Judicial Influence on Incarceration

The United States incarcerates a greater proportion of its population than any other nation in the world—almost one out of every 160 adults as of 2012 (Walmsley Reference Walmsley2006; Carson and Golinelli Reference Carson2013). This mass incarceration has disrupted family life (Western Reference Western2006), displaced public resources (Mauer Reference Mauer2006), reinforced cycles of poverty and crime (Pager Reference Pager2007), disenfranchised a significant portion of African American citizens (Mauer Reference Mauer2006), and inspired negative orientations toward government (Weaver and Lerman Reference Weaver2010). Accordingly, understanding the causal factors behind this remarkable phenomenon is of critical importance for understanding American social and political life.

Numerous factors may have contributed to the rise of mass incarceration in the United States, such as economic inequality (Garland Reference Garland1990; Gordon Reference Gordon1994), the legacy of racial cleavages (Smith Reference Smith2004; Alexander Reference Alexander2011), and a proliferation of out-of-wedlock births (Jacobs and Helms Reference Jacobs1996). Social movements (Gottschalk Reference Gottschalk2006) and public opinion (Nicholson-Crotty, Peterson, and Ramirez Reference Nicholson-Crotty2009; Enns Reference Enns2014) also may have played important roles. Increased incarceration may be a strategy employed by “tough-on-crime” politicians to curry electoral favor (Jacobs and Helms Reference Jacobs1996; Davey Reference Davey1998; Yates and Fording Reference Yates2005). Previous research has found that elected officials, including presidents (Jacobs and Helms Reference Jacobs1996), governors (Davey Reference Davey1998), and state legislators (Smith Reference Smith2004), influence the incarceration rate, and even elected state judges tend to become more punitive as reelection contests approach (Huber and Gordon Reference Huber2004). Many studies find that increased Republican Party power or conservative ideology is associated with increased incarceration (Jacobs and Helms Reference Jacobs1996; Smith Reference Smith2004; Yates and Fording Reference Yates2005). Others cite specific policy agendas, such as the “war on drugs,” limitations on parole, and the establishment of sentencing guidelines and mandatory minimums, as potential causal factors (Nicholson-Crotty and Meier Reference Nicholson-Crotty2003; Mauer Reference Mauer2006).

Several studies suggest that courts also influence incarceration rates (Yates and Fording Reference Yates2005; Simon Reference Simon2007, Reference Simon2013). US courts ultimately resolve a wide range of questions related to criminal justice, including the interpretation of common law principles, statutes, and constitutional rights. By expanding or contracting the protections afforded to criminal suspects and defendants, courts may significantly alter the rate at which Americans are put behind bars. For example, the Warren Court is frequently credited with initiating a “revolution in the constitutional law of criminal procedure” (Keck Reference Keck2004, 72) that “remade the entire American system of criminal justice” (Powe Reference Powe2000, 412). Similarly, some claim that the Burger and Rehnquist Courts instituted a “counterrevolution” by restricting defendant rights and advancing conservative criminal justice policies (Clayton and Pickerill Reference Clayton2006; see Keck Reference Keck2004, 177–79). Moreover, courts often exercise considerable influence over policy outcomes (Hume Reference Hume2009, Reference Hume2013; Keck Reference Keck2009; Howard and Steigerwalt Reference Howard2011), and SCOTUS is particularly successful at implementing criminal justice rulings, even when those decisions are unpopular (Hall Reference Hall2011). A systematic analysis of new admissions to federal prison from 1950 to 2008 indicates that SCOTUS decisions are strongly associated with long-term shifts in incarceration (Hall Reference Hall2015).

SCOLRs may play an even larger role in the criminal justice process. “The great preponderance of trials, both civil and criminal, take place in the state courts” (Galanter Reference Galanter2004, 506), and SCOLRs devote a substantial and increasing portion of their docket to criminal cases (Kritzer et al. Reference Kritzer2007). Even more important, the vast majority of prisoners in the United States reside in state penitentiaries (Carson and Golinelli Reference Carson2013). Accordingly, Yates and Fording (Reference Yates2005) find that shifts in SCOLR ideology are associated with changes in state incarceration rates.

High-court judges probably influence incarceration rates by altering the behavior of other actors in the criminal justice system. First, high courts influence decision making by lower-court judges, who generally adhere to the judicial hierarchy (Gruhl Reference Gruhl1980; Songer and Sheehan Reference Songer1990; Songer, Segal, and Cameron Reference Songer1994; Brent Reference Brent1999). Although lower-court compliance is not perfect (Baum Reference Baum1978; Scott Reference Scott2006), defying precedent tends to be the exception rather than the rule (Westerland et al. Reference Westerland2010). Because these judges oversee the process of trying and sentencing criminal defendants, their decisions directly affect who goes to jail and who does not. In fact, some scholars have used the influence of trial court judges on incarceration in order to study recidivism rates (Berube and Green Reference Berube2007; Green and Winik Reference Green2010).

Second, the changing behavior of lower-court judges likely affects the behavior of prosecutors. These actors strategically anticipate the probability of success when deciding whether to prosecute a case, appeal a decision, or plea bargain (Albonetti Reference Albonetti1987; Lederman Reference Lederman1999; Taha Reference Taha2010; Wedeking Reference Wedeking2010). Accordingly, as high courts expand or contract criminal rights and lower courts comply, prosecutors alter their behavior. For example, SCOTUS’s influence on incarceration is mediated by the behavior of lower-court judges and prosecutors (Hall Reference Hall2014a). Specifically, SCOTUS’s criminal justice policy is associated with case filing rates, plea bargaining rates, and conviction rates in federal district courts. The behavior of these judges and prosecutors is associated with subsequent shifts in the federal incarceration rate.

Finally, as the behavior of lower-court judges and prosecutors changes, law enforcement officials, defendants, and defense attorneys should adjust their behavior in response to the new legal landscape. For example, SCOTUS’s decision in Mapp v. Ohio (367 U.S. 643 [1961]) prompted changes in the frequency of motions to suppress, guilty pleas, and the issuing of search warrants (Hall Reference Hall2011, 51–56). That decision may have also influenced arrest and crime rates, as police officers struggled to comply with the new rules (Nagel Reference Nagel1965; Canon Reference Canon1974a). Similarly, the Court’s ruling in Miranda v. Arizona (384 U.S. 436 [1966]) altered police and arrestee behavior (Hall Reference Hall2011, 62–70) and may have reduced the ability of law enforcement officials to solve crimes (Cassell and Fowles Reference Cassell1998a, Reference Cassell1998b; Donohue Reference Donohue1998).

In short, we argue that high courts influence the behavior of judges, prosecutors, and law enforcement officials, which subsequently affects how many people go to prison. Of course, the behavior of these actors is also likely influenced by their own and each others’ preferences; however, in the aggregate, judicial preferences should substantially shape incarceration rates (Yates and Fording Reference Yates2005; Hall Reference Hall2015).

We expect the causal chain between judicial preferences and new prison admissions to unfold rather slowly as justices influence the behavior of lower-court judges, prosecutors, and law enforcement officials. Considering the complexity and delays inherent in the American legal process, an immediate effect of high-court preferences on policy outcomes would seem implausible. If high-court preferences affect incarceration, they most likely do so by producing only long-term shifts in new prison admissions. Standard conceptions of judicial ideology define decisions that are supportive of criminal suspects and criminal defendants as liberal decisions.Footnote ¹ Accordingly, the extant literature suggests our first hypothesis:

The Limits of Judicial Influence

Previous research suggests that high courts influence the criminal justice system in general (e.g., Powe Reference Powe2000; Keck Reference Keck2004) and incarceration in particular (e.g., Yates and Fording Reference Yates2005; Hall Reference Hall2015), yet many scholars emphasize the limits of judicial impact. US high courts are traditionally viewed as severely constrained in their ability to influence policy outcomes (e.g., Horowitz Reference Horowitz1977; McCloskey Reference McCloskey2010), even those related to criminal justice. For example, Wasby finds that, for many law enforcement officials, SCOTUS’s most prominent rulings simply “did not matter much” (Reference Wasby1976, 8), and Rosenberg claims the criminal rights revolution “failed” because judges ultimately depend on other actors to implement their rulings (Reference Rosenberg2008, 335).

To be sure, most scholars agree that courts influence social change (e.g., Rosenberg Reference Rosenberg2014); however, even the most expansive views of judicial power concede that courts are constrained by various contextual factors (e.g., Hall Reference Hall2011). Accordingly, the critical task for understanding judicial impact is to identify these constraining factors. As described above, high courts probably affect incarceration rates by altering the behavior of street-level bureaucrats, such as trial court judges, prosecutors, and law enforcement officials. Therefore, judicial impact is likely moderated by factors that inhibit high-court control over these bureaucrats. Specifically, we argue that two factors condition judicial impact on incarceration: political climates and institutional structures.

Analysis

Our goal is to identify factors that constrain high-court influence; accordingly, we are interested in assessing judicial power. In his seminal work on the topic, Jack Nagel defines power as follows: “a power relation, actual or potential, is an actual or potential causal relation between the preferences of an actor regarding an outcome and the outcome itself” (1975, 29). Accordingly, we assess high-court influence by measuring the association between high-court ideology and incarceration. We test two factors (one climatic and one institutional) that might constrain judicial influence. Specifically, we test whether divided government and institutional thickness moderate the relationship between high-court ideology and incarceration.

Our methodological approach offers several advantages. First, rather than examine the effects of individual rulings on specific behaviors (e.g., the effect of Mapp v. Ohio on the exclusion of evidence), we evaluate judicial impact by testing the aggregate influence of judicial ideology on a broad policy outcome: the rate of new admissions to state prisons. This aggregate approach accounts for the possibility that broad policy outputs may be gradually influenced by numerous judicial decisions over time or the anticipation of judicial decisions. Second, we employ a multilevel error correction model, which captures the dynamic effects of judicial preferences across states and over time. Most importantly, the error correction framework allows us to distinguish between short- and long-term effects and assess the speed with which judicial preferences affect policy outcomes. Third, our analysis exploits the political and institutional variation across the American states to test the factors that condition judicial impact. Accordingly, our study offers the first systematic assessment of factors that condition the aggregate influence of judicial preferences on policy outcomes over time and across political and institutional contexts.

Model Estimation

We use the error correction model (ECM) framework to estimate our models. ECMs represent a general class of regression models appropriate for both stationary and nonstationary data (De Boef and Keele 2008).Footnote ¹² This approach is particularly appropriate in this context because it distinguishes between immediate and long-term effects. One way to express a single-equation ECM is as follows (excluding a constant term for simplicity):

The critical feature of an ECM is that for each independent variable X we include up to two parameter estimates: $\beta 1$ for the differenced variable and $\beta 2$ for the lagged level of the variable. If either of these coefficient estimates indicates a statistically significant relationship, then it is appropriate to conclude that the explanatory variable is associated with the dependent variable. In this simple bivariate example, $\beta 1$ provides an estimate of the initial change in the dependent variable produced by a shock to the independent variable. This is called the “short-term” effect, not meaning that the effect is impermanent but rather that the effect occurs wholly at a specific point in time. The error correction rate ( $\alpha 1$ ) and $\beta 2$ provide the information needed to estimate the slightly more complicated “long-term” effect (the long-run multiplier). The long-term effect is the portion of the connection between X and Y that does not occur at one particular point in time but is distributed temporally such that a portion of the impact is distributed over a period of time.

A multilevel (or hierarchical) modeling framework is particularly well suited for analyzing time-series–cross-sectional data such as ours. Hierarchical data structures contain more than one level of analysis, where units from one level are nested within units from a higher level. This approach facilitates the analysis of observations that are likely to be correlated within levels of analysis (Gelman and Hill 2006). Therefore, we estimate a multilevel ECM employing a two-level hierarchical structure: state-years (level 1 units) nested within states (level 2 units).Footnote ¹³ In order to facilitate intuitive and comparable interpretation, we standardize all independent variables. State-level variables were standardized on the basis of the average within-state variation of each variable.

Judicial Influence on Incarceration

We begin with an analysis of whether high-court ideology is associated with changes in incarceration rates across all states. If high courts influence this important policy output, we expect incarceration rates to be negatively associated with judicial liberalism. The results of the multilevel ECMs are reported in table 1. In this model, both SCOLR and SCOTUS liberalism are associated with long-term decreases in a state’s incarceration rate. In other words, liberal judicial preferences are associated with long-term decreases in state incarceration rates. Most of the control variables perform as expected. Higher levels of conservatives, economic inequality, and the crime rate are associated with long-term increases in incarceration. The coefficients for elite liberalism are signed in the expected direction, but neither reaches conventional levels of statistical significance. Consistent with our expectations, none of the variables are significantly associated with immediate changes in incarceration. These results support our first hypothesis: judicial preferences appear to influence new admissions to state prisons over the long term.Footnote ¹⁴

Table 1.Judicial Ideology and Incarceration

Note.—Multilevel error correction model of changes in new state incarcerations; standard errors in parentheses.

^* p < .05 (two-tailed test).

In order to understand the total long-term effects of our explanatory variables, we estimate the long-run multipliers (LRMs) for SCOLR and SCOTUS liberalism on the basis of the model in table 1.Footnote ¹⁵ The LRM estimates the total effect of a change in an independent variable in 1 year as it continues to influence incarceration at a decaying rate in future years. Because PAJID scores and JCS scores are not bridged to a common metric, we cannot directly compare the influence of these courts; however, we can estimate the effects of a typical shift in each court’s ideology (recall that SCOLR liberalism is standardized on the basis of within-state variation). A 1 standard deviation (SD) increase in SCOLR liberalism is associated with a total long-term decrease of 17.1 in the incarceration rate. A 1 SD increase in SCOTUS liberalism is associated with a total long-term decrease of 22.4 in the incarceration rate.

To put those figures in context, consider the state of Georgia, which had an incarceration rate of 131.0 in 2004.Footnote ¹⁶ In other words, of the 8,769,252 people living in Georgia in 2004, 11,489 of them were sentenced to prison that year. On the basis of the results in table 1, if the liberalism of the median justice on the Supreme Court of Georgia were to increase 1 SD, we would expect the incarceration rate to decrease 17.1 in the long term. Stated another way, a 1 SD increase in the court’s liberalism in that year would be associated with 1,500 fewer people being sentenced to prison over time. A 1 SD increase in SCOTUS liberalism would be associated with 1,964 fewer incarcerations over time.

Consistent with our expectations, the effects of judicial preferences on incarceration appear to unfold rather slowly. The error correction rate indicates the speed at which the long-term effects take place. For example, the error correction rate of −0.11 in table 1 indicates that out of the total long-term effect of SCOLR liberalism (−17.1 or 17.1 fewer admissions per 100,000 people), 11% of this effect takes place at time t + 1 (−1.9), an additional 11% of the remaining effect takes place at time t + 2 (−1.7), and so on, until the total long-term effect has been distributed. Accordingly, it takes more than a decade for 75% of the total long-run impact of judicial preferences to be realized in the incarceration rate. Recall that the SCOLR ideology scores are not dynamic (meaning that each judge gets the same score for his or her entire career). Therefore, a shift in SCOLR liberalism indicates that a new judge has been elected or appointed to the court, shifting the court’s median ideology. Accordingly, the long-term effects are consistent with the idea that a judge’s arrival on the court has gradual effects as the new court hears cases over time, those cases filter through the judicial system, and those rulings affect the behavior of street-level bureaucrats.

Divided Government and Institutional Thickness

We now turn to an analysis of factors that condition high-court influence on incarceration. We hypothesized that courts would enjoy greater influence in states with divided government and thin judicial hierarchies. To test these hypotheses, we segregate the data into four groups on the basis of whether the partisan control of the state government was divided and whether the state judicial system had intermediate appellate courts.Footnote ¹⁷ We then run the multilevel ECM from table 1 on each of these four subsets of data. The results of the models are presented in table 2.

Table 2.Divided Government, Institutional Thickness, and Judicial Ideology

Note.—Multilevel error correction models of changes in new state incarcerations for states with unified (Unf Gov) versus divided (Div Gov) government and states that have intermediate courts (Int Courts) versus those that do not (No Int Courts); standard errors in parentheses.

^* p < .05 (two-tailed test).

The first model in table 2 includes states with unified government and intermediate courts. Consistent with our expectations, SCOLR liberalism is not significantly associated with decreases in the incarceration rate in this model.Footnote ¹⁸ The second model in table 2 includes states with unified government and no intermediate courts. Once again, judicial ideology has no effect on incarceration in these states; however, in this model, elite ideology is associated with long-term shifts in incarceration. The third and fourth models in table 2 examine states with divided government. The third model includes states with intermediate courts, and the fourth model includes states that lack these courts. In these models, both SCOLR and SCOTUS liberalism are negatively associated with long-term changes in incarceration, suggesting that high courts influence new admissions in these states. The coefficients for judicial ideology in these models are not only statistically significant but also substantially larger than in the models for states with unified government. This finding supports our second hypothesis: the negative relationship between judicial liberalism and state incarceration rates is stronger in states with divided government.

We estimate the substantive impact of judicial preferences on incarceration by calculating the LRMs on the basis of the third and fourth models in table 2. Figure 1 illustrates the total long-term effect of a 1 SD increase in SCOLR liberalism (fig. 1A ) and SCOTUS liberalism (fig. 1B ) on changes in state incarceration rates.Footnote ¹⁹ As shown in figure 1, the long-term effects of judicial ideology are significantly larger in states with divided government than in states with unified government.

Figure 1.

Total long-term effects on incarceration: A, effects of 1 SD increase in SCOLR liberalism; B, effects of 1 SD increase in SCOTUS liberalism.

Contrary to our third hypothesis, the total long-term effect of judicial preferences is not larger in states that lack intermediate courts; if anything, the long-term effects appear to be smaller in these states. However, a closer look at the data reveals a more complicated picture. The difference between these two groups is not simply the magnitude of the effects but also the speed at which these effects manifest themselves. The error correction rate in states with divided government and intermediate courts is only −.06, compared to −.33 in states with divided government and no intermediate courts. This means that it takes more than 2 decades for 75% of that effect to be realized in states with divided government and intermediate courts. In states with divided government and no intermediate courts, 75% of judicial impact is realized in just 4 years.

In figure 2, we plot the annual impact of the long-term effects in states with divided government to show how vastly different these processes are. The figure presents the effect of a 1 SD increase in SCOLR liberalism (fig. 2A ) and SCOTUS liberalism (fig. 2B ) in 1 year on new prison admissions in subsequent years. As the figure indicates, the long-term effects of judicial liberalism in states with intermediate courts is a drastically slower process than it is in states with thin judicial hierarchies. The initial effects of judicial ideology are actually much stronger in states that lack intermediate courts, but those effects dissipate within a few years. Therefore, we find only partial support for our third hypothesis.

Figure 2.

Annual decay of long-term effects on incarceration in states with divided government: A, effects of 1 SD increase in SCOLAR liberalism; B, effects of 1 SD increase in SCOTUS liberalism.

In sum, high-court judges appear to exert substantially greater influence on incarceration rates when state governments are divided, and they exert influence more quickly when they sit at the top of thin judicial hierarchies.