Scholars of a variety of stripes are concerned today with if and how ordinary citizens change their views on various aspects of politics. For example, Levendusky (Reference Levendusky2023) reports the results of concerted framing efforts to induce change in partisan affective polarization. Persuasibility experiments in political intolerance research have long sought to convince people to give their initial position a “sober second thought” in hopes that intolerance could be converted to tolerance (e.g., Gibson Reference Gibson1998). And scholars of the U.S. Supreme Court have studiously investigated whether unpopular Court rulings could cause people to alter their views toward the institution (e.g., Christenson and Glick Reference Christenson and Glick2015).
While these efforts are often successful at documenting attitude change, nearly all face a critical limitation—their inability to show that the attitude change persists over time. Indeed, as Druckman (Reference Druckman2022, 75) observed:
The flip side of “what happened before” is “what happens after”: how long a given persuasive effect lasts. Although the question has been considered since [1951], it is far from settled ….
Thus, a critical unanswered question for those investigating attitude change is whether changes created by various interventions linger.
Judicial scholars have been especially interested in trying to understand how attitudes toward legal institutions evolve. Research has examined the effects of contentious nominations to the U.S. Supreme Court (e.g., Krewson Reference Krewson2022); other work focuses on the consequences of unpopular Court rulings (e.g., Bartels and Johnston Reference Bartels and Johnston2013). State court studies have investigated whether judicial elections reshape attitudes (e.g., Gibson Reference Gibson2012), while research on the legitimacy of the police has considered the role of experiences with unfair treatment by legal authorities in altering support for law enforcement (e.g., Gibson Reference GibsonForthcoming; Rengifo and Slocum Reference Rengifo and Slocum2020).
Many are skeptical that changes persist (e.g., Santoro and Broockman Reference Santoro and Broockman2022). For instance, earlier research has suggested that a drop in support for courts can be reversed through a process dubbed “values-based regeneration” (Mondak and Smithey Reference Mondak and Smithey1997). The theory is straightforward: a shock to a system (judicial attitudes) dissipates over time (via forgetting, the addition of new information, public attention to events wanes, or because the experiment is over), allowing the system to revert to its previous state. Regarding institutional support, that state is typically one of allegiance to the institution because such allegiances are learned at an early age, are reinforced by exposure to judicial symbols, and are therefore resistant to change (Gibson and Caldeira Reference Gibson and Caldeira2009). In short, democratic values regenerate diffuse support for judicial institutions. If the theory is correct, then the significance of attitude changes is limited indeed: events may come along that disrupt attitudes, but once those happenings fade in memories, attitudes revert to their pre-event state.
When it comes to Supreme Court attitudes, one of the most consequential policy changes in recent memory is that associated with the Dobbs ruling abrogating the abortion rights first announced in Roe v. Wade. Footnote 1 In a recent article, I (Gibson Reference Gibson2024b) reported that the Court’s ruling overturning abortion rights did unprecedented damage to the institution’s popular legitimacy. Based on a one-shot survey fielded right after the decision was announced, I concluded “that Dobbs produced a sizable dent in [the Court’s] institutional support, perhaps an unprecedented degree” (1). Importantly, my analysis was unable to address the crucial issue of whether that “dent” has persisted over time.
This letter’s purpose is to assess whether the Dobbs decision had a lasting effect on Supreme Court support. The most telling way of determining whether the consequences of a Court decision linger is with panel data (although panel data also present their own significant analytical challenges—e.g., typically very large attrition rates).Footnote 2 My analysis cannot and does not rely on panel surveys, however. Absent data on individual-level change, the next best analytical strategy requires several steps, with the initial task involving establishing whether the aggregate-level change in support for the Supreme Court exists, based mainly on comparing a new 2023 survey with my 2022 results. Next, I examine changes in the effects of a host of micro-level correlates of support.
I also significantly expand my predictive theory. Missing from his post-Dobbs analysis is any consideration of the connections between democratic values and institutional support. I justified the exclusion of the values variables via the assumption that values are largely orthogonal to the other predictors of diffuse support.Footnote 3 However, according to the values-based regeneration mechanism, Court allegiance after a time should become closely reconnected to one’s degree of support for democratic values. While a reasonable hypothesis posits that the Dobbs decision weakened these relationships—in the short term, highly salient and controversial contemporary events could well have had more influence on diffuse support than long-standing value commitments—over time, the values/support relationship may have reasserted itself, according to the theory. Consequently, I also expect a strong relationship between democratic values and Court support in the post-Dobbs period, including in my analysis measures of the values that have been found in previous research (e.g., Gibson and Caldeira Reference Gibson and Caldeira2009) to predict Supreme Court support.
Generally, I find that the effects of the Dobbs decision on Court support have persisted over time, although perhaps at a slightly weakened level. I also discover in the 2023 data a close connection between democratic values and institutional support, which may signal values-based regeneration. At the most theoretical level, I conclude that the question of whether attitude changes persist is more complicated than it appears at first glance, which means that additional research on mechanisms of decay and persistence is essential.
THE UPDATING MODEL
The model that most scholars embrace goes something like this. Events occur, and then, people perceive and assess them. The assessments are used to update overall judgments of the performance of the institution (see Lodge, Steenbergen and Brau Reference Lodge, Steenbergen and Brau1995). For some, the updating is to the ideological distance between themselves and the institution (e.g., Bartels and Johnston Reference Bartels and Johnston2013). For others (e.g., Strother and Gadarian Reference Strother and Gadarian2022), the updated assessments pertain to the degree of perceived politicization of the institution. For Easton and many others (e.g., Haglin et al. Reference Haglin, Jordan, Merrill and Ura2021), the updating is to specific support (general assessments of the performance of the institution). After a time, it is even possible (if not likely) that the original reasons for updating one’s running tally of institutional performance assessments are forgotten even though the residue of the events (the updated tally) remains in place. When events are still fresh in the minds of people, assessments can affect institutional support; over time, however, events’ effects get filtered through specific support. Of course, diffuse support is shaped by other factors (e.g., democratic values) as well, but the influence of short-term events is captured by measures of awareness and assessments of those events, and the cumulative effects of events are captured by updated specific support. Figure 1 depicts a model of how these various processes apply to the Dobbs decision and to other decisions and events in general.
Figure 1. A Directed Acyclic Graph of Event Evaluations, Specific Support, and Institutional Support
Notes: Support for linkages from prior research: (a) Ansolabehere and White (Reference Ansolabehere and White2020); (b) Ansolabehere and White (Reference Ansolabehere and White2020); (c) Gibson (Reference Gibson2024b); (d) Gibson and Nelson (Reference Gibson and Nelson2015); (e) Gibson (Reference Gibson2024a; Reference Gibson2024b). See also Supplementary material E. “Awareness” and “Assessments” are included as a reminder of the difficulty (but not impossibility) of events about which people are unaware influencing their Court attitudes and that, even when aware, events are in no sense “objective” but are instead filtered through understandings and assessments (which, of course, vary across the population).
I specifically define “lingering” in this research as the consequences of a decision once updating to specific support is taken into account. That is, I hypothesize that, over time, the effects of most decisions lose some of their efficacy because they get incorporated into broader institutional performance evaluations. If, after some time, individual ruling assessments still have a direct impact on diffuse support, then the case will be judged to be especially influential—and the effect to have “lingered.” My expectation is that the influence of few decisions lingers and the effects of no decisions linger over a lengthy period of time.
So, as a dynamic process, unwanted decisions influence support in the short term, but in the longer term most decisional assessments get incorporated into specific support, which can drive down diffuse support, until diffuse support is (or may be) resuscitated by the psychological need for consistency between general democratic values and attitudes toward one of the most important democratic institutions. Put more succinctly, many salient decisions may temporarily undercut the Court’s legitimacy, but only a small handful will do so in the long term.
TRACKING AGGREGATE-LEVEL DIFFUSE SUPPORT
Following my earlier research (Gibson Reference Gibson2024b), I used three indicators of institutional support for the Supreme Court. (Supplementary material B reports the measurement of all the concepts used in this analysis). I provided 2022 evidence of both the validity and reliability of this item set. In my new 2023 nationally representative survey (see Supplementary material A),Footnote 4 the set is also quite valid and reliable. As a measure of the latent construct “diffuse support for the Supreme Court,” I use an index that is the average response to these propositions.
The first analytical question I address is whether aggregate-level Court support has changed. Figure 2 adds my results to my findings for four earlier surveys. For simplicity, the figure reports the percentages of respondents within each survey giving no supportive replies to the three diffuse support indicators.
Figure 2. Change in Supreme Court Legitimacy, Pre- and Post-Dobbs
Notes: The difference in the average index of diffuse support (not shown in this figure) across the five waves is statistically significant at p < 0.001. The difference between the July 2022 and February 2023 index scores is not statistically significant (p > 0.05). The difference in the average index of support in July 2020 compared to February 2023 is significant at p = 0.006.
Source: Surveys prior to February 2023, Gibson (Reference Gibson2024b).
While extant research generally shows that the Supreme Court’s legitimacy changes little (e.g., Nelson and Tucker Reference Nelson and Tucker2021), these data support a quite different conclusion. During the first three surveys, the percentages expressing no Court support averaged somewhere around 30%; after the Dobbs ruling, the average percentage climbed to the low 40% range. As noted in the figure, for the institutional support index, the difference between the 2022 and 2023 surveys is not statistically significant, while the difference between the 2020 and 2023 surveys is quite significant (and even more so for the difference between the surveys in 2020 and 2022). Some might be tempted to conclude that the 2023 results suggest an eased dip in Court support, but the minimalist statistical conclusion is that the dip persisted and that the data reveal substantially less diffuse support in the post-Dobbs era than in the pre-Dobbs era.
CHANGING PREDICTORS OF DIFFUSE SUPPORT
Table 1 replicates my earlier analysis, with one exception: Equation VI adds three measures of democratic values to Equation V, which, of course, included no democratic values indicators.Footnote 5
Table 1. Predictors of Supreme Court Institutional Support, Post-Dobbs, 2023
Notes: Significance of unstandardized OLS regression coefficients (b): *** p ≤ 0.001 ** p ≤ 0.01 * p ≤ 0.05. All variables are scored to range from 0 to 1. For their distributions, see Supplementary material C. The coefficient shown in parentheses after the predictor’s name is the bivariate correlation with diffuse support. s.e., the standard error of the unstandardized regression coefficient.
The first notable aspect of these results is that adding the democratic values variables to Equation V (shown in Equation VI) has little effect on the substantive conclusions about the other predictors of institutional support.Footnote 6 The largest change in a coefficient is associated with the respondent’s level of education. My Equation V seems to overestimate education’s effect in part because better-educated people are more likely to embrace democratic values. All other coefficients change very little from Equation V to Equation VI, corroborating my claim that my 2022 equations, without measures of democratic values, do not produce biased estimates.
At the same time, support for democratic values is a powerful covariate of diffuse support for the Supreme Court, as has been found in essentially all previous research studies. The strongest predictor of diffuse support in Equation VI is rule-of-law attitudes, although both open-mindedness and prioritizing individual liberty are also useful predictors.
Returning to the main objective of assessing the persistence of the effects of the Dobbs ruling, Equation I reports (somewhat limited) evidence of an interaction between awareness of the decision and approval of it, just as I found. However, the 2023 interaction is dramatically weaker: the interactive coefficient is 0.15 in 2023 but 0.43 in 2022. This indicates that, in 2023, at the highest awareness level (fairly widespread), approval and support are connected at 0.23 (0.08 + 0.15), but at the lowest awareness level (fairly rare) the coefficient is indistinguishable from zero (0.08). The much-weakened role of assessments and awareness for Court support is also signaled by the finding that, in the last three equations reported in Table 1, none of the three variables achieve statistical significance, which differs from my earlier findings.Footnote 7 Even though self-reported awareness of the Dobbs decision changed little from 2022 to 2023, the variability in awareness levels is much less closely connected to Court attitudes, possibly because even those reporting low awareness in 2023 had absorbed at least some information about the ruling from the widespread discussion throughout the latter half of 2022.
Notably, I also reported the results of this interactive model based on awareness and approval of the 2020 Barrett nomination/confirmation. In that analysis, the interactive coefficient was 0.18, which is quite similar to the 2023 Dobbs interactive coefficient of 0.15. This clearly suggests that the observed conditional effects of awareness are greatest shortly after the event’s occurrence. These findings imply that the conditioning effect of awareness of the decision may have reverted to normalcy.
Similarly, the connection between abortion attitudes’ moral content and diffuse support dissipated considerably over time. In 2022, the interaction of abortion preferences and the degree of their moral content were quite strong; in 2023, the interaction coefficient never achieved statistical significance. Similarly, in 2022, the Court’s views of those whose abortion preferences were grounded in moral concerns were closely connected, whereas in 2023 that connection weakened almost entirely. These findings are even more intriguing because aggregate abortion preferences changed little between 2022 and 2023, in terms of either support for abortion rights or the degree to which abortion preferences are infused with moral content (see Supplementary material C). Perhaps, for some, abortion shifted from being primarily a moral issue to being largely a political issue—also suggesting that an issue’s moral content must always be measured rather than assumed.
In the 2022 and 2023 equations, neither partisan nor ideological identifications were connected to institutional support (see also Supplementary material D). However, in 2023, the degree of ideological proximity between the respondent and the Court is weakly but significantly related to Court support, unlike in 2022. Perhaps some small portion of Dobbs’ effects is getting filtered through this specific support variable.
As I have noted, Table 1 reports that democratic values and institutional support are closely connected in the 2023 survey. The best evidence for values-based regeneration would be from a survey conducted immediately after the Dobbs decision but that included measures of democratic values. The expectation would be that the relationship between diffuse support and democratic values would be weakened because the support would reflect contemporary assessments of the decision more than long-standing values. To my knowledge, however, no survey data are available to directly test that hypothesis.
Values-based regeneration would suggest that after the “dust” from Dobbs settles, the strong “normal” relationships will reassert themselves. It may be useful, therefore, to determine whether the standard predictors of Court support change from their pre-Dobbs role. Some evidence on that score is available.
In my Supplementary material F, I report an analysis of pre-Dobbs institutional support using a July 2020 survey. That appendix sought to determine the effects pre-Dobbs of including or excluding measures of democratic values from an equation predicting diffuse support. The democratic values I consider are support for the rule of law, political tolerance, and a preference for liberty over order.
Considering the difference in the measures of democratic values used, a strict comparison of the 2020 results with the 2023 results is not possible. Still, it is noteworthy that for the measure nearly identical in the two surveys (support for the rule of law), the regression coefficient for 2020 was 0.19, and for 2023, it was 0.34. For a multi-item indicator of support for liberty over order, the 2020 coefficient was 0.18; for the single-item indicator in 2023, the coefficient was 0.10. For political tolerance in 2020, the coefficient was 0.10; for open-mindedness in 2023, the coefficient was 0.25. I reiterate that a strict comparison of the individual coefficients is ill-advised. But in the 2020 survey, the addition of the three measures of democratic values to the base equation raised R2 by 10 percentage points (see Table F.1); in 2023, in a more fully specified model, the addition of the three measures of democratic values raised R 2 by 18 percentage points. The minimalist conclusion I draw from these results is that the connection between democratic values and institutional support for the Court is at least as strong in 2023 as it was in 2020. Unfortunately, we simply do not know how strong the connection was right after the decision was announced. To reiterate, perhaps Table 1’s most important finding is that reactions to the ruling play a much smaller role in shaping institutional support than do democratic values.Footnote 8
My analysis obviously provides no dispositive test of the values-regeneration hypothesis. However, if values regenerate support, then the connections between values and support after the Dobbs controversy abated a bit should look like the connections between values and support prior to the ruling. The available data show pretty much that.
CONCLUDING COMMENTS
This analysis’s most obvious shortcoming is that it is not based on panel data. While panel data have their own limitations, most agree that the best way to study individual-level change is with individual-level data.
Nevertheless, this survey’s findings are compelling. The negative knock on the Supreme Court’s legitimacy associated with its Dobbs ruling persisted for at least six months after the decision. I conclude this based on several empirical findings:
-
• Court support did not change between the 2022 and 2023 surveys, with the 2023 results being significantly worse for the Court than the pre-Dobbs evidence. For those believing that the Court must maintain a deep “reservoir of goodwill,” these findings are ominous.
-
• The association between Dobbs assessments and diffuse support persisted in 2023, although considerably weakened, even when assessments and specific support are included in the same equation.
-
• My earlier results do not seem to have been compromised by his inability to include measures of support for democratic values.
At the same time, however, my analysis has produced some important caveats and conundrums about the persistence of the Dobbs effect.
-
• First, nearly all 2023 findings are weaker than those from 2022. As the simplest illustration, in 2022, Equation V accounted for 40% of the variance in institutional support. In 2023, the same equation was able to account for only 26% of the variance in support.
-
• Because I found, like my earlier research, a significant interaction between awareness of the decision and assessments of it, awareness levels matter. If a decision is not salient, the decision’s effect will be weakened. Highly salient decisions have a much better chance of producing significant consequences. Over time, the salience of any given decision undoubtedly wanes.
-
• Several sub-lessons are associated with this finding. First, awareness of decisions ought to be measured, not assumed. Second, when a decision is made, sufficient time must elapse for people to learn about the decision (and “sufficiency” could well be measured in months, not days). Third, awareness surely diminishes over time, perhaps as an individual decision’s effects get incorporated into overall performance assessments (specific support) and as its independent influence therefore dissipates. Conclusions about the effects of events on court attitudes may therefore be dependent upon the timing of the “post” survey.
-
• Some of the moral urgency of the abortion issue seems to have subsided by 2023. Perhaps abortion politics (e.g., political battles in various states over abortion rights) overtook morality considerations. If so, then the degree of an issue’s moral content can and does vary over time—and therefore must be measured, not assumed.
-
• Finally, I note that Supreme Court support is strongly grounded in democratic values in the 2023 survey. Indeed, as Table 1 reported, the addition of the measures of democratic values to Equation V increased R2 in the 2023 analysis from 26% to 44%. This may indicate that Court support by 2023 was becoming more closely aligned with democratic values than in 2022, although, obviously, democratic values measures in 2022 are not available. If Mondak and Smithey (Reference Mondak and Smithey1997) are correct about “values-based regeneration”—that an unwanted decision can knock the relationship of values and support off-kilter, but, over time, the relationship rights itself—then this empirical finding takes on greater theoretical significance. Without a substantial correlation of values and support in the 2023 data, the regeneration hypothesis would become less plausible. We do not know how long complete values-based regeneration requires; perhaps it needs more time than the six months considered here.
At a more widely applicable theoretical level, these findings suggest that whether and how attitude change persists is complicated. Change may be associated with levels of support for a policy or institution, but it may also be associated with the criteria upon which people base their support, as well as with whether the matter remains salient. Answering the simple question of whether attitude change persists is anything but simple—and certainly requires additional research.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit http://doi.org/10.1017/S0003055424000169.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/XNJSV3.
ACKNOWLEDGMENTS
I am deeply indebted to the Weidenbaum Center at Washington University in St. Louis and its director, Professor Andrew Reeves, and to the American Social Survey (TASS), directed by Professor Jacob Montgomery, for support for this project. Katie O’Quinn and Madeline Mader provided invaluable assistance on this research. I am also much indebted to Jeffrey Mondak, Matthew Levendusky, Linda Skitka, Jeffrey Yates, and Michael Nelson for comments on an earlier version of this article. This research was funded by the Weidenbaum Center, Washington University, in St. Louis.
CONFLICT OF INTEREST
The author declares no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The author declares the human subject research in this article was reviewed and deemed exempt from review by the Washington University in St. Louis IRB (#202309091). The author affirms that this article adheres to the principles concerning research with human participants laid out in APSA’s Principles and Guidance on Human Subject Research (2020).
Open access
