Printer Friendly

The public's conditional response to Supreme Court decisions.

In the years leading up to the Supreme Court's decision in Roe v. Wade (1973), debates over abortion became quite heated, but judicial scholars would have predicted a diminution in the high level of public controversy after the Court ruled. Indeed, the prevalent theory of the day was that when the Supreme Court made decisions, the public simply accepted them as legitimate. This theory - established by Dahl (1957) and labeled the positive response hypothesis by Franklin and Kosaki (1989) - suggests that there was nothing more to discuss after Roe because the Court had established a clear constitutional right to abortion on demand during the first trimester of pregnancy. Whether or not an individual agreed with this position before Roe, she would at least accept it as law after the decision because the Supreme Court was seen as the ultimate arbiter of the law. If this were the rule and not the exception, however, then the public debate concerning abortion should have dissipated in the wake of the Roe decision. Clearly, history does not support this conclusion.

While Roe provides only one example of how the positive response hypothesis fails to capture the Court's ability to affect public opinion, it suggests to many scholars (e.g., Franklin and Kosaki 1989; Hoekstra and Segal 1996) that Court decisions do not necessarily decrease public controversy. Sometimes public debate increases in the wake of Supreme Court decisions. This leads us to believe that the traditional theory of the relationship between Supreme Court decisions and public opinion is suspect.

We offer a theoretical framework that approaches the puzzle of how the Supreme Court affects public opinion in a novel manner. Specifically, we seek to answer when, if ever, the Court affects public attitudes when it makes decisions. We derive our framework from theories of Supreme Court credibility as well as from a variation on a social-psychological model that suggests that (1) individuals form opinions about an issue when they are provided with information about it, and (2) once individuals form opinions, they do not readily change their mind. With this foundation, we generate predictions about how and when the Court will influence public attitudes about two specific issues: abortion and the death penalty.

EXISTING THEORIES

Judicial specialists offer two competing theories to explain the Supreme Court's effect on public opinion. The first, initially advanced by Dahl (1957), focuses almost exclusively on how the Court legitimizes its own policies or those of the other two branches of government. Building on Dahl's theory, a number of scholars hypothesize that when the Court takes a position on an issue, overall public support for that position inherently increases (for a complete review of this literature, see Hoekstra and Segal 1996). In other words, because the Court is perceived as the ultimate arbiter of the law, its decisions are viewed as legitimate, credible, and therefore correct. This theory, labeled the positive response hypothesis by Franklin and Kosaki (1989), predicts that the public listens to the Court and always supports its decisions.

Franklin and Kosaki (1989) empirically test the predictions of the positive response hypothesis and demonstrate that it cannot account for how the structure of public opinion toward abortion changed in the wake of Roe v. Wade (1973). In its stead they offer the structural response hypothesis, which predicts that in highly salient cases the Court's policy choices do not necessarily translate into an aggregate increase in positive attitudes toward its position on that issue. Rather, while citizens often accept the Court's choices, their reactions are affected more strongly by the political context within which they live (see, e.g., Huckfeldt and Sprague 1995). In addition, the structural response hypothesis predicts that the Court has a microlevel effect on public opinion because it increases the intensity of within-group opinions about particular issues. The general point is that the Court does not necessarily legitimate particular policies (although it may, if aggregate public opinion favors a particular policy choice before a decision) but affects attitudes by bringing issues into the public discourse.

The empirical evidence presented by Franklin and Kosaki (1989) is quite compelling. They find that levels of support for abortion on demand diverged between groups (e.g., Catholics versus non-Catholics; whites versus nonwhites) in the wake of Roe. More specifically, the difference between these groups changed by approximately 50% as a result of the Court's decision (p. 762). This leads Franklin and Kosaki to posit that the effect of Roe was further crystallization of issue preferences and greater homogeneity of within-group beliefs (p. 762). From this they conclude, in line with the structural response hypothesis and contextual theory, that the Court's decision polarized group attitudes toward abortion.

Despite these strong findings concerning public opinion toward Roe, two key questions arise about the applicability and generality of Franklin and Kosaki's model. Does the Supreme Court continue to affect public attitudes when it rules on a salient issue more than once? If so, do between-group attitudes become more polarized each and every time the Court rules on that issue? In order to understand the extent to which the Court affects public opinion, we must come to terms with how and when the public reacts to decisions in more than one case and in more than one issue area. To find the answers we turn to theories about Court credibility (e.g., Caldeira 1986; Caldeira and Gibson 1992) and to social psychological theories of attitude formation and change (e.g., Chaiken 1980; Fiske and Taylor 1991; Petty and Cacioppo 1981; Petty and Krosnick 1995).

INFORMATION, CREDIBILITY, AND ATTITUDE CHANGE

We provide an alternative to the positive response hypothesis and an extension of the structural response hypothesis. More specifically, we offer a variation of the social-psychological elaboration likelihood model to explain how the Supreme Court affects public opinion. Our model is based on two different theories. First, because the public generally views the Court as a highly credible institution (Caldeira 1986; Caldeira and Gibson 1992; Hoekstra and Segal 1996), individuals more clearly elaborate their attitudes toward an issue after a ruling (although they do not necessarily agree with the decision). Unlike the positive response hypothesis, however, our account suggests only that the public pays attention to and discusses major decisions because of the Court's high degree of legitimacy. In other words, as a credible source of information, when the Court makes its first major decision on a particular issue, the structure of public opinion changes in a manner consistent with the structural response hypothesis. That is, even if individuals disagree with a particular decision, their opinions on that issue will crystallize (become stronger) for or against the Court's policy choice.(1)

To support this part of our model, we draw on research demonstrating that even people who know very little about the Supreme Court often hold it in high regard. For instance, Caldeira and Gibson (1992, 635) argue that "the Supreme Court has traditionally fared well in the estimations of the public." Other works suggest that individuals perceive the Court as highly legitimate simply because it is the ultimate arbiter of the law (Adamany 1973; Adamany and Grossman 1983). While we take issue with the conclusions of these latter works, the authors do make a point with which we agree: The public would never even pay attention to the Supreme Court if it were not viewed as a legitimate institution. Thus, as Hoekstra and Segal (1996, 1097) point out, individuals may hold well-defined attitudes toward an issue, but the Court will still have an effect on those opinions because of its high degree of legitimacy.

The first part of our model also draws on two key studies from social-psychological theory. Chaiken (1980) studies the persistence of attitude change based on different source and message characteristics. Specifically, she is interested in the likelihood of attitude change when an individual is presented with the same information more than once. To test this process Chaiken uses an experimental setting to explore how attitudes change when individuals are presented with similar information on two separate occasions. She finds that, after subjects heard the same information a second time, "there was a trend for all subjects to show less persuasion" (quoted in Petty and Cacioppo 1981, 259). Furthermore, those presented with information about an issue of high consequence (highly salient for them) were even less likely to change attitudes.

Chaiken's work is consistent with the elaboration likelihood model, a commonly used model of persuasion, which demonstrates that once an issue is elaborated to the extent that individuals form opinions, further elaboration, even if it is in conflict with their opinion, will tend not to change that opinion (Petty and Cacioppo 1981). Petty and Cacioppo (p. 263) find that the intensity of cognitive processes influences the propensity for attitude change: "In studies where issue relevant cognitive activity was likely to be intense . . . the attitude changes produced have been found to be relatively enduring." Together with the theories about Court credibility, this work establishes the first part of our model.

The second part of our model relies exclusively on the elaboration likelihood model. We argue that once the Supreme Court helps individuals elaborate their opinions, subsequent decisions within the same issue area - even if they overrule an initial landmark decision-will have little effect on public opinion. In short, even though the Court is still viewed as a credible institution, people will not change opinions formed on the basis of the Court's prior decision.

In sum (1) the Supreme Court is viewed as a credible and legitimate institution; (2) individual's opinions on certain issues crystallize when the Court decides its first major case in a highly salient issue area because those opinions are affected by the Court's policy choices; and (3) opinions about the Court's policy choices are unlikely to change once they are set, even when the Court rules again on the same issue. We call this the conditional response hypothesis and posit the following two hypotheses:

HYPOTHESIS 1. When the Supreme Court makes its first major (landmark) decision in a salient issue area, increased cleavages will occur in between-group attitudes about an issue, and there will be an overall change in public opinion toward that issue.

HYPOTHESIS 2. When the Supreme Court decides subsequent cases within the same issue area, little if any change will occur in the structure of group attitudes toward that issue regardless of the direction, intensity, or breadth of the decision, and there will be no overall change in public opinion toward that issue.(2)

Each hypothesis is tested separately. Hypothesis 1 involves cases in which the Court makes its first landmark decision in a given issue area. Hypothesis 2 involves subsequent cases concerning the same issue. The findings of Franklin and Kosaki (1989) serve as an initial test of Hypothesis 1, since Roe was the Supreme Court's first major and highly visible abortion decision. Another highly visible abortion case (Webster v. Reproductive Health Services [1989]) is used to evaluate Hypothesis 2. This analysis is meant to discern whether the Court had a similar effect on public attitudes toward abortion in both cases. We argue that Webster is an ideal case with which to test our second hypothesis since it was the first (and arguably the best) chance to strike Roe, thus making it highly controversial.(3)

DATA AND MODELING STRATEGIES

Data for our test are drawn from two CBS News/New York Times polls (1989a, 1989b) concerning public attitudes toward abortion during summer and autumn 1989.(4) The first poll was conducted two months prior to the Webster decision, and the second three months after. With data of this nature (about the same issue and of approximately the same importance as Roe), we can test Hypothesis 2.(5)

Our dependent measure is the number of circumstances, from a list of possible choices, in which a respondent would permit a legal abortion. Because the first survey listed five possible scenarios for abortion, and the second listed seven, we construct a scale from 0 to 2 and match only the common reasons from each survey, ranging from 0 (no instances of legal abortions) to 2 (two or more instances of legal abortions). While this is not a perfect solution, it still captures the circumstances under which an individual believes abortion should be legal. With these data, we construct two separate abortion scales: One reports responses to Health-related reasons for abortion, and the other reports responses to Discretionary abortions (see the Appendix for definitions of these two scales).(6)

To explain how group attitudes may have changed after Webster, we specify a model with several independent measures (Education, Gender, Race, Catholic Adherents, and Frequency of Church Attendance) to capture the extent to which attitudes toward abortion are affected by the Court. Not only are these variables theoretically pleasing, but also, because Franklin and Kosaki (1989) invoke them, we can make direct comparisons to their findings. Furthermore, these variables allow us to draw out the relationships among group membership, individual characteristics, and attitudes toward abortion.

We also include a dummy variable, [d.sub.1] (which we interact with all the independent variables), to discern whether attitudes prior to the decision change afterward. This Postdecision Dummy equals 0 before the decision and 1 afterward, so that it may interact with the [[Beta].sub.2,k] coefficients (the postdecision individual characteristics). As Franklin and Kosald note (1989, 7556): "The coefficients in this model are specified so that they may vary between the predecision and postdecision samples." The equation we estimate, then, is the following:

y = [[Alpha].sub.0] + [[Alpha].sub.1][d.sub.1] + ([[Beta].sub.1,1] + [[Beta].sub.2,1][d.sub.1])[x.sub.1] + ... + ([[Beta].sub.1,k] + [[Beta].sub.2,k][d.sub.1])[x.sub.k] + [Epsilon]. (1)

The second equation we estimate constrains [d.sub.1] to 0, thereby suppressing the interaction effect with the [[Beta].sub.2,k] coefficients.

y = [[Alpha].sub.0] + [[Beta].sub.1,1][x.sub.1] + ... + [[Beta].sub.1,k][x.sub.k] + [Epsilon]. (2)

We only invoke this constrained model (and report the results) for use in a likelihood ratio test, which helps us explain whether an overall change occurs in aggregate group feelings toward abortion after the Court's decision.

Because our dependent variable consists of discrete ordinal numbers (0, 1, or 2), traditional linear regression is an inappropriate method for testing our hypotheses. Indeed, Greene (1993, 672) notes that OLS errs by taking the discrete rankings and treating the differences as the same, and he argues that a standard logit or probit model also cannot account for the ordinal nature of the dependent variable. He contends that (under certain conditions) an ordered probit model is the best way to estimate models with discrete dependent variables (p. 673). We therefore turn to this model, first offered by McKelvey and Zavonia (1975), because it takes into consideration the ordinal nature of the data by treating each observation only as a rank ordering. We estimate these models using maximum likelihood estimation (see McKelvey and Zavonia 1975 or Greene 1993 for the derivation of this model and its likelihood function). In each model, a positive coefficient indicates an increased likelihood that a respondent believes more instances for abortion should be legal, while a negative sign indicates that she thinks there should be fewer instances of legal abortion.

A comparison of the constrained and unconstrained models allows us to test whether the postdecision characteristics are jointly constrained to 0 (meaning the overall structure of opinion did not change because of Webster). Therefore, the difference in the [[Beta].sub.1,k] and [[Beta].sub.2,k] coefficients before and after the decision tests the extent to which the structure of between-group attitudes changes as a result of a particular Court decision. As Kmenta (1986, 491) suggests, "if we wish to test the validity of nonlinear restrictions rather than assuming it a priori, we can use . . . the likelihood test." We can observe the difference in the likelihood function between the constrained and unconstrained models, which is -2 times the change of the log of the likelihood function, distributed as chi-squared, with degrees of freedom equal to the number of constraints. The results of this test allow us to make different claims about each hypothesis. When testing Hypothesis 1 we expect at least some of the postdecision [[Beta].sub.2,k] coefficients to be statistically significant, [[Alpha].sub.1] to remain small, and the likelihood ratio test between the constrained and unconstrained models to reach statistical significances.(7) If this happens, then we can reject the implicit null hypothesis that the structure of opinion did not change. For Hypothesis 2, however, we expect few if any of the postdecision [[Beta].sub.2,k], coefficients to be statistically significant, [[Alpha].sub.1] still to remain small, and the likelihood ratio test not to reach statistical significance. These results would maintain Hypothesis 2 - shifts in public attitudes did not occur when the Court ruled in subsequent cases in the same issue area.

RESULTS

As stated above, we accept the findings of Franklin and Kosaki (1989) concerning the public's reaction to Roe as support for Hypothesis 1. That is, the public in general reacted to the Court's decision, and the views of several social groupings (such as Catholics and non-Catholics) became strongly polarized. Our first test, then, focuses on Hypothesis 2. We estimate a model for those respondents who had heard of the Court's decision in Webster (see the Appendix for the distributions of all variables used for the model).(8) Table 1 presents the constrained and unconstrained models for both the discretionary and health-related abortion scales. The sample includes all the respondents in the predecision sample but only those in the postdecision sample who said that they had heard of Webster. Note that while some of the predecision signs may seem counterintuitive (Catholics have a positive sign in the discretionary model), only education and church attendance are significant predictors of polarized attitudes toward abortion prior to Webster. Those with higher education are more supportive of abortion than those with lower education, and those who attend church more frequently are less supportive than those who attend less frequently.(9)

To demonstrate whether attitudes toward abortion changed as a result of the Court's action, we turn to the postdecision coefficients in the unconstrained model. As our second hypothesis predicts, there is little or no change in the estimated coefficients from the predecision to the postdecision attitudes. The most striking change is that the sign on Catholics who had heard of Webster becomes negative after the ruling in the discretionary model. This change is marginal evidence, at [TABULAR DATA FOR TABLE 1 OMITTED] best, of polarization between Catholics and non-Catholics (t = 1.79), which accords with Hypothesis 2. In general, these models perform quite well, reaching highly significant chi-squared values of 258.2 and 172.1 for the discretionary and health scales, respectively (d.f. = 10). No reliable evidence exists that the Court affected intergroup attitudes toward abortion with its Webster decision, however, because no postdecision coefficient is significantly different from zero.

To test whether the overall structure of opinion changed, we next compare the constrained and unconstrained models (refer to Table 1). The first test compares the discretionary abortion scale models, and it is evident that the overall structure of opinion did not change - the chi-squared value of the difference between the constrained and unconstrained models only reaches 6.66 with four degrees of freedom (equal to the number of constraints). This indicates little, if any, overall shift in attitudes toward abortion. The same holds for the health scale, for which the chi-squared value only reaches 2.26. Thus, in neither instance did the Court significantly alter the structure of public opinion toward abortion. While there is some evidence of the positive response effect in the health-scale model (all the signs become positive in the wake of Webster), the fact that no coefficient is significantly different from zero suggests little aggregate increase in support for health-related abortions.

The most interesting finding here is that Catholics (who might be expected most likely to oppose abortion, see Gilbert 1993) did not seem to crystallize their beliefs in the wake of Webster. Even on the health-related abortion scale - where one may predict a general increase in the level of support for abortion - the Court elicited scant reaction. Furthermore, those groups most likely to support abortion rights (Protestants, those with high education, and females) barely reacted as well. Attitudes seem to have neither changed in the aggregate nor polarized between groups as a result of Webster.

In general, these findings clearly demonstrate that people who had heard of Webster responded significantly less to the Court's action than did people who had heard of Roe. While Catholics who had heard of Webster became slightly more opposed to discretionary abortions after the ruling, the evidence in Table 1 still indicates little support for the notion that the Court significantly altered group attitudes in general. Rather, the public hardly responded, which suggests that when the Court rules following a major decision (Roe), the public is not likely to react to its policy choice.

A FURTHER TEST: THE DEATH PENALTY

The results of the previous test suggest that the structural response hypothesis (explicated here as Hypothesis 1) cannot account for the insignificant shifts in public opinion toward abortion after Webster. This tells us that the Court's ability to affect public opinion toward abortion may be conditional. In other words, the Court may only affect public opinion under specific circumstances. At this point, however, we cannot claim that our approach has explanatory power beyond this specific issue. It may be that the Court's effect is unique to abortion because it is such a salient issue for so many people. To explore the general applicability of our hypotheses, we must test them in another issue area. If the results of additional tests mirror those from the abortion cases, then we have compelling evidence that the Court's ability to affect public opinion is conditioned on the temporal ordering of its decisions. To discern whether this effect holds, we turn to another salient issue area, the death penalty.

We begin with an analysis of how the Supreme Court affected public opinion toward the death penalty with its first landmark (and highly publicized) ruling on the constitutionality of this punishment - Furman v. Georgia (1972). Then we analyze whether the Court affected public opinion in two of its many subsequent capital punishment cases: Gregg v. Georgia (1976) and McCleskey v. Kemp (1987). We expect that Furman will elicit significant change in group attitudes and also will alter overall public opinion toward the death penalty (Hypothesis 1), but that after Gregg and McCleskey there will be little change on either level (Hypothesis 2).

Data for the death penalty analyses are drawn from the General Social Survey cumulative file, 1972-94 (Davis and Smith 1994). The GSS conducted interviews from February to April in 1972, 1973, 1976, 1977, 1987, and 1988.(10) This allows for a comparison of public attitudes toward the death penalty several months before each of the decisions and then within a year afterward. The GSS also provides a number of useful measures for the purpose of testing our theory. Thus, we specify models that are very similar to the models used to test attitude change in the wake of Webster. The independent measures in these models are Education, Gender, Race, Party Identification, Jewish Adherents,(11) and Frequency of Church Attendance. These variables allow us to draw out the relationships among group membership, individual characteristics, and attitudes toward the death penalty. Again, we include a dummy variable to distinguish between pre- and postdecision attitudes.

The dependent measure is whether a respondent Favors the Death Penalty for convicted murderers.(12) The dependent variable is coded 1 (favors the death penalty) or 0 (opposes this punishment). Given the dichotomous nature of this variable, an alternate to regression analysis is again warranted. We use binary probit models, which attempt to account fully for the relevant causal factors (see Greene 1993, 635-51 for the derivation and discussion of the probit model), and we estimate these models using maximum likelihood estimation. As discussed in the abortion section, the comparison of the constrained and unconstrained models allows us to discern whether the Court affected overall public opinion toward the death penalty.(13)

OPINION CHANGE AND DEATH PENALTY CASES

In Furman v. Georgia, the Court issued a short per curium decision holding that Georgia laws governing the death penalty violated the Eighth Amendment protection against cruel and unusual punishment (see Epstein and Walker 1995). Table 2 presents the constrained and unconstrained models for public attitudes before and after Furman. Note that prior to this decision, females and nonwhites were significantly more opposed to capital punishment than were males or whites, and Republicans favored the death penalty more than did Democrats. These findings are intuitive and conform to existing theories concerning different group attitudes toward the death penalty (see, e.g., Epstein and Kobylka 1992).

The postdecision coefficients in the unconstrained model help determine whether the structure of public opinion changed as a result of Furman. As Hypothesis 1 predicts, the decision did elicit specific social cleavages in two areas: Those with higher education clearly became more opposed to the death penalty than did those with less education, while partisan attitudes toward capital punishment became more polarized. These findings support Hypothesis 1, as [[Alpha].sub.1] remains small, while some of the [[Beta].sub.2,k] coefficients are large. Both models are significant, with chi-square values of 205.21 and 216.57, respectively.
TABLE 2. Estimated Capital Punishment Responses for Furman v.
Georgia (Probit Estimates)

                              Unconstrained         Constrained
Parameter                         Model                Model

Constant                            .34(***)             .34(***)
                                   (.08)                (.07)
Postdecision Dummy                  .14                  .15(**)
                                   (.11)                (.05)
Education                          -.02                 -.08(**)
                                   (.04)                (.02)
Education                          -.13(*)              -
Postdecision                       (.05)
Female                             -.40(***)            -.40(***)
                                   (.07)                (.05)
Female Postdecision                 .00                 -
                                   (.10)
Nonwhite                           -.74(***)            -.69(***)
                                   (.10)                (.07)
Nonwhite Postdecision               .11                  -
                                   (.15)
Party Identification                .04(*)               .61(***)
                                   (.02)                (.01)
Party Identification                .05(*)               -
Postdecision                       (.02)
Jewish                              .41(*)               .32(*)
                                   (.20)                (.15)
Jewish Postdecision                -.23                  -
                                   (.30)
Church Attendance                   .02(*)               .02(*)
                                   (.01)                (.01)
Model [x.sup.2]                  216.57               205.21
log (L)                       -1,713.19            -1,718.87
Likelihood test [x.sup.2]          -                   11.36
Significance                       -             p [less than] .05

Source: General Social Surveys 1972-94 (November 1994).

Note: N = 2,716. * p [less than] .05; ** p [less than] .01; *** p [less than]
.001. Standard errors are in parentheses.


To discern any overall shift in group attitudes toward capital punishment after Furman, we also compare the constrained and unconstrained models using the likelihood ratio test. The chi-square value for this test reaches 11.36 with five degrees of freedom (refer to Table 2). This suggests (with p [less than] .05) that overall public opinion toward the death penalty changed as a result of Furman, the first major case in which the Court ruled on the constitutionality of the death penalty, which conforms to Franklin and Kosaki's (1989) finding based on Roe.

Since 1972 the Court has issued many additional opinions covering the death penalty, two of which are particularly important and salient: Gregg v. Georgia (1976) and McCleskey v. Kemp (1987).(14) In Gregg the Court declared that Georgia's new death penalty law, which stipulated a separate trial and sentencing hearing for specific crimes like murder, did not per se violate the Eighth Amendment. Several justices in the majority thought that this law was "a model death penalty scheme" (Epstein and Walker 1995, 631). Although this ruling did not overturn Furman, it affected death penalty laws throughout the states; many legislatures simply conformed current laws to the Georgia scheme. According to Hypothesis 2, we predict little or no response after Gregg, just as after Webster. Table 3 shows how public opinion changed after this ruling.
TABLE 3. Estimated Capital Punishment Responses for Gregg v. Georgia (Probit
Estimates)

                              Unconstrained     Constrained
Parameter                          Model           Model

Constant                            .75(***)       .77(***)
                                   (.09)          (.07)
Postdecision Dummy                  .15            .12
                                   (.12)          (.05)
Education                          -.10(**)       -.09(***)
                                   (.03)          (.02)
Education                           .16(**)         -
Postdecision                       (.05)
Female                             -.40(***)      -.35(***)
                                   (.07)          (.05)
Female Postdecision                 .04             -
                                   (.10)
Nonwhite                           -.77(***)      -.78(***)
                                   (.12)          (.08)
Nonwhite Postdecision              -.02             -
                                   (.17)
Party Identification                .04(*)         .02(*)
                                   (.02)          (.01)
Party Identification               -.03             -
Postdecision                       (.03)
Jewish                              .18            .09
                                   (.27)          (.17)
Jewish Postdecision                -.18             -
                                   (.36)
Church Attendance                   .00            .00
                                   (.00)          (.00)
Model [x.sup.2]                  153.49         152.06
log (L)                       -1,619.47      -1,620.19
Likelihood test [x.sup.2]         -               1.44
Significance                      -               ns

Source: General Social Surveys 1972-94 (November 1994).

Note: N = 2,784. * p [less than] .05; ** p [less than] .01;
*** p [less than] .001. Standard errors are in parentheses.


It is clear from an initial inspection of the model that there was little shift in group attitudes toward the death penalty after Gregg. The only group split is between those with high education and those with lower education.(15) No other postdecision independent measure indicates group cleavage. Although one might expect an exacerbation of the cleavage between Democrats and Republicans, and possibly between whites and nonwhites, none occurred.

The above finding, combined with the results of the likelihood ratio test, suggests that the Court had little effect on public opinion as a result of Gregg. Indeed, the chi-square for the likelihood test is nowhere near significant (refer to Table 3). We therefore conclude that the Court's decision had no significant effect on overall public attitudes, or on specific group attitudes, toward the death penalty.

In McCleskey v. Kemp (1987) the Court was again asked to rule on the constitutionality of Georgia's death penalty law. This time the issue focused on whether it violated the equal protection of African Americans, since statistical evidence documented that they were four times more likely to be given a death sentence in Georgia than were whites (see Epstein and Walker 1995, 634). Despite the strong evidence of a clear racial disparity in death sentences, the justices upheld the constitutionality of the Georgia law. Even though this case was highly controversial, we do not expect a pronounced public response. Table 4 explicates the results.

Compared to Gregg, there is even less evidence of social cleavage concerning the death penalty after McCleskey. There is marginal evidence that females became slightly more opposed to the death penalty in the wake of this ruling (t = 1.81), but no other postdecision cleavage is statistically different from zero. This indicates that the model in Table 4 comports with our conditional response hypothesis. Indeed, because the Court had already spoken about the death penalty issue, we expect little change in public attitudes despite the salience of this decision.

As a final test of Hypothesis 2, we perform the likelihood ratio test for the constrained and unconstrained models in McCleskey to determine whether overall public opinion moved after this ruling. The chi-square is quite small and therefore not significant (refer to Table 4). This suggests even more strongly that there was no discernible effect of this decision on public attitudes toward the death penalty.

In sum, our results indicate that the first time the Court rules on issues such as abortion and the death penalty, it affects between-group attitudes, and the overall structure of public opinion toward these issues.
TABLE 4. Estimated Capital Punishment Responses for McCleskey v.
Kemp (Probit Estimates)

                              Unconstrained       Constrained
Parameter                         Model              Model

Constant                             .67(***)          .73(***)
                                    (.09)              (.08)
Postdecision Dummy                   .21                .05
                                    (.13)              (.05)
Education                           -.04               -.05(*)
                                    (.03)              (.02)
Education                           -.02              -
Postdecision                        (.04)
Female                              -.06               -.14(**)
                                    (.07)              (.05)
Female Postdecision                 -.20(*)           -
                                    (.11)
Nonwhite                            -.67(***)          -.70(***)
                                    (.08)              (.06)
Nonwhite Postdecision               -.09              -
                                    (.13)
Party Identification                 .10(***)           .10(***)
                                    (.02)              (.02)
Party Identification                 .00              -
Postdecision                        (.03)
Jewish                               .53                .16
                                    (.40)              (.22)
Jewish Postdecision                 -.55              -
                                    (.49)
Church Attendance                   -.02(*)            -.02(*)
                                    (.01)              (.01)
Model [x.sup.2]                   317.41             311.64
log (L)                        -1,553.88          -1,556.76
Likelihood test [x.sup.2]          -                   5.76
Significance                       -                  ns

Source: General Social Surveys 1972-94 (November 1994).

Note: N = 2,920. * p [less than] .05; ** p [less than] .01;
*** p [less than] .001. Standard errors are in parentheses.


This structural response is conditioned, however, as predicted by Hypothesis 2. Indeed, our analysis of Webster, Gregg, and McCleskey indicates that when the Court speaks more than once, its effect is minimal in later cases. These results strongly support the predictions of the conditional response hypothesis.

CONCLUSION

The Supreme Court can and does influence public attitudes toward highly salient issues, but its effect is conditional. At times the public will react (when an issue is initially brought to the forefront of political discourse by a landmark Court decision), but at others it will not (when the Court rules on an issue again). Despite the limitations of our data (different wording for the death penalty dependent variable, and the hard-to-match questions for the abortion dependent variables), our findings are quite consistent with our theory. Indeed, they suggest that the conditional response hypothesis has much more explanatory power than previous views of how the Supreme Court affects public opinion in the United States.

These results indicate that judicial scholars should adjust the way they approach this political phenomenon. Franklin and Kosaki (1989) demonstrate that neither the positive response hypothesis, nor simple time-series analyses are productive. We agree. But there are three avenues of research that may provide even better insight into the influence of controversial Court decisions. First, studies should focus on attitude change after the Court hands down a first landmark decision in a particular issue area (such as US Term Limits v. Thornton, term limits; Romer v. Evans, equal protection for homosexuals; and Reno v. ACLU, pornography on the Internet), as this is the time when the greatest effect can be expected.

Second, our results complement Hoekstra and Segal's (1996) finding that changes in public opinion are likely to occur in close proximity to case origination. They find that the Court's ruling in Lamb's Chapel v. Center Moriches Union Free School District (1993) probably affected attitudes in the local community, where the public had abundant information about the decision, even though the case had national implications as well. Thus, attention should be given to areas in which prior elaboration has not occurred but in which information about the Court's decision will be readily available, highly relevant, and clearly salient.

The third avenue adds to the first two. Interrupted time-series analysis may demonstrate more clearly how specific group attitudes change over time. For example, an analysis of an issue on which the Court has decided many cases - such as school prayer - may determine whether the shock from each ruling influences aggregate public attitudes toward that issue. In this way, we may be able to discern the movement of opinion and how it is changed by both landmark and subsequent decisions within the same issue area, while controlling for other macropolitical factors. This would allow a comparison of how different decisions in different issue areas elicit similar or dissimilar public responses.

At the end of the day we believe that the conditional response hypothesis should be viewed as a general explanation for how the Supreme Court affects public opinion. Indeed, it suggests that, as a credible institution, the Court can highlight issues in the public discourse (as it did with Roe and Furman), thereby changing the structure of public attitudes. Despite its high degree of legitimacy, however, the Court's ability to affect public attitudes is muted when it alters, overturns, or reiterates its initial policy choices in subsequent cases.

APPENDIX

Abortion Cases

A. Dependent Variable

The pre-Webster sample was asked the following question: Please tell me whether or not you think it should be possible for a pregnant woman to obtain a legal abortion if:

there is a strong chance of serious defect in the baby. (used in Health Scale)

the woman's own health is seriously endangered by the pregnancy. (used in Health Scale)

she is not married and does not want to marry the man. (used in Discretionary Scale)

the family has a very low income and cannot afford any more children. (used in Discretionary Scale)

the pregnancy interfered with work or education. (used in Discretionary Scale)

For the Post-Webster sample, the abortion question was as follows: Should it be legal or should it be against the law for a pregnant woman to obtain an abortion if:

the woman's health is seriously endangered by the pregnancy. (used in Health Scale)

the pregnancy is the result of rape. (used in Health Scale)

there is a chance of a serious defect in the baby. (used in Health Scale)

she is not married and does not want to marry the father. (used in Discretionary Scale)

the family has a very low income and cannot afford any more children. (used in Discretionary Scale)

the pregnancy would force the woman to drop out of school or interrupt her career (combined variable). (used in Discretionary Scale)

B. Scales for dependent measures

Discretionary abortions:

0 = Abortion should never be legal as a discretionary measure.

1 = Abortion should be legal in only one instance as a discretionary measure.

2 = Abortion should be legal in two or more instances as a discretionary measure.

Health-related abortions:

0 = Abortion should never be legal even as a health measure.

1 = Abortion should be legal in only one instance as a health measure.

2 = Abortion should be legal in two or more instances as a health measure.

C. Question used to truncate sample into those who had heard and those who had not heard about the Supreme Court's decision in Webster (and marginal distributions)

In July, the Supreme Court decided Webster v. Reproductive Health Services, a major case about abortion. Do you approve or disapprove of the Supreme Court's decision in this case, or haven't you heard enough about it to have an opinion?

Respondents who either approved or disapproved: 1,251

Respondents who did not have enough information to approve or disapprove: 1,587

D. Questions used for independent variables in abortion models (both samples)

Catholic: What is your religious preference today?

Nonwhite: Are you white, black, or some other race?

Gender: Respondent's sex.

Church Attendance: Do you attend religious services every week, almost every week, once/twice a week, once a month, a few times a year, or never?

Education: What is the last grade in school you completed?

E. Distributions of independent variables for those who heard of Webster

Church Attendance (0 = Never; 1 = Few times a year; 2 = Once/twice a month; 3 = Almost every week; 4 = Every week): 0 = 177; 1 = 379; 2 = 200; 3 = 143; 4 = 352

Education (0 = No high school degree; 1 = High school degree; 2 = Some college; 4 = College graduate and beyond): 1 = 122; 2 = 414; 3 = 340; 4 = 375

Gender (0 = Male; 1 = female): 0 = 570; 1 = 681

Nonwhite (0 = White; 1 = Black): 0 = 1,142; 1 = 109

Catholic (0 = Other; 1 = Catholic): 0 = 940; 1 = 305

F. Distributions of independent variables for those who had not heard of Webster

Church Attendance (0 - Never; 1 = Few times a year; 2 = Once/twice a month; 3 = Almost every week; 4 = Every week): 0 = 192; 1 = 475; 2 = 275; 3 = 211; 4 = 434

Education (0 = No high school degree; 1 = High school degree; 2 - Some college; 4 = College graduate and beyond): 1 - 184; 2 = 630; 3 = 419; 4 = 354

Gender (0 = male; 1 = female): 0 = 692; 1 = 895

Nonwhite (0 = White; 1 = Black): 0 = 1,434; 1 = 153

Catholic (0 = Other; 1 = Catholic): 0 = 1,179; 1 = 408

Capital Punishment Cases

A. Dependent variables (see footnote 12 for complete wording of questions used for this variable):

0 = Do not favor capital punishment for murderers.

1 = Favor capital punishment for murderers.

Furman: 0 = 1,151; 1 = 1,750

Gregg: 0 = 847; 1 = 2,002

McCleskey: 0 = 847; 1 = 2,211

B. Questions used for independent variables for capital punishment models

Jewish: What is your religious preference? Is it Protestant, Catholic, Jewish, some other religion, or no religion?

Nonwhite: What race do you consider yourself?

Gender: Respondent's sex.

Church Attendance: How often do you attend religious services?

Party ID: Generally speaking, do you think of yourself as a Republican, Democrat, independent, or what?

Education: Respondent's degree.

C. Distributions of independent variables for capital punishment cases

Church Attendance (0 = Never; 1 = Less than once a year; 2: Once/twice a year; 3 = Several times a year; 4 = About once a month; 5 = 2-3 times a month; 6 = Nearly every week; 7: Every week):

Furman: 0 = 354; 1 = 254; 2 = 372; 3 = 452; 4 = 201; 5 = 267; 6 = 214; 7 = 768; 8 = 213

Gregg: 0 = 406; 1 = 263; 2 = 407; 3 = 421; 4 = 219; 5 = 238; 6 = 117;7 = 626;8 = 256

McCleskey: 0 = 450; 1 = 221; 2 = 408; 3 = 466; 4 = 270; 5 = 358; 6 = 206; 7 = 650; 8 = 255

Education (0 = Less than high school; 1 = High school degree; 2 = Associate/junior college; 3 = Bachelor's degree; 4 = Graduate):

Furman: 0 = 1,187; 1 = 1,482; 2: 37; 3 = 256; 4 = 117

Gregg: 0 = 1,064; 1 = 1,469; 2 = 58; 3 = 291; 4 = 135

McCleskey: 0 = 835; 1 = 1,719; 2 = 142; 3 = 417; 4 = 176

Party Identification (0 - Strong Democrat; 1 = Not very strong Democrat; 2 = Independent, close to Democrat; 3 = Independent; 4 = Independent, close to Republican; 5 = Not very strong Republican; 6 = Strong Republican):

Furman: 0 = 557; 1 = 819; 2 = 352; 3 = 302; 4 = 241; 5 = 449; 6 = 249

Gregg: 0 = 500; 1 = 803; 2 = 407; 3 = 416; 4 = 235; 5 = 440; 6 = 203

McCleskey: 0 = 658; 1 = 708; 2 = 391; 3 = 385; 4 = 288; 5 = 541;6 = 299

Gender (0 = male 1 = female):

Furman: 0 = 1,508; 1 = 1,609

Gregg: 0 = 1,362; 1 = 1,667

McCleskey: 0 = 1,416; 1 = 1,884

Nonwhite (0 = White; 1 = Black):

Furman: 0 = 2,656; 1 = 444

Gregg: 0 = 2,700; 1 = 305

McCleskey: 0 = 2,456; 1 = 730

Jewish (0 = Other; 1 = Jewish):

Furman: 0 = 3,021; 1 = 96

Gregg: 0 = 2,867; 1 = 62

McCleskey: 0 = 3,249; 1 = 51

A previous version of this manuscript was presented at the 1996 annual meetings of the Southern Political Science Association, Atlanta, Georgia, and won honorable mention in the 1996 Congressional Quarterly competition for the best graduate-level paper in the field of law and courts. The authors thank Lee Epstein, Valerie Hoekstra, Christina Wolbrecht, and the anonymous reviewers for their comments and helpful insights throughout the writing of this manuscript. However, all remaining mistakes are our own.

1 Of course, there are other reasons a first major decision may affect public opinion. One alternative explanation is that the media are much more inclined to cover a new landmark decision (see, e.g., Slotnick, Segal, and Compoli 1994), and the public response may be related to the great amount of coverage.

2 Implicit in both of these hypotheses is the following null hypothesis: When the Supreme Court makes a decision, there will be no changes in the structure of public opinion. The null hypothesis affects how we can interpret our results. In our tests of Hypothesis 1, statistically significant results can lead us to reject the implicit null hypothesis, thus yielding credence to Hypothesis 1. For our tests of Hypothesis 2, statistically insignificant results allow us to maintain the null (fail to reject the null), thus yielding credence to Hypothesis 2. We have worded our hypotheses positively to enhance the readability of the paper.

3 There were other cases before Webster (e.g., Akron v. Akron Center for Reproductive Health [1983]), but Webster was still the best chance because the Court finally had a conservative majority. As Justice Blackmun said prior to Webster: "Will Roe v. Wade go down the drain? I think there's a very distinct possibility that it will, this term. You can count the votes" (Epstein and Walker 1995, 462). Aside from being the best opportunity to overturn Roe, Webster was important for showing how the Court would view specific state regulations on abortion. More specifically, in Webster the Court had to decide the constitutionality of a Missouri law which stated that (1) life begins at conception; (2) public employees could not perform abortions not necessary to save the mother; (3) public employees could not counsel women to choose abortion; and (4) physicians must test women to be sure the gestational age of the fetus is less than 20 weeks (adapted from Epstein and Walker 1995, 463). Finally, Webster was so controversial that 78 amicus curiae briefs were filed, representing more than 5,000 interested parties. All this suggests that Webster was highly salient and controversial.

4 These data were provided by the Inter-university Consortium for Political and Social Research. The two studies used are the CBS News/New York Times Monthly Poll (1989a); and the CBS News/New York Times Abortion Polls (1989b). The recoded data may be obtained from the authors to replicate this study.

5 Because the interest is in the short-term effect of the decision, it is important and intuitive to use data as close to the decision date as possible. We argue that our data are quite good because they were obtained within months of the Court decision.

6 Because we do not have exact matches for the pre- and postdecision samples, and because there are fewer health questions in the predecision sample, we had to modify Franklin and Kosaki's (1989) technique. Thus, our index combines the responses, so that we had a scale ranging from 0 (abortion should not be permitted under any circumstances), to 1 (in only one instance should abortion be legal), to 2 (in two or more instances abortion should be legal). Although this is not a perfect solution, our marginal distributions on both abortion scales are quite similar to those of Franklin and Kosaki (1989). Thus, this technique does not seem to bias our results.

7 The null hypothesis is that none of the coefficients will be different from zero, so if more than one is statistically significant, then there is support for our hypothesis.

8 We follow Franklin and Kosaki (1989) in this respect because, if we did not distinguish between those who heard of the Webster decision and those who did not, then we would be left to answer the same questions, namely, "How do we sort out the impact of the decision from the spurious influence of other factors? To what do we attribute the change?" (Franklin and Kosaki 1989, 754-5). By truncating the sample and only using those who heard of the decision, we should be able to parcel out the effect of the Court's decision. Thus, we only run this model for those who had heard of the decision.

9 As do Franklin and Kosaki (1989), we fix the church attendance coefficient because preliminary analysis showed no tendency whatsoever for this effect to change in the wake of Webster.

10 These data were provided by the Inter-university Consortium for Political and Social Research (Davis and Smith 1994). Recoded data may be obtained from the authors.

11 We include Jewish adherents for reasons similar to including Catholic adherents in the abortion model. Just as Catholics tend to have very conservative views on abortion, Jews tend to have very liberal views opposing capital punishment (see, e.g., Melton 1989).

12 The wording of this question in the GSS changed after 1973. At the time of Furman, the question was: "Are you in favor of the death penalty for persons convicted of murder?" At the time of Gregg and McCleskey, the question was "Do you favor or oppose the death penalty for persons convicted of murder?" Since we are not comparing individual responses over time, the difference does not matter for our purposes.

13 Since the GSS does not provide a question for the death penalty that asks whether a respondent had heard of the Court's decision, it may be difficult to sort out the influence of the decision from the spurious influence of other factors (see Franklin and Kosaki 1989, 754-5). By not excluding individuals who had not heard about the particular death penalty cases, our models incur some degree of bias, specifically against Hypothesis 1 and toward Hypothesis 2. This means that our findings in support of Hypothesis 1 (an effect from Furman) could be stronger than they appear if these individuals were excluded from our sample. In other words, if we could include only those who had heard of the decision, then we would expect much stronger changes in group attitudes in the wake of the decision, because we would capture only the attitudes of those who probably cared about the decision. Our results are still strong, which means we overcome this bias against Hypothesis 1. Unfortunately, the bias toward Hypothesis 2 (no effect from Gregg and McCleskey) cannot be easily overcome because when all those who had not heard of the decision are included in the sample, it is much easier to find no effect. In short, if someone had not heard of the decision, there is little reason to assume an attitude change as a result of the decision (see Franklin and Kosaki 1989, 760). Nonetheless, since these cases were quite salient and were part of the public discourse, we are confident that this bias is not severe.

14 We use Gregg because it seemed to settle the debate over the death penalty, mainly by ruling that it does not violate the Constitution (Epstein and Walker 1995, 633). This change in Court policy, just two years after Furman, makes Gregg the next highly significant death penalty case. We use McCleskey because of its highly controversial issue of whether African-Americans receive the death penalty more often than others. This case was highly controversial and highly publicized. Since both cases were likely to elicit major reactions by the public, they are excellent tests for our conditional response hypothesis.

15 The positive coefficient of the postdecision dummy on the education variable substantively implies that, after Gregg, individuals with more education supported the death penalty at a higher rate than before the decision. This may be due to the decision itself or to other macropolitical changes during the period between the surveys (such as a salient criminal trial that fostered pro-death penalty sentiments among the well educated). This statistically significant coefficient does not affect the explanatory power of the unconstrained model, however, as the likelihood ratio test between the unconstrained and constrained models does not approach statistical significance, which implies negligible change in the overall structure of public opinion toward the death penalty.

REFERENCES

Adamany, David. 1973. "Legitimacy, Realigning Elections, and the Supreme Court." Wisconsin Law Review 1973(3):790-846.

Adamany, David, and Joel Grossman. 1983. "Support for the Supreme Court as a National Policy Maker." In American Court Systems: Readings in Judicial Process and Behavior, ed. Sheldon Goldman and Austin Sarat. New York: Longman.

Akron v. Akron Center for Reproductive Health. 1983. 462 U.S. 416.

Caldeira, Gregory A. 1986. "Neither the Purse nor the Sword: Dynamics of Public Confidence in the Supreme Court." American Political Science Review 80(December):1209-60.

Caldeira, Gregory, and James L. Gibson. 1992. "The Etiology of Public Support for the Supreme Court." American Journal of Political Science 36(August):635-64.

CBS News/New York Times. 1989a. CBS News/New York Times Monthly Poll, April 1989 [computer file] (Study #9234). ICPSR version. Ann Arbor, MI: CBS/News/New York Times [producers], 1989. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1990.

CBS News/New York Times. 1989b. CBS News/New York Times Abortion Polls, September-October 1989 [computer file] (Study #9234). ICPSR version. Ann Arbor, MI: CBS News [producer]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1991.

Chaiken, Shelly. 1980. "Heuristic versus systematic information processing and the use of source message cues in persuasion." Journal of Personality and Social Psychology 39(November):752-66.

Dahl, Robert. 1957. "Decision Making in a Democracy: The Supreme Court as a National Policy Maker." Journal of Public Law 6(Fall):279-95.

Davis, James A., and Tom W. Smith. 1994. General Social Survey, 1972-1994 [machine-readable data file] (Study #6492). ICPSR version. Chicago, IL: National Opinion Research Center [producer]. Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut [distributor], 1994.

Epstein, Lee, and Joseph Kobylka. 1992. The Supreme Court and Legal Change: Abortion and the Death Penalty. Chapel Hill: University of North Carolina Press.

Epstein, Lee, and Thomas Walker. 1995. Constitutional Law for a Changing America: Rights, Liberties, and Justice. Washington, DC: Congressional Quarterly Press.

Fiske, Susan, and Shelly Taylor. 1991. Social Cognition. New York: McGraw-Hill.

Franklin, Charles, and Liane Kosaki. 1989. "Republican Schoolmaster: The U.S. Supreme Court, Public Opinion and Abortion." American Political Science Review 83(September):751-71.

Furman v. Georgia. 1972. 408 U.S. 238.

Gilbert, Christopher. 1993. The Impact of Churches on Political Behavior. Westport, CT: Greenwood.

Greene, William. 1993. Econometric Analysis. 2d ed. Englewood Cliffs, NJ: Prentice-Hall.

Gregg v. Georgia. 1976. 428 U.S. 153.

Hoekstra, Valerie, and Jeffrey Segal. 1996. "The Shepherding of Local Public Opinion: The Supreme Court and Lamb's Chapel." Journal of Politics 58(November):1079-102.

Huckfeldt, Robert, and John Sprague. 1995. Citizens, Contexts, and Social Communication. Cambridge: Cambridge University Press.

Kmenta, Jan. Elements of Econometrics. 1986. 2d ed. New York: MacMillan.

Lamb's Chapel v. Center Moriches Union Free School District. 1993. 124 L.Ed. 2d 352.

McCleskey v. Kemp. 1987. 481 U.S. 279.

McKelvey, Richard, and William Zavoina. 1975. "A Statistical Model for the Analysis of Ordinal Level Dependent Variables." Journal of Mathematical Sociology 4(Summer):103-20.

Melton, Gordon. 1989. The Churches Speak on Capital Punishment: Official Statements from Religious Bodies and Ecumenical Organizations. New York: Gale Research.

Petty, Richard, and John Cacioppo. 1981. Attitudes and Persuasion: Classic and Contemporary Approaches. Dubuque, IA: Wm. C. Brown.

Petty, Richard, and Jon A. Krosnick. 1995. Attitude Strength: Antecedents and Consequences. Hillsdale, NJ: Lawrence Erlbaum.

Roe v. Wade. 1973. 410 U.S. 113.

Slotnick, Elliot, Jennifer Segal, and Lisa Campoli. 1994. "Television News and the Supreme Court: Correlates of Decisional Coverage." Presented at the annual meetings of the American Political Science Association, New York.

Webster v. Reproductive Health Services. 1989. 492 U.S. 490.

Timothy R. Johnson is Assistant Professor, Department of Political Science, Southern Illinois University-Carbondale, Carbondale, IL 62026. Andrew D. Martin is Assistant Professor, Department of Political Science, SUNY at Stony Brook, Stony Brook, NY 117944392.
COPYRIGHT 1998 Cambridge University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Johnson, Timothy R.; Martin, Andrew D.
Publication:American Political Science Review
Date:Jun 1, 1998
Words:9041
Previous Article:The illusion of power and the disruption of moral norms: Thycydides' critique of Periclean policy.
Next Article:A political explanation of variations in Central Bank independence.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters