For a decade now, I have felt certain that the Court's contraction of federal rights and remedies on grounds of federalism should be interpreted as a plain invitation to state courts to step into the breach … the diminution of federal scrutiny and protection out of purported deference to the states mandates the assumption of a more responsible state court role … The Fourth Amendment has been most clearly targeted for attack …. (Justice William Reference BrennanBrennan, James Madison Lecture on Constitutional Law at New York University School of Law, November 18, 1986)
Advocates of federalism, both in the United States and elsewhere, often cite the potential for enhanced protection of individual civil liberties as an emerging rationale for a federal system dividing governmental responsibilities between central and regional governments and central and regional judiciaries (Reference Katz and TarrKatz & Tarr 1996). One scholar argues that the European community's increasing federalism has led to greater individual rights (Reference Lenaerts, Katz and TarrLenaerts 1996:139), while another argues that the Canadian Charter of Rights and Freedoms judicialized civil rights protection, which has led to greater security of individual rights (Reference Cotler, Katz and TarrCotler 1996).
Of course, this idea of greater protection runs counter to accepted justifications for federalism, which is often viewed as an instrument of group or state's rights (Reference Katz and TarrKatz & Tarr 1996). However, regional courts relying on regional constitutions can offer enhanced protection for civil liberties beyond the baselines established by a federal constitution and a national court (see Reference LindeLinde 1984). It is that possibility for increased protection that led Justice William Brennan, confronting an increasingly conservative U.S.Supreme Court (USSC), to encourage state courts to rely on state constitutional grounds as a strategy to increase and insulate the protections of the Bill of Rights. A state high court's (or SCOLR, for State Court of Last Resort) role as final arbiter of its own state constitution provides a useful, albeit limited, tool through which courts can accomplish this goal. During the last few decades, some judicial scholars and officials have viewed this position as an opportunity to enhance civil liberties. Given this impetus and some subsequent liberal state court rulings, many have asserted that an era of “new judicial federalism” was at hand (see Reference SolimineSolimine 2002; Reference PulliamPulliam 1999).
However, subsequent research has produced little evidence that such a revolution has occurred in the United States (Reference RosenfeldRosenfeld 1988; Reference CauthenCauthen 2000). Undoubtedly this is because the use of regional constitutional protection is not as straightforward as advocates of federalism would suggest. In the United States, success depends on the actions of others to whom state court justices are accountable both within and outside their respective states. State law decisions can lead to responses from other state actors. Such responses may mean personal consequences for the state court justices or a policy outcome far removed from the state court's own preference. In addition, under current doctrine, the USSC reserves the right to review the independence of state law conclusions and potentially find them inadequate. Thus, the power of other actors to check state judicial innovation will moderate use of the state constitutional doctrine.
This article examines the political, institutional, and legal environment of state courts, including the effect of the USSC. Prior state court research examines the institutional and political factors that may either encourage or discourage SCOLRs' aggressive pursuit of their policy preferences. We add to this literature by examining how the SCOLR takes into account the ideological position of the USSC, in addition to state factors.
In examining SCOLR decisions from 1981 to 1993, we discover that state court justices do use the state grounds doctrine as Justice Brennan urged them, but only when those justices have the desire to do so and there is a favorable political and legal context within their state. We also find that SCOLR decisions are influenced by the relative position of the USSC.
Civil Liberties Protection and State Constitutional Grounds
The USSC has the power to review state court applications of federal law, but SCOLR reserve the right to interpret state law authoritatively. Put more simply, as defined by the USSC in the nineteenth-century case of Murdock v. City of Memphis (1874), SCOLR decisions based on state law independent of federal interpretation are outside the jurisdiction of the federal courts and, thus, are nonreviewable by the USSC (Reference HaasHaas 1981).
In the context of civil liberties, incorporated federal rights provide a minimum level of protection from which states cannot subtract,Footnote 1 therefore, deviations based on state constitutional law can only expand civil liberty guarantees. Each state retains the “sovereign right to adopt in its own Constitution individual liberties more expansive than those conferred by the Federal Constitution” (Pruneyard Shopping Center v. Robins, 447 U.S. 74, 81 [1980]). Jurisdiction remains unclear, however, as long as the basis of the state court's decision is unclear, because civil liberties issues implicate federal as well as state law. Alexander Hamilton noted long ago that in cases of concurrent jurisdiction an appeal would lie from the state courts to the national court (Reference RossiterRossiter 1961).
Many argue that a conservative USSC now uses concurrent jurisdiction to overturn state court decisions favoring criminal defendants. These scholars and commentators cite the 1983 case of Michigan v. Long to buttress this claim. In this case, police questioned David Long after Long drove his car off a road and into a ditch. Officers said he “appeared to be under the influence of something.” Noticing a hunting knife on the floor of the car, they conducted a Terry v. Ohio (1968)Footnote 2“protective search” of the car, searching for weapons. The police found a bag of marijuana, and Long was arrested for possession. The Michigan court argued that the search violated both federal and state constitutions, suggesting that if the federal ruling was overturned the more rigorous ruling from the Michigan constitution would survive.
The USSC, however, not only ruled that Michigan misapplied Terry v. Ohio and the Fourth Amendment, but also ruled that Long had an insufficient adequate and independent state ground. The USSC required that a state court resting a decision on independent state grounds, and therefore not subject to USSC review, must “make clear by a plain statement in its judgment or opinion that the federal cases are being used only for the purpose of guidance, and not themselves compel the result that the court has reached” (463 U.S. 1032, 1041 [1983]). Scholars and observers criticized this “plain statement” rule as a conservative attempt to rein in liberal state court decisions (Reference Segal and SpaethSegal & Spaeth 2002:74–5), pointing out that the Long decision adopted a presumption of federal jurisdiction unless the state grounds were made explicit.Footnote 3
While the decision permitted the USSC to use concurrent jurisdiction to overturn a liberal decision that favored defendant rights,Footnote 4 it allowed SCOLRs the option of using plain language to buttress protections of civil liberties through adequate and independent state grounds, and there have been several high-profile liberal SCOLR case rulings supporting individual rights explicitly based on state constitutions. For example, the Georgia Supreme Court struck down that state's anti-sodomy law, after the USSC refused to do so in Bowers v. Hardwick (1986), citing state constitutional right to privacy grounds (Powell v. State 1998). Similarly, the California Supreme Court overturned a state statute requiring minors to obtain parental consent or court approval before having an abortion by relying on the privacy provisions of the California Constitution (American Academy of Pediatrics v. Lungren 1997), effectively avoiding the USSC's precedent upholding similar laws (Planned Parenthood v. Casey 1992). Several states have ruled in favor of equalized public school financing after the USSC refused to do so in 1973 (San Antonio School District v. Rodriguez 1973; see Reference FriedelbaumFriedelbaum 1992).Footnote 5
Despite these examples, scholars suggest that there has not been a burst of state court activism on behalf of civil liberties (Reference SolimineSolimine 2002), pre- or post-Long. The potential for federalism to enhance civil liberties has simply not occurred in the United States. One scholar, examining First Amendment protections, notes that “what has become apparent during the past three decades of the new judicial federalism is the narrow framework of expressive freedom in which state courts have elected to function” (Reference FriedelbaumFriedelbaum 2004:657–8). In an early systematic analysis, Reference RosenfeldRosenfeld (1988) examined more than 500 state cases and found no pattern for the use of the state grounds doctrine; she concluded that most state courts had not properly used the Michigan v. Long decision (Reference RosenfeldRosenfeld 1988:1049–50). Other scholars have reported not only a similar confusion in state courts' use of the doctrine but also little evidence of any increased reliance upon state grounds (Reference SolimineSolimine 2002; Reference Fahlbusch and GonzalezFahlbusch & Gonzalez 1987).
Constraints on Judicial Federalism
One potential explanation for why this federalism revolution has not materialized is that the state courts have become increasingly conservative (see Reference LatzerLatzer 1991). Since SCOLR decisions are, to some extent, a function of state judicial policy preferences (e.g., Reference LangerLanger 2002; Reference HallHall 1992),Footnote 6 a conservative state high court would not deviate from a conservative USSC in expanding defendants' search-and-seizure rights.
Another explanation lies in the constraints imposed by the political and institutional settings of a SCOLR. Each SCOLR exists in a different political and institutional context than the other SCOLRs and the USSC. While the literature acknowledges that courts seek to influence policy toward personal preferences, emerging state court literature also shows that a SCOLR's ability to shape a decision to achieve its outcome is dependent upon the broader political landscape and institutional structure. Comparative studies of constitutional courts have found that the capacity for independent action by judiciaries increases with the fragmentation of other political actors (Reference Ferejohn, Boix and StokesFerejohn et al. n.d.). Thus, state courts should be freer to develop state law when political division in the state is greater. Institutional factors such as the structure of the judicial system, the level and nature of the court's workload, and methods of judicial selection and retention have all been demonstrated to affect judicial behavior (Reference Glick and EmmertGlick & Emmert 1987; Reference Brace and HallBrace & Hall 1990, Reference Brace and Hall1995; Reference Hall, Brace, Clayton and GillmanHall & Brace 1999; Reference HallHall 1985, Reference Hall1992; Reference LangerLanger 2002). In particular, unlike the federal judicial system, many states require justices to face periodic elections of one form or another, and these elections can induce a substantial degree of accountability (Reference HallHall 2001).
Accountability and electoral reprisal are particularly relevant in regard to expanding the rights of criminal defendants (Reference SolimineSolimine 2002). Decisions sustaining defendant rights are not as popular as decisions supporting the police (see, e.g., Reference HallHall 1992). Previous research demonstrates that state justices and judges alter their behavior in criminal cases in a more conservative, law-and-order direction to avoid electoral retribution (Reference HallHall 1995; Reference Huber and GordonHuber & Gordon 2004). These demonstrations of electoral accountability suggest that the preferences of state citizens and the state political actors (STATE) may condition the development of legal principles in state judicial systems. Previous studies of state court justices have found that competitive elections and the brevity of judicial terms of office inspire strategic behavior on the part of justices to forestall opposition (Reference HallHall 1992). Because the Long decision deprives state judges of “the ability to immunize themselves with ambiguity” (Reference AlthouseAlthouse 1993:989), it is reasonable to posit that justices confronting retention and reelection may be less likely to support defendant rights.
Besides political pressure, the structure of the state court system can contribute to the opportunities a SCOLR has for judicial innovation. For example, the presence or absence of an intervening appellate court is routinely noted as affecting the behavior of SCOLR judges, including time-consuming behavior such as registering dissent (Reference Brace and HallBrace & Hall 1990). A SCOLR without an intermediate appellate court typically has less discretion over case selection, which can affect the quantity and content of the court's docket (Reference Atkins and GlickAtkins & Glick 1976). An intervening court can shoulder much of the burden of mandatory error correction and allow the SCOLR the freedom and time to choose apt vehicles for legal innovation. An intermediate appellate judiciary also allows litigants more time to develop constitutional issues for high court disposition.
Another influence on the development of state legal doctrine is the content of state law. The content of state constitutions varies considerably, often incorporating different language to articulate the legal rights of citizens or extending overt protections that the U.S. Constitution does not. Justice Reference BrennanBrennan (1977) notes that even when state constitutions use language similar to the U.S. Constitution to describe the rights of citizens, state judges are not bound by federal interpretation of similar language when interpreting their own constitutions. Several states also recognize a right of privacy in their constitutions, making explicit what the USSC has extended tentatively over the course of many controversial decisions. In the criminal justice context, a right to privacy may be used to recognize additional protection against police searches.
States can also directly address the relationship between state constitutional doctrine and federal rights by amending their constitutions. In 1982, two years after adoption of an amendment articulating a state right to privacy, Florida revised its constitutional search-and-seizure provision to bind construction directly to the USSC's reading of the Fourth Amendment. Florida's amendment prevents the SCOLR from using Florida's constitution to extend protection to criminal suspects further than the baseline required by federal law. Chief Justice Warren Burger appeared to encourage such developments in a subsequent concurring opinion: “… [w]hen state courts interpret state law to require more than the Federal Constitution requires, the citizens of the state must be aware that they have the power to amend state law to ensure rational enforcement” (Florida v. Casal, 462 U.S. 637, 639 [1983]). California's constitution was similarly amended in 1990 to ensure that “[t]his Constitution shall not be construed by the courts to afford greater rights to criminal defendants than those afforded by the Constitution of the United States …” (California Constitution, Article 1, Section 24). Under such circumstances, the SCOLR should be constrained by the language of its constitution from recognizing greater rights than the U.S. Constitution recognizes.
Finally, the USSC is not a passive actor in this process. The Long decision signals that the USSC will remain a key player in the oversight of state criminal justice jurisprudence (Reference SolimineSolimine 2002). Given concurrent jurisdiction and the removal of the ability of state judges to “immunize themselves with ambiguity” (Reference AlthouseAlthouse 1993), the state court must take into account the preferences of the USSC when choosing whether and how to establish legal doctrine and its subsequent policy consequences. Because of their position between the justices of the USSC, who reserve the right to judge the adequacy of independent state law decisions, and state political preferences with various methods of holding judges accountable, the relative positions of the SCOLR, STATE, and USSC are relevant to identifying when pressures will constrain the development of state constitutional law.
A Framework of State Court Protection of Civil Liberties
Stepping back, we can combine ideological, institutional and contextual factors, including the relative ideological positions of the state, the SCOLR, and the USSC, to construct a framework suggesting which factors and under what circumstance would lead state courts to use state grounds doctrine to expand Fourth Amendment protections.
Assuming a unidimensional policy space for search-and-seizure decisions corresponding to a liberal-conservative continuum, we anticipate that the relative ideological positions of the public and judicial actors affect SCOLRs' production of legal doctrine that departs from the applicable federal requirements under the Fourth Amendment. Rational litigation strategies (Reference SongerSonger et al. 1995), strategic auditing (Reference CameronCameron et al. 2000), and the norms of federal supremacy and vertical stare decisis support the conclusion that SCOLRs will not depart from USSC doctrine when they would prefer to extend less protection than applicable federal law requires.Footnote 7 Thus, we confine our analytic attention to situations in which the SCOLR favors more expansive civil liberties than the USSC.
Since ambiguity will no longer suffice, more-liberal state courts only enjoy the independence and opportunity to provide greater protection using state grounds in certain contexts. To illustrate, we analyze three regimes, or spatial models, varying the distribution of preferences among the three primary actors while keeping the SCOLR to the left of the USSC. In each model, the line represents a continuum reflecting the degree of intrusiveness. The positions of the actors (STATE, SCOLR, USSC) are akin to “indifference points” and indicate the degree of intrusiveness a search must surpass before it is deemed “unreasonable.” Thus, positions to the left, which set an “unreasonable search” at lower levels of intrusiveness, are more “liberal” or expansive toward defendants' rights than those to the right.
Figure 1 shows three alternate regimes. In the first, which we label Regime 1, the SCOLR is more liberal than the USSC, but less so than the state citizenry. While challenged searches falling between the SCOLR and USSC preferences in intrusiveness need not be suppressed under the federal constitution, the SCOLR may be inclined to invalidate them through a more liberal reading of civil liberties under the state constitution. The citizens of the state, more liberal than the SCOLR, should not disapprove and may even approve of such action. Under these circumstances, state grounds decisions are politically safe for the SCOLR, but the opportunities to produce such decisions may be limited. State political officials, likely to be even more sensitive to the political preferences of the citizenry than the courts, have less to gain from an aggressive crime control agenda and may act on their own to limit the intrusiveness of searches. Nevertheless, conditions in this regime are comparatively favorable to the production of state law extension of search-and-seizure protection.
Figure 1. Three Ideological Preference Regimes.
Regime 2 switches the positions of the state and SCOLR. In this regime, the SCOLR is the most liberal actor in the policy issue. As in Regime 1, the state citizens and SCOLR are both more liberal than the USSC, but the citizenry is comparatively less liberal than its high court. In Regime 2, we expect that the SCOLR will have considerable freedom politically to move criminal justice policy to the left of the USSC baseline and may have some opportunity to produce state law even closer to its ideological position due to slippage in oversight of the judiciary.
Regime 3 is the last arrangement that preserves the liberal-conservative relationship between the SCOLR and the USSC. In this regime, both judicial actors are more liberal than the state citizenry. Because the SCOLR prefers a less-intrusive standard for unreasonable searches than the USSC, the state justices would prefer to craft an independent ground for invalidating searches, but the state citizenry prefers no deviation from the requirements imposed by federal supremacy. The SCOLR, more liberal than the USSC, may want to move search-and-seizure policy left of the federal minimum but is limited by the possible reaction of within-state political forces. Thus, the SCOLR will have only as much political freedom to pursue state law protections against searches and seizures as the state's degree of oversight permits.
Hypotheses
In each of the three specified regimes, the SCOLR is more liberal than the USSC and, thus, more inclined to find state grounds for a pro-defendant civil liberties decision than a SCOLR more conservative than the USSC. However, each regime also identifies circumstances that could more or less militate against too much state legal activism. Our first three hypotheses address these expectations:
Hypothesis 1: A SCOLR has a positive probability of producing pro-defendant state grounds decisions in Regime 1.
Hypothesis 2: A SCOLR has the greatest probability of producing pro-defendant state grounds decisions in Regime 2 compared to any other regime.
Hypothesis 3: A SCOLR has a low positive probability of producing pro-defendant state grounds decisions in Regime 3.
Previous research demonstrates that competitive judicial elections, those in which judges can face opposing candidates, are just as contested as elections in the U.S. House of Representatives (Reference HallHall 2001). Also, despite fears that elections provide little real accountability, elected judges do appear to alter their behavior in response to a perceived risk of electoral reprisal (Reference HallHall 2001; Reference Huber and GordonHuber & Gordon 2004). We expect, therefore, that justices subject to competitive elections will be less responsive to the political opportunities represented by the above regimes than justices who do not face competitive elections.
Hypothesis 4: A SCOLR whose justices are subject to competitive reelection is likely to produce fewer pro-defendant criminal justice decisions based on independent state grounds in general than one retained through other mechanisms.
Hypothesis 5: A SCOLR whose justices are subject to competitive reelection is less likely to produce decisions based on independent state grounds than a SCOLR in an equivalent regime whose justices are not subject to competitive election.
Institutional factors can further enhance or impede the recognition of greater protection under state constitutions. Shorter terms of office increase the sensitivity of justices to retention considerations and lead to more moderate behavior. Our next hypothesis relates to this structural constraint:
Hypothesis 6: A SCOLR whose justices serve longer terms is likely to produce more pro-defendant criminal justice decisions based on independent state grounds.
State court justices use the legal materials available to them in order to produce state grounds decisions independent of federal criminal justice doctrine. Thus, the presence or absence of explicitly recognized rights or language restricting judicial innovation is relevant to the production of state court doctrine, as reflected in our next hypotheses:
Hypothesis 7: The SCOLR of a state with an explicit right to privacy in its constitution is likely to produce more pro-defendant criminal justice decisions based on independent state grounds.
Hypothesis 8: The SCOLR of a state with a constitutional provision binding the interpretation of state legal rights to the interpretation of the U.S. Constitution by the USSC is unlikely to produce any pro-defendant criminal justice decisions based on independent state grounds.
The presence or absence of intermediate appellate courts has been demonstrated to influence patterns of state court activity. SCOLR justices without appellate courts to serve as a buffer between themselves and trial courts typically have higher caseloads composed predominantly of routine cases that leave little time or opportunity for the selection and careful consideration of cases that will yield new legal doctrine. Reference LangerLanger (2002:90) finds that the presence of an intermediate appellate court significantly increases the likelihood of a SCOLR docketing a judicial review case in a politically salient issue. This leads to our next hypothesis:
Hypothesis 9: The SCOLR of a state without an intermediate appeals court is less likely to have the resources and attention necessary to produce decisions based on independent state grounds.
Judicial activity also depends on the freedom of the courts from political reprisal of the state government. Comparative studies of courts find consistently that judicial independence results in part from the fragmentation of the political system (Reference Bednar and FerejohnBednar et al. 2001). Courts take a stronger policymaking role in countries where political divisions prevent the legislature and executive from reprisal or override. Political fragmentation can encourage judicial activity as well, when governments fail to reach policy choices and courts fill the void (see Reference FerejohnFerejohn 2002). Our last hypothesis is suggested by these findings:
Hypothesis 10: The SCOLR in a less politically fragmented state is unlikely to have the resources and attention necessary to produce decisions based on independent state grounds.
Data, Method, and Model
Dependent Variable
In order to test our hypotheses, we created a database of all SCOLR cases decided in favor of the defendant and involving Fourth Amendment search-and-seizure issues from 1981 through 1993 in all 50 states. The 13-year period covered provides an adequate time series to control for alternate environmental and institutional explanations. Including a lagged dependent variable means that our term of analysis began in 1982, just before the Long decision as well as the USSC's restrictive Fourth Amendment decisions Illinois v. Gates (1983) and U.S. v. Leon (1984).
To find cases, we used a keyword search in Westlaw by using the phrase search and seizure. We used those cases that listed search and seizure as the first or main issue. We used this phrase to search for cases for each individual state for the years 1981 to 1993. In Texas and Oklahoma, there are different high courts, one for civil cases and one for criminal cases. In those states, we only examined the criminal courts of appeal. This search resulted in a total of 1,112 cases. The cases were then aggregated by state and by year. That is, if one state had five search-and-seizure cases, the number five was entered for that state for that year. This left us with a database of 650 data points (state-years). A one-year lagged dependent variable eliminated one year, leaving 600 observations.
From these data, we first created two variables that indicated whether the SCOLR decided the case citing state case law or federal case law.Footnote 8 These variables were mutually exclusive. In other words, cases could only be coded as positive (1) for state case law or positive (1) for federal case law, not for both. While cases could have more than one issue, we focused on the major issue of each case.Footnote 9 We then aggregated the variables by year to give us an annual count of how many cases each SCOLR decided based on state or federal case law. We then used the state count for the dependent variable, which is a count of the number of cases per year decided by each SCOLR in favor of a criminal defendant and based on state grounds.
Independent Variables
For independent variables, we use several control variables suggested by previous literature and created our own key variables. One explanation is judicial ideology. To determine if judicial ideology is a primary factor in shaping SCOLR decisions, we used the “party-adjusted judge ideology” (PAJID) scores developed by Reference BraceBrace et alia (2000). These scores measure liberalism and range from 1 (conservative) to 100 (liberal). In the area of rights of the accused, in which a good deal of attitudinal research has been conducted (see Reference Segal and SpaethSegal & Spaeth 2002), judicial liberalism favors broad readings of such rights, so that a more conservative justice or court should less inclined to favor a claim based on the constitutional rights of the accused, all else equal.
However, as we have argued, policy preference measures alone are not sufficiently attentive to the varying political circumstances of SCOLRs. Thus, we produced a set of variables to indicate whether a given SCOLR falls into one of the three regimes detailed above. To create these variables, we needed measures of SCOLR, USSC, and STATE ideology that were strictly comparable to insure both cross-institutional and cross-time comparability of preference measures (Reference Bailey and ChangBailey & Chang 2001). To do so, we used the state ideology scores created by Reference BerryBerry et al. (1998) of state citizen ideology,Footnote 10 the Brace et alia PAJID measures of state judicial ideology (2000), which are themselves derived from the Reference BerryBerry et al. (1998) state ideology scores (the adjustment incorporating the partisan affiliation of the individual justice) and thus strictly comparable, and we also derived a comparable measure of USSC ideology.
The two leading measures of USSC justice preferences, the Reference Segal and CoverSegal-Cover (1989) and Reference Martin and QuinnMartin-Quinn (2002) scores, have their own scales and are not compatible with the other measures used in this study. We do not know if the most conservative position on the Segal-Cover or Martin-Quinn scales corresponds to the most conservative position on the Berry et alia or PAJID scales and vice-versa. To fit values of USSC ideology into the ideology space defined by the Berry et alia measures, we harmonized the scales of the Martin-Quinn ideology scores with the Berry measures using two “bridge” points: Justices Sandra Day O'Connor and Clarence Thomas, two contemporary Supreme Court justices, also served in appointed positions in state government.Footnote 11 We assigned each of our bridge justices the Berry et alia government ideology score for the year the future justice was appointed to the position in that state (O'Connor=49.62, Thomas=32.85). The scales were harmonized by attributing the ideology of the state government, not the state court, in which they served at that time. As only two fixed points are needed to define a linear scale, we calculated “stretch” and “shift” parameters similar to the adjustment performed by Reference GrosecloseGroseclose et alia (1999) for ADA scores.Footnote 12 This allowed us to translate the other justices into the same metric while retaining the relative placements of each justice estimated by Reference Martin and QuinnMartin and Quinn (2002).Footnote 13 This method assumes that justices' ideologies are consistent through time and measured accurately by the state governments that appointed them. The PAJID scores make the same assumptions, since they are constant and based on the ideology of the state.
While there are criticisms of the Berry et alia measures, they have been used in several studies of policymaking (see, e.g., Reference Scholz and WoodScholz & Wood 1999; Reference Howard and NixonHoward & Nixon 2002), and using these measures allowed us to examine comparable scales for all the institutions in our regimes. We then created three dummy variables indicating comparative ideological placements of the concerned institutions.Footnote 14 As described above, in each of the three regimes of interest, the SCOLR is to the left of the USSC and thus would prefer to articulate a state law doctrine more protective of defendants. In each regime, however, the ideology of the state citizenry is in a different position relative to the SCOLR and USSC. State-years in which the SCOLR was to the right of the USSC fell into none of the regimes of interest above. For these observations, all three of the regime indicators were coded zero, so that the effects of the regime variables included as covariates were in comparison to a baseline of state-years in which the SCOLR was more conservative than the USSC.
As our discussion of prior research shows, institutional constraints associated with judicial retention also play a role constraining judicial activism. Thus we included variables to capture differing judicial independence based on the length of judicial terms and the method of judicial selection. We expect that judges who are subject to recurring competitive elections are more responsive to public sentiment on criminal justice issues and less inclined toward judicial activism than justices retained through other methods. In a related note, we recognize that judges, like all political actors, assume that the public has a very short memory span, and previous research suggests that judges' responses to electoral pressures heighten as reelection nears (Reference Huber and GordonHuber & Gordon 2004). We suspect that judges (elected or appointed) with longer terms may be more likely to practice activism than those with shorter terms. We thus included a variable indicating the length in years of the judicial term.
Environmental factors native to the state may also influence a court's behavior. An important factor that could contribute to a state court's production of state law decisions is the rate of case production in the judicial system. States that generate more criminal cases may produce more state law decisions for that reason alone. Thus, using data from the Uniform Crime Reporting Program, we introduced a variable measuring the yearly arrest rate in the state to control for this “exposure” factor.Footnote 15 We also included a measure of the percentage of population living in urban areas as a variable to control for the possibility that criminal justice issues may be more salient in states with proportionately greater urban populations. In such an environment, a SCOLR may be reluctant to extend the rights of the criminally accused.
Our data are composed of separate time series for each state in the study. In order to account for persistence in the production of state law decisions, we included a lagged dependent variable among the covariates. As indicated above, this reduced our data set to 12 years, or 600 observations. Missing items for various data reduced the set analyzed to 574.
Methodology
Since our dependent variable was a discrete count of the number of cases in each year produced by the SCOLR grounded in state law, an event count model was appropriate. Inspection of the data presented us with two major problems, however. The standard method for discrete, nonzero count outcomes, the Poisson regression model, makes the restrictive assumption that the conditional mean and variance of the model are equal (Reference Cameron and TrivediCameron & Trivedi 1998:21). However, our data were produced by 49 separate judicial systems,Footnote 16 each with qualities likely to cause more dissimilarity in state law production than the Poisson model assumes. In fact, the observed variance of the outcome was more than three times its mean. Of separate concern is the fact that 240 of our 574 observations, more than 40%, were zeros, when we should expect merely 113 under the Poisson assumption. The preponderance of zero observations suggests that separate factors may keep certain courts from ever producing pro-defendant criminal search-and-seizure cases decided on an independent state law basis.
To deal with both of these problems, we estimate a zero-inflated negative binomial regression model. This model deals with overdispersion, the greater than expected variation evident in the data, by modifying the conditional mean and variance functions, and it specifies a separate data-generating process for some of the observed zero counts. The model is discussed in more detail in the Appendix.
The notion that certain conditions may increase the likelihood of a SCOLR failing to produce state grounds decisions seems especially appropriate in attempting to capture and understand the influence of certain legal, contextual, and institutional variables on the production of legal doctrine. Language in the state constitution explicitly tying its interpretation to that of the U.S. Constitution might rule out the possibility of state law innovation completely, rather than merely dampen it. Thus, we introduced a variable indicating the presence of such language in the zero-inflation equation.
Comparative studies suggest that political fragmentation is a necessary, but not sufficient, cause of judicial independence. Therefore, we also included a measure of state political competitiveness, that of Reference Holbrook and Van DunkHolbrook and Van Dunk (1993), to the zero-inflation equation.Footnote 17 Finally, the effect of intermediate courts on the size and composition of SCOLR caseloads and the consequent effects on the work that the justices are able to do leads us to expect that the presence of at least one intermediate court of appeals in a state judicial system will provide the conditions necessary for a SCOLR to engage in state constitutional interpretation.Footnote 18
We list the complete set of the variables in our model, with descriptive statistics, in Table 1. The overall average rate of production was 1.62 state law cases per year, with Regime 2 having the highest yearly average of the regimes. All three of the coded regimes, taken together, constituted only 20% of the observations, indicating that in most state-years the SCOLR is more conservative than the USSC. Justices in 51% of our observations retained their seat via competitive election, the average length of a judicial term was slightly less than 11 years, and more than two-thirds of the states have intermediate appellate courts.
Table 1. Descriptive Statistics
Notes: N = 574. Observations are state-years, 49 states over 12 years with isolated observations deleted due to missing data.
Results
We present the results of our zero-inflated negative binomial model in Table 2. Since we have directional hypotheses, our hypothesis tests are one-tailed. Our first three hypotheses concern SCOLRs in the three target regimes and were tested with indicator variables included for each. The hypotheses are substantially confirmed. As we anticipated, all three coefficients are positive and Regime 2, which we expected to have the most powerful effect, is largest in magnitude and statistically significant, even using conservative robust standard errors clustered to account for the nonindependence of the observations.Footnote 19 The coefficient for Regime 2 (0.670, p<0.01) is positive and significant, producing an incidence rate ratio of 1.95. The incidence rate ratios allowed us to compare the expected number of counts for different hypothetical observations in a manner similar to odds ratios for binary outcomes. Thus, state courts in Regime 2 are expected to produce nearly twice as many decisions based on state law than courts in the baseline regimes, in which the USSC is to the left of the SCOLR. We also reasoned that courts in Regime 3 are ideologically inclined to produce state law decisions but unlikely to have the political opportunity to do so, and the variable carries a positive coefficient, but it is small substantively and insignificant in relation to its standard error.
Table 2. Event Count Model for the Production of State Law Decisions by State Courts of Last Resort, 1982–1993
* p<.05,
** p<.01,
*** p<.001 (one tailed).
Standard errors are robust, corrected for clustering on states. N=574.
Our fourth hypothesis, that justices subject to competitive election are less likely to produce state law–based decisions, was tested with a variable indicating that condition and interactions of that variable with the regime indicators. Although the competitive elections indicator itself does not produce a statistically significant coefficient, the effect of competitive elections can be seen in its interactions with the regime variables. All of them carry negative coefficients, opposite the constituent regime terms. The election interactions for Regimes 1 and 2 are not significant statistically, but that of Regime 3 is negative and significant, and the incidence rate ratio reveals that judges in that regime who could face election challenges are predicted to produce half as many state law decisions as those who do not.
Turning to other variables, we note that many produce coefficients that, while not significant statistically, carry expected signs. We expected that length of term would have a positive effect on the production of state grounds decisions, but while the coefficient on the relevant variable is positive, it is not significant statistically. We hypothesized that the presence of an explicit, constitutional right to privacy would have a positive effect on production of state law decisions, but its coefficient is negative. It is quite small, however, and statistically not distinguishable from zero, so it appears to have no real effect. A more urbanized population carries the expected negative coefficient but is also not significant statistically. The coefficient for the state's arrest rate is positive but vanishingly small and not significant. Inclusion of the lagged dependent variable indicates substantial persistence in the production of state law decisions, which could be interpreted as the effect of unspecified state-specific political conditions or, more substantively, as the influence of stare decisis. Previous state law decisions that identify legal protections against searches provide fertile ground for additional decisions.
The zero-inflation equation estimates the effect of several variables on the likelihood that courts in the sample will produce no state law decisions. Thus, the coefficients are interpreted as the effect on the probability that the SCOLR will be in a “zero state.” All three of the variables in the equation are significant and have the effects anticipated. State constitutional language binding the interpretation of rights to that of the U.S. Constitution increases the likelihood that the state court will remain in a zero-producing state. Judicial structure also matters. The impact of having an intermediate appellate court is clearly significant, decreasing the likelihood of a SCOLR producing no state law decisions in a given year. The same effect is observed for political competitiveness. Increases in political competitiveness reduce the probability of a state court remaining inactive in state law decisions. Finally, the joint Wald statistic for both equations is highly significant (χ2=325.22, p<0.001), indicating that the covariates are not jointly zero.
Discussion and Conclusion
Modern federalism can offer greater protection of individual rights through regional courts and regional constitutions. However, political, institutional, and other conditions contribute substantially to realizing greater judicial protection of civil liberties. In addition, the variation in legal materials matters. Explicit state constitutional language binding interpretation of rights to the U. S. Constitution matters tremendously, as do previous state law decisions. The results of our analysis reveal SCOLR justices who, rather than having misunderstood or neglected Michigan v. Long (1983), as many scholars have concluded, can and will rely on state constitutional grounds when they have the incentive and a favorable context in which to do so.
Following Reference LatzerLatzer (1991) and other scholars, our data indicate that the conservatism of the state courts themselves is a primary reason for the lack of evidence for a substantial embrace of the “new judicial federalism.” Only about 20% of the state courts in our 13-year, 49-state sample fell to the left of the USSC according to our ideology measures, corrected to occupy the same scale. Thus, not many SCOLRs are ideologically inclined to extend greater legal protection to accused criminals. However, when they are, state court justices appear to do so in a predictable, rational fashion.
When SCOLRs have the intent to craft state law protections for civil liberties in search and seizure and are not inhibited from doing so by political forces within the state, they do try to enhance this protection. Specifically, we find that when the citizenry and court both fall to the left of the USSC, SCOLRs are substantially more likely to move under their own law. State court justices are sensitive to the attitudes of the citizenry, and their sensitivity is enhanced by the use of competitive elections to retain justices. Our results also indicate that SCOLR justices are mindful of reactions from the USSC as well, choosing to rely explicitly on state constitutional grounds and thus deter federal review when the circumstances permit. We find support for the judicialization of politics stressed by some comparative scholars as well. Greater political fragmentation in the other branches increases the likelihood that the SCOLR will take action through the constitution of the state.
We also find a powerful effect of constraints placed on SCOLRs by the language of the constitutions they expound. Although articulated privacy rights do not have a significant effect on state law doctrine in search-and-seizure decisions, specific provisions binding the interpretation of state legal rights to the interpretation of coordinate federal rights appear to prohibit the production of independent state law doctrine. This result supports the notion that the legal environment in which law is produced influences the basis upon which state court decisions are grounded and the subsequent content of legal doctrine.
Our analysis has been confined to a specific issue, search-and-seizure decisions, which are likely to be of high salience and more conducive to the constraints of within-state accountability. Search-and-seizure law has also been substantially “federalized” since the 1960s. We believe that our findings can be generalized to other civil liberties issues that also conform to these descriptions, but production of state law doctrines in other areas that have seen less activity at the federal level may depart from these observations. Our research should serve as a valuable starting point for future study of federalism and of judicial innovation in a federal framework.
Appendix: The Zero-Inflated Negative Binomial Count Model
Use of the zero-modified, or zero-inflated, negative binomial (ZINB) model addresses two separate problems. Our dependent variable, the number of state law–based decisions produced per year by SCOLRs, is a discrete outcome conforming roughly to a Poisson distribution, but it demonstrates variance much greater than its mean and has a proportion of zero counts much larger than what one would expect from a Poisson distribution. Thus, we use a model that accounts for heterogeneity in the variance function and a separate process that generates zero counts exclusively.
The negative binomial model modifies the variance estimation of the Poisson exponential regression model to reflect individual heterogeneity often observed in cross-sectional data (Reference Cameron and TrivediCameron & Trivedi 1998:100–1). The distribution itself is similar to Poisson:
where
The vector of covariates is denoted x′i, and β is a vector of coefficients to be estimated. The variance (ω i) of the negative binomial mean dispersion model estimated by Stata (referred to by Reference Cameron and TrivediCameron and Trivedi (1998) as the NB2 model), however, is a quadratic function of the conditional mean with an estimated parameter (α) on the squared term (Reference Cameron and TrivediCameron & Trivedi 1998:63):
As the negative binomial model reduces to Poisson when α=0, a test of this hypothesis can be conducted to assess whether the data are better described by a Poisson or negative binomial distribution. We provide such a test in Table 2 and find that the data are significantly overdispersed, as expected from our cross-section of American states.
The second departure is the specification of a separate process producing zeros. Such a process can cause overdispersion, which will cause excess zeros (Reference Cameron and TrivediCameron & Trivedi 1998:99). However, even when excluding the zero counts from our data, the variance of the dependent variable remains more than twice the mean of the remaining observations. In addition, the overall number of observed zeros is 212% of the number expected from a Poisson distribution with the parameters of our data set.
The ZINB model posits that two separate processes are at work modeled by two different probability distributions, one producing zero counts and another producing nonnegative counts according to a negative binomial distribution. So the realized outcome is a combination of the negative binomial process and a separate dichotomous process that produces zeros with probability φ. The probability of being in the zero-only state can itself be parameterized with covariates as a probit equation:
where z′i is a vector of covariates, γ is a vector of coefficients to be estimated and Φ is the standard normal cdf.
The zero-inflated count model may produce zeros as a realized negative binomial outcome. In other words, the model does not assume that all zeros are a result of the zero-producing process, as indicated below:
The probability of observing a zero is the sum of the first line and the second, with r=0.
The standard negative binomial model is not nested within the zero-inflated version, but a normally distributed test statistic for non-nested models, called the Vuong statistic, can be adapted to compare the models. In this instance, the Vuong test cautiously recommends the zero-inflation model. The statistic is positive (1.22), favoring the zero-inflation, and rejects the null at an alpha of 0.10.
The conditional probability structure of the ZINB model recalls the Heckman approach to solving sample selectivity bias in regression models (Reference HeckmanHeckman 1979). The negative binomial, zero-modified, and sample selection models all deal in some way with unmodeled heterogeneity of the outcome, and both zero-inflation and sample selection specify the data as generated by two processes. Reference GreeneGreene (1997:13–4) and Reference Cameron and TrivediCameron and Trivedi (1998:342–3) address the similarities and differences between zero-modified count models and models for sample-selected data. In the sample selection context, biased estimates from the count model result from correlation of the unmodeled heterogeneity with a selection mechanism. Greene, however, notes that zero-modified models are directed at modeling deviations of the observed data from the Poisson distribution. The zero-modified approach specifies the production of a specific outcome, zero, but this approach does not consider possible correlation between the two functions.
Reference Cameron and TrivediCameron and Trivedi (1998) discuss the reinterpretation of a “hurdle” model as a sample selection model. The hurdle model is a zero-modified count model in which all zeros are produced by the zero-only state as in 1.5 above, but with r strictly positive. Once an observation overcomes the zero-state, its outcome is governed by a Poisson or negative binomial distribution truncated at zero. Like the zero-inflated version, use of this model might be motivated by a theoretical expectation that the zero counts in data are produced by a different process, although in the hurdle model they are produced exclusively by that process. This model would be appropriate for our data if we had cause to believe that all state courts would produce state law decisions unless another process restrained them from doing so, but our model allows zeros to result from the “normal” count process as well.
