![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
Little is known about the predictors of sentencing for the typical female offender—one who commits a misdemeanor or lesser offense. Moreover, although ample discussions of racial/ethnic disparity in sentencing may be found in the extant literature, most researchers have focused on what happens to males who commit felonies. Thus, to help fill a void I examine the likelihood of receiving a jail sentence among a sample of cases for female misdemeanants. All were convicted in New York City's Criminal Court. I account for direct and indirect effects by estimating a causal model that predicts the sentencing outcome. Race/ethnicity did not directly affect sentencing. Indirect effects, however, were found. Black and Hispanic females were more likely to receive jail sentences than their White counterparts due to differences in socio‐economic status, community ties, prior record, earlier case processing, and charge severity.
Acknowledgments
I am especially grateful to Alan J. Lizotte for assistance with earlier versions of this manuscript and for longstanding support of my work. Cassia Spohn and Dennis W. (Denn) Roncek also provided helpful comments on earlier drafts and I thank them for their guidance. Thanks are also due to the New York City Criminal Justice Agency (CJA), the New York Division of Criminal Justice Services, the New York City Police Department, the New York State Office of Court Administration, and the New York City Department of Correction, who provided the data for this study but have no responsibility for the methods of analysis used in this paper or its conclusions. An earlier version of this manuscript was prepared with the help of a generous Junior Faculty Research Grant from the University of North Carolina at Charlotte. Finally, I would like to thank the editors of Justice Quarterly and three anonymous reviewers for their valuable comments and suggestions.
Notes
1. The exclusion of Hispanics (or the failure to separately control for Hispanics apart from Blacks) has been a major criticism of many studies of sentencing. In Spohn's (Citation2000, p. 454) recent review of the extant sentencing literature, only 10 of 32 state‐level studies included Hispanics in their samples, and five of these investigations were based on the same or similar data. Of the remaining studies, Hispanics were either excluded from the analyses or were included with Black offenders (i.e., in a non‐White category). Similarly, Steffensmeier and Demuth (Citation2000, p. 706) argue that the lack of attention to Hispanic or Latino defendants is a “glaring omission,” given that Hispanic Americans make up roughly 10 percent of the US population and are projected, by the year 2010, to become the largest minority group in the United States (see also Demuth, Citation2003). As we shall see, there is reason to believe that sentencing outcomes may differ for Blacks and Hispanics (Holmes, Hosch, Daudistel, Perez, & Graves, Citation1996; LaFree, Citation1985; Spohn & DeLone, Citation2000; Zatz, Citation1984, Citation1985, Citation2000). In this study, cases for Hispanics and Blacks are controlled separately.
2. Although the results of studies that have tested the liberation hypothesis are not entirely consistent, many find that the effect of race/ethnicity is either confined to or is more pronounced in less serious cases (Crawford, Chiricos, & Kleck, Citation1998; Spohn & Cederblom, Citation1991; Spohn & DeLone, Citation2000; Wooldredge, Citation1998) or more ambiguous cases (Baldus, Woodworth, & Pulaski, Citation1990; Unnever & Hembroff, Citation1988).
3. There is, however, a growing body of literature that suggests that sentencing disadvantage, and possibly negative stereotypes, apply to minority males, especially Hispanic males (see, for example, Kramer & Ulmer, Citation2002; Spohn & Holleran, Citation2000; Steffensmeier & Demuth, Citation2001). Because the focus of this research is on female offenders, the literature review concentrates on negative stereotypes related to Black and Hispanic females. For the same reason, the majority of empirical findings that are noted pertain to female offenders.
4. Interestingly, the effect of race/ethnicity was confined to men. Kruttschnitt (Citation1984) found that minority males received more severe sentences than their White counterparts. Black men received longer prison sentences than White men in Crew's (Citation1991) investigation. And, Spohn and Beichner's (Citation2000) sex‐specific analyses showed that Black and Hispanic men were more likely than White men to be incarcerated.
5. The total number of females included in the sample was 266. This sample size is relatively small when one considers that a three‐way interaction effect was considered (race × gender × age) for its effect on the likelihood of receiving a downward departure. Given this reality, the reported results should be considered tentative.
6. Two studies simultaneously considered female felons and misdemeanants (i.e., Steffensmeier et al., Citation1993, Citation1998).
7. It may also be that race/ethnicity and gender jointly condition the effects of legal and extra‐legal variables. In a recent study of the length of imprisonment for federal offenders, Albonetti (Citation2002) found that the effects of the “mechanisms of discretion,” guilty pleas, and legally relevant variables (e.g., offense, prior record) were not the same for Black, White, and Hispanic females. White females were more likely to benefit from a substantial assistance departure than Hispanic and Black females. For non‐substantial assistance departures, White and Hispanic females received similar sentence reductions. Black females, however, received the smallest sentence reduction. At the same time, increases in the “guidelines offense level” resulted in a greater sentencing disadvantage for White females than for Black females.
8. Few researchers have examined the effect of social class on sentencing, which Zatz (Citation2000) argues is a serious void. The present research responds to this criticism. With regard to the few studies that have examined the concept, some have found that unemployed and less educated individuals are more inclined to receive more severe sentences. Theoretically, “unemployed offenders will be seen as a threatening population that needs to be controlled and one form of control is through incarceration” (Spohn & DeLone, Citation2000, p. 29). Turning only to female‐specific discussions of sentencing, Farrington and Morris (Citation1983) and Bickle and Peterson (Citation1991) found that employed and unemployed females were treated similarly. However, Crew (Citation1991) notes that the probability of receiving a longer sentence increased for females who were employed. Inconsistent with that finding, Kruttschnitt (Citation1980–81) found that more lenient sentences were likely for employed females, when compared to those who were welfare recipients or those who were temporarily unemployed. Interestingly, however, housewives fared better than employed women. With regard to educational attainment (i.e., another measure of social class), a number of researchers have found that better educated offenders received more lenient sentences than their less educated counterparts (Albonetti, Citation1997; Nagel & Hagan, Citation1982; Peterson & Hagan, Citation1984; Steffensmeier & Demuth, Citation2000). However, Kautt and Spohn's (Citation2002) recent analysis of federal drug offenders failed to produce the effect; regardless of the race of the defendant, the defendant's educational level had no effect on the length of sentence, for three broad categories of drug offenses (i.e., mandatory minimum, hybrid, or guidelines based). In terms of female‐specific examinations, only one study was identified where the researcher tested for the effect of educational attainment. In that study, Kruttschnitt (Citation1984) failed to observe a statistically significant relationship between the number of years a female spent in school and the severity of the sentence received.
9. Type of conviction (i.e., whether a defendant pleaded guilty) may also serve as an intervening variable. Albonetti (Citation1999) found that among white‐collar offenders race operated through the mode of conviction to affect the likelihood of receiving a suspended sentence; race did not directly affect the type of sentence received. However, in another examination, Albonetti (Citation1998b) found that race directly influenced sentence length for white‐collar offenders. And, in that earlier study, race did not operate through a guilty plea to influence this outcome. In this study, the mode of conviction is not analyzed as an intervening variable because almost all the females sentenced for misdemeanors pleaded guilty. Among cases for felons this is less likely to be the case.
10. Steffensmeier and his colleagues (Citation1993) found that older females were less likely to be incarcerated, but were more likely to receive longer prison terms.
11. Harris' (Citation1977) functional theory of deviant type‐scripts also points to the significance of being married and having children (for women). As wives and mothers, women are expected to remain in the home. Such an assignment with its affiliated roles serves to keep society functioning as desired. Type‐scripting women as criminal, however, would threaten current arrangements. According to Harris, there are negative consequences to removing a woman from the home and placing her, instead, in a jail. For example, the nuclear family may dissolve, financial burdens may arise, or affected males may need to leave their jobs in order to maintain the family (Harris, Citation1977, p. 13).
12. The New York City Criminal Justice Agency, Inc. (CJA) maintains a computerized database containing arrest and case‐processing information about New York City arrestees. Data are collected during a pre‐arraignment interview, which is used to ascertain community‐ties information and make a recommendation for release on recognizance at the defendant' first court appearance. Court information on all interviewed defendants is gathered from the Criminal and Supreme Court calendars.
13. This sample of cases is derived from a random sample of 15,359 summary arrests and a random oversample of 998 female arrests that occurred during 1989. Approximately 15 percent of the arrests in the random sample were for females. Of these arrests, 2,720 cases for females were designated as Criminal Court (i.e., lower court) cases. Of these, 1,815 females were sentenced at some point during lower‐court processing. Preliminary analyses showed that 62.6 percent were sentenced at arraignment, the first appearance in lower court. Those cases are not analyzed here because the effects of pretrial release, case‐processing time, and pretrial failure to appear do not apply.
14. It should be noted that the incarceration variable measured sentences to actual time, rather than to “time served.” An initial examination of a variable ascertaining the most severe sentence received showed that many of the cases for females resulted in sentences to incarceration. However, a subsequent examination of a variable that measured “the number of days sentenced to jail” showed that few of the cases resulted in sentences of actual jail time. Most females, who initially appeared to have received jail sentences, actually ended up receiving sentences of “time served.” For this reason, the latter group of females was included with those who did not receive an incarcerative sanction.
15. Age is measured as a continuous variable in this study. Steffensmeier and his colleagues (Citation1995) found that age had a non‐linear (i.e., curvilinear) effect on incarceration and sentence length. For that reason, in a preliminary analysis, I examined the possibility that the youngest and oldest females would receive the most lenient treatment. The results indicated no differences between females in varying age categories.
16. Each of these community‐ties items had five response categories: yes, yes verified, no, no verified, and unresolved conflict. The “yes” and “no” responses indicated the defendant's response and that it had not been verified. The “yes verified” and “no verified” categories were used when information provided by the defendant had been verified by a third‐party contact. “Unresolved conflict” meant that the information provided by the interviewed defendant did not match the information given by the verifier.
17. Other prior‐record measures were examined. Prior arrests did not predict the likelihood of incarceration as well as prior convictions, and both were highly correlated. Less than 4 percent of the females had served time in prison or were on probation. Consequently, neither of the latter two variables provided a good measure of prior record for this sample.
18. Other measures of charge severity were also examined, including the number of arrest charges, a continuous measure of the severity of the top arrest charge, a dichotomous measure of the severity of the top arrest charge, and a continuous measure of the severity of the top charge at disposition. None of these measures predicted the likelihood of incarceration as strongly as the dichotomous measure retained for analysis. Indeed, neither severity at arrest nor the number of arrest charges predicted the sentencing outcome.
19. In terms of the other major offense categories, about a third (33.5 percent) were charged with a property offense, 10.6 percent were charged with a violent offense, 9.1 percent had a top charge for a prostitution offense, and 13.6 percent had a top charge for another type of offense. A series of additional analyses that are not reported here revealed that substitution of the drug/non‐drug dichotomy with one of the others did not alter the study's findings.
20. A continuous measure of case‐processing time and a logged measure were also assessed for their effects on the sentencing outcome. Neither predicted the likelihood of incarceration as strongly as the dichotomous measure.
21. In traditional path analysis, models are constructed on the basis of a series of OLS equations. From these equations the regression coefficients that are affiliated with each variable are included in the models. These coefficients are then used to estimate the magnitude of the indirect effects.
22. Logistic regression is used for the analyses of this study's endogenous variables because they are inherent dichotomies and each is measured by only one variable. It is well known that dependent variables that are dichotomous should be analyzed with probit or logistic regression. Full information maximum likelihood covariance structure models are linear in functional form and unsuited for dichotomous endogenous variables. In SEM models, weighted least squares (WLS) or generalized least squares (GLS) may be used to solve some of the efficiency problems but the functional form of such models does not produce predicted values that resemble the curve of an ogive, making both techniques undesirable. Most researchers who use SEM object to models with only one indicator for each of the latent variables (see, for example, Bollen, Citation1989; Hoyle, Citation1995). Moreover, endogenous variables that are inherently dichotomies (e.g., had a warrant issued for pretrial failure to appear or not) do not even crudely represent an inadequately measured continuous variable (e.g., liberalism measured as a dichotomy). Therefore, using tetrachoric and polychoric correlations as inputs to an SEM model for analyzing such variables may misrepresent the relationships of these dichotomous variables to other dichotomous endogenous variables.
23. Researchers frequently use the mean to make these estimates. However, it is important to note that the selection of a different point value would result in different estimates. Because probabilities cumulate as an “S”‐shaped curve, evaluating probability changes at the mean will generally result in large probability changes (the steep part of the function), with the estimates of changes becoming smaller in the flat tails of the “S.” The probabilities will be derived from the following formula (Peterson, Citation1985):
where: L 0 = ln(P/1 − P); L 1 = L 0 + B; B = the logit coefficient for a relevant independent variable; and P = the mean of the dependent variable.
24. There was no indication of multicollinearity among the variables. Therefore, all were retained in the analysis that follows.
25. Recall that the effect of unemployment operated indirectly through the warrant‐ordered variable. The effect of educational attainment operated primarily through two intervening variables—pretrial release and employment status.
