Public Responsiveness to Supreme Court Decisions

Transcription

1 Public Responsiveness to Supreme Court Decisions Joseph Daniel Ura ABSTRACT Much research concludes that public reactions to Supreme Court decisions are marginal, conditional, or highly localized. This paper reevaluates that result in light of substantial support for the theory of thermostatic public responsiveness to changes in public policy and the macro economy. This macro political theory suggests that there will be negative feedback from the ideological direction of Supreme Court decisions in public mood. I assess this expectation, modeling Stimson s public mood index as a function of cumulative Supreme Court liberalism, controlling for congressional policymaking and the state of the economy. Consistent with a thermostatic theory of public opinion, I find significant evidence of a negative relationship between Supreme Court liberalism and public mood. This result indicates that Supreme Court decisions may loom larger in the public mind than current scholarship suggests and have consequences for American national elections and patterns of public policymaking. Keywords: macro politics, public opinion, Supreme Court, thermostatic models, time series. Paper prepared for presentation at the 2010 Annual Meeting of the Midwest Political Science Association. Joseph Daniel Ura is Assistant Professor, Department of Political Science, Texas A&M University, Box 4348 TAMU, College Station, TX (jura@politics.tamu.edu).

2 Over the last six decades, federal courts in general, and the Supreme Court in particular, have become increasingly prominent and institutionalized components of American national government, seeking to control matters at the heart of contemporary politics (Kramer 2004, p. 227; see also Burns 2009; McGuire 2004). At a minimum, the federal courts now define the limits of state and national policymaking on a variety of salient issues. 1 More profoundly, though, the modern Supreme Court has increasingly positioned itself as the final and authoritative voice on constitutional interpretation. As Chief Justice Rehnquist wrote in United States v. Morrison (2000), No doubt the political branches have a role in interpreting and applying the Constitution, but ever since Marbury this Court has remained the ultimate expositor of the constitutional text (see also Friedman 2009; Starr 2002; Tushnet 2000; and also Tate and Vallinder 1995 for a comparative perspective). The growing importance of the judiciary in American national politics is evident in several ways, including the emergence of judicial nominations as a salient issue in presidential election campaigns (Nemacheck 2004; Pickerill and Clayton 2004; Stephenson 1999) and the growing role of interest groups and public campaigns in the confirmation of federal judges and Supreme Court justices (Caldeira and Wright 1998; Grossman and Wasby 1972; Maltese 1998; Segal, Cameron, and Cover 1992). 2 1 These include the death penalty (Furman v. Georgia 1972; Gregg v. Georgia 1976; Atkins v. Virginia 2003), affirmative action (Regents of the University of California v. Bakke 1978; Gratz v. Bollinger 2003; Grutter v. Bollinger 2003), abortion (Roe v. Wade 1973; Harris v. McCrae 1980; Planned Parenthood v. Casey 1992; Gonzales v. Carhart 2007), the conduct of the War on Terror (Hamdan v. Rumsfeld 2006), and even standards for counting contested ballots in presidential elections (Bush v. Gore 2000). 2 Though the federal judiciary has emerged as an important issue in one form or another in presidential election campaigns throughout American history, attention to Supreme Court appointments has become a steady theme of campaigns for the White House during the last three decades. This pattern has been especially prominent in recent elections. In 2004, Senator John Kerry rallied supporters for his presidential campaign by pointing to the election s implications for ideological balance on the nation s highest court: You want three words of motivation?... The Supreme Court. That alone is the motivation (Johnson 2004). In 2008, Senator John McCain promised to use the power of the presidency to correct the one great exception to the balanced functioning of the federal system of checks and balances represented by the abuse of our federal courts by the people we entrust with judicial power who legislate from the bench... [against the] expressed will of the voters (Toobin 2008). Endorsing then-senator Obama s presidential candidacy in 2008, the editors of Esquire magazine wrote, The best argument for the election of Barack Obama as president of the United States is written quite clearly in the peaks and squiggles of [89 year old Associate Justice] John Paul Stevens s EKG (Esquire 2008). Similarly, the process confirming Supreme Court justices has become high political drama in its own right. Confirmation votes now represent a seismic... struggle between the President and the US Senate over the ideological makeup of the nation s highest court (Johnson and Roberts 2004, p. 663). The magnitude of the struggle is evident in the (increasingly) intense efforts of organized interest groups to generate and amplify public support for and opposition 1

3 Despite these trends, the scholarly literature on public reactions to Supreme Court decisions generally concludes that the Court has, at most, marginal, conditional, or highly localized effects on public opinion (e.g. Baas and Thomas 1984; Franklin and Kosaki 1989; Hoekstra 2000, 2002; Hoekstra and Segal 1996; Johnson and Martin 1998; Marshall 1989; Mondak 1991). These conclusions stand in sharp contrast with the role of the federal courts in salient policy domains and the attention given to courts by presidential candidates and interest groups. Indeed, the juxtaposition of growing judicial entanglement with ordinary politics and scholarly conclusions that the Supreme Court is generally inconsequential for national public opinion raises a critical problem for students of judicial politics and mass political behavior. Are accounts of the growing role of the Supreme Court in normal politics much ado about nothing in the minds of ordinary Americans, or have scholars of linkages between Supreme Court decisions and public opinion missed something important about how the Court s decisions influence public sentiment? In this paper, I reconsider the role of the Supreme Court in shaping public opinion. I argue, first, that a lack of evidence of attitude changes in the wake of Supreme Court decisions that is, changes in the distribution of absolute preferences on an issue or of ideal points in some policy space does not necessarily imply a general lack of public reaction to the Court. Next, I draw on the macro politics literature to identify an alternative theory of public responsiveness to the Supreme Court. In particular, I show that the thermostatic model of public opinion predicts that the ideological direction of Supreme Court decisions should be negatively related to public preferences over the ideological direction of future policymaking. I assess this expectation by modeling Stimson s (1999, 2009) mood index as a function of liberalism in Supreme Court decisions, controlling for policymaking by Congress and the state of the macro economy. As predicted, the model indicates a significant, negative relationship between Supreme Court decision-making and public mood. to individual nominees on the basis of the expected policy implications of their membership on the Court (Caldeira and Wright 1998; Grossman and Wasby 1972; Maltese 1998; Segal, Cameron, and Cover 1992). 2

4 The Supreme Court and Public Opinion: Legitimation The foundation of contemporary political science s consideration of the relationship between the Supreme Court and public opinion is the legitimation hypothesis due to Dahl (1957). Dahl argued that the Supreme Court s ability to make policy is highly constrained, and that its principal function is to attach the unique legitimacy attributed to its interpretations of the Constitution... [to] the fundamental policies of the successful [governing] coalition (1957, pp ). Generally, Dahl s statement has been interpreted as a theoretical prediction that the Supreme Court s decisions are persuasive, shap[ing] policy attitudes on even the most controversial issues (Hoekstra 2002, p. 90). 3 In other words, the legitimation hypothesis has been presented as a prediction that the Supreme Court s decisions have the ability to influence individuals absolute policy preferences their ideal points in a policy space changing people from segregationists to integrationists, from pro-lifers to pro-choicers (Caldeira 1991, p. 305). However, empirical support for this proposition has been limited. For example, Marshall (1989) rejects the legitimation hypothesis. Examining eighteen instances over a 45-year period where survey data exist on aggregate public opinion on issues shortly 3 Dahl s own statement of the legitimation hypothesis is actually more circumspect than its characterization in subsequent literature suggests. For his part, Dahl writes that the Court may attach the unique legitimacy of its interpretations of the Constitution... [to] the fundamental policies of the successful [governing] coalition. (1957, pp ). Yet, Dahl also notes, In the absence of substantial agreement within the [national governing] alliance, an attempt by the Court to make national policy is likely to lead to disaster, as the Dred Scott decision and the early New Deal cases demonstrate (1957, p. 293). Dahl is careful to confine his speculation on the Court s powers to the domain of legitimacy rather than the realm of preference. Thus, Dahl s essay does not suggest that the Supreme Court can manufacture support for this or that policy out of thin air. Rather, Dahl suggests that the Court may legitimize some critical policies which have already been adopted by a persistent, national lawmaking majority, such as the Jeffersonian alliance, the Jacksonian, the extraordinarily long-lived Republican dominance of the post-civil War years, and the New Deal alliance shaped by Franklin Roosevelt (1957, p. 293). This begs the question, If legitimacy is not support, what is it? Caldeira (1991) suggests several alternatives: including compliance with policies at odds with one s preferences and the opportunity to persuade the [disagreeing] public of governmental agencies rights to undertake action (p. 305). While it is difficult to discern the original meaning, as it were, of Dahl s 1957 essay, Caldeira s latter suggestion seems closest to the mark. Elsewhere, Dahl argues that a political system does not depend on a widespread belief that certain... rules... are preferable to other rules; in some circumstances a democratic system could be highly stable if a substantial part of the electorate merely accepted them (Dahl 1961, p. 314). Though it is not stated explicitly, it is reasonable to infer that Dahl s sense of legitimation relates to the notion of acceptance, that his concept of legitimizing is more strongly associated with deescalating political opposition to a majority policy from the realm of constitutional objections to the domain of mere disagreement. 3

5 before and after relevant Supreme Court decisions, Marshall finds that public opinion is just as likely to shift away from the Supreme Court s positions as toward them. He concludes that there is no systematic relationship between Supreme Court decisions and the public s absolute issue preferences, let alone a consistent pattern supporting the legitimation hypothesis. Franklin and Kosaki s (1989) study of public reaction to Roe v. Wade (1973) also rejects the legitimation hypothesis, showing that the decision had only a marginal influence on the aggregate distribution of support for abortion rights. In particular, they find that exposure to Roe had a polarizing effect on abortion attitudes, catalyzing both abortion supporters and opponents to become more extreme in their issue positions. Using data from the General Social Survey, Franklin and Kosaki show that the predicted probability of a white, Catholic respondent opposing all discretionary abortions increased from 0.51 in 1972 (pre-roe) to 0.53 in 1973 (post-roe) while the probability of a Protestant respondent opposing all discretionary abortions declined from 0.34 in 1972 to 0.28 in Johnson and Martin (1998) also reject the legitimation hypothesis, finding evidence that some Supreme Court decisions have a marginal but significant polarizing effect on abortion and death penalty attitudes. However, Johnson and Martin also refine Franklin and Kosaki s structural response hypothesis, showing that the Supreme Court may affect public opinion when it initially rules on a salient issue, but that subsequent decisions on the same issue will have little influence on opinion (1998, p. 299). Evidence of the Supreme Court s limited powers to change the nation s mind on salient policy questions has catalyzed research on more focussed public responses to the Court s decisions. Hoekstra s (2000, 2003; Hoekstra and Segal 1996) analyses of the Supreme Court s influence on local public opinion begin with the premise that individual-level responsiveness to Supreme Court decision-making is conditioned by the strength of pre-existing attitudes and exposure to information about the Court and its decisions, which corresponds with individual s geographic proximity to a community in which a Supreme Court case has originated. Hoekstra finds much empirical support for this position, demonstrating that support for policy positions adopted by the Supreme Court increases marginally among individuals who live in communities nearby those 4

6 in which Supreme Court cases originated and not at all in communities far removed from the specific controversy addressed by the Court (2000, 2002). Experimental studies, have generally confirmed this argument, showing that sufficient information about Supreme Court decisions can be persuasive in some non-salient issue domains for which respondents may not have strong prior attitudes (Mondak 1994; Hoekstra 1995; Unger 2008; but see Bass and Thomas 1984). Taken as a whole, the literature suggests that public reactions to Supreme Court decisions are, at most, modest and conditional. On the one hand, observational evidence of legitimation is limited to local attitudes on a small number of issue domains. On the other, polarizing structural responses to the Court s decisions are limited to a handful of landmark cases. Indeed, it appears that Caldeira s now-classic critique of legitimation that we have relatively few well-documented instances when the Supreme Court has shaped the aggregate distribution of public support for this or that policy remains substantially accurate (1991, p. 312). 4 A Different Perspective: Thermostatic Responsiveness Evidence of the Supreme Court s limited ability to legitimize policy attitudes does not necessarily imply that public opinion is unresponsive to the Supreme Court in other ways, and it is a mistake to generalize that the absence of one kind of public reaction to the Supreme Court entails a general lack of engagement with the judiciary. Indeed, the public may observe and react to Supreme Court decisions in politically relevant ways even if the Court does not regularly lead Americans to change their attitudes about this or that policy. For example, the degree of ideological convergence between aggregate patterns of Supreme Court decision-making and public mood predicts changes in the level of public confidence in the Court (Durr, Martin, and Wolbrecht 2000), which, in turn, 4 This conclusion is more than an empirical curiosity. The public s apparent lack of responsiveness to judicial decision-making along with some evidence that it is generally ignorant of the Supreme Court (Delli Carpini and Keeter 1996; Morin 1989; Kritzer 2001; but see Caldeira and McGuire 2005; Gibson and Cladeira 2009) have combined to create a folk theorem among many judges, legal academics, and political scientists that ordinary citizens cannot (and perhaps should not) meaningfully assess or sensibly respond to the work of judges and courts. Gibson and Caldeira (2009, p. 429) summarize this stark syllogism, [I]f the American people are irretrievably ignorant of law and courts, elites preferably legal elites should be given the task of [evaluating courts and] determining who should, and should not, be a judge. 5

7 influences congressional support for the Supreme Court (Ura and Wohlfarth N.D.) and judicial self-restraint (Clark 2009). This result suggests that, despite the supposed imperceptibility of the Court... the public perceives... and evaluates the Court (Durr, Martin, and Wolbrecht 2000 p. 775). More generally, this ideological divergence result suggests that (at least some) individuals are capable of comparing the political content of Supreme Court decisions with their own policy preferences to render a judgment about the fitness and faithfulness of the Court as a governmental agent. The ability to compare the Court s decisions with personal policy preferences suggests the potential for thermostatic responsiveness to Supreme Court decision-making. Thermostatic models of public opinion posit a signalling relationship between the public and the government: when policy deviates from the public s ideal position, it communicates a desire for policymakers to adjust policy in a corrective direction (Wlezien 1995, 1996). 5 In general, thermostatic models posit that as public policy changes, individuals amend their preferences for future policy changes in response, demanding relatively more of some policy as an outcome moves away from their preferred state of the world and relatively less of the policy as the observed state of the world converges with their preferred states. These individual changes in relative preferences aggregate up into a macro-level signal of the mass public s demand (or lack thereof) for future (corrective) policy change. To visualize this process in the context of judicial decision-making, consider a hypothetical uniform distribution of citizens in a liberal-conservative policy space where there exists some status quo policy, Q 1 (Figure 1). Each individual in this space has some absolute policy preference, an ideal point. Each also has a relative preference for the direction of future policymaking those to the left of the status quo prefer more liberalism in the future; those to the right would prefer less 5 Thermostatic models of the linkages between public opinion and government have been applied broadly to study the national government s responsiveness to changes in public opinion including the Supreme Court s responsiveness to changes in public mood (Durr 1993a; Ellis, Ura, and Ashley-Robsinson 2006; Erikson, MacKuen, and Stimson 2002; McGuire and Stimson 2004; Stimson, MacKuen, and Erikson 1995; Ura and Ellis 2008; Wlezien 1995, 1996) and the public s responsiveness to policymaking by Congress and the President (Enns and Kellstedt 2008; Erikson, Stimson, and MacKuen 2002). Yet, their application to the study of public responsiveness to the Supreme Court has been more limited, as I discuss below. 6

8 New Policy (Q 2 ) Old Policy (Q 1 ) Liberal Conservative Consistent supporters of "more liberalism." Consistent supporters of "less liberalism." Prefer "more liberalism" under the old policy, "less liberalism" under the new policy. Figure 1: The Supreme Court and Relative Policy Preferences liberalism. Suppose a Supreme Court decision had the effect of changing public policy in a liberal direction, creating a new policy at Q 2. As the literature suggests, the Court s decision is not likely to provoke much change in the underlying distribution of policy preferences. However, the change in policy would produce a reduction in the proportion of citizens expressing a preference for more liberalism in future policymaking. Citizens whose ideal points fall between the old policy and the new policy would have expressed a preference for more liberalism before the Court acted and less liberalism afterward. So, while the Court may have no effect on the aggregate distribution of absolute policy preferences (policy attitudes), it may have an important influence on the distribution of relative preferences. As the Court makes more and more liberal or conservative decisions over time, some individuals will continue to exhibit changes in their relative preferences for future policy changes in the opposite direction of policymaking. In other words, there will be negative feedback from policymaking by the Supreme Court in the individuals relative policy preferences. This micro theory also has a clear macro implication: the accumulation of public policy changes emanating from the Court should produce dynamic, negative responses in the public s aggregate preferences for the direction of future policy change. While there is much evidence showing this type of mass responsiveness to important policy changes promulgated in the elected branches of national government (Enns and Kellstedt 2008; Erikson, Stimson, and MacKuen 2002), general public reactions to Supreme Court decisions have 7

9 not been incorporated into the literature on thermostatic public opinion. 6 Yet, the Supreme Court s ability to make public policy that markedly rearranges the prior distribution of political benefits, either material or symbolic, for various segments of the population suggests the potential for similar dynamics in public opinion in the wake of important Supreme Court decisions (Flemming, Bohte, and Wood 1997, p. 1247; but see Rosenberg 1992). The emergent prediction that relative public opinion liberalism is affected by Supreme Court decision-making can be subjected to empirical scrutiny. Assessing Thermostatic Responses to Supreme Court Decisions The extant literature indicates a relatively straightforward set of predictive influences on public mood. These influences establish a useful baseline model of mass policy sentiment into which Supreme Court decision-making may be integrated. The first of these is public policy created in the elected branches of national government, usually measured as cumulative legislative enactments (Erikson, MacKuen, and Stimson 2002; Kelly 2009) or federal spending (Wlezien 1995, 1996; Wlezien and Soroka 2010). Regardless of the measurement approach employed, these models regularly show negative feedback in public opinion from changes in public policy. The second baseline predictor of policy sentiment is the state of the macroeconomy, usually indicated by the rates of inflation and unemployment (Enns and Kellstedt 2008; Erikson, MacKuen, and Stimson 2002). Increased inflation predicts greater conservatism in mood while greater unemployment predicts more public opinion liberalism. The thermostatic theory of public responsiveness to Supreme Court decision-making may be assessed by incorporating estimates of the effects of Supreme Court decisions on public mood into this baseline model. 6 Two studies, though, do find negative relationships between the ideological direction of decision making of federal courts and public opinion on some issues. Page, Shapiro, and Dempsey (1987) find that the direction of federal courts decision-making has a negative relationship with public opinion on several issues. They conclude that the unpopular federal judiciary of the period they analyzed was a point of negative reference, and that When their [the federal courts ] statements and actions push in one direction... public opinion tends to move in the opposite direction (1987, p. 32). Similarly, Wlezien and Goggin (1993) find that public opinion in support of abortion policy as it is now increased during the 1980s as the Supreme Court permitted the states to regulate abortion more strictly over that decade and interest groups became more active in that policy domain. 8

10 Measurement and Data I measure aggregate public opinion liberalism using the 2008 estimates of Stimson s (1999; 2009) annual mood index. Inflation is the the percentage change in the Consumer Price Index (January to December) in each year (Bureau of Labor Statistics 2009a). Unemployment is the average annual rate of unemployment (Bureau of Labor Statistics 2009b). Since the Supreme Court s outputs are discrete decisions, which are more analogous to individual pieces of legislation rather than budget authorization, I measure public policy based on counts of major legislative enactments rather than budget authorizations. Following Erikson, MacKuen, and Stimson (2002) and Kelly 2009(), I begin with Mayhew s (1991, 2008) list of major or important pieces of legislations passed in each year (selected based on media coverage of Congress), code each law as liberal or conservative, and take the net number of important liberal laws created in each year as an indicator of the annual amount of policy liberalism created by Congress. Next, these values are used to construct a cumulative a measure of cumulative policy liberalism produced by Congress and the president by scoring each year s policy outputs as the difference between its value and the mean of the annual series and then taking the sum of the resulting series at each point in time. Erikson, MacKuen, and Stimson refer to the resulting series as the policy index, a convention I also adopt. The policy liberalism series, beginning in 1953 and extended through 2006, is illustrated in Figure 2(a). 7 To measure the ideological content of Supreme Court decisions, I compute a cumulative index that mirrors the basic approach employed in the construction of the policy index. I begin by identifying a set of important or salient Supreme Court cases from 1953 to Following Epstein and Segal s (2000) approach, I define salient cases as those with a decision mentioned on the front page of The New York Times. 8 Next, I identify the ideological direction of each case using the Original United States Supreme Court Database (2007). This allows me to compute the 7 Mayhew has extended his list of important legislation passed in each year since from 1990 to 2008 (1991, 2008). Ideological codings are based on the procedure described by Erikson, MacKuen, and Stimson (2002). A list of important legislation from 1990 to 2006 and ideological codings are reported in the appendix. 8 Epstein and Segal (2000) report cases appearing on the front page of The New York Times through 1996, and I extend the list through The list of salient cases from is included in the Appendix. 9

11 Important Laws Aggregating to Policy 15 Important Laws: Mean = 1.54 Std. dev. = 3.59 Policy Index: Mean = Std. dev. = Salient Cases Aggregating to Caselaw 26 Salient Cases: Mean = 3.11 Std. dev. = 6.82 Caselaw Index: Mean = Std. dev. = Note : Bars are the net number of liberal important laws enacted by Congress (Mayhew 1989, 2010) and salient Supreme Court decisions (Epstein and Segal 2000) in each calendar year from 1953 to Black bars correspond to labeled years. Bar values correspond to the left vertical axes, which are labeled with the observed maximum and minimum of each series. Lines are the policy and caselaw indices described in the text and represent the cumulative policy liberalism created by Congress and the Supreme Court. Line values are scaled on the right vertical axes, which are also labeled with the observed extreme values in each series. Figure 2: Measuring Cumulative Liberalism in Important Legislation and Salient Supreme Court Cases net number of salient or important Supreme Court decisions in each year. Finally, I construct a cumulative measure of liberalism in the Supreme Court s decisions by rescaling the net number of liberal decisions in each period as its deviation from the mean value of the annual Supreme Court liberalism series and taking the sum of the series at each point in time. For convenience s sake, I refer to the resulting time series, which is shown in in Figure 2(b), as the caselaw index. Model Specification and Estimation With these data in hand, I test the theory of thermostatic responsiveness to Supreme Court decisionmaking by incorporating the caselaw index into a model of public mood. Among alternative statistical models, the error correction model (ECM) which explicitly models short-run and long-run effects for each independent variable (DeBoef and Keele 2008; Beck 1993; Durr 1993; Smith 1993) stands out as an appropriate choice for assessing the dynamic linkages between Supreme Court decisions and changes in public mood. In the bivariate case, the Bardsen (1989) single 10

12 equation ECM takes the form: Y t = α 0 + α 1Y t 1 + β 1 X t + β 2X t 1 + ɛ t, (1) where α 1 indicates the speed of the reequilibration of Y to a deviation from its equilibrium with X, β 2 reflects the long run effect of changes in X on Y, and β 1 indicates the contemporaneous relationship between a change in X and a change in Y. Thus, in addition to indicating the direction and magnitude of the effect of each independent variable on changes in public mood, the ECM can also reveal the temporal dynamics of the specified predictive relationships. Though ECM was originally developed for investigating cointegrated time series, DeBoef and Keele (2008) note that it may also be applied in a variety of time series contexts in the absence of cointegration with either stationary or nonstationary data. (Indeed, they show that the ECM is equivalent to the more familiar autoregressive distributed lag model.) In addition to these attractive analytic properties, Monte Carlo experiments indicate that an ECM implemented through OLS capably recovers the data generating process even in small samples (DeBoef and Keele 2008). Thus, I implement the ECM approach, estimating a pair of models of the first difference of mood. The first is a baseline model, and estimates the effects of the first lag of mood (error correction) as well as a the first difference (short run effect) and first lag (long run effect) of policy (congressional liberalism), inflation, and unemployment. The second model builds on this baseline, estimating short run and long run effects associated with the caselaw index (Supreme Court liberalism). Since annual estimates of mood for the early 1950s are unstable due to the small number of underlying survey marginals available during that period (Erikson, Stimson, and MacKuen 2002; Stimson 1999, 2004), models are estimated for years from 1956 forward. Estimates and diagnostics for the baseline model and the expanded Supreme Court model are reported in Table 1. 11

13 Table 1: An Error Correction Model of Annual Mood ( ) Predictors (Expected Sign) Baseline Model Supreme Court Model Short Run Effects Policy Index t (+) (0.08) (0.08) Caselaw Index t ( ) -0.14* (0.04) Inflation t ( ) (0.14) (0.14) Unemployment t 1 (+) 0.63* 0.68* (0.33) (0.31) Long Run Effects Policy Index t 1 ( ) -0.05* -0.07* (0.02) (0.02) Caselaw Index t 1 ( ) 0.01 (0.01) Inflation t 1 ( ) -0.25* (0.14) (0.16) Unemployment t 1 (+) (0.24) (0.22) Error Correction, Constant, and Diagnostics Error Correction (Mood t 1 ) -0.30* -0.28* (0.08) (0.08) Constant 18.75* (5.78) (5.20) R Adjusted R Breusch-Pagan Test for Heteroskedasticity a Breusch-Godfrey LM Test for Autocorrelation b Augmented Dickey-Fuller Test for Unit Root c -8.49* -6.75* Note: OLS Estimates. Standard errors in parentheses. *p < 0.05; One-tailed tests. N = 29. a The Breusch-Pagan statistic tests the null hypothesis of constant error variance. b The Breusch-Godfrey Lagrange multiplier tests the null hypothesis of uncorrelated residuals. c The augmented Dickey-Fuller statistic tests the null hypothesis of a unit root (integrated) process in the model s residuals using MacKinnon s (1994) critical values. 12

14 Estimates of the the Baseline Model First, the baseline model recovers the key substantive results of Erikson, MacKuen, and Stimson (2002), though the error correction specification provides novel insight about the dynamics which govern public responsiveness to public policy and the state of the macro economy. Consistent with a thermostatic theory of public opinion, the model indicates a negative and significant long-run relationship between the ideological direction of policy change and public opinion liberalism. In particular, the model predicts that each additional piece of important liberal legislation passed at time t would begin to filter into mood, beginning at time t + 1, at a rate indicated by the error correction parameter (-0.30). (The estimated short run effect of policy is incorrectly signed and not significantly different than zero.) This error correction estimate indicates that the system of congressional responsiveness to public support for the Supreme Court adjusts at a modest pace, predicting that 30 percent of the predicted long run effect would appear in the mood series in year t+1, with 30 percent of the remainder appearing in year t+2, and so forth until mood has increases a total of 0.05 points. This rate of error correction predicts that half of the predicted long effect of a change in policy will appear in mood over a two year period (yielding a median lag length of 2), and that about 75% of this total effect will filter into the mood time series within four years. Of course, the magnitude of the estimated long run effect of public policy for mood is quite modest expressed in terms of a unit change. However, the baseline effect of public policy for mood is better understood relative to the observed range of the policy index. The mean of the absolute value of the policy index s year-to-year changes is 2.64, and the standard deviation of the policy index itself is This means that the change in policy liberalism produced in a typical year predicts a long run change in of 0.13 points in mood. A standard deviation increase in policy liberalism yields an expected decrease in mood of 0.92 points (roughly 0.21 standard deviations). In addition to the effects of public policy for mood, the baseline model also shows the influence of macroeconomic currents on public opinion liberalism. Again, the error correction model estimates confirm the substantive conclusions of Erikson, MacKuen, and Stimson (2002) while revealing new information about the dynamics at work in mood s over-time changes. The baseline 13

15 model shows that public mood is negatively associated with the rate of inflation in the long run and positively associated with the rate of unemployment in the short run. (Neither the short run effect of inflation nor the long run effect of unemployment are statistically significant, though both are correctly signed.) Each point increase in the rate of inflation predicts a long run decrease of 0.25 points in mood. Thus, higher inflation predicts greater conservatism in public mood. As with public policy, the model indicates that changes in inflation affect mood in a manner consistent with the estimated error correction dynamics. A one-unit increase in the public s confidence in Congress in year t disrupts the equilibrium relationship between public opinion and the macro economy, predicting a long-run decrease of 0.25 points in mood, which filters into mood at the estimated rate of error correction, This predicts that 30 percent of the predicted long-run effect would appear in the institutionalization series in year t + 1, with 30 percent of the remainder appearing in year t + 2, and so on until the full long-run decrease in mood has transpired. In contrast to the long run effects associated with public policy and inflation in the baseline model, public opinion responds to changes in unemployment in the short run. Higher unemployment is associated with greater public opinion liberalism, and each point increase in the rate of inflation predicts a short increase of 0.63 points in the mood index. In the jargon of error correction models, short run effects reflect instantaneous adjustment in contrast to the reequilibration represented by long run effects. Here, the significant short run effect associated with a unit change in unemployment at time t predicts an increase of 0.63 points in mood, also at time t. As with the effects of public policy, the practical magnitudes of the estimated effects of macroeconomic conditions for mood are somewhat obscured by expressing them in terms of single unit changes. Though the size of the effects associated with one-unit changes in inflation and unemployment are superficially much larger than the effect of a single important law, macro economic conditions are obviously measured on percentage scales that are not directly comparable to the metric of policy index. The standard deviations of inflation and unemployment are 3.04 and 1.53, respectively. The mean absolute value of annual changes in inflation is 3.81 and

16 in unemployment. Thus, the expected values of the changes in inflation and unemployment from one year to the next respectively predict a long run decrease of 0.95 points in mood and a short run increase of 1.13 in public opinion liberalism. Alternatively, a standard deviation increase in inflation yields an expected long run decrease in mood of 0.76 points (0.17 standard deviations); a standard deviation increase unemployment predicts a yields short run increase of in mood of 1.00 points (0.23 standard deviations). Taken together, the estimates in the baseline model confirm what is already known about public opinion s reactions to changes in the macro political-economy. There is negative feedback in public mood from policymaking by the elected branches of national government. Inflation generates greater public demand for conservative policies (shrinking the size and scope of government) while unemployment catalyzes greater public opinion liberalism (demand for a bigger, more active government). Additionally, the standardized magnitudes of each of these effects are roughly comparable (about one fifth of a standard deviation s total change in mood for each standard deviation s change in policy, inflation, or unemployment), indicating some balance in the collective public mind with respect to the weight assigned to these forces as they bear on mass policy sentiment. Additionally, differences in the dynamics associated with these effects (long run versus short run effects) provide interesting insight into the behavioral aspects that motivate changes in aggregate mood. Whereas public responses to public policy and inflation play out only in the long run, responses to unemployment are immediate. This may reflect the long and variable lags to borrow Friedman s (1969) evocative phrase with which policies enacted into law actually become effectual (or are even implemented) and the relatively distant time horizon at which the (usually slowly) diminishing value of savings, wages, and annuities due to inflation filters into the nation s political consciousness compared to the relatively stark realities of lost jobs and downward wage pressures. 15

17 Estimates of the Expanded Supreme Court Model Against this backdrop, I proceed to the second stage of the analysis, incorporating the the cumulative ideological content of Supreme Court decision-making, measured by the caselaw index, into the baseline model described above. Estimates of this expanded model are reported in the second column of Table 1. Consistent with expectations derived from the thermostatic theory of public responsiveness to Supreme Court decision-making, the model shows that changes in the caselaw index are negatively and significantly associated with changes in public mood. Thus, growing cumulative liberalism in Supreme Court decision-making predicts increasing conservatism in public opinion, and vice versa. This effect takes the form of a short run effect, or instantaneous adjustment. (The long run effect associated with the caselaw index is insignificant and incorrectly signed.) All else equal, each additional salient liberal Supreme Court decision at time t significantly predicts a conservative change in mood of 0.14 points at time t. This effect indicates that a typical year s change in the caselaw index (5.29 net Supreme Court decisions in one ideological direction or the other) is expected to alter public mood by 0.74 points, and that a one standard deviation increase in the caselaw index (38.13 points) predicts a short run decrease of 1.21 standard deviations in mood (5.34 points). In addition to information about the dynamic effects of Supreme Court decisions for public mood, estimating the two model parameters associated with the caselaw index (i.e. the short run and long run effects) improves the overall fit and explanatory power of the baseline model. In terms of explanatory power, the R 2 statistic for the baseline model is 0.35 while the R 2 of the expanded model including Supreme Court decisions is The effects associated with the caselaw index, therefore, account for nearly 17 percent of the observed variance in mood and improve the overall proportion of variance explained by the model by nearly 49 percent. This improvement in overall model fit corresponds to a preference for the expanded model over the baseline model even after accounting for the parsimony lost by two additional model parameters. This preference is evident 16

18 in the adjusted R 2 reported for each model (0.24 for the base model and 0.41 for the expanded model). In relation to other included predictors, the estimated effects of Supreme Court decisionmaking do not significantly alter parameter estimates associated with any other model component. This generally means that the results of hypothesis tests and subsequent inferences about the effects of public policy and the state of the macroeconomy are the same in either mode. The only substantive difference between the models with respect to other predictors of mood relates to the rate of inflation. A modest increase in the variance estimate associated with the long run effect of inflation for public mood in the expanded model tips the t statistic associated with that parameter estimate below its critical value. Thus, the results of the second model do not support rejection of the null hypothesis of no relationship between changes in public mood and the lagged value of the rate of inflation. Substantively, these results indicate that public opinion responds systematically to Supreme Court decisions and that this responsiveness takes the form of negative feedback. When the Court s decision s push in one ideological direction, public opinion is likely to push back in the opposite direction. Moreover, the model indicates that the magnitude of the impact of Supreme Court decision-making for public mood is relatively large. The total predicted effect of a single salient Supreme Court decision is three times as large the effect associated with an important piece of legislation, and and the standardized (short run) effect of the caselaw index is roughly five times the size of the standardized (long run) effect of the policy index. Though alternative data may yield different estimates of the relative strength of the various predictive effects discussed here, the data analyzed here strongly suggest that the Supreme Court s and its decisions weigh more heavily on public opinion than most extant literature suggests. Discussion At least two factors perhaps account for estimated dynamics with which Supreme Court decisions influence mood and the size of their effects relative to policy changes and economic dynamics. 17

19 First, the Court s decisions (salient and otherwise) are much more likely to be given in the first half of each year than the second half. This pattern arises as a result of the Supreme Court s term, which opens in October and is usually concluded by June. As a result, the public regularly has at least six months to register its political reaction to a typical Supreme Court decision in the same calendar year as the decision itself. Though this process is dynamic in the real world unfolding over days, weeks, and months both the Supreme Court decisions and public responses to them are likely to register in the same year, appearing contemporaneous in a matrix of annual data. In contrast, structural features of the policymaking process in Congress may exert downward bias on any empirical estimate of the effect of public policy on public opinion. Unlike Supreme Court decisions, the lack of temporal regularity and the availability of advanced information about future congressional policy changes may diminish the estimated influence of other policy changes for public mood. The legislative agendas of presidents and party leaders in Congress are well known. Proposed legislation, committee hearings, proposed amendments, markups, and floor debates are all public records. The partisan breakdown of each chamber is known, and a variety of official and unofficial whip counts often enter public discussions of proposed legislation. Thus, it is often clear well before major legislation receives a final up or down vote whether it is likely to become law. As such, information about the likely state of future public policy changes can often be incorporated into public opinion well before changes have actually been enacted. In other words, the general outline of a future policy change may be well known to the public well before a president hands out ceremonial signing pens. Public responses to congressional policymaking, then, may be spread out in a way that makes them difficult to discern or systematically estimate since it is likely that some part of the response occurs before legislation becomes law. Similar dynamics are also likely to be at work in estimates of the economy s effects. Advanced information about the state of the future economy is also widely available in to form of economic forecasts. There is considerable evidence that (in the aggregate) the public assimilates forward-looking economic data and forms reasonable judgements about the prospective state of the economy (citations) and that these prospective evaluations of economic conditions influence 18

20 current public opinion (e.g. MacKuen, Erikson, and Stimson 1992). Again though, the availability of information about likely future conditions may erode the magnitude of the effects of economic conditions for public mood estimated for any given point in time. Second, the timeline of the Supreme Court s decision process also makes the nature of information about those decision s relatively unique in the system of public opinion dynamics. Supreme Court decisions offer news only three times in their life cycle: when the Court grants certiorari, when a case is argued, and when a decision is rendered. Though public debate about cases can (and often does) accompany them throughout their lifecycle, actual news coverage of cases is necessarily punctuated. All conference votes and memoranda, all debate and deliberation among and between the justices, and all other processes and procedures that take place within the Court between the public events associated with each case are kept secret. Therefore, much preliminary information about the likely direction of case outcomes remains unknown to ordinary citizens. The actual announcement of decisions, therefore, amounts to a sudden release of information which can be absorbed into public opinion only after the Court has acted. This pattern may, again, contribute to the estimated dynamic form of public responsiveness to the Court. Conclusions I began by identifying a puzzle in the supposed imperceptibility of the [Supreme] Court in national public opinion (Durr, Martin, and Wolbrecht 2000, p. 775). While historians, legal scholars, and some political scientists have noted the increasing prominence of the Supreme Court in American national politics during the last half-century, most scholarly analysis of the relationship between the Supreme Court and public opinion has found, at most, marginal and conditional relationships between the Court s decisions and the distribution of policy attitudes in the mass public. Yet, the Supreme Court is increasingly engaged with salient questions of public policy, the object of growing attention by interest groups and candidates for public office, a subject of public commentary and debate, and assertive of a uniquely authoritative voice in matters of constitutional 19

21 interpretation. The juxtaposition of growing judicial entanglement with our ordinary politics and scholarly conclusions that the Supreme Court is generally inconsequential for national public opinion raises the critical puzzle: Are accounts of the growing role of the Supreme Court in normal politics merely much ado about nothing in the minds of ordinary Americans, or have scholars of linkages between Supreme Court decisions and public opinion missed something important about how the Court s decisions influence public sentiment? There is, indeed, considerable evidence that the Supreme Court does not legitimize policy attitudes at the national level and that structural responses to the Court s decisions are marginal and conditional. I have argued, however, that it is a mistake to infer that an absence of particular kinds of public responsiveness implies general public disengagement with the judiciary. Moreover, there is, in fact, important evidence showing sensible and systematic relationships between the Supreme Court s decisions and public evaluations of the Court and some policy attitudes. Existing evidence of mass responsiveness to the Supreme Court suggests the public s capacity for thermostatic reactions to the Court s decisions. The theory of thermostatic responsiveness to Supreme Court decisions predicts negative feedback in the public s relative policy preferences from changes in public policy resulting from the Court s actions. I assess this theory by modeling public mood as a function of the cumulative liberalism in salient Supreme Court decisions, controlling for changes in public policy created by the elected branches of national government and the state of the macro economy. Single-equation error correction model estimates show a significant negative relationship between public mood and Supreme Court liberalism in the short run. When the Supreme Court s decisions accumulate in one ideological direction, public mood presses back in the other direction. Moreover, the estimated effects of Supreme Court liberalism are especially strong. The predicted effect of a single salient Supreme Court decision is about three times the size of the effect of a single important piece of legislation enacted by Congress, though there are some cogent reasons to think that estimates of the effects of ordinary legislation may be depressed by the availability of information about the legislative pro- 20