LABELING THEORY CAPSTONE PAPER

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 1 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS AND ITS CONSEQUENCES FOR RECIDIVISM*

TED CHIRICOS KELLE BARRICK WILLIAM BALES

College of Criminology and Criminal Justice Florida State University

STEPHANIE BONTRAGER Justice Research Associates

KEYWORDS: labeling, felony conviction, recidivism

Florida law allows judges to withhold adjudication of guilt for individ- uals who have been found guilty of a felony and are being sentenced to probation. Such individuals lose no civil rights and may lawfully assert they had not been convicted of a felony. Labeling theory would predict that the receipt of a felony label could increase the likelihood of recidi- vism. Reconviction data for 95,919 men and women who were either adjudicated or had adjudication withheld show that those formally labeled are significantly more likely to recidivate in 2 years than those who are not. Labeling effects are stronger for women, whites, and those who reach the age of 30 years without a prior conviction. Second-level indicators of county characteristics (e.g., crime rates or concentrated disadvantage) have no significant effect on the adjudication/recidivism relationship.

Traditional labeling theory explains the potential “escalating” conse- quences of a criminal or delinquent labeling experience in two ways (Lof- land, 1969; Sherman et al., 1992). The first consequence involves a

* The authors would like to thank the anonymous reviewers as well as Carter Hay, Dan Mears, Brian Stults, and especially Xia Wang for their helpful comments on an earlier version of this article. Direct correspondence to Ted Chiricos (e-mail: [email protected]).

CRIMINOLOGY VOLUME 45 NUMBER 3 2007 547

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 2 22-AUG-07 10:10

548 CHIRICOS, BARRICK, BALES & BONTRAGER

transformation of identity,1 and the second emphasizes structural impedi- ments to conventional life that result from a labeling event.2 Although labeling events have been variably operationalized to include police con- tact, arrest, conviction, and imprisonment, it is arguable that felony convic- tion is the most consequential in relation to the development of structural impediments. The label of “convicted felon” strips an individual of the right to vote, serve on juries, own firearms, or hold public office. In many states, convicted felons are prohibited from obtaining student loans, employment in state-licensed occupations, or employment with state- licensed companies. In addition, the label of convicted felon may contrib- ute to various informal exclusions that can make access to noncriminal activities more difficult and criminal alternatives more attractive.3

The state of Florida has a law that allows individuals who have been found guilty of a felony, either by a judge, jury, or plea, to literally avoid the label of convicted felon. Judges have the option of “withholding adju- dication” of guilt for convicted felons who are being sentenced to proba- tion. The consequence of this unique labeling event is that offenders who are equivalent in terms of factual guilt can either be labeled a convicted felon or not. For those offenders who have adjudication withheld—half of the felony probationers in Florida in recent years—no civil rights are lost and such individuals may legitimately say on employment applications and elsewhere that a felony conviction did not occur. For those offenders who are formally adjudicated, all of the structural impediments of being a con- victed felon are possible.

Labeling theory would hypothesize that being formally adjudicated should increase the chances of recidivism. The research question that we address here is whether the labeling event of adjudication or the withhold- ing of adjudication has any effect on subsequent recidivism.4 Drawing on

1. Early research involving self-esteem or “self-typing” was reported by Ageton and Elliott (1974), Gibbs (1974), Harris (1976), Jensen (1972), and Tittle (1972), among others. More recently, “reflected appraisals of self” (Heimer and Mat- sueda, 1996; Matsueda, 1992) have been implicated in identity-related labeling research.

2. For an excellent discussion of structural impediments related to labeling and sup- portive research, see Sampson and Laub (1997: 147–52).

3. Discussions of the collateral consequences of felony conviction have primarily focused on the loss of voting rights (Hull, 2006; Manza and Uggen, 2006; Uggen and Manza, 2002). But a broader range of consequences from imprisonment, which presumes conviction, has also been described (Mauer and Chesney-Lind, 2002; Western, 2002; Western and Pettit, 2005).

4. It is also possible, as specific deterrence theory would hypothesize, that being formally adjudicated could have the opposite effect of reducing the chances of recidivism. For discussions of specific deterrence, see, for example, Paternoster (1987), Nagin (1998), and Stafford and Warr (1993). At the same time, a positive relationship between formal sanctioning and subsequent criminal activity could

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 3 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 549

recent theory and research, we also examine whether any labeling effects on recidivism are contingent on characteristics of the defendant, such as sex, race, or prior record, as well as on characteristics of the county to which the sentenced defendant returns, such as rates of crime or levels of social disadvantage. To assess these questions, we examine the post- sentencing recidivism experience of 95,919 men and women who were found guilty of a violent, property, or drug felony and sentenced to proba- tion in Florida between 2000 and 2002. This study is the first to examine the potential labeling effects of adult felony conviction.

Hierarchical linear modeling (HLM) is used to assess the direct effect of having adjudication applied or withheld, as well as to assess the effects of other individual-level variables. In addition, we assess the direct effect of county-level characteristics on recidivism and the conditioning relevance that these second-level county characteristics may have on the relationship between adjudication status and recidivism.

LABELING EFFECTS: CONCEPTUAL ISSUES

A recurrent theme in the recent renewal of interest in labeling effects (Bernburg and Krohn, 2003; Bernburg, Krohn, and Rivera, 2006; Sampson and Laub, 1997) is the recognition that consequences of formal sanctions may be contingent on the characteristics of offenders. As noted by Pater- noster and Iovanni (1989: 381), “we should not expect labeling effects to be invariant across societal subgroups.” With this in mind, four individual- level contingencies will inform parts of our analysis. They are race, sex, prior record, and “stakes in conformity.” Their conceptual relevance for potentially observed labeling effects is reviewed briefly here.

Race was one of the first issues to be raised in discussions of contingent labeling outcomes. Harris (1976: 433) argued that the effects of official labeling would be least consequential for those individuals most excluded from “ascriptive membership” in dominant groups—particularly minori- ties. Earlier, Jensen (1972) as well as Ageton and Elliott (1974) hypothe- sized that higher status persons—presumably including non-minorities in most contexts—may be more susceptible to labeling effects. Similarly, Paternoster and Iovanni suggested that persons who grant less “legitimacy to the legal order and its agents” may, as a result, be “more impervious to labeling effects” than others (1989: 382).

Alternatively, although not referencing race per se, Sampson and Laub (1997: 153) discussed the potential life-course consequences of official sanctions for the disadvantaged urban poor. They noted that “among the

be the consequence of factors other than those attributed to labeling. Among these alternatives could be general strain (Agnew, 2006), defiance (Sherman, 1993), and control–balance (Tittle, 1995).

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 4 22-AUG-07 10:10

550 CHIRICOS, BARRICK, BALES & BONTRAGER

disadvantaged, things seem to work differently. Deficits and disadvantages pile up faster and this has continuing negative consequences for later development. . . .” Building on that argument, Bernburg and Krohn (2003) hypothesized that labeling effects would be strongest among “disadvan- taged offenders” as distinguished by minority and poverty status. Their expectations were further supported by the observation that official label- ing of disadvantaged youth could be “enhanced by the negative stereo- types that are already associated with these youths in mainstream culture” (2003: 1290).

The relevance of sex for labeling outcomes has been substantially over- looked in criminological discourse. Even a book with the promising title of Labeling Women Deviant (Schur, 1983) makes no mention of whether or how labeling effects for women would be different from those for men. Regarding juveniles, Ageton and Elliott (1974) hypothesized that “males are more likely to be affected negatively by police contact than females,” but no rationale was associated with that expectation. Somewhat later, Ray and Downs (1986), also in the context of adolescents, articulated the opposite expectation. On the premise that “females are expected to be more attentive to interpersonal relationships than men,” they hypothe- sized that “labels may exert more of an influence on behavior for females than males” (1986: 171).

The possibility that women could be more adversely affected by labeling than men has also been noted in passing references that were concerned primarily with broader issues. For example, in his discussion of “gender blind criminology,” Messerschmidt (1993) conjectured how females might be factored into traditional criminological theories if one took gender into account. In this context, he speculated that because men exercise greater power in society than women, “men should therefore have increased opportunity to counteract official labeling . . .” (1993: 4). Similarly, Gior- dano, Cernkovich, and Lowery (2004) were developing alternative expec- tations concerning the relative prospects of women as compared with men for desistance from crime, without specific reference to formal labeling. In that context, they observed that “the greater social stigma attached to female in contrast to male involvement in antisocial behavior” could in the long run “be more limiting to life chances/opportunities for a return to conventional roles” (2004: 189).

Interestingly, the antithesis to labeling, deterrence, which anticipates that official sanctions will reduce rather than amplify criminal involve- ment, has generally argued that women are more likely to be deterred than men because they have more to lose if caught committing a crime. The argument has been made that women have been socialized to feel more responsible for social relationships than men (Blackwell and Eschholz, 2002) and to have a “greater investment in conformity” than is

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 5 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 551

true for men (Richards and Tittle, 1981), both of which could presumably be lost as a consequence of a formal criminal sanction. If sanctions are expected to have stronger deterrent effects for women than for men, then logically, labeling effects should be weaker for women than they are for men. However, these arguments have been made in the context of general deterrence and not specific deterrence, which would be more appropriate here.

Expectations concerning the relevance of prior record for potential labeling effects have not been extensively developed, but a limited consen- sus favors the prospect that the consequences of formal sanctions will be greater for naı̈ve, as opposed to repeat offenders. Noting that some ana- lysts have speculated that novice criminals would be more susceptible to deterrence, Horwitz and Wasserman (1979) offered “an alternative hypothesis.” They suggested that punishments of “increasing severity will lead to a greater increase in the rate of subsequent criminal offenses for primary than for secondary deviants” (1979: 57). Smith and Gartin (1989) articulated a similar hypothesis, although like Horwitz and Wasserman, they provided no supporting rationale. What that rationale could be is sug- gested by Paternoster and Iovanni (1989: 386) who observed that once an individual has been labeled, “it is doubtful that further increments in label- ing will continue to produce further deviance.” They attributed this occur- rence to a “leveling off” of labeling effects on the premise that once an individual has experienced social exclusions on the basis of a label, “it is not unreasonable to assume that further exclusion would have little addi- tional meaning . . .” (1989: 386).

Regarding “stakes in conformity,” Sherman et al. have shown that the issue has actually been used to hypothesize contradictory labeling effects. What they characterize as the “greater vulnerability” version of labeling theory expects that those individuals who “care more about the opinions of conventional society” will be more vulnerable to the escalating effects of labeling (1992: 682). From this point of view, nonminorities and higher status individuals would presumably be more vulnerable to labeling effects because they have more to lose. But it has been argued that stakes in con- formity could have the opposite consequence. Specifically, persons with a higher stake in conformity, such as those who are employed and/or mar- ried, could be more insulated from the negative consequences of official labels. Sherman et al. (1992) refer to this position as the “less vulnerabil- ity” approach to labeling effects. In this view, those individuals who have more to lose because of their informal social bonds may as a consequence have “other social resources that overcome the impact of labeling” (1992: 682).

This “less vulnerability” position is implicit in the observation by Samp- son and Laub (1997: 152) that “sanctioning tends to aggravate crime

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 6 22-AUG-07 10:10

552 CHIRICOS, BARRICK, BALES & BONTRAGER

among populations with low ‘stakes in conformity’.” They also note, with Braithwaite (1989) and Sherman (1993), that criminal sanctions may actu- ally promote future defiance of the law when applied to those individuals with weaker social bonds to conventional society (Sampson and Laub, 1997: 153). Such a position is consistent with recent ethnographic (Ander- son, 1999) and autobiographic (Shakur, 1993) accounts of the effects of official sanctions on individuals from highly disadvantaged inner-city envi- ronments who presumably have fewer stakes in conformity than others.

COMMUNITY-LEVEL CONTEXTS

However much individuals with particular characteristics may vary in their vulnerability to labeling effects, Paternoster and Iovanni (1989: 373) remind us that “the effect of status characteristics on labeling outcomes is not invariant, but varies substantially across different social contexts.” In a related manner, Sampson and Laub (1997: 153) noted that in some social environments, “deficits and disadvantages pile up faster” and as a result “one cannot ignore the effects of larger social contexts” in studying the life-course development of individuals who may have experienced official labeling and some of its negative consequences. Kubrin and Stewart (2006: 172) have similarly argued that “where ex-offenders live greatly affects their ability to reintegrate into society” after being formally sanctioned.

For these reasons, we have chosen to examine several social contexts aggregated at the county level that may interact with individual traits in producing greater or lesser labeling effects. To our knowledge, this study is the first study of labeling outcomes to do so. Specifically, we include in our analyses measures of concentrated disadvantage (CD), concentrated extremes (affluence and poverty), index crime rates, and police presence. The conceptual relevance of each measure for potentially influencing labeling outcomes is briefly reviewed here.

As noted by Sampson, Raudenbush, and Earls (1997), among others, CD is a measure of “underclass concentration,” which involves high levels of minority presence, poverty, welfare, and female-headed families. Such places not only have frequently increased levels of crime (Rountree, Land, and Miethe, 1994; Sampson, Raudenbush, and Earls, 1997; Silver, 2000), but they also have low levels of neighborhood resources and social ser- vices (Sampson, Morenoff, and Gannon-Rowley, 2002; Wilson, 1987). In addition, as Sampson and Laub (1993: 295) have noted, “counties charac- terized by. . .a large concentration of the ‘underclass’. . .are more likely than other counties to be perceived as containing offensive and threaten- ing populations” and so are more likely to be more heavily policed. Kubrin and Stewart (2006) hypothesized and found higher levels of recidivism among felons released to community supervision in places with higher levels of CD. For these reasons, we might expect that those individuals

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 7 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 553

who are labeled as convicted felons would see “deficits and disadvan- tages”—specifically recidivism—“pile up” more often in places that have higher levels of CD.

Whatever the value of taking account of the negative consequences of CD, Sampson, Morenoff, and Gannon-Rowley (2002: 446) have suggested that the common tactic of focusing on concentrated disadvantage may “. . .obscure the potential protective effects of affluent neighborhoods.” Morenoff, Sampson, and Raudenbush (2001: 528) have also noted that “the resources that affluent neighborhoods can mobilize are theoretically relevant to understanding the activation of social control . . . .” They found that a measure of concentrated affluence relative to poverty (index of con- centrated extremes, ICE) was related to lower levels of homicide. Consis- tent with that, Kubrin and Stewart (2006) hypothesized and showed that ICE was negatively related to the likelihood of recidivism for ex-offenders in Portland, Oregon. On that basis, we might expect that the negative effects of a felony convict label could be diminished in places with higher levels of ICE.

However, the ICE is really a measure of “relative inequality” (Kubrin and Stewart, 2006: 177) and not of “relative affluence” as operationalized by Sampson, Morenoff, and Earls (1999). It is reasonable to imagine that with extremes of poverty and affluence coexisting in the same county, the “protective effects of affluent neighborhoods” (Sampson, Morenoff, and Gannon-Rowley, 2002) could lead to “the activation of social control” (Morenoff, Sampson, and Raudenbush, 2001) in ways that result in more aggressive policing and prosecution, especially in poorer sections of the county. In that context, one might expect higher levels of ICE to result in higher levels of formal social control and recidivism generally and in an amplification of the effects of being labeled as a convicted felon.

We also expect that convicted felons will be more likely to recidivate in places with higher index crime rates and with a stronger law enforcement presence. The rationale for the former expectation is that places with higher levels of crime, which will also likely have increased levels of CD, have more criminal opportunities, learning environments, and role models (Akers, 1998; Anderson, 1999; Warr, 2001). Individuals who are labeled as convicted felons and who may experience social exclusions not exper- ienced by their nonlabeled counterparts could find it easier to resume criminal activity in such a context, rather than in places with lower levels of crime. This occurrence may explain why, for example, Smith and Pater- noster (1990) found that the likelihood of recidivism for those individuals who had been referred to juvenile court in Florida was higher in counties with higher crime rates.

The expectation concerning higher recidivism for convicted felons in places with a stronger police presence is predicated on Lofland’s (1969)

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 8 22-AUG-07 10:10

554 CHIRICOS, BARRICK, BALES & BONTRAGER

concept of “prevalent sensitivity.” His argument is that labels—in this case, recidivating felon—are more likely to be “imputed” in places where a higher prevalent sensitivity to misbehavior exists because more “special- ists” are deployed to detect and respond to it. In his terms: “as the number of imputational specialists increases . . . it is likely that the number of peo- ple imputed as deviant will also increase” (1969: 136). So if individuals labeled as convicted felons resume criminal activity, it seems more likely that they would be caught and officially recidivate where more police officers are present.

As a final conceptual issue, it can be noted that every adult facing a felony conviction, whether or not adjudication is withheld, has been arrested at least once and prosecuted. Assuming that those experiences constitute labeling events, the ideal situation described by Bernburg and Krohn (2003) of comparing samples from the general (unlabeled) popula- tion to assess “absolute labeling” effects is not available for the label of convicted felon. However, when it comes to generating the structural impediments that can accrue from labeling, the event of felony conviction, although “relative,” may nonetheless be relatively more important than any other for adults. Certainly, police contact, arrest, and prosecution do not cause individuals to lose their civil rights or to be legally excluded from the great variety of occupational opportunities denied to those indi- viduals who are convicted felons. By controlling for prior convictions, both adult and juvenile, an analysis of the consequences of having adjudication withheld or applied for convicted felons affords an estimate of the inde- pendent effect of “being labeled or not” in a relative labeling context that is most critical.

LABELING EFFECTS: PRIOR RESEARCH

Several systematic reviews of labeling effects research (Barrick, 2005; Bernburg, 2002; Palamara, Cullen, and Gertsten, 1986; Paternoster and Iovanni, 1989) underscore several things. First, much evidence that can be inferred as relating to labeling is actually sanction effects research, which is often primarily concerned with deterrence. More important, extraordi- nary methodological variation exists in this research, in terms of both the specification of independent and dependent variables and the elaboration of contingent or subsample analyses. This variation makes a brief sum- mary of any research results or any patterns of support or nonsupport for labeling hypotheses all but impossible. Thus, we limit our discussion of prior research here to that which is most relevant to the analyses that we have undertaken. Specifically, we review studies that have examined “con- viction” as the key labeling event, and we review some contingent results

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 9 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 555

of labeling research that anticipate the kinds of analyses that we describe below.

CONVICTION EFFECTS RESEARCH

We are aware of five studies that isolate conviction as a key labeling event. Two studies address misdemeanor charges of domestic violence (Thistlethwaite, Wooldredge, and Gibbs, 1998; Ventura and Davis, 2005), another deals with drunk driving (Taxman and Piquero, 1998), and two are concerned with the “conviction” of juveniles for serious offenses (Fagan, Kupchik, and Liberman, 2003; Hagan and Palloni, 1990). To our knowl- edge, no research to date has examined the labeling effect of a felony con- viction for adults, which is the focus of the current study.

The two domestic violence studies that examine conviction effects involved midwestern cities and used rearrest within 1 year for another domestic violence charge as a measure of recidivism. Multivariate logistic analyses in both studies showed that conviction significantly reduced the likelihood of recidivism, thus supporting a deterrent effect more than a labeling effect (Thistlethwaite, Wooldredge, and Gibbs, 1998: 394; Ventura and Davis, 2005: 270).

Taxman and Piquero (1998) estimated the consequences of more than 3,500 “drunk driving convictions” in Maryland between 1985 and 1993. They measured recidivism by reconviction for drunk driving within 3 years. Their primary concern was to distinguish between the effects of punitive as opposed to rehabilitative responses by the court. But they also made an important distinction—from a labeling perspective—between those who received a “guilty disposition” and those who received “proba- tion before judgement” (PBJ). The latter seems to be similar to the “adju- dication withheld” option that we examine for felony convictions, inasmuch as getting PBJ meant that one’s conviction was “stayed” and effectively expunged from the public record if probation was successfully completed (Taxman and Piquero, 1998: 133).

The authors report that for all offenders, conviction increased the risk of recidivism by 12 percent but that effect was not significant with other con- trols included. Separate analyses for first-time offenders showed that a guilty disposition increased the risk of recidivism by 27 percent, and this was the only significant punishment or rehabilitation effect for such offenders (Taxman and Piquero, 1998: 135–7). This finding supports an interpretation that conviction for drunk driving had a labeling effect that was contingent on prior record, which is an issue that we discuss more fully in the next section.

Hagan and Palloni (1990) reanalyzed data from the Cambridge Study in

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 10 22-AUG-07 10:10

556 CHIRICOS, BARRICK, BALES & BONTRAGER

Delinquency Development that were originally collected by West and Far- rington (1973). The primary dependent variables were self-reports of delinquency and crime (38 items) that were generated by interviews with randomly chosen London youth. The interviews were conducted over a period of years when the boys were between 8 and 24 years of age. For those who were interviewed, official data from the Criminal Record Office were used to create matched samples of youth who had been “convicted” by the age of 15 years and those who had not (Farrington, 1977: 114).5

In looking at self-reported delinquency subsequent to the experience of conviction, Hagan and Palloni (1990) controlled for individual and family attributes that could be taken as indicators of “characterological and cul- tural” traits that might predict delinquent involvement. They also con- trolled for prior reported delinquency. A key finding was that being convicted before the age of 15 years significantly increased the likelihood of delinquency at ages 16–17, 18–19, and 21–22 years independent of the effects of all other controls. This result clearly supports the expectation that the labeling experience of conviction contributes to subsequent delin- quent and criminal involvement (1990: 278–9).6

Fagan, Kupchick, and Liberman (2003) compared the consequences for juveniles of being convicted in an adult criminal court with those of being convicted in a juvenile court for felony assault, robbery, or burglary. To do this, they matched counties from northern New Jersey and from New York, which were all part of the same New York metro SMSA. From the New Jersey counties, they examined the consequences of juvenile court conviction, and from the New York counties, the same was done for crimi- nal court conviction. They tracked rearrest and reincarceration outcomes for at least 2 years after the completion of punishment for those convicted.

An outcome called “any court action,” which the authors describe as “analogous to conviction,” was found to significantly increase the chances of rearrest for violent, property, and weapons crimes (not drugs) for youth convicted in either the juvenile or the criminal courts. However, it was also found that adolescent offenders convicted in criminal court were rear- rested and reincarcerated more often and more quickly for violent, prop- erty, and weapons offenses than those convicted in juvenile court. The reverse was true for drug crimes (Fagan, Kupchick, and Liberman, 2003: 69). In sum, the available evidence on whether conviction is related to

5. Farrington (1977: 114) describes the conviction data in this manner: “searches of the Criminal Record Office” uncovered a number of youth who “had been found guilty in court. . . .” The crimes for which these youth had been convicted “were mainly burglaries, thefts and unauthorized takings of motor vehicles.”

6. The authors also report a significant interactive effect of parent and son convic- tion on subsequent delinquency, which was the primary theoretical focus of their study (Hagan and Palloni, 1990: 278–9).

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 11 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 557

recidivism is limited to juveniles, misdemeanor domestic violence offend- ers, and first-offending drunk drivers. How felony conviction plays out in the life course of adults remains to be seen.

CONTINGENT LABELING EFFECTS RESEARCH

It should be noted that some studies reporting what we call contingent labeling effects are not labeling studies per se. In fact, several of them are primarily concerned with specific (as distinct from general) deterrence, but if they report a relationship with recidivism subsequent to a sanctioning experience, we designate that as a potential labeling effect.

One consistent contingency seems to be prior record. As noted, Taxman and Piquero’s study (1998: 135–7) of drunk driving convictions found sub- stantially stronger labeling effects (increased recidivism) for first-time offenders than for their full sample. Similarly, Fagan, Kupchick, and Liber- man (2003: 41) report that the relatively stronger labeling effects (recidi- vism) for youth convicted in criminal as opposed to juvenile court are especially pronounced for those with no record of prior arrests. Horwitz and Wasserman (1979) found that more severe sanctions were related to higher rates of rearrest for juveniles in Newark, but that effect held only for those identified as first-time offenders on the basis of arrest records. DeJong (1997) analyzed the rearrest likelihood of those either sentenced to jail or not for misdemeanors7 in New York City. She reported that the experience of incarceration increased recidivism risk but only for first-time arrestees (1997: 571). The pattern of observed labeling effects being con- centrated among first offenders is consistent with the suggestion by Pater- noster and Iovanni (1989) that incremental labeling events may be less consequential than those that occur earlier in the life course of an individual.

Stakes in conformity have also been studied in relation to sanction effects. In general, what has been found is that apparent labeling effects are stronger for those who have a low stake in conformity or weak ties to conventional society. DeJong (1997) reported that incarceration increased the likelihood of rearrest especially for those with “low stakes” as indi- cated by having at least two of these traits: unmarried, unemployed, and not a high-school graduate. Sherman et al. (1992: 686) found that formal arrest for domestic violence significantly increased the likelihood of rear- rest in Milwaukee among those individuals who had a low stake in con- formity as measured by marital status and employment. Two additional

7. Her sample was taken from “criminal court” as opposed to “Supreme Court” cases, which in New York involves primarily misdemeanors or felony cases that are negotiated down. Felonies that are not negotiated down are tried in the Supreme Court.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 12 22-AUG-07 10:10

558 CHIRICOS, BARRICK, BALES & BONTRAGER

domestic violence studies focused on those individuals with “high” stakes and concluded that sanctions for these people are associated with less recidivism, which the authors conclude is a stronger deterrent effect (Pate and Hamilton, 1992; Thistlethwaite, Wooldredge, and Gibbs, 1998). The negative relationship between stakes and recidivism for those sanctioned at least implies that those with lower stakes would experience higher recidivism (labeling effect) as reported by Dejong (1997) and Sherman et al. (1992).

A third contingency that has received some attention is race. Several early studies found that labeling was more consequential for white juveniles than for others. Ageton and Elliott (1974) reported the race-spe- cific effect for police contact and self-reported “delinquency orientation” among California youth. Harris (1975) found that New Jersey youth who had been incarcerated for longer periods expressed a higher “relative expected value of criminal choice” but that relationship held only for whites. Harris (1975: 97) hypothesized that the label might mean more to white youth because it came from a system “dominated by white . . . authorities,” whereas minority youth might perceive the label as “created and applied by ‘outsiders.’ ” In contrast, the most recent and sophisticated labeling work with juveniles has shown that juvenile justice intervention increased the chances of subsequent adult criminality only for blacks (Bernburg and Krohn, 2003: 1314).

Two domestic violence intervention studies report conflicting results concerning race in relation to the effects of arrest. Sherman et al. (1992: 680) report that race did not condition the effect of arrest on subsequent domestic violence in Milwaukee. However, Berk et al. (1992: 705) found that for pooled data from four cities, the labeling effect of arrest on subse- quent offending was 11 percent higher for blacks than for whites and 47 percent higher for blacks who were unemployed than for employed whites.

Two reviews of the recidivism literature conclude that males are consist- ently more likely than females to resume criminal activity after formal sanctioning (Baumer, 1997; Gendreau, Little, and Goggin, 1996). Consis- tent with this generalization, Giordano, Cernkovich, and Lowery (2004) reported longitudinal data from Ohio showing that previously institution- alized women were much less likely to recidivate than their male counter- parts. Similarly, Smith and Paternoster (1990) found that male youth who had been formally processed by the juvenile justice system in Florida were significantly more likely to recidivate. However, Taxman and Piquero’s (1998) study of drunk driving convictions, Sung’s (1993) analysis of sen- tenced drug offenders, and Thistlethwaite, Wooldredge, and Gibbs’s (1998) research on individuals arrested for domestic violence each found that sex was not a significant predictor of recidivism.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 13 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 559

On the basis of conceptual considerations that have been raised and prior research that has been reported, there are reasonably consistent expectations that we can bring to our analysis concerning the relevance of prior record and stakes in conformity for potential labeling effects. But the same cannot be said for sex or race. Specifically, both conceptual and empirical support exists for expecting stronger labeling effects for less experienced offenders and for those with lower stakes in conformity. Con- cerning the sex of offenders, the conceptual arguments favor greater label- ing consequences for women than for men, but the empirical evidence suggests otherwise. Finally, expectations concerning the relative effect of labeling by race are mixed on both conceptual and empirical grounds.8

RESEARCH METHODOLOGY

In this study, we assess the consequence for recidivism of having adjudi- cation applied or withheld for 71,548 male offenders and 24,371 female offenders found guilty of a felony and sentenced to probation in Florida between 2000 and 2002. To the best of our knowledge, this is the first labeling research to examine the effect of adult felony conviction, and the first to examine whether community-level characteristics, such as rates of crime and CD, influence the relationship between the felony conviction and recidivism.

Our dependent variable, recidivism, is operationalized by whether (yes = 1) a felony probationer is convicted of another felony within 2 years of being sentenced. This variable involves a new offense for which an individ- ual is sentenced to prison, probation, or jail, and it does not include techni- cal violations of the terms of probation. This measure of recidivism only captures those actually found guilty of another offense, and not those who may have been arrested and charged but had charges dropped or were found not guilty. Overall, 19 percent of our sample recidivated within the 2-year follow-up period.9

8. It should be noted that several studies have examined recidivism among felony probationers, and with relative consistency, they report that recidivism is more likely for males, minorities, and persons with a more extensive prior record (Ben- edict and Huff-Corzine, 1997; Kruttschnitt, Uggen, and Shelton, 2000; Petersilia, 1985; Spohn and Holleran, 2002; Whitehead, 1991). However, these were not studies of labeling effects and the contingencies thereof, because they involve only individuals who have been labeled. Our concern is whether these factors, as well as stakes in conformity, have a conditional consequence for observed label- ing effects.

9. Arrest as an alternative measure of recidivism was considered, but those data are not available to us. The inclusion of probation revocation was also considered but rejected because the vast majority of those involved technical violations and not the commission of a new crime. Thus, the most complete and homogeneous mea- sure available to us in these data is reconviction.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 14 22-AUG-07 10:10

560 CHIRICOS, BARRICK, BALES & BONTRAGER

LEVEL 1 PREDICTORS

Our primary independent variable is whether (yes = 1) a person being sentenced to probation for a felony has adjudication formally applied. This outcome certifies the felony convict label. A total of 40 percent of our felony probationers were convicted in this manner.

Among the other factors used to model recidivism are demographic attributes of the defendant, his or her crime, and prior record. The means and standard deviations of the level 1 variables used are described in table 1, which also shows the bivariate correlation matrix for all variables of interest. At the individual level, the Florida Sentencing Guidelines database was used to identify offenders as either non-Hispanic blacks (yes = 1) or Hispanic of whatever race (yes = 1). The surnames of all offenders who were not identified as Hispanic from the Guidelines data were checked against the U.S. Census list of Hispanic surnames (Word and Per- kins, 1996), and any individual whose name matched one of those on the list was coded as an Hispanic defendant. The use of the surname list afforded a more comprehensive identification of Hispanics, which raises their proportion in our sample from 11.5 percent to 14.4 percent. Unfortu- nately, we cannot know how many females are identified as Hispanic or not by virtue of marriage. Also included as a predictor was the defendant’s age in years at sentencing.

The defendant’s crime was coded as either property (yes = 1) or drug related (yes = 1), with violent crime as the reference category.10 This choice was made because violent offenders were the least likely to recidi- vate11 and we expected the other types of crime to be more consequential in our recidivism models. The seriousness of the crime is indicated by a score generated by the Florida Sentencing Guidelines. Points ranging from 4 to 116 are assigned to the primary offense, and additional points are assigned for additional offenses and a variety of other factors, including victim injury or the use of a firearm.12 A related variable, which partially reflects crime seriousness, is length of supervision in months that is speci- fied by the probation sentence.

Prior record is indicated by points that are assigned by the Florida

10. “Other” crimes was such an eclectic category that we have not included it in the analysis. The most common of these crimes were Traffic (53 percent); Weapons, Possession (16 percent); Escape (15 percent); and DUI, No Injury (4 percent).

11. The relative frequency of recidivism was 25 percent for property offenders, 23 percent for drug offenders, and 15 percent for violent offenders.

12. Additional points can be assigned on the basis of legal status violations, commu- nity sanction violations, prior serious felony points, violation of the Law Enforce- ment Protection Act, drug trafficking, grand theft of a motor vehicle, commission of the primary offense by a street gang member, and involvement of domestic violence.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 15 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 561

T ab

le 1

. C

o rr

el at

io n

s, M

ea n

s, a

n d

S ta

n d

ar d

D ev

ia ti

o n

s am

o n

g S

tu d

y V

ar ia

b le

s

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19

1. R

ec o

n vi

ct io

n 1

2. A

d ju

d ic

at ed

.1 12

* 1

3. M

al e

.0 67

* .0

82 *

1

4. H

is p

an ic

–. 03

5* –.

04 4*

.0 77

* 1

5. B

la ck

.1 02

* .0

97 *

–. 03

1* –.

30 9*

1

6. W

h it

e –.

07 3*

–. 06

2* –.

02 4*

–. 40

7* –.

74 3*

1

7. A

ge –.

10 3*

.1 45

* –.

04 3*

–. 02

5* –.

08 9*

.1 03

* 1

8. V

io le

n t

o ff

en se

–. 08

3* –.

01 4*

.0 35

* .0

62 *

–. 03

2* –.

01 3*

.0 48

* 1

9. P

ro p

er ty

o ff

en se

.0 31

* –.

03 9*

–. 09

9* .0

38 *

–. 04

1* .0

12 *

–. 09

9* –.

42 2*

1

10 .

D ru

g o

ff en

se .0

38 *

.0 51

* .0

70 *

–. 09

1* .0

68 *

–. 00

1 .0

59 *

–. 41

2* –.

65 3*

1

11 .

P ri

o r

su p

er vi

si o

n v

io la

ti o

n .1

54 *

.4 27

* .0

39 *

–. 08

9* .1

42 *

–. 07

3* .1

37 *

–. 09

5* –.

03 6*

.1 15

* 1

12 .

P ri

o r

re co

rd p

o in

ts (

L n

) .1

74 *

.4 87

* .1

16 *

–. 12

8* .1

35 *

–. 04

0* .1

49 *

–. 09

3* –.

06 4*

.1 42

* .4

86 *

1

13 .

C ri

m e

se ri

o u

sn es

s (L

n )

–. 00

2 .0

68 *

.1 11

* .0

56 *

–. 01

1* –.

02 9*

–. 04

7* .4

61 *

–. 28

0* –.

10 4*

–. 02

5* –.

05 0*

1

14 .

S u

p er

vi si

o n

l en

gt h

( L

n )

–. 01

1* .1

96 *

.0 21

* –.

04 6*

–. 03

5* .0

66 *

.0 06

* .1

10 *

.0 39

* –.

13 1*

.0 71

* .0

77 *

.3 02

* 1

15 .

In ve

rs e

M il

ls r

at io

–. 16

7* –.

58 0*

–. 14

9* .0

67 *

–. 15

4* .1

01 *

–. 26

0* –.

00 1

.0 74

* –.

07 4*

–. 67

7* –.

88 7*

–. 16

0* –.

18 7*

1

16 .

IC E

–. 00

4 –.

03 7*

.0 01

–. 11

6* –.

05 3*

.1 33

* .0

19 *

–. 05

6* –.

03 7*

.0 83

* .0

35 *

.0 35

* –.

04 6*

.0 18

* .0

90 *

1

17 .

D is

ad va

n ta

ge .0

15 *

.0 04

.0 16

* .2

01 *

.1 66

* –.

30 0*

–. 01

4* .0

60 *

.0 32

* –.

08 3*

–. 00

6 –.

06 6*

.0 76

* –.

10 7*

–. 02

4* –.

68 4*

1

18 .

C ri

m e

ra te

.0 27

* –.

00 5

.0 14

* .2

35 *

.1 11

* –.

27 2*

–. 01

4* .0

41 *

.0 41

* –.

07 6*

.0 10

* –.

04 0*

.0 57

* –.

13 7*

–. 01

0* –.

32 3*

.6 41

* 1

19 .

P o

li ce

p re

se n

ce .0

04 –.

00 3

.0 30

* .2

02 *

.0 67

* –.

20 6*

.0 22

* .0

36 *

–. 00

7* –.

02 3*

–. 00

6 –.

04 9*

.0 59

* –.

08 3*

.0 03

–. 21

8* .4

12 *

.5 95

* 1

M ea

n 0.

19 0.

40 0.

75 0.

14 0.

36 0.

49 30

.9 6

0. 21

0. 40

0. 38

0. 22

–2 .4

6 2.

99 3.

21 1.

08 –0

.0 4

0. 01

56 61

23 1.

5

S D

0. 39

0. 49

0. 44

0. 35

0. 48

0. 50

10 .5

5 0.

41 0.

49 0.

49 0.

41 3.

99 0.

68 0.

48 0.

56 0.

03 0.

88 15

62 31

.6 9

*p <

. 05

.

N O

T E

: N

= 9

5, 91

9 w

it h

in c

o u

n ty

; N

= 6

7 b

et w

ee n

c o

u n

ti es

.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 16 22-AUG-07 10:10

562 CHIRICOS, BARRICK, BALES & BONTRAGER

Guidelines to offense-specific prior convictions for a felony or misde- meanor as an adult or juvenile by a state, federal, military, or foreign court. As with other research using Guidelines data, this affords an unusu- ally comprehensive measure of an individual’s prior criminal record. Offenses that occurred more than 10 years before the current offense are not scored if the individual has been conviction free for that time. Also included among our predictors of recidivism is whether the individual had previously violated the terms of supervision while on probation or commu- nity control (yes = 1), which can be regarded as another measure of prior record. To correct for skewness, we used natural log values for crime seri- ousness, prior record, and supervision length.

It is possible that being adjudicated or not may be the consequence of uncontrolled factors that also predict an outcome of interest such as recidi- vism. In such a situation, the judge’s decision to adjudicate the offender is not independent of the residual terms in the equation for recidivism, which would result in biased estimates. To control for this “treatment effect” (Moffitt, 1999), which is somewhat analogous to “selection bias,”13 we computed inverse Mills ratios for inclusion in each model estimating recidivism.

The inverse Mills ratios were computed through a two-stage process. First, we estimated a probit model to predict adjudication, including all level 1 independent variables shown in table 1. We also included exclusion restrictions, which are “variables that affect the selection process but not the substantive equation of interest” (Bushway, Johnson, and Slocum, 2007: 152). For this purpose, whether the offender went to trial (1 = yes) is used to predict adjudication but not recidivism, because this variable is arguably relevant for adjudication and not recidivism. In addition, since our data are drawn from 67 counties, we included 66 county dummy vari- ables and only kept those significant county dummy variables for a reduced probit model. The second step calculated the inverse Mills ratio using the predicted value for each offender in the probit equation for adju- dication from the first step, Zδ. Following Bushway, Johnson and Slocum

13. Selection bias (Berk, 1983; Heckman, 1979) refers to a situation wherein not all members of a population are represented inasmuch as some have been “selected” in terms of variables that need to be accounted for when predicting subsequent outcomes. This process applies, for example, to people who are sen- tenced to prison (as opposed to probation) when one is interested in the factors that would predict differential sentence length. In other instances, all members of a population are represented, but some are accorded one “treatment” as opposed to another (Moffitt, 1999). This process applies to a situation such as our own in which all persons found guilty of a felony are either adjudicated or have adjudi- cation withheld. The object of controlling for “treatment effect” is to account for the influence of variables that may predict both the treatment experienced and its presumed consequence—in this case, recidivism.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 17 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 563

(2007: 161), the inverse Mills ratio was estimated for each case by dividing the normal density function evaluated at –Zδ by 1 minus the normal cumulative distribution function estimated at –Zδ.

One primary objective of this research is to determine whether “larger social contexts” (Sampson and Laub, 1997) play a role in shaping the potential labeling effects of felony conviction. The social contexts that we examine are characteristics of the county in which offenders experienced their probation supervision. We would prefer to have a lower level of aggregation, but data limitations relative to the situating of offenders by place preclude that possibility. Our second-level variables include total crime rate per 100,000 residents and a measure of police presence, which is the number of police per 100,000 residents. Both of these measures are for the year 2000 (Florida Department of Law Enforcement, 2005). Our mea- sure of CD uses 2000 Census data for percent black, percent of families receiving public assistance, percentage of persons living below the poverty level, and the percentage of families headed by a single mother. Factor loadings were used to weight the variables included in the CD index. Sev- eral other variables failed to load on this factor, including unemployment rate, percent Hispanic, and percent of the population under 18 years of age.

To operationalize the ICE, we follow Kubrin and Stewart (2006) and Massey (2001). For a given county, the ICE index is computed as [(number of affluent families – number of poor families) / total number of families]. In this context, “affluent” is defined as families that are 2 standard devia- tions above the mean family income and “poor” is defined as families below the officially designated poverty level. We expect each of the second level variables to potentially interact with felony conviction in the specifi- cation of labeling effects. The means and standard deviations for these variables, as well as their bivariate correlations with first-level and second- level predictors are described in table 1.

ANALYTIC STRATEGY

HLM represents an improvement over traditional regression techniques when analyzing “nested” or multilevel data. According to Raudenbush and Bryk (2002), multilevel analysis using traditional regression modeling risks violation of two regression assumptions, the independence of error terms and homoscedasticity. Individuals nested within second-level con- texts—in our case, counties—will no longer represent independent obser- vations because they may share many of the same characteristics. This analysis results in correlated error terms and biased estimates of standard errors. HLM corrects for this problem “by incorporating into the statistical model a unique random effect for each organizational [county level] equa- tion” (Bryk and Raudenbush, 1992: 84). Second, tests of significance for

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 18 22-AUG-07 10:10

564 CHIRICOS, BARRICK, BALES & BONTRAGER

second-level variables are accomplished in HLM by adjusting the degrees of freedom to represent the number of second-level units in the analysis. HLM also allows researchers to test the cross-level effects of second-level factors on individual-level relationships. The improved estimation of inde- pendent effects and cross-level interactions are of primary interest in the current research.

As discussed by Raudenbush and Bryk (2002), the use of traditional HLM with a dichotomous dependent variable such as recidivism would not be appropriate for two reasons. First, the assumption that error terms are normally distributed is violated with binary outcome data. Also vio- lated is the assumption of linearity in the relationship between the inde- pendent and the dependent variables and the assumption that the variance of the error terms be distributed equally across all values of the dependent variable. To overcome these problems, hierarchical generalized linear modeling (HGLM) is used with these data. All level 1 and level 2 variables in the analysis are grand mean centered with the exception of dichoto- mous measures. Both Raudenbush and Bryk (2002) and Hox (2002) observe that grand mean centering as opposed to group mean centering facilitates the interpretation of coefficients, especially at level 1 and that, without serious theoretical justification for doing otherwise, is the pre- ferred choice. Hox further notes that not using grand mean centering in models with cross-level interactions can lead to “serious interpretation problems” (2002: 56). In all analyses, unit-specific models with robust stan- dard errors are used. Nonsignificant variance components are specified as fixed.

RESEARCH FINDINGS

UNCONDITIONAL MODEL

The unconditional model is estimated primarily to measure the magni- tude of variation in recidivism across counties. Our variance component estimate, .037, is significant (p < .001), which is a common indicator that additional modeling using second-level predictors is warranted. However, the intraclass correlation of .006 suggests that although between county variation is significant, it is clearly not substantial. Nonetheless, we will proceed with second-level analyses.

LEVEL 1 RESULTS

The second stage of analysis estimates recidivism as a function of indi- vidual-level predictors. Results of the level 1 conditional modeling are described in table 2, which shows the effect of individual-level predictors on recidivism and the degree to which those effects vary across counties

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 19 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 565

(variance component). All individual-level predictors are significantly related to recidivism except for crime seriousness.

In table 2, we see that odds of recidivism for property offenders are 72 percent greater than violent offenders and that the odds that black proba- tioners will fail are 36 percent higher than for whites. Those who had vio- lated the terms of a previous probation have odds of recidivism that are 22 percent greater than for those who had not, and males are 38 percent more likely to fail than females. The odds of recidivism are lower for Hispanics than for non-Hispanic whites and are reduced for those with longer terms of probation supervision. Most importantly, table 2 shows that indepen- dent of the effects of all other predictors, having been convicted of a fel- ony increases the odds of recidivism by 17 percent when compared with those who had adjudication withheld.

Table 2. Level 1 Conditional HGLM Model of Recidivism

Odds Variance Independent Variable Coefficient SE Ratio Component

Intercept –2.158 .061 .081*** .092*** Adjudicated .155 .032 1.167*** .016* Male .320 .035 1.377*** .028*** Age at sentencing –.044 .002 .957*** .000* Hispanic –.082 .041 .921* a

Black .308 .034 1.360*** .020** Property .543 .037 1.721*** .018* Drug .386 .046 1.470*** .042*** Prior supervision violation .199 .049 1.220*** .033* Crime seriousness (Ln) .037 .027 1.037 a

Prior record points (Ln) .087 .022 1.091*** .007* Supervision length (Ln) –.218 .031 .804*** .017** Inverse Mills ratio –.472 .107 .624*** .156**

*p < .05; **p < .01; ***p < .001. a Nonsignificant variance components have been fixed. NOTE: N = 95,919 within county; N = 67 between counties.

As noted, both theory and prior research (sometimes contradictory) have raised the possibility that labeling effects may vary by a person’s race, sex, prior record, and stakes in conformity. To assess this possibility, we repeat the level 1 modeling described above with interaction terms involving adjudication status and the aforementioned individual attributes. Typically, stakes in conformity is operationalized in terms of things that an individual could potentially lose, such as a marital bond or employment as

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 20 22-AUG-07 10:10

566 CHIRICOS, BARRICK, BALES & BONTRAGER

the result of criminal behavior. Unfortunately, those particular data (as well as educational attainment) are not available for Florida probationers.

However, a theoretical analog to stakes in conformity is Hirschi’s (1969) “commitment to conventional lines of action.” It is presumed to work the same way as stakes in conformity, inasmuch as it provides an individual with “something to lose” in the event of delinquent or criminal behavior. It seems reasonable to suggest that individuals who have achieved the age of 30 years and have avoided a criminal or delinquent record (up to the current instance) have established more of a commitment to conventional behavior than those who are younger with a criminal record. As such, they have acquired a stake in conformity that threatens at least the loss of repu- tation and status as a consequence of criminal behavior. Thus, we measure “commitment” or stakes in conformity in terms of age and prior record. For our purpose, “high stakes” individuals are those who are 30+ years of age without a prior conviction (felony or misdemeanor) as an adult or juvenile.

Table 3 shows the recidivism model with all level 1 estimators and with the addition of four interaction terms involving adjudication and offender characteristics. All level 1 predictors, except Hispanic, and crime serious- ness in three of four models, have a significant relationship with recidi- vism. The effect of adjudication on recidivism is significantly higher for females than for males. Table 3 also shows that the likelihood of recidi- vism, given adjudication is significantly higher for white offenders than for others. Prior record (yes/no) has no significant interaction with adjudica- tion status in predicting recidivism. With regard to our measure of high stakes in conformity, table 3 shows that probationers who had achieved the age of 30 years without a criminal or delinquent conviction are signifi- cantly more likely to recidivate given adjudication than those with a prior record before the age of 30 years.14

SECOND-LEVEL RESULTS

In this stage of our analysis, the slope for the relationship between recidivism and adjudication is modeled using county-level contextual vari- ables in conjunction with individual-level predictors. Strong collinearity problems are encountered when the ICE and CD are included in the same

14. In separate models (not shown, results available on request), the effect of adjudi- cation is significant for men as well as for women; for Hispanics, blacks, and whites; for those with and without a prior record; and for those with and without high stakes. This finding suggests a robustness in the consequences of adjudica- tion for recidivism, notwithstanding the significance of interaction terms in table 3 showing the effects of adjudication to be stronger for women, whites, and those with high stakes.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 21 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 567

T ab

le 3

. L

ev el

1 C

o n

d it

io n

al H

G L

M M

o d

el w

it h

A d

ju d

ic at

io n

a n

d O

ff en

d er

C h

ar ac

te ri

st ic

In te

ra ct

io n

s

S ex

P ri

o r

R ec

o rd

S ta

ke s

in C

o n

fo rm

it y

R ac

e

In d

ep en

d en

t V

ar ia

b le

C o

ef fi

ci en

t O

d d

s R

at io

C o

ef fi

ci en

t O

d d

s R

at io

C o

ef fi

ci en

t O

d d

s R

at io

a C

o ef

fi ci

en t

O d

d s

R at

io

In te

rc ep

t –2

.5 75

.0 76

** *

–2 .5

46 .0

78 **

* –2

.4 77

.0 84

** *

–2 .5

43 .0

79 **

* A

d ju

d ic

at ed

.2 85

1. 33

0* **

.1 52

1. 16

4* **

.0 67

1. 06

9 .0

57 1.

05 9

M al

e .4

08 1.

50 3*

** .3

16 1.

37 2*

** .2

66 1.

30 5*

** .3

17 1.

37 2*

** A

ge a

t se

n te

n ci

n g

–. 04

4 .9

57 **

* –.

04 5

.9 56

** *

–. 05

9 .9

43 **

* –.

04 4

.9 57

** *

H is

p an

ic –.

07 9

.9 24

–. 08

1 .9

23 –.

02 9

.9 72

–. 01

8 .9

82 B

la ck

.3 09

1. 36

2* **

.3 16

1. 37

2* **

.3 58

1. 43

0* **

.3 80

1. 46

2* **

P ro

p er

ty .5

46 1.

72 6*

** .5

64 1.

75 8*

** .5

28 1.

69 5*

** .5

45 1.

72 5*

** D

ru g

.3 87

1. 47

2* **

.4 08

1. 50

4* **

.3 90

1. 47

7* **

.3 90

1. 47

7* **

P ri

o r

su p

er vi

si o

n v

io la

ti o

n .2

18 1.

24 4*

** .1

80 1.

19 8*

* .3

61 1.

43 5*

** .2

17 1.

24 2*

** C

ri m

e se

ri o

u sn

es s

(L n

) .0

43 1.

04 4

.0 36

1. 03

7 .1

20 1.

12 7*

** .0

40 1.

04 1

P ri

o r

re co

rd p

o in

ts (

L n

) .0

95 1.

10 0*

** .0

60 1.

06 1*

* .1

02 1.

10 7*

* .0

93 1.

09 7*

** S

u p

er vi

si o

n l

en gt

h (

L n

) –.

21 8

.8 04

** *

–. 24

6 .7

82 **

* –.

26 0

.7 71

** *

–. 21

6 .8

06 **

* In

ve rs

e M

il ls

r at

io –.

45 0

.6 38

** *

–. 59

1 .5

54 **

* –.

15 2

.8 59

–. 45

6 .6

34 **

* M

al e

* A

d j

–. 20

4 .8

16 **

* N

o p

ri o

r p

ts *

A d

j .0

20 1.

02 0

H ig

h s

ta k

es *

A d

j .2

03 1.

22 5*

* W

h it

e *

A d

j .1

45 1.

15 6*

**

*p <

. 05

; **

p <

. 01

; **

*p <

. 00

1. a T

h es

e re

su lt

s ar

e fr

o m

a r

ed u

ce d

s am

p le

i n

cl u

d in

g o

n ly

t h

o se

w it

h h

ig h

o r

lo w

s ta

k es

i n

c o

n fo

rm it

y (N

= 4

1, 62

7) .

N O

T E

S :

N o

n si

gn if

ic an

t va

ri an

ce c

o m

p o

n en

ts h

av e

b ee

n f

ix ed

; N

= 9

5, 91

9 w

it h

in c

o u

n ty

; N

= 6

7 b

et w

ee n

c o

u n

ti es

.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 22 22-AUG-07 10:10

568 CHIRICOS, BARRICK, BALES & BONTRAGER

models. Like Kubrin and Stewart (2006), we resolve that issue by develop- ing estimates that keep these variables separate. However, whereas Kubrin and Stewart only examined those level 2 predictors, our estimates also include overall crime rates and police presence. These models origi- nally include all level 1 and level 2 predictors. The latter are introduced sequentially in two different estimates in order to keep ICE and CD sepa- rate. The sequence for introduction is as follows: ICE, crime rate, and police presence and, alternatively, CD, crime rate, and police presence. Following the criteria articulated by Raudenbush and Bryk (2002: 268), we first added the level 2 variables sequentially to the intercept model and retained only those that were significant. Then we sequentially added the level 2 variables to the adjudicated equation and again only retained sig- nificant variables.15

Table 4a. Combined Level 1 and Level 2 HGLM Model of Recidivism (ICE)

Independent Variable Coefficient SE Odds Ratio

Intercept –2.546 .061 .078*** ICE 1.298 .517 3.662*

Adjudicated .154 .032 1.167*** Male .314 .035 1.369*** Age at sentencing –.045 .002 .956*** Hispanic –.083 .041 .921* Black .312 .034 1.366*** Property .552 .036 1.738*** Drug .390 .046 1.477*** Prior supervision violation .184 .054 1.202** Crime seriousness (Ln) .030 .028 1.031 Prior record points (Ln) .078 .024 1.082** Supervision length (Ln) –.216 .032 .806*** Inverse Mills ratio –.509 .118 .601***

*p < .05; **p < .01; ***p < .001. NOTES: Nonsignificant variance components have been fixed; N = 95,919 within county; N = 67 between counties.

15. Raudenbush and Bryk note that “the analyst will usually want to develop a tenta- tive model for the intercept . . . before proceeding to fit models for the random slopes” (2002: 267) and then, “the most direct evidence of whether a level-2 pre- dictor should be included is the magnitude of its estimated effect and related t-ratio. Predictors with t-ratios near or less than 1 are obvious candidates for exclusion from the model” (2002: 268).

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 23 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 569

Table 4b. Combined Level 1 and Level 2 HGLM Model of Recidivism (CD)

Independent Variable Coefficient SE Odds Ratio

Intercept –2.530 .061 .080*** Concentrated Disadvantage –.060 .026 .942*

Adjudicated .157 .031 1.170*** Male .311 .036 1.364*** Age at sentencing –.044 .002 .957*** Hispanic –.082 .042 .921* Black .314 .035 1.369*** Property .549 .036 1.731*** Drug .389 .046 1.475*** Prior supervision violation .192 .051 1.211** Crime seriousness (Ln) .035 .026 1.036 Prior record points (Ln) .083 .022 1.087** Supervision length (Ln) –.218 .031 .804*** Inverse Mills ratio –.486 .107 .615***

*p < .05; **p < .01; ***p < .001. NOTES: Nonsignificant variance components have been fixed; N = 95,919 within county; N = 67 between counties.

Tables 4a and b show the results of estimating recidivism using all level 1 predictors, including adjudication status and any significant level 2 predictors. We are concerned with the direct effects of county-level con- texts on recidivism (intercept model) and more importantly, with any pos- sible interactions of level 2 variables and adjudication. Crime rates and police presence have no significant relationships either directly with recidi- vism or interactively with adjudication status. Neither the ICE nor CD interacts significantly with adjudication in the prediction of recidivism. However, concentrated extremes of affluence and poverty are associated with higher levels of recidivism across Florida counties, whereas CD is associated with slightly lower levels of recidivism.

The lack of interactive effects involving adjudication and level 2 predictors suggests that being convicted as a felon increases the likelihood of recidivism regardless of the characteristics of counties in which proba- tioners are supervised. As with the individual-level traits discussed above (table 3), this further illustrates the apparent robustness of the effects of a felony convict label inasmuch as those effects are not conditioned either by individual or “larger social contexts.” In addition, although not a focus of this research, our data for Florida counties show opposite direct effects of ICE and CD on recidivism than what Kubrin and Stewart (2006) report

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 24 22-AUG-07 10:10

570 CHIRICOS, BARRICK, BALES & BONTRAGER

for Oregon census tracts.16

DISCUSSION As noted, this study is the first one to examine the effects of being

labeled a convicted felon on the recidivism of adults. The option of with- holding adjudication of guilt and the label of convicted felon for those who are legally guilty provides a unique opportunity to explore the conse- quences of labeling within the criminal justice system. The decision to withhold adjudication involves a situation wherein legally equivalent adults—in terms of guilt—are either labeled as a felon or not. In terms of social and legal significance, a more consequential label is not available in the criminal court system. Our principal findings can be briefly summarized.

1. Being adjudicated guilty as a felon significantly and substantially increases the likelihood of recidivism in comparison with those who have adjudication withheld.

2. The effect of being adjudicated guilty on recidivism is stronger for whites than for blacks and Hispanics and for females as opposed to males. It is also stronger for those who reach the age of 30 years without any prior convictions compared with those with priors before turning 30 years old.

3. The community-level contexts that we examined have no conse- quence for the relationship between adjudication of guilt and recidivism.

At a practical level, these results underscore the potential benefits that

16. There are several methodological differences between the Kubrin and Stewart (2006) study and our own. Their study was performed in Portland, Oregon and ours in Florida. The proportion of non-Hispanic whites in Multnomah County (Portland) was substantially higher (79.4 percent) in the 2000 Census than in Florida (66.5 percent). Their sample, which included misdemeanants as well as felons, was 68 percent white, 25 percent black, 4 percent Hispanic, and 3 percent other. Our sample is 49 percent white, 36 percent black, and 14 percent Hispanic (others excluded). Their measure of recidivism is arrest, whereas ours is reconvic- tion for a felony. More important, perhaps, their data for ICE and CD were aggregated at the census tract level and ours at the county level. At the lower level of aggregation, it makes sense to expect the relative presence of affluence to provide for both social capital and collective efficacy in ways that might reduce recidivism overall. But at the county level, the affluent and poor are likely not living near one another and the potential benefits of having affluence nearby may not work in the same way. In fact, the social and physical distance that may char- acterize the affluent and poor in a county may work in ways that afford the afflu- ent the opportunity and means to more vigorously police poor neighborhoods and prosecute those who may be apprehended. This would not only mitigate the recidivism retarding effects of affluence noted by Kubrin and Stewart, but it may also aggravate the prospects of criminal activity resulting in a formal charge and conviction—our measure of recidivism.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 25 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 571

can accrue to individuals and to society—in this case, from less criminal activity—when a policy of minimizing harm (Clear, 1995; Rubin, 1999) is implemented in the administration of justice. The withholding of adjudica- tion for those who have been found guilty of a felony is a policy that these data suggest is directly related to lower levels of recidivism than are found among those who are formally labeled as a convicted felon. In this case, minimizing harm at the individual level has consequences for reducing harm in the broader community.

At a theoretical level, the pattern of findings suggests that the conse- quences of labeling are greater for those who would otherwise be consid- ered less likely to recidivate. As noted, research has generally shown that recidivism is more likely for males, minorities, and persons with a more extensive prior record (Benedict and Huff-Corzine, 1997; Kruttschnitt, Uggen, and Shelton, 2000; Spohn and Holleran, 2002; Whitehead, 1991). These data suggest that those who are generally more likely to recidivate are less disadvantaged by a formal label than are women, whites, and those without an early prior record. It could also be argued that women, whites, and those without an early prior record have more to lose from criminal stigma because crime is more often associated with males, minori- ties, and those with extensive criminal experience.

It also happens that the pattern of results reported here is consistent with some of the theoretical expectations raised earlier in this article. Spe- cifically, the diminished relevance of the felony convict label for black and Hispanic offenders was anticipated, even though prior research is both limited and inconsistent in this regard. For example, Harris (1976: 433) hypothesized that the effects of official labeling would be least conse- quential for those most excluded from “ascriptive membership” in domi- nant groups, whereas Paternoster and Iovanni (1989: 382) hypothesized that those who “grant less legitimacy to the legal order and its agents”— presumably including racial and ethnic minorities—could be “more imper- vious to labeling effects.” Assuming that, on average, whites are less disad- vantaged than blacks, our results run counter to the expectation raised by Sampson and Laub’s (1997: 153) that deficits would “pile up faster” for those who are already disadvantaged. These results also diverge from those of Bernburg and Krohn (2003) who found that labeling effects were stronger for African-American youth than for others. It is possible that the race-specific effects of formal labeling are different for adults than for juveniles, which is an issue that future research should carefully explore.

The fact that a felony convict label more strongly predicts recidivism for females than for males is consistent with the scant theorizing that has been done about labeling effects for women, but it is inconsistent with a limited body of empirical evidence showing that labeling effects are more conse- quential for males. Messerschmidt (1993: 4) hypothesized that because

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 26 22-AUG-07 10:10

572 CHIRICOS, BARRICK, BALES & BONTRAGER

men exercise greater power in society than women, “men should . . . have greater opportunity to counteract official labeling.” Giordano, Cernkovich, and Lowery (2004: 189) hypothesized that the greater social stigma attached to “antisocial” behavior by females could “be more limit- ing to life chances/opportunities for a return to conventional roles” and thus more conducive to recidivism. Although our results are consistent with those expectations, the process whereby labeling comes to have more consequence for women than for men cannot be known from these data.

The finding that labeling effects are stronger for those who, in Hirschi’s (1969) terms, have established a more substantial “commitment to conven- tional action” by having reached at least 30 years of age without a prior record is consistent with the “greater vulnerability” labeling hypothesis (Sherman et al., 1992). In that view, it is expected that labeling would be more consequential for those with a higher stake in conformity and who, on that account, presumably “care more about the opinions of conven- tional society” (1992: 682). Another way to interpret this result is that, as Paternoster and Iovanni (1989) hypothesized, labeling experiences have diminishing consequences for those persons who have been formally labeled earlier in life.

Although not part of our original theoretical framework, it is worth not- ing that finding less recidivism for those having adjudication withheld is consistent with expectations from a “reintegrative shaming” perspective. As described by Braithwaite (1989: 55), reintegrative shaming combines “expressions of community disapproval” for a behavior with “gestures of reacceptance into the community of law-abiding citizens.” One who has been found guilty of a felony and sentenced to probation has clearly received an expression of community disapproval. At the same time, not being labeled a convicted felon (having adjudication withheld) reduces the prospect of formal and informal exclusions from the community of law- abiding citizens that traditionally accompany the label of convicted felon.

In fact, Braithwaite (1989: 101) juxtaposes and contrasts reintegrative shaming with the central labeling concept of “stigmatization,” which he describes as “disintegrative shaming in which no effort is made to recon- cile the offender with the community.” What the decision to impose or withhold adjudication of guilt may amount to is a choice between stigmati- zation and reintegration—or at least diminished disintegration. The avoid- ance of the label in this instance not only reduces the potentially stigmatizing effects of labeling, but it may at the same time increase the reintegrative potential of the punitive response for the individual involved.

The lack of an interaction between second-level variables and adjudica- tion status in these analyses could mean one of three things. On the one hand, it could indicate that the effect of adjudication on recidivism is suffi- ciently robust that it holds in a variety of community contexts. Thus, it is

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 27 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 573

seemingly unaffected by levels of crime, police presence, CD, and an ICE of affluence and poverty. Alternatively, as the intraclass correlation sug- gests, there may be insufficient variation in recidivism across counties to matter for these purposes. Finally, it may be the case that community con- text is inadequately operationalized with data that are aggregated at the county level. Ideally, an assessment of the impact of community character- istics on the relationship between adjudication as a felon and recidivism would make use of neighborhood-level data, aggregated at the census tract or block level. Unfortunately, we cannot locate our probationers more precisely than by the county in which their probation supervision took place.

Another limitation of the current research is the potential for omitted variable biases at both levels of analysis. At the individual level, it would have been helpful if available data included socioeconomic status as well as three additional indicators of stakes in conformity— employment, mari- tal status, and educational attainment. At the second level of aggregation, measures of probation caseload could have relevance for the likelihood of recidivism, as could measures of social capital within different social contexts.

In sum, this research has shown that when adults are adjudicated guilty of a felony, a significant and substantial increase occurs in the likelihood of reconviction within 2 years. When combined with the recent results reported by Bernburg and his colleagues (Bernburg and Krohn, 2003; Bernburg, Krohn, and Rivera, 2006) involving the labeling of juveniles, these data provide additional encouragement for the renewed empirical assessment of labeling theory. Moreover, research moving forward in this area should pay particular attention to whether and how labeling effects vary for different kinds of individuals and for broader social contexts that may be operationalized more finely than was possible in the current research.

REFERENCES

Ageton, Suzanne S., and Delbert Elliott. 1974. The effects of legal process- ing on delinquent orientations. Social Problems 22:87–100.

Agnew, Robert. 2006. Pressured Into Crime: An Overview of General Strain Theory. Los Angeles, CA: Roxbury Publishing.

Akers, Ronald L. 1998. Social Learning and Social Structure: A General Theory of Crime and Deviance. Boston, MA: Northeastern University Press.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 28 22-AUG-07 10:10

574 CHIRICOS, BARRICK, BALES & BONTRAGER

Anderson, Elijah. 1999. Code of the Street. New York: W. W. Norton and Company.

Barrick, Kelle. 2005. Labeling and recidivism: A quantitative assessment of the evidence. Presented at the annual meeting of the American Soci- ety of Criminology, Toronto, Ontario, Canada, November.

Baumer, Eric. 1997. Levels and predictors of recidivism: The Malta experi- ence. Criminology 35:601–28.

Benedict, W. Reed, and Lin Huff-Corzine. 1997. Return to the scene of the punishment: Recidivism of adult male property offenders on felony probation, 1986-1989. Journal of Research in Crime and Delinquency 34:237–53.

Berk, Richard A. 1983. An introduction to sample selection bias in socio- logical data. American Sociological Review 48:386–98.

Berk, Richard A., Alec Campbell, Ruth Klap, and Bruce Western. 1992. The deterrent effect of arrest in incidents of domestic violence: A Bayesian analysis of four field experiments. American Sociological Review 57:698–708.

Bernburg, Jon Gunnar. 2002. State reaction, life-course outcomes and structural disadvantage: A panel study of the impact of formal criminal labeling on the transition to adulthood. PhD Dissertation. University at Albany, State University of New York.

Bernburg, Jon Gunnar, and Marvin D. Krohn. 2003. Labeling, life chances and adult crime: The direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology 41:1287–318.

Bernburg, Jon Gunnar, Marvin D. Krohn, and Craig J. Rivera. 2006. Offi- cial labeling, criminal embeddedness, and subsequent delinquency: A longitudinal test of labeling theory. Journal of Research in Crime and Delinquency 43:67–88.

Blackwell, Brenda Sims, and Sarah Eschholz. 2002. Sex differences and rational choice: Traditional tests and new directions. In Rational Choice and Criminal Behavior, eds. Alex R. Piquero and Stephen G. Tibbetts. New York: Routledge.

Braithwaite, John. 1989. Crime, Shame and Reintegration. Cambridge, UK: Cambridge University Press.

Bryk, Anthony S., and Stephen W. Raudenbush. 1992. Hierarchical Linear Models. Newbury Park, CA: Sage.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 29 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 575

Bushway, Shawn, Brian D. Johnson, and Lee Ann Slocum. 2007. Is the magic still there? The use of the Heckman two-step correction for selection bias in criminology. Journal of Quantitative Criminology 23:151-78.

Clear, Todd R. 1995. Harm in American Penology. Albany, NY: SUNY Press.

DeJong, Christine. 1997. Specific deterrence and survival analysis: Inte- grating theoretical and empirical models of recidivism. Criminology 35:561–76.

Fagan, Jefferey A., Aaron Kupchick, and Akiva Liberman. 2003. Be care- ful what you wish for. The comparative impacts of juvenile versus crim- inal court sanctions on recidivism among adolescent felony offenders. New York: Columbia Law School Public Law and Legal Theory Work- ing Group.

Farrington, David. 1977. The effects of public labeling. British Journal of Criminology 17:112–25.

Florida Department of Law Enforcement. 2005. Crime in Florida, Florida Uniform Crime Report, 2005. Tallahassee, FL: Florida Statistical Anal- ysis Center. http://www.fdle.state.fl.us/fsac/data_statistics.asp.

Gendreau, Paul, Tracy Little, and Claire Goggin. 1996. A meta-analysis of the predictors of adult offender recidivism: What works! Criminology 34:575–607.

Gibbs, Jack. 1974. The effects of juvenile legal procedures on juvenile offenders’ self-attitudes. Journal of Research in Crime and Delinquency 11:51–5.

Giordano, Peggy C., Stephan A. Cernkovich, and Allen R. Lowery. 2004. A long-term follow-up of serious adolescent female offenders. In Aggression, Antisocial Behavior and Violence among Girls, eds. Martha Putallaz and Karen L. Bierman. New York: The Guilford Press.

Hagan, John, and Alberto Palloni. 1990. The social reproduction of a crim- inal class in working class London. American Journal of Sociology 96:265–99.

Harris, Anthony. 1975. Imprisonment and the expected value of criminal choice: A specification and test of aspects of the labeling perspective. American Sociological Review 40:71–87.

Harris, Anthony. 1976. Race, commitment to deviance and spoiled iden- tity. American Sociological Review 41:432–42.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 30 22-AUG-07 10:10

576 CHIRICOS, BARRICK, BALES & BONTRAGER

Heckman, James J. 1979. Sample selection bias as a specification error. Econometrica 47:153–161.

Heimer, Karen, and Ross L. Matsueda. 1996. Role taking, role commit- ment and delinquency: A theory of differential social control. Ameri- can Sociological Review 59:365–90.

Hirschi, Travis. 1969. Causes of Delinquency. Berkeley: University of Cali- fornia Press.

Horwitz, Allan, and Michael Wasserman. 1979. The effect of social control on delinquent behavior: A longitudinal test. Sociological Focus 12:53–70.

Hox, Joop. 2002. Multilevel Analyses: Techniques and Applications. Mah- wah, NJ: Lawrence Erlbaum Associates.

Hull, Elizabeth A. 2006. The Disenfranchisement of Ex-Felons. Philadel- phia, PA: Temple University Press.

Jensen, Gary. 1972. Delinquency and adolescent self-conceptions: A study of the personal relevance of infraction. Social Problems 20:84–103.

Kruttschnitt, Candace, Christopher Uggen, and Kelly Shelton. 2000. Predictors of desistence among sex offenders: The interaction of for- mal and informal social controls. Justice Quarterly 47:61–87.

Kubrin, Charis E., and Eric A. Stewart. 2006. Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology. 44:165–98.

Lofland, John. 1969. Deviance and Identity. Englewood Cliffs, NJ: Pren- tice-Hall.

Manza, Jeff, and Christopher Uggen. 2006. Locked Out: Felon Disen- franchisement and American Democracy. New York: Oxford Univer- sity Press.

Massey, Douglas S. 2001. Ecology comes back to sociology. In Does It Take a Village? Community Effects on Children, Adolescents and Fami- lies, eds. Alan Booth and Ann Crouter. Mahwah, NJ: Lawrence Erlbaum.

Matsueda, Ross L. 1992. Reflected appraisals, parental labeling and delin- quency: Specifying a symbolic interactionist theory. American Journal of Sociology 97:1577–611.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 31 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 577

Mauer, Marc, and Meda Chesney-Lind. 2002. Invisible Punishment: The Collateral Consequences of Mass Imprisonment. New York: The New Press.

Messerschmidt, James W. 1993. Masculinities and Crime. Lanham, MD: Rowman and Littlefield Publishers, Inc.

Moffitt, Robert A. 1999. New developments in econometric methods for labor market analysis. In Handbook of Labor Economics, vol. III, eds. Orley Ashenfelter and David Card. North Holland, NY: Elsevier Science.

Morenoff, Jeffrey D., Robert J. Sampson, and Stephen W. Raudenbush. 2001. Neighborhood inequality, collective efficacy and the spatial dynamics of urban violence. Criminology 39:517–59.

Nagin, Daniel S. 1998. Criminal deterrence research at the outset of the twenty-first century. In Crime and Justice: A Review of Research, vol. XXIII, ed. Michael Tonry. Chicago, IL: University of Chicago Press.

Palamara, Frances, Francis T. Cullen, and Joanne C. Gertsten. 1986. The effect of police and mental health intervention on juvenile deviance: Specifying contingencies in the impact of formal reaction. Journal of Health and Social Behavior 27:90–105.

Pate, Antony M., and Edwin E. Hamilton. 1992. Formal and informal deterrents to domestic violence: The Dade County spouse assault experiment. American Sociological Review 57:691–97.

Paternoster, Raymond. 1987. The deterrent effect of the perceived cer- tainty and severity of punishment: A review of the evidence and issues. Justice Quarterly 4:173–217.

Paternoster, Raymond, and Leeann Iovanni. 1989. The labeling perspec- tive and delinquency: An elaboration of the theory and an assessment of the evidence. Justice Quarterly 6:359–93.

Petersilia, Joan. 1985. Probation and felony offenders. Federal Probation 49:4–9.

Raudenbush, Stephen, and Anthony Bryk. 2002. Hierarchical Linear Mod- els: Applications and Data Analysis Methods. Thousand Oaks, CA: Sage.

Ray, Melvin C., and William R. Downs. 1986. An empirical test of labeling theory using longitudinal data. Journal of Research in Crime and Delin- quency 23:169–94.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 32 22-AUG-07 10:10

578 CHIRICOS, BARRICK, BALES & BONTRAGER

Richards, Pamela, and Charles R. Tittle. 1981. Gender and perceived chances of arrest. Social Forces 59:1182–99.

Rountree, Pamela Wilcox, Kenneth C. Land, and Terance D. Miethe. 1994. Macro-micro integration in the study of victimization: A hierar- chical logistic model analysis across Seattle neighborhoods. Criminol- ogy 32:387–414.

Rubin, Edward L. 1999. Minimizing Harm: A New Crime Policy for America. Boulder, CO: Westview Press.

Sampson, Robert J., and John H. Laub. 1993. Structural variation in juve- nile court processing: Inequality, the underclass and social control. Law and Society Review 27:285–311.

Sampson, Robert, and John H. Laub. 1997. A life-course theory of cumu- lative disadvantage and the stability of delinquency. In Developmental Theories of Crime and Delinquency, ed. Terence P. Thornberry. New Brunswick, NJ: Transaction Publishers.

Sampson, Robert J., Jeffrey D. Morenoff, and Felton Earls. 1999. Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review 64:644–60.

Sampson, Robert J., Jeffrey D. Morenoff, and Thomas Gannon-Rowley. 2002. Assessing neighborhood effects: Social processes and new direc- tions in research. Annual Review of Sociology 28:443–78.

Sampson, Robert J., Stephen Raudenbush, and Felton Earls. 1997. Neigh- borhoods and violent crime: A multilevel study of collective efficacy. Science 277:918–24.

Schur, Edwin M. 1983. Labeling Women Deviant. Philadelphia, PA: Tem- ple University Press.

Shakur, Sanyika. 1993. Monster: The Autobiography of an L.A. Gang Member. New York: The Atlantic Monthly Press.

Sherman, Lawrence W. 1993. Defiance, deterrence and irrelevance: A the- ory of the criminal sanction. Journal of Research in Crime and Delin- quency 30:445–73.

Sherman, Lawrence W., Douglas A. Smith, Janell D. Schmidt, and Dennis P. Rogan. 1992. Crime, punishment and stake in conformity: Legal and informal control of domestic violence. American Sociological Review 57:680–90.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 33 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 579

Silver, Eric. 2000. Extending social disorganization theory: A multi-level approach to the study of violence among persons with mental illnesses. Criminology 38:1043–73.

Smith, Douglas A., and Patrick Gartin. 1989. Specifying specific deter- rence: The influence of arrest on future criminal activity. American Sociological Review 54:94–106.

Smith, Douglas A., and Raymond Paternoster. 1990. Formal processing and future delinquency: Deviance amplification as selection artifact. Law and Society Review 24:1109–31.

Spohn, Cassia, and David Holleran. 2002. The effect of imprisonment on recidivism rates of felony offenders: A focus on drug offenders. Crimi- nology 40:329–57.

Stafford, Mark, and Mark Warr. 1993. A reconceptualization of general and specific deterrence. Journal of Research in Crime and Delinquency 30:123–35.

Sung, Hung-En. 2003. Differential impact of deterrence vs. rehabilitation as drug interventions on recidivism after 36 months. Journal of Offender Rehabilitation 37:95–108.

Taxman, Faye S., and Alex Piquero. 1998. On preventing drunk driving recidivism: An examination of rehabilitation and punishment approaches. Journal of Criminal Justice 26:129–43.

Thistlethwaite, Amy, John Wooldredge, and David Gibbs. 1998. Severity of dispositions and domestic violence recidivism. Crime and Delin- quency 44:388–98.

Tittle, Charles. 1972. Institutional living and self esteem. Social Problems 20:65–77.

Tittle, Charles. 1995. Control Balance: Toward a General Theory of Devi- ance. Boulder, CO: Westview.

Ulmer, Jeffery T., and Brian Johnson. 2004. Sentencing in context: A mul- tilevel analysis. Criminology 42:137–77.

Ventura, Lois A., and Gabrielle Davis. 2005. Domestic violence: Court case conviction and recidivism. Violence Against Women 11:255–77.

Warr, Mark. 2001. The social origins of crime: Edwin Sutherland and the theory of differential association. In Explaining Crime and Criminals, eds. Raymond Paternoster and Ronet Bachman. Los Angeles, CA: Roxbury Publishing Company.

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 34 22-AUG-07 10:10

580 CHIRICOS, BARRICK, BALES & BONTRAGER

West, Donald, and David Farrington. 1973. Who Becomes Delinquent? London: Heinemann Educational.

Western, Bruce. 2002. The impact of incarceration on wage mobility and inequality. American Sociological Review 67:477–98.

Western, Bruce, and Becky Pettit. 2005. Black-white wage inequality, employment rates and incarceration. American Journal of Sociology 111:553–78.

Whitehead, John T. 1991. The effectiveness of felony probation: Results from an eastern state. Justice Quarterly 8:525–43.

Wilson, William J. 1987. The Truly Disadvantaged. Chicago, IL: The Uni- versity of Chicago Press.

Word, David L., and Colby R. Perkins. 1996. Building a Spanish surname list for the 1990s—a new approach to an old problem. U.S. Census Bureau Technical Working Paper No. 13. Washington, DC: U.S. Department of Commerce.

Ted Chiricos is the William Julius Wilson Professor of Criminology in the College of Criminology and Criminal Justice at Florida State University. His research interests include the effects of race and social threat on jus- tice outcomes, as well as the various factors that contribute to punitiveness in American culture.

Kelle Barrick is a doctoral student in the College of Criminology and Criminal Justice at Florida State University. Her current interests include criminological theory; the links among race, class, gender, and crime; and corporate crime. She is also the Richard Rachin Fellow for the Journal of Drug Issues.

William Bales is an associate professor in the College of Criminology and Criminal Justice at Florida State University. His interests include sentenc- ing research, assessing the effectiveness and consequences of punishment and correctional strategies, and community reentry issues among incarcer- ated adult and juvenile populations.

Stephanie Bontrager recently received her Ph.D. degree in criminology from Florida State University. She is a senior research analyst with the

\\server05\productn\C\CRY\45-3\CRY305.txt unknown Seq: 35 22-AUG-07 10:10

THE LABELING OF CONVICTED FELONS 581

Justice Research Center in Tallahassee, Florida and serves as the project director for the National Council on Crime and Delinquency Tracking and Follow-up Evaluation. Her research interests include the relationship between social threat and social control, the interaction between macro forms of social control, and the effects of economic, racial, and gender inequality on crime.