perspective posits, theorists have tended to overlook the fact that incarceration is not a self-contained process; locking a person up cause many individual and social effects that are not typically witnessed or examined. These processes, some generated by incarceration itself, can serve to undermine the efficacy of incarceration in reducing crime [ 10,24 ]. These counter- tendencies can be summarized in three broad categories. First, even without looking at the social effects of incarceration, the mainstream perspective ignores other individual level effects. These include scarring, school of crime and the relocation of crimes within the prison. Scarring refers to the idea that when offenders are released from prison they are scarred by their experience inside (lack of health care, sexual abuse, institutionalization, and exposure to a hyper-violent environment) and discriminated agai nst upon their release, their punishment 524 G. Dhondt seemingly never-ending as it continues with employment discrimination, housing discrimination, denial of voting rights, and denial of student loans [23]. This, in turn, does the opposite of rehabilitation and instead creates a more anti-social individual.

Another set of reasons why incarceration can actually serve to increase criminal behavior in the individual is the idea that prison is a ‘school of crime ’in which criminals first learn and then improve their skills at criminal behavior and create connections with other criminals. This account implies that incarceration removes prisoners from social networks connected with employment and instead connects them to social networks associated with criminal activity. Finally, some scholars have argued that incarceration does not necessarily reduce crime but merely ‘relocates ’it behind bars. What is considered crime, such as rape and gang activity, is much more prevalent inside a prison than outside [ 10,21 ]. A second set of considerations arises from the recognition that incarceration is not a self-contained process and that the removal of the so-called ‘bad apple ’has larger social effects on families, neighbor- hoods and extended networks. Formal social control (police, prison, employment, school, welfare, and so on) is thought to only partly contribute to public safety. An equally important contribution is informal social control, such as the social relation- ships and collective efficacy among neighbors. In theory, formal social control could supplement informal social control, but in a series of articles, Todd Clear and co- authors have suggested why this might not be the case [ 10,11 ,13 ,16 ,17 ]. One central theme in this set of articles is that prison is a blunt form of control which creates an unintended impact of increasing crime through various mechanisms. One of such mechanisms, coercive mobility, argues that increasing formal social control weakens the informal social control which is often more important in preventing crime than formal social control. Social relationships, ties and interaction among neighbors, together with the willingness to intervene on behalf of their neighbors is called collective efficacy [ 28 ]. Clear et al. [ 13] argue that high rates of residential mobility contribute to higher crime rates. Residential mobility creates neighbors who are isolated from each other with a low degree of integration, and this reduces collective efficacy and neighbor- hood stability. 4Incarceration can be theorized as a form of involuntary or coercive mobility. High rates of incarceration undermine the collective efficacy and social stability of neighborhoods, which could lead to a situation in which formal social control undermines informal social control and thus in which incarceration could lead to higher crime rates in marginalized and vulnerable neighborhoods. Riley ([29 ], at page 4) summarizes the thesis concisely:

The coercive mobility hypothesis suggests that increased rates of incarceration may weaken the families and communities that offenders leave behind and actually reduce effective social control efforts and increase crime in neighbor- hoods characterized by high rates of incarceration. In effect, high rates of incarceration decrease residential stability as prisoners and family members are forced to relocate. Family members of prisoners often relocate to be nearer to an institutionalized loved one or in response to economic and child care 4Researchers have also pointed to other related processes that exacerbate coercive mobility including the exacerbation of inequality and the crowding-out of preventative funding.

Bluntness of incarceration: crime and punishment in Tallahassee 525 contingencies associated with the removal of an incarcerated family member.

Neighbors may also move to escape what is perceived as a dangerous environ- ment. Those who remain in communities experiencing high rates of coercive mobility are left to cope without the assistance of those who have been relocated and in a context that features disrupted social networks and rising levels of alienation and distrust.

Finally, the tendency of incarceration to decrease crime assumes that the individual offender has incentives but ignores that individuals in institutions which search, process, and lock up criminals also have incentives. If an institution such as the criminal justice system is created to solve a social problem and the institution is successful, one might expect the disappearance of this particular institution. However, this is rarely the case as those whose welfare is dependent on maintaining the system continue to advocate for it. Elected officials - such as prosecutors, judges, politicians, and sheriffs - campaign on tough-on-crime platforms. Police and their unions struggle for more political power and strive for more police. Prison guards and their unions struggle for more political power and strive for more prison guards. Given the fact that there are well identified mechanisms that suggest reasons why incarceration can decrease crime (the mainstream perspective) as well as increase crime (the critical perspective), empirical examinations become all the more impor- tant for testing out these alternative hypotheses and providing some guidance for public policy. There have been empirical examinations which both support the mainstream and critical perspective. Devine et al. [ 30] and Marvell and Moody [ 31 ] find strong support of the mainstream perspective using a national time series empirical study. Levitt [ 2], Spelman [ 3], and Spelman [ 4] find strong support for the mainstream perspective at the state and county level while also taking into account of simultaneity. Liedka et al. [ 5] find some empirical support of critical perspective using state level data by comparing states with “low ”and “high ”levels of incarcer- ation. They find that incarceration decreases crime but less so as the scale of incarceration increases. Kovandzic et al. [ 32], Kovandzic and Vieraitis [ 33], and Kovandzic et al. [ 34] provide evidence in support of the critical perspective at the city and county level. Clear et al. [ 13] directly test the coercive mobility at the neighborhood level and also find some limited support for it. Dhondt [ 35] uses a state level panel analysis with marijuana and cocaine mandatory minimums as instrumental variables and finds support for the critical perspective. How do we make some sense of the wide variety of estimates of the relationship between crime and incarceration? The largest negative (i.e. mainstream perspective) estimates of incarceration on crime typically come from national time-series analyses [ 30 ,31 ]. County level estimates are on average lower than state level estimates, which in turn are lower than national level estimates [ 2, 3, 5, 36 ]. While a strong negative relationship is present at the national level, this relationship mostly disap- pears at the county level [ 32,33 ].

The issue of aggregation is a particularly vexed issue in the empirical literature: the problem of aggregation in crime and punishment statistics and the appropriate level at which to undertake the analysis. If one takes an atomistic approach where the only impacts of incarceration on crime are felt by and through the incarcerated individual, then the level of aggregation does not matter. This is implicitly the position taken by 526 G. Dhondt most scholars who have estimated this incarceration-crime relationship at the national or state level [2, 36 ]. These scholars geared their level of aggregation to the level at which legal frameworks are decided, thus implicitly assuming a constant implemen- tation and an absence of effects at the level of the community. Many of the counter-tendencies that were examined earlier (i.e. factors which lead incarceration effects to have a positive impact on crime), however, have the most salience at the neighborhood level. For example, the institutions that search, process, and lock up criminals operate at the local level and differ across neighborhoods.

These institutions will thus mediate the legal framework differently depending on their particular location. Similarly, incarceration impacts directly the social fabric of the communities where the people who are locked away come from. Greater levels of prison cycling (what Clear et al. term ‘coercive mobility ’–admissions and releases from prisons) will destabilize the informal organization of the neighborhood which is an effective form of maintaining public safety. Problems of poverty and inequality can be exacerbated by higher levels of incarceration, notably via the dampening of the employment prospects of the people incarcerated. The imagination of the people living in communities with high rates of incarceration is also affected by this state of affairs. If many people go in and out of prison all the time, the experience of incarceration may become ‘normalized ’[10 ]. This may have contradictory effects on the deterrence factor in specific neighborhoods. On the one hand, higher rates of incarceration make the threat more salient and real. On the other hand, if the experience is normalized, the fear of incarceration may be dampened. Consequently, the deterrence factor may not play out in the same way in different neighborhoods, depending on which tendencies dominate. The combined weight of these arguments suggests that a disaggregated approach is best in examining the impact of incarcer- ation on crime. Similarly, while most incarceration policy is set at the federal or state level, incarceration impacts certain neighborhoods highly disproportionately. The Justice Mapping Center has created maps of various cities which show the spatial concen- tration of incarceration. For example, in New York City, the historical Black neigh- borhoods of Harlem, Bedford-Stuyvesant, Brownsville and their surrounding areas supply over 50 % of all prisoners while only housing 17 % of New York ’s male population aged 16 to 59. 5 Clear [ 16] argues that there exist four loci of concentration: socio-economic status, gender/age, race/ethnicity, and place. Socio-economic status concentration means that prison is reserved for the poor. The gender/age loci mean that prison is reserved for young men. 6The race/ethnicity loci mean that prison is reserved for Black, Latino and Native Americans. By aggregating to the level of county, state or nation, it is possible that such effects could be glossed over or missed. Conducting a statistical study at a higher level of aggregation (national, state, or county) makes it difficult to precisely examine the impacts occurring at the 5See Justice Mapping Center, http://www.justicemapping.org/6“Those policy choices [of increasing incarceration] have had distinct implications for the way prison populations have come to reflect a concentrated experience among certain subgroups of the U.S. population —in particular, young black men from impoverished places. This concentra tion is, in some ways, the most salient characteristic of incarceration policy in the United States, since the socia l consequences of incarceration are dominantly felt among those people and in those places of concentration ”([16 ], 49).

Bluntness of incarceration: crime and punishment in Tallahassee 527 community level and the channels through which a lot of the tendencies and counter- tendencies operate. As Spelman [4] says, “the heterogeneity of most local areas in income and social class, race, household structure, and age, suggests that variations within states [and counties] may be considerably greater than variation among them. ” Aggregating will thus tend to reduce the range of variation of many of the variables of interest. Despite these objections to aggregation, there may be problems using highly disaggregated data as well. Specifically, it is possible that crime ‘spills over ’from some identified neighborhoods to others. In such a scenario, focusing on the rela- tionship between crime and incarceration in a particular neighborhood mismeasures the actual effect. This noted, most work by sociologists and criminologists has long come to the conclusion that crime is predominantly a local phenomenon and that spillover effects are negligible [ 37–43 ].

Section “Results ”of this paper empirically examines the relationship between crime and incarceration at the neighborhood level and tests directly for the coercive mobility claim that in neighborhoods with high incarceration there is positive effect on future crime rates. Clear et al. [ 13] is the only previous study of this relationship at the neighborhood level, and it examines 80 Tallahassee neighborhoods in 1996 and 1997. Clear et al. find that releases from prison in 1996 has a positive relationship with crime rates in 1997 in all neighborhoods. They also find that prison admissions in 1996 have a positive relationship with crime in high incarceration neighborhoods but not in others. The Clear et al. study of the impact of incarceration on crime rates and crime rates on incarceration is limited in that it only examines the effect of incarceration of 1 year on crime the next year and does not illuminate long term patterns. One might expect patterns of social disorganization to accumulate over much longer time horizons; hence, a longer term approach is worthwhile. Accordingly, this paper uses the same neighborhoods but employs a longer time horizon in examining the relationship between crime and incarceration between 1995 and 2002. I find strikingly more compelling evidence in favor of the coercive mobility thesis. First, I show a very clear non-linear pattern where incarceration has a different impact on different neighborhoods depending on their level of incarcera- tion. Second, while Clear et al. [ 13] find only weak mixed support for the coercive mobility theory because of the lack of data points, I find much stronger and robust evidence. By examining differences in neighborhoods in different ways, I am able to show that the dynamics of incarceration work very differently in high incarceration neighborhoods and marginalized communities versus other neighborhoods. I also find strong support of differential incarceration where correlates of marginalization, especially the percentage of Blacks in the neighborhood, has a strong correlation with incarceration over and above what may be merited by increased crime rates in such neighborhoods. The next section provides greater detail on data collection and methodology.

Data and methodology At the outset, it is useful to briefly comment on another particularly vexed issue in the empirical literature: the problem of simultaneity bias. Incarceration levels are likely 528 G. Dhondt influenced by crime levels, and thus probably not exogenous. This gives rise to a system of simultaneous equations, where it is not even clear that both of the relation- ships go in the same direction. From the viewpoint of deterrence, crime should increase incarceration,ceteris paribus, but incarceration in turn should decrease crime. Estimating an equation linking the two variables without taking this into account can lead to a biased estimate. I adopt a standard procedure to deal with this by using the lag of incarceration or prison cycling as the dependent variable. Equation 1below provides the baseline specification of the empirical model. I use a fixed-effects model to regress crime per capita reported in a neighborhood and the variable of interest on the percentage of people incarcerated in a neighborhood, controlling for several correlates. Formally, the basic specification used to estimate the effect of incarceration on crime is:

Y it¼b 0þb 1XitþbZ itþ a iþg tþ" it ð1 Þ where the subscript icorresponds to neighborhoods, and tindexes years. Yis the dependent variable of interest, crime per capita, X is the regressor of interest (prison admissions or coercive mobility) and Zis a vector of controls including demographic and social characteristics. Year and neighborhood fixed effects are included in all models.

While most studies use random effects, this is not appropriate in these types of empirical studies because there are large stable differences in crime, admissions, and releases across neighborhoods, counties, or states. We also do not have a random sample of states, counties, or this in this particular study of neighborhoods. Because of these two facts - both that we do not have a random sample and that there are large differences across neighborhoods which are stable through time - fixed effect is the appropriate analysis to use. The data are organized in a panel format for a seven-year period between 1995 and 2002. 1997 was dropped from the panel since it was missing crime data. Data were collected from different sources: The neighborhoods were created by Clear et al. [ 13], prison admissions and prison releases were provided by the Florida Department of Corrections for the years 1995 to 2002, the Tallahassee Police Department provided the crimes known to police, and all other data comes from the 1990 and 2000 censuses. 7 These neighborhoods were mapped in three steps. 8In the first step, completed in 1997, a survey was conducted of all local neighborhood associations registered with the city of Tallahassee Neighborhood Services. All registered associations were asked to identify the boundaries of their association. Responses were mapped and coded and were compared with the boundaries determined by the Tallahassee Neighborhood Services. The second step involved a case-by-case review of geographic features in problematic areas. The third step was completed with the assistance of the Tallahassee 7The Tallahassee dataset which included total crime, prison admissions and releases by neighborhood was constructed by Kristin Scully and emailed to me by Natasha Frost. Todd Clear explained how the neighborhoods were constructed and gave me information on how the neighborhoods were matched to the 1990 census track and block groups.

8For a more detailed description of the construction of the Tallahassee neighborhoods, see Clear et al. [ 13].

Bluntness of incarceration: crime and punishment in Tallahassee 529 Police Department and compared neighborhood boundaries to established police crime-reporting areas and 1990 U.S. census tract and block groups. This resulted in the identification of 80 different neighborhoods in the city of Tallahassee with populations in those neighborhoods varying between 249 and 4,538 [ 13].

The 2000 U.S. census had some changes in the tract and block groups of Tallahassee. The 2000 census tract and block groups map were used with the 1990 census tract and block groups map to identify these changes and match the 2000 census reporting to the 80 Tallahassee neighborhoods.

Variables Crimes known to police, 1995 to 2002 The Tallahassee Police Department provided crime statistics by geographic location that were based on the Tallahassee Police Department ’s reporting areas. All offenses reported within the city limits were mapped by neighborhood. The offenses include homicide, sexual battery, other sex offenses, strong-arm robbery, armed robbery, commercial burglary, residential bur- glary, auto burglary, auto theft, aggravated battery with firearm only, aggravated assault with firearm only, loitering and prowling, and suspicious incident.

Prison admissions and prison releases The Florida Department of Corrections pro- vided data files for all offenders admitted to serve prison sentences who listed Leon County as their place of residence for each year between 1995 and 2002. These data files contained addresses which were mapped and matched with the different Talla- hassee neighborhoods. The Florida Department of Corrections also provided data files for all offenders ’Inmate Release Plans who were being released back into Leon County for each year between 1995 and 2002. The addresses were mapped and matched with the 80 Tallahassee neighborhoods. While cleaning up the data, I dropped two neighborhoods because they had missing prison admissions, leaving 78 neighborhoods. Demographic data comes from 1990 and 2000 U.S. census reports and the Summary File 3A was downloaded from the Inter-university Consortium for Political Table 1 Summary statistics Variable Observations Mean Std. Deviation Min Max Crime 560 0.04156 0.03531 0 0.3567 Admissions 546 0.002020 0.003088 0 0.02093 Black 560 0.3267 0.2919 0 1.000 Latino 560 0.03401 0.02864 0 0.1559 Employed 560 0.5219 0.11522 0.03910 0.7475 Poverty 560 0.2109 0.1577 0 0.7325 Single Mother 560 0.03311 0.02900 0 0.1643 No High School 560 0.06146 0.05594 0 0.3379 Male Youth 560 0.1494 0.1074 0.01795 0.5139 All variables are per capita 530 G. Dhondt and Social Research (ICPSR). Demographic variables included in the analysis were linearly interpolated values from the census of overall population, Black population, Latino population, employed population, median household income, population with no high school diploma, population with a bachelors degree, male youth (<18) population, population of single mothers, and population of residents living below the poverty level. These were all converted into per capita numbers (Table1).

Table 2 Crime, incarceration and prison cycling VARIABLES (1) (2) (3) (4) (5) Prison Admissions (%) 0.204 (0.555) Lagged Prison Admissions (%) 1.382** 0.943 (0.582) (1.238) Lagged Prison Admissions (%) squared 32.90 (81.93) Lagged Recycle Admissions (%) 0.644* 0.521 (0.382) (0.802) Lagged Recycle (%) squared 45.57* (27.60) Black (%) 0.00298 0.0145 0.0151 0.0125 0.0178 (0.0268) (0.0302) (0.0303) (0.0304) (0.0305) Latino (%) 0.166 0.168 0.166 0.170 0.160 (0.109) (0.124) (0.124) (0.125) (0.124) Employed(%) 0.0532 0.0347 0.0357 0.0406 0.0500 (0.0388) (0.0442) (0.0443) (0.0445) (0.0447) Low Income (%) 0.0438* 0.0203 0.0209 0.0216 0.0199 (0.0249) (0.0282) (0.0283) (0.0284) (0.0283) Single Mother (%) 0.449*** 0.297 0.307 0.293 0.321* (0.164) (0.187) (0.189) (0.188) (0.188) No High School (%) 0.544*** 0.336*** 0.332*** 0.344*** 0.344*** (0.0789) (0.0908) (0.0913) (0.0911) (0.0909) Male Youth (%) 0.254*** 0.171*** 0.172*** 0.179*** 0.184*** (0.0517) (0.0591) (0.0592) (0.0596) (0.0596) Constant 0.0605** 0.0602* 0.0606* 0.0504 0.0507 (0.0274) (0.0309) (0.0309) (0.0316) (0.0315) Observations 546 390 390 390 390 R-squared 0.793 0.818 0.818 0.817 0.818 Standard errors in parentheses *** p< 0.01, ** p< 0.05, * p< 0.1 All equations include year and neighborhood fixed effects. Columns (1) regresses contemporaneous crime and prison admissions for all neighborhoods. Column (2) regresses crime and lagged prison admissions for all neighborhoods. Columns (3) adds a quadratic term to lagged for all neighborhoods. Columns (4) and (5) repeat the exercise of columns (2) and (3) using prison cycling rather than admissions as the independent variable Bluntness of incarceration: crime and punishment in Tallahassee 531 Table2shows the summary statistics of the data set. Crime varies widely, ranging from zero reported crimes to 35 crimes per 100. Similarly, admissions vary as well, from a low of zero admissions in the neighborhood in that year to over 2 per 100.

Results A cursory examination of Tallahassee in the period of the mid-1990s to 2002 might provide a bright picture of criminal justice policy in support of the mainstream perspective. Much like the rest of the U.S. and Canada, the crime rate dropped drastically. Figure 1illustrates how the crime rate in Tallahassee decreased between 1995 and 2002. This was accompanied with an increase in incarceration rates in the city. Figure 2shows that the incarceration rate in Tallahassee increased between 1995 and 2002. At first glance, it would seem that incarceration was working during this period to reduce crime systematically. However, for the reasons alluded to above, there may be neighborhood effects that are not seen when one looks at aggregate trends. One can see this in Table 3. The first column of Table 3shows the results of running equation (1) to examine the contem- poraneous relationship between crime and prison admissions for all neighborhoods in the city of Tallahassee. The coefficient of prison admissions is positive but it is not statistically significant and not negative as one might expect from the aggregate trends shown in Fig. 1. Column (2) shows the result of regressing crime on incarcer- ation in the previous year. First, and somewhat surprisingly, there is a significant positive coefficient: a 1 percentage point increase in prison admissions per capita in the past year is associated with an increase in per capita crime of 1.4 percentage points in the current period. Column (3) tests for non-linearities in the relationship by introducing a squared term for the lagged admissions per capita. The coefficients remain positive, but not statistically significant.

As the discussion in Section I suggests, while admissions or incarceration is an important variable, one may be equally interested in the relationship between prison .035 .04 .045 .05 .055 (mean) crimeperc 1994 1996 1998 2000 2002 year Fig. 1 Crime (per capita) 532 G. Dhondt cycling and crime according to the coercive mobility thesis. Figure3provides some basic evidence for the contention that admissions in 1 year is strongly positive correlated ( r0 0.65) with releases in the previous year, suggesting some level of cycling among populations. Accordingly, columns (4) and (5) in Table 3repeat the exercise for columns (2) and (3), except that I replace the admissions per capita with the prison cycling population per capita (i.e. admissions plus releases per capita). While I find a positive correlation in column (4), column (5) provides the most intriguing result. Specifically, there is some suggestive evidence of a non-linear relationship, with higher incarcer- ation having a crime-reducing effect at low levels of prison cycling but a crime- increasing effect at higher levels. This is important evidence in favor of the coercive mobility thesis. It is also interesting to look at the impact of other correlates. As might be expected, in all specifications many of the correlates have a significant and positive relationship with crime rates in accordance with our priors. These include the percentage of households in the neighborhood which are headed by females with children, the percentage of residents who are lacking a high school diploma, the percentage of residents who are both young and male, the percentage of residents who live in poverty, and the percentage of residents who are Black. Table 4is a more direct test of the coercive mobility thesis. To review, such an argument suggests that high mobility in marginalized neighborhoods has a destabilizing effect and decreases the collective neighborhood efficacy of maintaining social order through informal control. In order to operationalize the difference between marginalized and other neighborhoods, I stratify the sample in three different ways. First, I define a marginalized neighborhood as one that is in the top quintile of average incarceration over the period 1995 to 2002. 9Columns (1) and (2) in Table 4examine the relation- ship between crime rates and incarceration rates in low incarceration and high incarceration neighborhoods respectively, using this cut-off. There is a positive but insignificant correlation between crime rates in 1 year and incarceration rates in the previous year; a 1 percentage point increase in incarceration rates is associated with a 9I drop two neighborhoods that do not have information on incarceration rates in some years. .0018 .002 .0022 .0024 .0026 (mean) admperc 1994 1996 1998 2000 2002 year Fig. 2 Prison admissions (percentage) Bluntness of incarceration: crime and punishment in Tallahassee 533 0.7 percentage point increase in crime rates in the following year. However, there is a strong and statistically significant correlation between crime in 1 year and lagged incarceration in the previous year in high incarceration neighborhoods: a 1 percentage point increase in incarceration in 1 year is associated with a 2.0 percentage point increase in crime rates in the following year. This is strong evidence in favor of the coercive mobility thesis. Columns (3) and (4) and columns (5) and (6) repeat this exercise using the percentage of the population that is Black and the percentage of single mothers in the population. Once again, the effects are dramatically different between the two sub- samples. In the more marginalized communities, incarceration in 1 year leads to a rise in crime in the next, in direct contradiction to the thesis put forth by the mainstream perspective. In non-marginalized communities, there is no strong statistical evidence in either direction. In high Black population neighborhoods, a 1 percentage point increase in incarceration in 1 year is associated with a 2.1 percentage point increase in crime in Table 3Regression results, crime rates and prison admissions VARIABLES (1) (2)(3) (4)(5) (6) Low incarceration neighborhood High incarceration neighborhoodLow black population neighborhood High black population neighborhoodLow single mother neighborhood High single mother neighborhood :lagged Admissions Per Capita 0.738 2.080** 0.244 2.140** 1.142 1.918* (1.035) (0.840) (0.919) (0.904) (0.711) (1.018) Black (%) 0.00312 0.139 0.0164 0.241** 0.00597 0.0377 (0.0284) (0.196) (0.0335) (0.0960) (0.0339) (0.0724) Latino (%) 0.0185 1.984*** 0.0346 1.497*** 0.287** 0.133 (0.115) (0.473) (0.117) (0.542) (0.137) (0.321) Employed(%) 0.0250 0.257** 0.0130 0.0200 0.0956* 0.0875 (0.0448) (0.124) (0.0479) (0.127) (0.0512) (0.103) Low Income (%) 0.0104 0.0437 0.0121 0.0455 0.00570 0.109 (0.0267) (0.179) (0.0273) (0.141) (0.0325) (0.0771) Single Mother (%) 0.347* 0.925 0.402** 0.639 0.648** 0.685* (0.180) (0.832) (0.187) (0.777) (0.261) (0.388) No High School (%) 0.535*** 0.326 0.552*** 0.167 0.626*** 0.106 (0.0991) (0.399) (0.107) (0.353) (0.125) (0.182) Male Youth (%) 0.192*** 0.0455 0.126** 0.215 0.206*** 0.0351 (0.0555) (0.240) (0.0574) (0.165) (0.0675) (0.191) Constant 0.0437 0.171 0.0750** 0.156* 0.0217 0.0635 (0.0351) (0.173) (0.0360) (0.0867) (0.0394) (0.0588) Observations 315 75300 90310 80 R-squared 0.821 0.860 0.803 0.843 0.835 0.813 Standard errors in parentheses All equations include year and neighborhood fixed effects. Column (1) limits the sample to the neighbor- hoods in the bottom 4 quintiles of incarceration rates over the period. Column (2) limits the sample to the neighborhoods in the top quintile of incarceration rates over the period. Column (3) limits the sample to the neighborhoods in the bottom 4 quintiles of black population over the period. Column (4) limits the sample to the neighborhoods in the top quintile of black populations over the period . Column (5) limits the sample to the neighborhoods in the bottom 4 quintiles of rates of single mothers over the period. Column (6) limits the sample to the neighborhoods in the top quintile of rates of single motherhood over the period 534 G. Dhondt the next year. In low Black population neighborhoods, by contrast, a 1 percentage point increase in incarceration in 1 year is associated with a statistically insignificant 0.2 percentage point increase in crime in the following year. Similarly, in neighborhoods with a high percentage of single mothers, a 1 percentage point increase in incarceration is associated with a 1.9 percentage point increase in crime the following year, while a 1 percentage point increase in incarceration in neighborhoods with low percentages of single mothers is associated with a statistically insignificant 1.1 percentage point increase in crime the following year. The positive impacts of lagged admissions per capita on crime rates thus range from roughly twice to eight times as high in margin- alized neighborhoods as compared to non-marginalized areas, depending on the defini- tion of marginalization. In short, it would be reasonable to suggest that over the period studied, rising incarceration rates were ineffectual in reducing crime in the next period and may indeed have served to undermine forms of informal social control that could have been beneficial in this regard.In other recent work [ 44], I have examined a large literature which suggests that the logic of mass incarceration is better seen as a political management device whereby disaffected populations are disciplined ‘outside ’the normal methods avail- able under neoliberal capitalism. Such an argument would suggest that incarceration is far more likely to be correlated with restive populations than with actual crime on the ground. Table 4provides additional evidence for such a position. It shows the correlation between admissions and all the correlates I have examined so far. As is evident, crime rates are only weakly correlated with incarceration ( r0 0.31), especial- ly when compared with other social and demographic correlates of admissions such the percentage of Black population, ( r0 0.69) the percentage of the population with no high school diploma ( r0 0.68) and the percentage of female headed households with children ( r0 0.44).

Moreover, I looked at the correlation between the difference in the admission and crime per capita in neighborhoods with high Black percentages and high percentages of people with no high-school diplomas. Across the data set, crime rates are between 0 .005 .01 .015 .02 0 .005 .01 .015 .02 relperc admperc Fittedvalues Fig. 3Scatter plot of admissions per capita against lagged releases per capita Bluntness of incarceration: crime and punishment in Tallahassee 535 1.5 and 2 times higher in the neighborhoods with the highest Black percentage.

Prison admission rates by contrast are higher by between 5 and 9.5 times in these neighborhoods. Similarly, crime rates are between 1.5 and 2.5 times higher in the neighborhoods with the highest percentage of individuals with no high school diploma. Admission rates, by contrast, are higher from 4 to 7.5 times. This evidence supports those scholars which argue that incarceration has little to do with crime.

Dhondt [44] provides a more extensive discussion on why this might be the case.

Conclusions and further research This paper attempts to estimate the effect of adding prisoners on crime per capita and the effect of prison cycling on crime per capita at the neighborhood level in the city of Tallahassee. I provide evidence that increasing incarceration does not have a negative effect on crime in neighborhoods but instead has a positive effect in high incarcer- ation neighborhoods. Similarly, prison cycling —the mill of adding and releasing prisoners— also has a positive effect on crime in marginalized neighborhoods. These results are consistent with the other specifications where I find that increased incar- ceration leads to increased crime. 10 They provide support for the theory that there are crime-enhancing counter-tendencies which operate at the neighborhood level such as the coercive mobility effect. These results should caution against the continued use of incarceration as the first response to crime and other social problems since incarcer- ation might have the opposite of the officially stated intended effect. This paper points to the need for further research. Further research should test the external validity of the positive impact o f incarceration rates on crime rates in different localities at the neighborhood level. Equally important, there is little existing research as to why certain neighborhoods see disproportionately high levels of incarceration even though crime rates themselves may not be extraordinary. Such examinations will go some way to help create more humane and effective approaches to criminal justice in the United States of America. 10Other regressions are available from the author on request, including Arellano-Bond estimations. These regressions and the Arellano-Bond estimations show the robustness of the findings.

Table 4 Correlates of admissions per capita Prison admissions Black Crime Poverty Employed Bach.

DegreeNo High-school Single MotherLatino Prison Admissions(%) 1 Black (%) 0.69 1 Crime (%) 0.19 0.23 1 Poverty (%) 0.23 0.30 0.42 1 Employed (%) 0.40 0.56 0.08 0.16 1 Bachelor Degree (%) 0.42 0.61 0.25 0.55 0.69 1 No High School (%) 0.65 0.62 0.36 0.32 0.39 0.43 1 Single Mother (%) 0.47 0.55 0.039 0.14 0.23 0.27 0.438 1 Latino (%) 0.23 0.28 0.073 0.39 0.084 0.14 0.28 0.33 1 536 G. Dhondt References 1. Marvell, T. B., & Moody, C. E. (1994). Prison population growth and crime reduction.Journal of Quantitative Criminology, 10 (2), 109–140.

2. Levitt, S. (1996). The effect of prison population size on crime rates: evidence from prison over- crowding litigation. Quarterly Journal of Economics, 111 , 319–351.

3. Spelman, W. (2000). The limited importance of prison expansion. In A. Blumstein & J. Wallman (Eds.), The crime drop in America (pp. 97–129). New York: University of Cambridge Press.

4. Spelman, W. (2005). Jobs or jails?: The crime drop in Texas. Journal of Policy Analysis and Management, 24 (1), 133–165.

5. Liedka, R. V., Piehl, A. M., & Useem, B. (2006). The crime control effects of incarceration: does scale matter? Criminology & Public Policy, 5 (2), 245–276.

6. Useem, B., & Piehl, A. M. (2008). Prison state: The challenge of mass incarceration . New York:

Cambridge University Press.

7. Levitt, S. (2004). Understanding why crime fell in the 1990s: four factors that explain the decline and six that do not. Journal of Economic Perspectives, 18 (1), 163–190.

8. Reiman, J. (2007). The rich get richer and the poor get prison . New York: Wiley.

9. Zimring, F. E. (2007). The Great American Crime Decline (Studies in crime and public policy.) . New York: Oxford University Press.

10. Clear, T. R. (1996). Backfire: when incarceration increases crime. Journal of the Oklahoma Criminal Justice Research Consortium, 3 ,1–10.

11. Rose, D., & Clear, T. R. (1998). Incarceration, social capital, and crime: implications for social disorganization theory. Criminology, 36, 441–480.

12. Rose, D., & Clear, T. R. (2003). Incarceration, reentry and social capital: Social networks in the balance. In J. Travis & M. Waul (Eds.), Prisoners once removed: The impact of incarceration and reentry on children, families, and communities . Washington, D.C.: Urban Institute Press.

13. Clear, T. R., Rose, D., Waring, E., & Scully, K. (2003). Coercive mobility and crime: a preliminary examination of concentrated incarceration and social disorganization. Justice Quarterly, 20(1), 33–64.

14. Braman, D. (2004). Doing time on the outside: Incarceration and family life in America . Ann Arbor:

University of Michigan Press.

15. Western, B. (2006). Punishment and inequality in America . New York: Russell Sage Foundation.

16. Clear, T. R. (2007). Imprisoning communities: How mass incarceration makes disadvantaged neigh- borhoods worse . New York: Oxford University Press.

17. Clear, T. R. (2008). The effects of high imprisonment rates on communities. Crime and Justice, 37,97– 132.

18. Rotman, E. (1995). The failure of reform. In N. Morris & D. J. Rothman (Eds.), The Oxford history of the prison: The practice of punishment in western society (pp. 169–197). New York: Oxford University Press.

19. Rothman, D. J. (1995). Perfecting the prison: United States, 1789 –1865. In N. Morris & D. J. Rothman (Eds.), The Oxford history of the prison: The practice of punishment in western society (pp. 111–129).

New York: Oxford University Press.

20. Pew Center on the States. (2008). 1 in 100: Behind bars in America 2008 . Washington, DC: Pew Charitable Trusts.

21. Parenti, C. (1999). Lockdown America: Police and prisons in the age of crisis . New York: Verso.

22. Gilmore, R. W. (2007). Golden Gulag: Prisons, surplus, crisis, and opposition in globalizing Califor- nia . Berkeley: University of California Press.

23. Mauer, M., & Chesney-Lind, M. (Eds.). (2003). Invisible punishment: The collateral consequences of mass imprisonment. New York: The New.

24. Golash, D. (2005). The case against punishment: Retribution, crime prevention, and the law . New York: New York University Press.

25. Foucault, M. (1977). Discipline and punish. New York: Pantheon.

26. Becker, G. S. (1968). Crime and punishment: an economic approach. Journal of Political Economy, 76 (2), 169 –217.

27. Miller, J. G. (1996). Search and destroy: African-American males in the criminal justice system. New York: Cambridge University Press.

28. Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: a multilevel study of collective efficacy. Science, 277, 918–924.

29. Riley, J. (2004). Review essay: prisoners once removed. Alaska Justice Forum, 21(2), 4–5. Bluntness of incarceration: crime and punishment in Tallahassee 537 30. Devine, J. A., Sheley, J. F., & Smith, M. D. (1988). Macroeconomic and social-control policyinfluences on crime rate changes, 1948 –1985. American Sociological Review, 53 (3), 407–420.

31. Marvell, T. B., & Moody, C. E. (1997). The impact of prison growth on homicide. Homicide Studies, 1, 205 –233.

32. Kovandzic, T., Sloan, J. J., III, & Vieraitis, L. M. (2002). Unintended conssequences of politically popular sentencing policy: the homicide promoting effects of three-strikes in U.S. cities. Criminology & Public Policy, 1 (3), 399–424.

33. Kovandzic, T., & Vieraitis, L. (2006). The effect of county-level prison population growth on crime rates. Criminology & Public Policy, 5 (2), 213–244.

34. Kovandzic, T., Sloan, J. J., III, & Vieraitis, L. M. (2004). Striking out as crime reduction policy: the impact of three strikes laws on crime rates in US cities. Justice Quarterly, 21, 207–240.

35. Dhondt, G. (2012). The effect of prison population size on crime rates: Evidence from cocaine and marijuana mandatory minimum sentencing . New York: John Jay College of Criminal Justice, Eco- nomics Department, Working Paper.

36. Levitt, S., & Lochner, L. (2001). The determinants of juvenile crime. In J. Gruber (Ed.), Risky behavior among youths: An economic analysis (pp. 327–373). Chicago: National Bureau of Economic Research.

37. Bullock, H. A. (1955). Urban homicide in theory and fact. Journal of Criminal Law, Criminology and Police Science, 45 (5), 565–575.

38. Capone, D. L., & Nicholas, W. W. (1976). Urban structure and criminal mobility. American Behavioral Scientist, 20 , 199–214.

39. Baldwin, J., & Bottoms, A. (1976). The urban criminal: A study in sheffield. London: Tavistock.

40. Hakim, S. (1979). Inter-jurisdictional spillover of crime and police expenditure. Land Economics, 55 (2), 200 –212.

41. Hakim, S., & Rengert, G. F. (1981). Crime spillover. Beverly Hills: Sage.

42. Pettiway, L. (1982). Mobility of robbery and burglary offenders: ghetto and nonghetto spaces. Urban Affairs Quarterly, 18 , 255–270.

43. Taylor, R. B. (1998). Crime and small-scale places: What we know, what we can prevent, and what else we need to know. Paper presented at the Crime and Place: Plenary Papers of the 1997 Conference on Criminal Justice Research and Evaluation, Washington, D.C.

44. Dhondt, G. (2012). The logic of the whip: mass incarceration as labor discipline in the neoliberal social structures of accumulation . New York: John Jay College of Criminal Justice, Economics Department, Working Paper.

538 G. Dhondt