An Interventionist and Diversionary Rationale

The juvenile court adopted an explicit interventionist and rehabilitative rationale, providing positive programming to “protect the community and cure the child” simultaneously, as the child savers intended (Zimring Reference Zimring2005, 36). Nevertheless, Zimring argues that the court's “diversionary” rationale may have been even more salient, as the court could shield children from the long‐term negative impact of exposure to criminal punishment and criminal courts (Zimring Reference Zimring2005). According to this diversionary rationale, the juvenile court was “the lesser of evils” in relation to the criminal court (Zimring Reference Zimring2005, 41). Diverting youth would, in George Herbert Mead's terms, spare youth from the “retribution, repression, and exclusion” (Reference Mead1918, 590) of the punitive system of justice.

During the 1960s and 1970s, the US Supreme Court and many scholars questioned whether the juvenile court was in fact rehabilitative and the lesser of two evils (Feld Reference Feld1999; Zimring Reference Zimring2005, 41). By 1967, the Court decided In re Gault, which led to substantial changes in juvenile justice. After reviewing the punitive realities of the juvenile justice system, the Court mandated elementary procedural safeguards such as advance notice of charges, the right to a fair and impartial hearing, assistance of counsel, an opportunity to cross‐examine witnesses, and privilege against self‐incrimination of juvenile defendants (Feld Reference Feld1999). Although as Ward and Kupchik (Reference Ward and Kupchik2009) note, the Supreme Court rulings did not directly challenge the juvenile court's mission of rehabilitation, they did require accountability on the part of justice officials and limited subjective decision making, formalizing juvenile court processing.

On the heels of the 1960s and 1970s rulings requiring “system accountability” in the juvenile court (Ward and Kupchik Reference Ward and Kupchik2009), the late 1970s and 1980s marked a shift toward individual accountability, offender responsibility, and punitive sanctions for youth as well as adults (van den Haag Reference van den Haag1975; Feld Reference Feld1999; D. Garland Reference Garland2001). The juvenile court was clearly not immune from the punitive turn in criminal justice. If youth were now more deserving of punishment for their crimes, then legislatures could enact prosecutorial and programmatic changes that would require them to “deal with their commitment” of an offense before release (Ward and Kupchik Reference Ward and Kupchik2009, 103).

Juvenile transfer to adult jurisdiction is regarded as the most punitive response to juvenile crime. Yet how are we to understand blended sentencing legislation that seems to merge the juvenile court (as the lesser of the evils) and the criminal court? Considering the juvenile court's history and recent reforms, does blended sentencing legislation represent an attempt to strengthen the juvenile court's capacity to intervene in the best interest of the child? Or, does blended sentencing represent a punitive reform in the juvenile justice system, mirroring punitiveness in the criminal courts?

For purposes of this study, we are less concerned with the effectiveness or morality of blended sentencing laws than with their historical, political, and cultural underpinnings. We ask: Does a model that explains the increasing punitiveness of adult criminal sanctions also predict the rise of state adoption of blended sentencing? If so, it would provide evidence that blended sentencing signals a punitive turn toward crime control of juvenile offenders. We will therefore test whether the known drivers of harsh criminal punishment in the adult system also predict state adoption of blended sentencing.

The Rise of Blended Sentencing

Blended sentencing emerged almost three decades ago, with West Virginia being the first US state to adopt the practice in 1985. Texas and Rhode Island followed shortly thereafter, but only three states had adopted blended sentencing laws by 1990. State adoption then rose dramatically from 1994–1997, with twenty‐one states passing blended sentence laws. As shown in Figure 1, over half (twenty‐six) of the fifty states have now adopted a form of blended sentencing. Figure 1 shows some evidence of geographic clustering, with a majority of states in the Midwest (75 percent) adopting blended sentencing legislation and relatively few South Atlantic states (only Florida, Virginia, and West Virginia).Footnote ¹ At first glance, the lack of geographic clustering in the South Atlantic states may indicate that blended sentencing was not considered punitive enough, resulting in a lack of state adoption. Alternatively, it is notable that states with relatively low (Maine and New Hampshire) and relatively high (Louisiana and Mississippi) incarceration rates failed to adopt blended sentencing. Such patterns suggest that the blended sentencing movement is not a simple function of region or punitiveness, although our multivariate analysis will provide further insight into these factors.

Figure 1. Blended Sentencing in the United States by Census Regional Boundaries (1985–2008)

Blended sentencing legislation can be further divided according to which court—juvenile or criminal—has jurisdiction or sentencing authority. Juvenile blended sentencing laws in fourteen states allow the juvenile court to impose adult criminal sanctions on certain categories of crimes. Generally, the court is empowered to combine a juvenile disposition with a suspended adult sentence (Griffin Reference Griffin2003, Reference Griffin2008, Reference Griffin2010). On the other hand, twelve states allow the criminal court to sentence transferred juveniles to a juvenile court disposition; in some states the criminal court also suspends the adult sentence in hopes of motivating compliant behavior (Griffin Reference Griffin2003, Reference Griffin2008, Reference Griffin2010). As explained in the Appendix, blended sentencing legislation can be further divided by sentencing authority into five overlapping models, shown in Table A1 of the Appendix, by state, year of adoption, and court of jurisdiction.Footnote ²

The origins of the blended sentencing movement remain an open question. Some might have championed blended sentencing to moderate the effects of strict state transfer laws (Redding and Howell Reference Redding, Howell, Fagan and Zimring2000), shifting potentially adult‐certified youth back to the juvenile court. Other proponents, motivated by the perceived leniency of the juvenile court, might have intended to subject more youth to adult criminal sanctions. Still others see the legislative impact of blended sentencing as part of the power shift in the juvenile court from the hands of the judge to the prosecutor (Zimring Reference Zimring2005). Yet there is little direct evidence on the relative influence of such motivations. Moreover, despite calls for more robust attention to theory in juvenile sanctioning (Mears and Field Reference Mears and Field2000, 984), most research has adopted an instrumental cost‐benefit framework. While several excellent studies address the efficacy of juvenile justice reforms (see Redding and Howell Reference Redding, Howell, Fagan and Zimring2000; Podkopacz and Feld Reference Podkopacz and Feld2001; Cheesman et al. Reference Cheesman, Thomas, Denise, Matthew and Nicole2002; Cheesman and Waters Reference Cheesman and Nicole2008; Trulson et al. Reference Trulson, Haerle, DeLisi and Marquart2011; Brown and Sorensen Reference Brown and Sorensen2012), such work gives less attention to the connection between “day‐to‐day operations” and “an institution's self‐conceptions” (Garland Reference Garland1991, 117). To evaluate whether blended sentencing represents punitiveness in juvenile justice, an affirmation of historic rehabilitative goals, or a shift to prosecutorial power, we construct and estimate a conceptual model of its rise and adoption. Following David Garland (Reference Garland1990a, Reference Garland1991, 124), we draw from the sociology of punishment traditions of Durkheim, Marx, Foucault, and Elias.

Collective Conscience and Changing Sensibilities

Durkheim ([Reference Durkheim1893] 1933) emphasized the expressive nature of punishment—both as a representation of society's moral values and a mechanism to legitimize and reaffirm those values (Garland Reference Garland1990b, Reference Garland1991). From this perspective, changes in punishment should thus mirror broader shifts in the modern conscience collective (Durkheim [Reference Durkheim1893] 1933). If the collective conscience of society has shifted from rehabilitative to punitive in its orientation toward juvenile law violation, then what tangible variables account for these changes? In some respects, Michael Tonry applies a Durkheimian logic in Thinking About Crime, pointing to the “prevailing social values, attitudes, and beliefs” (2004, 5) driving adoption of punitive sanctions. Recent research by Enns (Reference Enns2014) supports Tonry by finding that from the mid‐1980s to 2009 there was a strong relationship between a rise in public punitiveness and the production of punitive policy. Although a limitation of Enns's study was the inability to identify specific mechanisms that produced increases in public punitiveness, Tonry emphasizes media attention and publicity, showing how US sensibilities to get tough on crime produce punitive policies even in the face of declining crime rates.

As Bernard (Reference Bernard1992) reports, public perceptions of youth crime are often untethered from actual juvenile offending. Thus, an increase in media coverage of a juvenile crime wave (Blumstein Reference Blumstein1995; Fox Reference Fox1996) and an explicit focus on high‐profile and exceptionally violent cases (Walker Reference Walker1994; Tonry Reference Tonry2004, 5) can create moral panics (Cohen Reference Cohen1972). These panics help shape public attitudes on crime, resulting in legal changes that encourage harsh punishment. Therefore, following Tonry (Reference Tonry2004), we suggest that increases in public attention to juvenile crime will be closely correlated with a state's adoption of blended sentencing. Unfortunately, we lack state‐level historical information about media or public attention devoted to youth crime. To test this idea, however, we constructed a variable assessing change in publicity of delinquency hearings over time, which ranged from generally closed to generally open to the public.

Although criminal justice policy bears an important relation to levels of crime, it rarely follows directly from crime rates. For example, Tonry (Reference Tonry2004) shows how crime rates were often declining when punitive policy changes were enacted. Following Durkheim ([Reference Durkheim1893] 1933) and others, the punitive crime control era represents a change in public sensibilities rather than an instrumental response to rising crime rates. Nevertheless, to assess how actual crime rates influence the adoption of blended sentencing, we also estimate the effects of direct measures of juvenile crime, such as the rate of juvenile arrests by offense type and the rate of youth confinement by race.

Marx and the “Economics and Politics” of Penal Policy

Scholars examining the structural determinants of crime policy often adopt a Marxian perspective, stressing the interests of the ruling class, which dominates economic production and imposes power in other social spheres (like politics).Footnote ³ In turn, political institutions adapt their conditions (such as punishment and criminal policy) to fit the dominant economic mode of production (Garland Reference Garland1990b). Although Marx wrote little on punishment, scholars in the Marxist tradition have linked economic production and political ideologies to the introduction or expansion of punishment (Garland Reference Garland1990b, Reference Garland1991; Beckett and Sasson Reference Beckett, Theodore and Simpson2000).

Rusche and Kirchheimer (Reference Rusche and Otto1939) built a foundation for research in this tradition, specifying several propositions relating labor market and class struggles with penal development. For example, they hypothesized that a surplus of labor led to an increase in harsh punishment, with elites wielding punishment as a managerial tool tied to the labor value of prisoners.Footnote ⁴ In times of labor surplus and high unemployment, punishments tend to become harsher for individuals (either in terms of physical conditions or sentence severity). The capitalist elite and their control of the distribution of resources is only part of Rusche and Kirchheimer's argument linking labor surplus and punishment strategies. By controlling the conditions of penal institutions relative to those of the poor, elites can further dominate the working class. As labor surplus grows, and with it the incentives to commit crime, punishments are thus stepped up.

Contemporary scholars have modified and tested Rusche and Kirchheimer's (1939) labor surplus and punishment theory. Today, labor surplus is typically operationalized as the unemployment rate, while punishment is operationalized as the imprisonment rate (Chiricos and Delone Reference Chiricos and Delone1992). Although studies by Inverarity and Grattet (Reference Inverarity and Ryken1989), Greenberg (Reference Greenberg1977), and Jankovic (Reference Jankovic1977) find a strong positive relationship between unemployment and prison commitments, other studies find little association (e.g., Parker and Horowitz Reference Parker and Horowitz1986) or apparently conflicting evidence (Inverarity and McCarthy Reference Inverarity and Daniel1988) that varies with model specification and methodological approach (Sutton Reference Sutton2000). In these instances, even Chiricos and Delone, whose meta‐analysis findings generally support the Rusche‐Kirchheimer hypothesis, state that “the research has left many if not most of the key theoretical issues unexamined” (1992, 432).

One of the key critiques of the Rusche‐Kirchheimer hypothesis is that it understates the importance of political forces that shape legislation of penal measures (Garland Reference Garland1990b). Contemporary research in this tradition considers both political and economic determinants of punitiveness. Most notably, Jacobs and Helms (Reference Jacobs and Helms1996) report that unemployment is not related to prison admission rates when controlling for changes in family structure, the percentage of young males in the population, and crime rates. With regard to politics, however, conservatism is a significant and positive predictor of punishment. Jacobs and Helms thus find little direct support for the Rusche‐Kirchheimer thesis but greater support for David Garland's (Reference Garland1990b) and Savelsberg's (Reference Savelsberg1994) understanding of the political drivers of criminal punishment. Similarly, Sutton finds that factors such as unemployment and homicide rates are not significantly related to imprisonment rates when structural political factors such as union density and left‐party dominance are simultaneously assessed. Strong unions and left‐party influence are significantly and negatively associated with imprisonment rates, suggesting that democratic parties “exert political influence in support of a range of ameliorative social policies, including less punitive responses to crime” (2004, 183). Based on these ideas and findings, we consider both economic measures (such as unemployment) and political measures (such as partisan legislative and gubernatorial control) in predicting adoption of blended sentencing.

Apparatuses and Instrumentalities of Punishment for “At‐Risk” Populations

David Garland's sociology of punishment perspective also emphasizes the “apparatus and instrumentalities” (1991, 124) of punishment, a reference to Foucault's (Reference Foucault1977, Reference Foucault1978, Reference Foucault1980, Reference Foucault1990) analysis of power relations in the penal process and controls such as surveillance, inspection, and normalization. Foucault's approach, moving from the institution outward, informs diverse theories of the evolving strategies and techniques of the penal field. For example, Feeley and Simon's (1992, 449) “new penology” adopts a Foucaultian perspective to describe the emergence of a “new strategic formation in the penal field.” This involves new discourses, such as the use of actuarial science and the standardization and use of efficient control mechanisms to target high‐risk groups of offenders. These include fixed sentences and guidelines to determine sentence type and length,Footnote ⁵ pretrial detention, and, more recently, pretrial bail assessments to estimate risk to public safety (Kempf‐Leonard and Peterson Reference Kempf‐Leonard and Peterson2000; Mamalian Reference Mamalian2011). Feeley and Simon maintain that the expansion of prison and community corrections (including alternatives such as electronic monitoring and boot camps) and the use of risk assessments are best understood in terms of “managing costs and controlling dangerous populations rather than social or personal transformation” (1992, 465). The instrumentalities and apparatuses in the new penology thus extend the continuum of control to populations deemed most at risk of re‐offense.

Feeley and Simon (Reference Feeley and Simon1992) do not address whether the new penology and actuarial justice has bled into the juvenile system. Kempf‐Leonard and Peterson (Reference Kempf‐Leonard and Peterson2000), however, point directly to developments in the juvenile court that could be attributed to actuarial justice, which informs our construction of variables to represent the new penology. For example, greater use of objective risk assessments in juvenile court parallels the use of prehearing detention to determine actuarial risk in the adult system.Footnote ⁶ States have increasingly adopted the use of detention risk assessments in juvenile court to identify youths eligible for detainment (Baird, Storrs, and Connelly Reference Baird, Gregory and Helen1984; Baird Reference Baird1985; Frazier Reference Frazier and Roberts1989; Weibush et al. Reference Weibush, Christopher, Barry, Onek, Howell, Hawkins, Krisberg and Wilson1995; Kempf‐Leonard and Peterson Reference Kempf‐Leonard and Peterson2000; Howell Reference Howell2003). This process de‐emphasizes individual characteristics and circumstances that could inform the best course of action for each juvenile and, instead, bases juvenile court decisions on offense severity and risk, which are markers of actuarial science and risk management. To assess the relationship between the new penology, community correctional control, and adoption of blended sentencing, we include measures of youth confinement for pretrial detention and adults on parole and probation.Footnote ⁷

The concept of objective risk assessment is embedded in the criminal justice system's use of sentencing guidelines and truth in sentencing policies. A similar structure is evolving in the juvenile court via new transfer mechanisms, such as direct file, that reduce judicial discretion and increase efficiency by standardizing and routinely processing juvenile cases. Direct file laws allow prosecutors to certify youth directly to adult court without judicial screening based on standardized criteria. The presumptive criteria reflect actuarial justice, by defining subpopulations (such as juveniles charged with specified offenses) as particularly threatening and in need of greater surveillance. In transferring control to prosecutors, direct file laws arguably save the juvenile court resources, since the cases that would require the most time and money to resolve are effectively transferred from the court (Kempf‐Leonard and Peterson Reference Kempf‐Leonard and Peterson2000). To assess the impact of reduced judicial discretion and the standardization of juvenile processing, we thus consider the effects of a direct file law.Footnote ⁸

Finally, Feeley and Simon (Reference Feeley and Simon1992) argue that the new penology represents a movement away from moral or clinical descriptions of the individual offender toward actuarial language that describes risk to public safety. This actuarial language reflects a “trend of the penal system to target categories and subpopulations rather than individuals” (Feeley and Simon Reference Feeley and Simon1992, 453). Class‐ and race‐based inequalities are deeply rooted in the actuarial language that defines particular groups as high‐risk offenders or career criminals. These links reinforce the idea that crime is a product of a marginalized and dangerous subpopulation—a “high risk group that must be managed for the protection of society” (Feeley and Simon Reference Feeley and Simon1992, 467). We thus consider race‐specific measures of incarceration and juvenile confinement.

In sum, our model suggests that the landscape of the contemporary crime control era looks something like this: states with high unemployment rates, conservative politics, and growing marginalized populations have higher incarceration rates. Further, moral panic fuels public fear and perceptions of crime, which engenders more severe punishments and the use of managerial techniques such as probation, risk assessments, and pretrial detention to manage the risk of particular subgroups to control crime.

Dependent Variable and Logic of Analysis

Our analysis is designed to identify whether the predictors of state adoption of blended sentencing parallel the known predictors of punitive justice in the adult system. The primary dependent variable is thus a time‐varying indicator of whether states adopt blended sentencing policies. Although states differ by type of adopted blended sentence, either juvenile or criminal court jurisdiction, we initially coded any state that adopted a blended sentence structure as a “blended sentence adopter” and created a dichotomous dependent variable for our event history analysis. We then conducted a more basic ANOVA comparison of blended sentencing adopted under juvenile versus criminal court jurisdiction. We employed this strategy because we are constrained by a small number of events (only twenty‐six state adopters over twenty‐four years). To maintain stability in our models while incorporating the predictor and control variables, we could not disaggregate the dependent variable into a multinomial dependent variable. This analysis thus follows prior research in aggregating various types of transfer mechanisms (e.g., direct file, judicial waiver, statutory exclusions) into a dichotomous variable despite procedural differences (see Bishop et al. Reference Bishop, Frazier, Lanza‐Kaduce and Lawrence1996; McNulty Reference McNulty1996; Lanza‐Kaduce et al. Reference Lanza‐Kaduce, Frazier, Lane and Bishop2002; Kurlychek and Johnson Reference Kurlychek and Johnson2004).

We use a discrete‐time logistic regression event history approach to predict state adoption of blended sentencing. Because event history analysis is concerned with time to “failure,” the risk set for this analysis includes all fifty states eligible to adopt blended sentencing from 1985 to 2008.Footnote ⁹ Event history analysis is advantageous in this setting because it appropriately models both time‐varying predictors (such as the unemployment rate) and censored cases that have yet to adopt blended sentencing (Allison Reference Allison1984; Yamaguchi Reference Yamaguchi1991).

Independent Variables

Each of our fixed and time‐varying predictors is described in Table A3 of the Appendix, so we focus discussion here on the key predictors. To assess political climate, we include a state‐ and year‐specific measure of the proportion of the legislature under Democratic or Republican control, as well as a dichotomous measure indicating a Democratic governor.Footnote ¹⁰ To assess how state punitiveness influences policy changes in the juvenile court, we include the African American incarceration rate,Footnote ¹¹ the rate of adults on probation to capture “criminal managerialism,”Footnote ¹² and the presence of the death penalty (Amnesty International n.d.). We further include measures of direct file lawsFootnote ¹³ (to indicate the standardization of juvenile case processing) and the openness of public hearings (to indicate publicity and public scrutiny of juvenile court operations) (see Symanski 2000, 2002, 2004, 2007).

Socioeconomic variables include unemployment rates and education levels, to estimate the impact of labor surplus and workforce education on adoption of blended sentencing.Footnote ¹⁴ As labor surpluses lead to increased punishment, communities of color have been punished most severely (Tonry Reference Tonry1996, Reference Tonry2004; Feld Reference Feld1999). To assess the influence of racial threat on the passage of blended sentencing reform, we include census information for the non‐Hispanic African American, Hispanic, and non‐Hispanic white juvenile population counts by state and year,Footnote ¹⁵ along with the rate of juveniles in confinement by race.Footnote ¹⁶ We caution, however, that Hispanic ethnicity was not consistently reported over the period (see, e.g., Liebler and Halpern‐Manners Reference Liebler and Halpern‐Manners2008). To assess how juvenile crime rates influence the adoption of blended sentencing, we created state‐ and year‐specific indicators of the rate of juvenile arrests by offense type (Puzzanchera and Kang Reference Puzzanchera and Kang2013). Finally, we controlled for region based on census indicators for Northeast, Midwest, South, and West.Footnote ¹⁷

Discrete‐Time Logistic Regression

Our discrete‐time logistic event history models take the following form:

where P_it represents the probability that blended sentencing passed in state i in time interval t, β signifies the effect of the independent variables, X₁, X₂ … X_k denote k time‐varying independent variables, and α_t represents a set of constants corresponding to each decade or discrete‐time unit. This approach allows us to employ time‐varying covariates to test how changing state characteristics affect the likelihood of state adoption of blended sentencing. Based on inspection of the hazard distribution and a statistical comparison of alternative time specifications, we specify time using a cubic model.Footnote ¹⁸

Figure 2 graphs the probability of state adoption of blended sentencing. Our cubic year model accounts for the few early adopters in the mid‐1980s, the sharp increase from 1992 to 1998, and the subsequent decline in the probability of adopting blended sentencing laws until the last state, Ohio, adopted its law in 2002.

Figure 2. Probability of State Adoption of Blended Sentencing

Bivariate Discrete‐Time Regression

After reviewing the timing of state adoption, we next consider state‐level predictors. We begin with bivariate analysis, which aids in model specification for the multivariate analysis. Table 1 presents the results from thirty‐four discrete‐time logistic event models predicting blended sentencing adoption. The first two columns (labeled Cubic Year and Exp(B)) show the relation between the independent variables and the passage of blended sentencing laws while controlling for time with the cubic year specification. We find that states with a Democratic governor are 60 percent less likely to pass blended sentencing laws than states where Democrats do not hold that office.Footnote ¹⁹ In contrast, the proportional makeup of the legislature, whether Democratic or Republican, is nonsignificant. We specified several legislative partisanship models (including partisan control as well as the proportion Democrat or Republican), but saw no significant effects. For measures of state punitiveness, the African American incarceration rate has a significant and positive effect on the passage of blended sentencing. A one standard deviation increase in the African American incarceration rate is associated with 35 percent greater odds of passing blended sentencing (using the unstandardized beta and standard deviation to calculate the effect size).Footnote ²⁰ Death penalty states had a significant negative relationship to blended sentencing, which likely reflects the regional patterning described above. At the bivariate level, the measures for socioeconomic conditions are not statistically significant, though these factors emerge more strongly in the multivariate analysis.

Table 1. Bivariate Predictors of Blended Sentencing 1985–2008 (Twenty–Six Events, N=872)

Variable	Model	Cubic Year	SE	Exp (B)
Political partisanship
Percent Democratic legislature	1	.006	(.012)	1.006
Percent Republican legislature	2	–.011	(.011)	.989
Democratic governor	3	–.927*	(.483)	.396
State punitiveness
African American incarceration rate	4	.004*	(.002)	1.005
Death penalty (vs. abolished states)	5	–.998**	(.443)	.368
Adult probation rate	6	.000	(.003)	1.000
Socioeconomic conditions
Unemployment rate	7	.250	(.166)	1.284
High school diploma rate	8	.370	(.290)	1.448
Juvenile court characteristics
Direct file (vs. no direct file)	9	.831*	(.448)	2.295
Open hearing (vs. closed)	10	.652	(.527)	1.920
Open hearing with provisions (vs. closed)		.128	(.510)	1.136
Upper age of jurisdiction	11	–.040	(.343)	.907
Juvenile population characteristics
White youth population (100,000s)	12	.055	(.294)	1.057
African American youth population (100,000s)	13	–.127	(.214)	.881
Hispanic youth population (100,000s)	14	.178*	(.086)	1.195
Juvenile confinement
Total confinement	15	–.205	(.214)	.815
White juvenile confinement	16	–.317	(.276)	.728
African American juvenile confinement	17	–.005	(.018)	.995
Hispanic juvenile confinement	18	.018	(.013)	1.019
Detention rate	19	.204	(.492)	1.227
Juvenile crime (arrests)
Total arrests	20	–.001	(.006)	.999
Part I arrests	21	.013	(.022)	1.013
Violent crime arrests	22	.044	(.118)	1.045
Property Part I arrests	23	.016	(.023)	1.016
Murder arrests	24	−2.070	(4.105)	.126
Rape arrests	25	–.967	(2.074)	.380
Robbery arrests	26	–.183	(.279)	.833
Aggravated assault	27	.274	(.200)	1.315
Burglary arrests	28	.088	(.129)	1.092
Larceny arrests	29	.010	(.029)	1.010
Motor vehicle theft	30	.333**	(.157)	1.359
Arson arrests	31	.282	(1.039)	1.326
Weapons violation arrests	32	.066	(.112)	1.068
Drug abuse/sale arrests	33	–.045	(.091)	.956
Census region
Midwest (vs Northeast)	34	.630	(.528)	1.758
South		–.066	(.648)	.936
West		.075	(.668)	1.077
Time
Year	35	.−1.316**	(.628)	.268
Year2		.214***	(.078)	1.239
Year3		–.008***	(.003)	.992

* Note: p<.10,

** p<.05,

*** p<.01. Standard errors in parentheses.

The next set of variables shows how a state's juvenile court characteristics, juvenile population, and juvenile crime characteristics are associated with the passage of blended sentence laws. Direct file laws have a significant positive impact, with direct file states being 2.3 times as likely as non‐direct‐file states to pass blended sentencing.Footnote ²¹ We observe nonsignificant effects for upper age limit of juvenile court jurisdiction and openness of public hearings.Footnote ²² For the measures of juvenile population characteristics, we find nonsignificant effects for the juvenile African American and white populations, but a positive and significant effect for the Hispanic population in the bivariate models. These race and ethnicity findings should be interpreted cautiously because reporting shifted over the period, allowing for multiple‐race responses in 2000 and subsequent years (Liebler and Halpern‐Manners Reference Liebler and Halpern‐Manners2008). Nevertheless, there appears to be geographic clustering of Hispanic youth in states that passed blended sentencing. The variables representing rates of confinement by race are all nonsignificant at the bivariate level.

The last cluster of variables considers juvenile arrests for serious and violent felony offenses. We analyzed all UCR Part I violent and property crime, in addition to UCR Part II crimes for weapons violations and drug abuse/sale. The only variable that reaches statistical significance is the juvenile arrest rate for motor vehicle theft. This finding is intriguing because arrest rates for motor vehicle theft peaked in 1990 and declined significantly during the mid‐1990s, when most states passed blended sentence laws (Griffin Reference Griffin2008). Other violent crimes such as murder, rape, robbery, and aggravated assault peaked during 1993 and 1994, closer to the time when most states enacted blended sentence laws, but they are not significantly related to adoption of these laws.

Discrete‐Time Multivariate Regression

Based on our bivariate analysis, we construct a set of nested multivariate models. We hypothesize that states with high unemployment rates, conservative politics, and high rates of African American incarceration and adult probation will be more punitive in their orientation to juvenile crime; they will pass blended sentencing not as a rehabilitative alternative to treat juvenile offenders, but as a crime control measure. In addition, states that allow the general public to attend delinquency hearings increase concerns that violent crime is on the rise, and these states will be more likely to pass blended sentencing laws as a means to punish offenders, particularly persons of color (Garland Reference Garland2001). If this pattern of results holds, the evidence would suggest that blended sentencing signals a punitive turn toward crime control in the juvenile court.

Table 2 presents three discrete‐time logistic regression models predicting state passage of blended sentencing. Model 1 examines juvenile crime while controlling for regional effects (relative to the Northeast), time, juvenile crime and confinement,Footnote ²³ and characteristics of the juvenile court. As is evident in the table, blended sentencing is well established in midwestern states. Because only those charged with serious crimes (e.g., murder, rape, robbery, aggravated assault, burglary, larceny, motor vehicle theft, and arson) are eligible for blended sentences, we hypothesized a significant positive relationship between arrest rates for violent and serious juvenile crime and the passage of blended sentence laws. As Table 2 shows, however, juvenile crime is generally not a significant predictor. In contrast, the rate of Hispanic youth in confinement emerges as significantly related to state passage of blended sentencing. A one standard deviation increase in the rate of Hispanic youth in confinement is associated with a 5 percent increase in the odds of a blended sentence law.

Table 2. Discrete‐Time Regression Predicting Blended Sentencing Law

	Model 1		Model 2		Model 3
Variable	B	Exp β	B	Exp β	B	Exp β
Census region
Midwest (vs. Northeast)	1.326*	3.766	.619	1.858	.284	1.328
	(.669)		(.830)		(1.059)
South (vs. Northeast)	–.386	.679	.111	1.117	.141	1.151
	(.744)		(.832)		(.895)
West (vs. Northeast)	.230	1.259	.101	1.106	–.007	.993
	(.961)		(1.025)		(1.044)
Juvenile crime rate
Juvenile violent arrests	.197	1.218	.140	1.151	.067	1.069
	(.183)		(.225)		(.253)
Juvenile property arrests	.029	1.030	.048	1.049	.048	1.050
	(.031)		(.034)		(.036)
Juvenile weapons violations	.120	1.127	.136	1.146	.154	1.166
	(.154)		(.217)		(.205)
Juvenile drug arrests	–.151	.860	–.162	.851	–.127	.881
	(.135)		(.153)		(.162)
Juvenile confinement rate
White confinement	–.697	.498	–.897*	.408	–.890*	.411
	(.465)		(.524)		(.510)
African American confinement	.034	.967	–.022	.979	–.029	.971
	(.034)		(.047)		(.064)
Hispanic confinement	.049**	1.050	.038	1.039	.033	1.033
	(.023)		(.027)		(.030)
Pretrial detention rate	.414	1.513	.643	1.903	.703	2.020
	(.706)		(.761)		(.805)
Juvenile court features
Direct file (vs. no direct file)	.898*	2.455	.808	2.243	.904	2.470
	(.531)		(.548)		(.552)
Public hearing
Open (vs. closed)	.287	1.332	.740	2.097	.394	1.483
	(.738)		(.785)		(.818)
Provisions (vs. closed)	–.134	.875	–.173	.841	–.553	.575
	(.651)		(.682)		(.772)
Socioeconomic
Unemployment rate			.438**	1.550	.475**	1.609
			(.221)		(.233)
High school diploma rate			.578	1.782	.588	1.801
			(.446)		(.480)
Political partisanship
Democratic governor			−1.162**	.313	–.899	.407
			(.553)		(.628)
State punitiveness
Adult African American incarceration					.006*	1.006
					(.004)
Adult probation rate					–.002	.998
					(.005)
Death penalty state					–.918	.399
					(.705)
Year	−1.899**	.150	−1.892**	.151	−1.890**	.151
	(.809)		(.872)		(.889)
Year2	.281**	1.324	.302***	1.352	.295***	1.343
	(.097)		(.104)		(.105)
Year3	–.010**	.990	–.011***	.989	–.011***	.989
	(.003)		(.003)		(.004)
Constant	−2.722	.066	−9.535**	.000	−9.446*	.000
	(2.110)		(4.698)		(4.901)
−2 log likelihood	180.9		169.4		164.6
Chi‐square (df)	52.9***	(17)	64.4***	(20)	69.2***	(23)
Events	26		26		26
N	871		871		871

* Note: p<.10,

** p<.05,

*** p<.01. Standard errors in parentheses.

While much research on the “criminology of the other” (Garland Reference Garland2001, 137) emphasizes African American youth, we find a significant positive effect only for Hispanic confinement. We report these results cautiously due to limitations in ethnicity data, but these findings align with research suggesting typification of Hispanics as criminals (Villarruel et al. Reference Villarruel, Nancy, Pamela, Omara, Susan and Kristen2004; Johnson et al. Reference Johnson, Stewart, Pickett and Gertz2011; Welch et al. Reference Welch, Payne, Chiricos and Gertz2011). Juvenile court characteristics are included to represent the new penology and the changing sensibilities of the US public (Feeley and Simon Reference Feeley and Simon1992; Tonry Reference Tonry2004). Model 1 of Table 2 supports our hypothesis that states enacting direct file laws are more likely to pass blended sentence laws. In fact, direct file states are 2.5 times as likely to pass blended sentencing laws. The estimate for open juvenile hearings is not statistically significant.

Model 2 of Table 2 introduces economic characteristics that represent labor surplus, which has been associated with punitiveness in previous research (Rusche and Kirchheimer Reference Rusche and Otto1939; Greenberg Reference Greenberg1977; Jankovic Reference Jankovic1977; Inverarity and Grattet Reference Inverarity and Ryken1989; Chiricos and Delone Reference Chiricos and Delone1992). If blended sentencing mirrors these crime control policies, then states with higher unemployment rates should be more likely to pass blended sentencing legislation. Consistent with this idea, we find that unemployment is a positive and significant predictor in multivariate models. For each 1 percent increase in the unemployment rate, the odds of passing blended sentencing laws increases by 55 percent. With regard to political partisanship, Model 2 shows a strong negative and significant relationship between Democratic leadership and adoption of blended sentencing; states with Democratic governors are approximately 69 percent less likely to adopt blended sentencing laws net of the other variables. This finding aligns with research associating punitive reforms with Republican rather than Democratic leadership (Sutton Reference Sutton1987, Reference Sutton2000, Reference Sutton2004; Jacobs and Helms Reference Jacobs and Helms1996; Jacobs and Carmichael Reference Jacobs and Carmichael2001, Reference Jacobs and Carmichael2002).

Finally, Model 3 of Table 2 introduces measures of state punitiveness. The African American incarceration rate is a significant and positive predictor. A standard deviation increase in the African American incarceration rate raises a state's odds of passing blended sentencing legislation by 59 percent. Moreover, the rate of white youth in confinement is a significant and negative predictor in Models 2 and 3. Net of the full set of covariates in the model, states are less likely to pass blended sentencing in states with large numbers of young whites in confinement—a relationship that was not apparent in the bivariate models.

In Model 3 of Table 2, the effect of unemployment persists, but the introduction of state punitiveness variables reduces the effect for Democratic governors and direct file laws to nonsignificance. This pattern is not unexpected, given the close association and endogeneity between incarceration and partisanship. This final model includes economic and political factors, juvenile crime rates, juvenile incarceration, juvenile court characteristics, and census region. Overall, the predictors of blended sentencing laws—racialized confinement, direct file, unemployment, and partisanship—appear more congruent with a punitive culture of control than with the historical treatment emphasis of the juvenile court (see Garland Reference Garland2001; see also Feeley and Simon Reference Feeley and Simon1992; Beckett and Herbert Reference Beckett and Steve2010; King, Massoglia, and Uggen Reference King, Michael and Christopher2012).

Although too few states passed blended sentencing laws to permit a disaggregated event history analysis, we conducted a simple ANOVA analysis to compare mean levels on our independent variables across different types of blended sentencing laws. We distinguished between those that passed blended sentencing under juvenile court jurisdiction, criminal court jurisdiction, or both juvenile and criminal court jurisdiction. We found few differences across these categories (as shown in Table A2 of the Appendix and reported in note 22), with the exception of two variables: states placing blended sentencing under criminal court jurisdiction had especially high rates of Hispanic juvenile confinement and an especially high likelihood of direct file transfer laws.