Lacking Legislative Experience: The Impact of Changing Justice Backgrounds on Judicial Review

Abstract

Though once commonplace, presidents no longer nominate individuals with legislative experience to the U.S. Supreme Court. What difference does this shift make? Drawing on theories that connect judicial background characteristics to decision-making, I test whether legislative background impacts federal judicial review. Using nonparametric matching and almost 150 years of judicial review decisions, I find that, while such experience decreases the likelihood of striking down a law, the effect is small. By contrast, partisanship has a much stronger impact, with justices more likely to strike down laws when the enacting Congressional majority is a different party from their appointing president.

After the wave election of 1994, Republicans in Congress promised to bring federal spending under control. The line-item veto, long desired as a tool for cutting pork-barrel spending from appropriation bills, was proposed almost immediately after Republicans took power, coming to fruition in 1996 as the Line-Item Veto Act. After its initial use by the Clinton administration led to a constitutional challenge, a six-justice majority invoked judicial review and struck down this law as a violation of the Presentment Clause in Clinton v. City of New York (1998). While the justices focused primarily on whether a line-item veto constituted a usurpation of legislative authority or was, instead, a constitutional grant of discretion to the executive branch, the statutory reform also raised deeper separation of powers questions. Could Congress meaningfully cut federal spending on its own, given constituent pressures and the desire to be reelected? Would the line-item veto manifest not as a tool for budget discipline, but instead as a shift in power to the executive branch, weakening Congress and threatening judicial independence (e.g., Fisher 1997)?

For much of its history, a legislative that considered similar issues would benefit from having multiple members with prior Congressional experience. In Clinton, by contrast, not a single member of the Court had Congressional experience, while only Justice O’Connor had served in a state legislature. Legislative experience was similarly lacking when the Court utilized judicial review to strike down the legislative veto in INS v. Chadha (1983) or limited Commerce Clause authority by striking down the Gun-Free School Zones Act of 1990 in United States v. Lopez (1995). Would these decisions have gone any differently had they been decided by a Court whose members had been legislators? The lack of legislative experience on the court is not a recent phenomenon, as Justice Ketanji Brown Jackson’s 2022 nomination to the Court continued a streak of appointees without any prior legislative service that, with the exception of Justice O’Connor, has not been broken since 1955. This trend stands in stark contrast to the late 18^th and 19^th century, where justices with legislative experience were the modal outcome, or in the first half of the 20^th century, where such a background was still common.

We have gone, then, from a Supreme Court whose membership contained significant legislative experience to one with little or none. How might this shift have affected the practice of judicial review? Might a Court that included former legislators, for example, have been more reluctant to disdain legislative rationales for maintaining the Voting Rights Act in Shelby County? More generally, judicial scholars have drawn attention to the modern Court’s increased willingness to appropriate power to itself (e.g., Keck 2004), with judicial review serving as a key tool for doing so. Could resurrecting a norm of choosing nominees with legislative experience affect this trend?

The development of new judicial databases now makes it possible to examine this question. Using 150 years of United States Supreme Court votes on judicial review, I test whether prior legislative experience reduces the likelihood of striking down a federal law. Using justice-votes as my unit of analysis and coarsened exact matching (Iacus, King, and Porro 2012) to preprocess my data, I find that, while there is evidence it does, the effect is modest, reducing the likelihood of a justice voting to strike down a law by about 3-4% (varying by the measure of legislative experience). By contrast, and in line with orthodox theories of how legal policy preferences drive Supreme Court decision-making (Segal and Spaeth 2002), the orientation between the party of a justice’s appointing president and the party of the enacting Congress has a more robust effect on judicial review votes, with justices 10-20% more likely to strike down a federal law enacted by the opposite party, depending on the baseline.

Besides the practical implications of my findings—that returning to the practice of nominating justices with legislative experience is unlikely to meaningfully impact the exercise of judicial review—my study also contributes to the literatures on judicial background and judicial attitudes. In line with their findings, this study adds to evidence of a judicial background impacting decision-making in cases where that background is theoretically relevant to the subject matter at hand (such as a judge’s gender in sexual harassment cases (Boyd, Epstein, and Martin 2010)). This study also supports one of the most common findings in judicial politics—that judicial policy preferences, ideology, and partisanship affect Supreme Court decisions—by providing systematic evidence of partisan effects over a time period quantitative studies rarely examine.

Judicial decision-making, social background, and legislative experience

Until the halfway point of the 20^th century, the Supreme Court was populated not only by former judges but also by former legislators: state legislators, members of the House, and U.S. Senators. The past seventy years, by contrast, have seen a Court whose members almost uniformly lack such experience. Figure 1 shows the striking nature of this change.

Figure 1. Supreme Court Justices With and Without Prior Legislative Experience by Year of Appointment.

Note: Figure 1 plots the appointment of Supreme Court justices, distinguishing between justices that had held and had not held legislative office prior to appointment.

As the figure suggests, the Eisenhower administration serves as a cut-point for change, with Justice O’Connor being the only justice with legislative experience after 1949 (Epstein, Knight, and Martin 2003; Epstein et al. 2009). At least two accounts for this shift have been offered by scholars, both of which have their roots in negative reactions to judicial appointments made by the FDR and Truman administrations. One account points to Congress as the catalyst for change, particularly among members opposed to the Court’s increasing role in advancing civil rights. Nominating more “professional” judges, civil rights opponents hoped, would select institutionally conservative judges less likely to use the courts to advance social change (Epstein et al. 2009).¹ Other scholars credit the Eisenhower administration, which desired to avoid the perceived “cronyism” associated with previous administrations by only nominating highly qualified judges, i.e., those with prior judicial experience (Goldman 1999).

Regardless of its origin, the norm of nominating federal appellate judges to the Court has held, perhaps because of the perceived advantages it brings to the politics of nominations. As the Court grew more important in both politics and policymaking, presidents have sought to protect their nominees from charges of partisan bias, lack of qualifications, or both. Choosing existing appellate court judges ameliorates both concerns, as these nominees are both more easily deemed qualified and less likely to have served in partisan politics. As Epstein, Segal, and Westerland (2008) found, perceived qualifications do affect the probability of a Senator voting to confirm a nominee, if less so than ideological congruence. Regardless of whether the strategy of choosing former appellate judges actually improves the chances of successful nomination, however, it remains a dominant strategy for presidential appointments to the Supreme Court.

The question here is how and to what degree this shift changed judicial behavior. In this study, both because it has a reasonable theoretical connection to legislative background and because it is behavior for which we now have high-quality contemporary and historical data, I focus on the judicial review of federal statutes. Most recent high-profile, systematic studies of judicial review have examined it in the context of separation-of-powers dynamics. This literature includes assessments of how increasing ideological or partisan distance between the Court and other institutions can decrease the incidence of judicial review (Harvey and Friedman 2005; Segal, Westerland, and Lindquist 2011), how the Court may use judicial review to support the goals or values of the current partisan regime (Gillman 2002; Whittington 2005; Lindquist and Solberg 2007), or how Congress uses jurisdiction-stripping proposals (Marshall, Curry, and Pacelle Jr. 2014) or “court-curbing” bills more generally (Clark 2009) to signal its discontent with the Court, leading to fewer laws being struck down thereafter. These studies, however, have more to say about interbranch relations or the degree to which Court majorities engage in strategic behavior and less about how an individual justice’s characteristics impact their proclivity to strike down laws.

At the individual or justice-vote level, judicial review is often subsumed within the broader context of how justices vote on the merits, such as serving as a control variable in studies of judicial decision-making. The predominant theory on merits decisions at the Court, of course, is that justices decide cases in line with their legal policy preferences, values, or ideological frameworks (e.g., Segal and Spaeth 2002). In the context of judicial review, this would mean that a liberal justice, for example, should be less likely to strike down a liberal law than their conservative fellows (Lindquist and Solberg 2007). Though the claim that ideology (however conceptualized) impacts judicial decision-making is, by now, orthodox social science, there remains ample room for other factors, such as social background, to play a role in explaining decision variance. This social background model—which can include both career background and demographic variables such as race, gender, or religion—was pithily defined by Gryski, Main, and Dixon as where “shared social and political traits reflect similar socialization processes and life experiences, which in turn produce similar attitudes and ultimately behavior (votes)” (Gryski, Main, and Dixon 1986, 528).

An earlier generation of judicial politics scholarship used quantitative social science techniques to connect social background—whether demographic or career—to outcomes, but these efforts garnered mixed results (Heise 2002), and were soon after supplanted by work focusing on a judge’s political or legal policy attitudes. The last twenty years, however, have seen a resurgence in research finding connections between social background factors and judicial decision-making. In contrast to the older generation of studies, contemporary scholars have focused on narrower slices of doctrine and policy, reducing incomparability, and leading to more limited and theoretically plausible hypotheses. Armed with better theories, better data, and more sophisticated methodological instruments, these studies have demonstrated the importance of an individual’s background characteristics in judicial decision-making. For example, we have findings that men decide sex discrimination cases differently when there are women on the appeals panel (Boyd, Epstein, and Martin 2010), that judges with daughters are more likely to take a liberal or feminist position in gender-related cases (Glynn and Sen 2015), that black judges are more likely to support affirmative-action programs than non-black judges (Kastellec 2013), that judges with prior criminal defense experience had different responses to the adoption of federal sentencing guidelines than those without (Sisk, Heise, and Morriss 1998), that judges with prior service in the executive branch were more likely to support the president in separation of powers cases (Robinson 2012), and that judges who had been public defenders differ in their sentencing practices from those who had not (Harris and Sen 2022).

Given this literature, what can one say about how prior service in a legislature might affect judicial review or other facets of judicial behavior? Relatively little. In the wake of Justice Souter’s retirement from the Court in 2009, for example, rumors swirled that President Obama might appoint Michigan Governor Jennifer Granholm, in part because of the “real-world” government experience she might bring to a court filled with former appellate judges (Saulny 2009). How such experience might change her behavior relative to other judges, however, was unclear, as there was almost no literature to either guide the formation of predictions or assess their empirical likelihood. The absence of large-scale quantitative research on legislative backgrounds was probably the result of data limitations, specifically that almost all Supreme Court datasets and associated measures had been left-censored at the start of the Warren Court. The period that we could study judicial background, then, was also the time period where justices with legislative experience vanished from the Court, leaving us with insufficient variation in the key independent variable. The development of new datasets, however, such as the “Legacy” United States Supreme Court database that extends back to 1791 (Spaeth et al. 2022) and Whittington’s Judicial Review of Congress dataset (Whittington 2022), now make it feasible to test the impact of legislative background on judicial review over the period where such experience varied.

How might we expect legislative background to affect judicial review? I suggest two potential mechanisms that might connect background to behavior. The first mechanism rests on a theory of information effects, where judges with particular backgrounds decide relevant cases differently than those without because they better understand the legal and policy issues before the court. Such information effects are also important because one judge’s experience may influence other judges on multi-member courts, such as where the presence of women on federal appellate panels deciding gender-rights claims led to a higher probability of male judges voting for that claim (Boyd, Epstein, and Martin 2010). A second potential mechanism comes from organizational sociology (e.g., Chao et al. 1994). Generalizing greatly, this literature suggests that members of organizations or institutions tend to adopt the values, goals, and norms shared by their fellow members, often in a relatively short period of time.² These effects persist even after one leaves the organization in question; however, for judges, we might expect both time and judicial socialization to interact with or diminish them.

My goal is not to test competing mechanisms, as I lack the data and measures to do so; rather, I only wish to show that there are viable bases for theorizing that prior legislative experience affects the exercise of judicial review on the Supreme Court. How might each mechanism operate? First, regardless of which mechanism was in play, we would expect that if there were effects from the loss of experience on the Court, they would be seen in decisions related to that legislative experience. Boyd et. al (2010), for example, found a relationship between a judge’s gender and their votes on cases involving gender, but not in other legal issue areas. A decision to exercise judicial review—in effect undoing the work of legislatures—is similarly related to prior experience as a legislator.

Second, each mechanism generates a straightforward expectation as to how legislative experience might impact judicial review votes. An information effects account would have former legislators possess a greater understanding of the complexities of lawmaking and the difficulty in successfully crafting legislation. Such awareness might make them more reluctant to strike laws down. Chao et al.’s framework suggests similar expectations. To varying degrees, legislators learn skills needed to pass legislation. Doing so might give them more insight into how difficult such a process might be, creating more sympathy or respect for lawmakers. Legislators join networks with other legislators while serving, developing personal relationships with other members. These relationships may affect judicial review, strongly if the law before the Court was developed by individuals within a justice’s networks, and weakly if not. Legislators may also come to identify with the goals and values of legislatures, such as representativeness, democratic accountability, and responsible policymaking. While support for these values might attenuate over time, justices who are former legislators may still be less likely to strike down laws other legislators create, relative to their fellows.

Regardless of which mechanism might operate here, each provides good reasons for hypothesizing that legislative experience should make justices less willing, on average, to strike down laws as unconstitutional.

Data, variables, and methods

Data

To build the database for this study, I first use Whittington’s Judicial Review of Congress Dataset, which currently ranges from 1792 to 2022 (Whittington 2022). Whittington’s dataset focuses on the review of Congressional authority, omitting cases, for example, where the Court reviews the actions of federal bureaucrats or judges.³ Using Whittington’s data limits my analysis to the judicial review of federal statutes. Without a similar database for state statutes, including judicial review of state law is infeasible, given the much larger number of state statutes reviewed and the challenges involved in something as simple as identifying the appropriate cases for inclusion in the data. The USSC database, for example, does not allow one to distinguish between judicial review of state statutes and judicial review of state administrative, executive, or judicial activity, meaning such identification would need to be done on a case-by-case basis for thousands of cases.

I left-censor Whittington’s data at December 1872 to avoid dealing with the modeling complexities introduced by the Whig Party.⁴ Using this list of cases, I generate a database of justice-votes drawing from both the United States Supreme Court (USSC) Legacy (1791-1945 terms) and Modern (1946-2022 terms) databases (Spaeth et al. 2022), combining my results into a single dataset with 10,837 observations nestled within 1,215 cases.

Variables

My dependent variable is the decision to strike down a federal law. Whittington’s database indicates whether the Court upheld or struck down a law but does not indicate how individual justices voted. Using the USSC database’s “majority” variable, I code a justice as voting to strike down the law (1) when they join a majority opinion that does so and not (0) when they are in dissent. I reverse the coding in the opposite scenario.⁵ I omitted observations where a justice did not participate in the case, reducing the effective size of the dataset to 10,461 observations.

My primary independent variable is a binary variable coded as 1 if the justice had prior career experience as a state or federal legislator, and 0 otherwise. I also create two continuous measures of legislative service, taking the total years of service (state and federal combined) as the first alternate measure and the square root of those total years as the second. I include the square root measure because there may not be a linear relationship between years of service and socialization, with earlier years instead being more important in establishing social background dynamics (e.g., Robinson 2012). To code the legislative experience variables, I relied on the backgrounds published in the Federal Judicial Center’s Biographical Directory of Federal Judges (2022), cross-checked against the U.S. Supreme Court Justices Database (Epstein et al. 2022).⁶ The organizational socialization literature suggests that socialization takes place relatively quickly (Thomas and Anderson 1998), and so I code the dichotomous experience variable as being met if a justice has a year or more of legislative service (the continuous variables have a similar threshold).

My model also contains several control variables. Most importantly, I control for the partisan orientation between the justice’s appointing president and the enacting Congress of the law under review. The historical reach of the data rules out contemporary measures of ideology or interbranch congruence that the judicial review literature regularly employs,⁷ forcing me to rely, instead, on cruder measures of partisanship. Specifically, and similar to Clark (2009), I construct a categorical measure of partisan orientation using the party of the justice’s appointing president and the makeup of the enacting Congress, coding 0 if they are the same, 1 if Congress is divided, and 2 if they are opposed. I also create an alternate measure of partisan orientation which codes whether the law before the Court was enacted by a Democratic or Republican “trifecta,” where one party controls not only Congress but the White House as well. This measure is coded as 0 if the law was passed by a trifecta of the same party as the justice’s appointing president, 1 if there is no trifecta, and 2 if the law was enacted by a trifecta of the opposite party. Given longstanding evidence that justices are influenced by their legal policy preferences (Segal and Spaeth 2002), I would expect justices to be more likely to strike down a law passed by a Congress of the other party.

I include several other controls suggested by the relevant literature. A law’s age is a common control in the judicial review literature (Harvey and Friedman 2005; Lindquist and Solberg 2007; Sala and Spriggs 2004), as relatively recent laws may be more likely to be challenged while older laws may be more vulnerable to challenge should it occur, given changes in social facts or other aspects of jurisprudence. Whittington’s database provides the age of the law in months at the time of its review by the Court, which I include. Drawing on findings that the Court might review laws differently depending on the policy issue area the law regulates (Lindquist and Solberg 2007; Marshall, Curry, and Pacelle Jr. 2014), I use Whittington’s issue area or subject matter variables.⁸ I control for general time trends by including the year the case was decided. Finally, I control for the law’s political or policy importance by adopting Whittington’s coding on whether the law under review was listed as a “landmark statute” in Landmark Legislation 1774-2012: Major U.S. Acts and Treaties, 2nd ed. (Stathis 2014).⁹

Methods

Given the binary nature of the dependent variable, logistic regression is an appropriate estimation method. To further improve estimation, I employ nonparametric matching (Ho et al. 2007), a preprocessing technique that draws on the logic of experiments to find “control” and “treatment cases” that are similar except for the presence of the treatment variable, creating a match. Using matching to improve estimation is now a commonplace technique in the social background literature (e.g., Boyd, Epstein, and Martin 2010; Kastellec 2013). Specifically, I use a matching method, Coarse Exact Matching (CEM), that does not require “exact” matches to balance the dataset, which prevents cases from being omitted at high enough rates to destroy a study’s statistical power. The CEM matching algorithm is well accepted and has been used successfully in other studies of judicial behavior (e.g., Black and Owens 2012). After using the CEM algorithm, my dataset retained 8,741 observations and reduced the degree of covariate imbalance (as shown by the L1 statistic) from. 453 to. 130—a sizable improvement.¹⁰ A fuller description of the matching protocols the study employs can be found in Supplementary Appendix B, while summary statistics for variables in the matched data are presented in Supplementary Appendix C.

Results

After preprocessing the data with CEM, I use logistic regression on the matched data, with robust standard errors clustered on the case.¹¹ Figure 2 provides the point estimates and 95% confidence intervals for each model variable.

Figure 2. Effect of Model Variables on Probability of Voting to Strike Down a Federal Law, Binary Measure of Legislative Experience.

Note: Figure 2 plots the point estimates (dots) and 95% confidence intervals (lines) of the model coefficients on the likelihood that a Supreme Court Justice votes to strike down a federal law. The reference category for the Divided Government and Opposite Congress codings for partisan orientation is where the law under review is passed by a Congress of the same party as the justice’s appointing president. The reference category for the issue area variables is Due Process.

The model results present several items of interest. First and most importantly, as theorized, prior experience as a legislative official reduces the likelihood of striking down a federal law. However, the impact is modest: holding all other model variables constant, the marginal difference between a justice with prior elected experience and one without is about 3% (specifically, with other variables held constant, legislative experience drops the predicted probability of a justice striking down a law from 30.0 to 27.2%). It’s true that small impacts can be substantively meaningful. In their study on whether background as a public defender impacts sentencing in federal courts, for example, Harris and Sen (2022) find a similarly small impact of career background on the probability of a defendant receiving incarceration (by contrast, they find a much larger impact of public defense background on the length of incarceration). However, the authors point out that given the sheer number of sentences handed out in federal courts, even small shifts in probability would have substantial real-world impacts. Judicial review of federal laws, by contrast, is uncommon, and it’s unclear whether the effect size found here would meaningfully impact judicial decision-making.

By contrast, the impact of partisanship is more substantial. Justices are ten percent more likely to vote to strike down a law enacted by a Congress of the opposite party of the justice’s appointing president as compared to a law enacted by a Congress of the same party. When using the baseline of a law passed under divided government compared to considering a law enacted by the opposite party, the likelihood of voting to strike increases to almost twenty percent. The results, thus, support the standard attitudinal story that justices factor partisanship or ideology into their decision-making. As for the other control variables, laws dealing with economic, separation of powers, or federalism concerns were less likely to be struck down than laws involving due process (perhaps because such laws were less likely to activate high-salience attitudinal concerns), and the likelihood of striking down a federal law increases over time, though the effect is small. A law’s age or landmark status, by contrast, have no discernable effect.

I present the model results for the continuous measures in Figure 3.

Figure 3. Effect of model variables on probability of voting to strike down a federal law, continuous measures of legislative experience.

Note: Figure 3 plots the point estimates (dots) and 95% confidence intervals (lines) of the model coefficients on the likelihood that a Supreme Court Justice votes to strike down a federal law. Model A uses years in office as the measure of prior legislative experience, while Model B employs the square-root of those years. The reference category for the Divided Government and Opposite Congress codings for partisan orientation is where the law under review is passed by a Congress of the same party as the justice’s appointing president. The reference category for the issue area variables is Due Process.

For all variables other than legislative experience, the results here are similar to the main model. For legislative experience, the linear measure is a poor fit, being minute in effect and statistically insignificant, while the square-root measure echoes the results shown in Figure 2, with the effect of legislative experience being both clearly distinguished from zero while also small in impact. Specifically, moving the square-root of years in legislative service from its minimum to maximum value while holding all other values constant reduces a justice’s probability of striking down a law by about 4.5%, an impact in line with the downward shift seen in the main model.

To further assess the robustness of my findings, I make three substantive changes in the model design and re-estimate the results. For each model variation, I preprocess the data using CEM as described in Supplementary Appendix B. First, I recode the binary legislative experience variable to include justices whose served in the executive branch (at the local, state, or federal levels), effectively creating a “former elected official” measure. Second, I test a model that restricts the binary experience variable to federal legislative experience. Third, I replace the divided government version of the partisan orientation variable in the main model with the trifecta measure, as described above.

In Figure 4, I present the coefficients and confidence intervals for the legislative experience variable used in each model (complete results for each model can be found in Supplementary Appendix Table D-1), as well as information about the matching process for each.¹²

Figure 4. Model comparisons of binary legislative experience measures and matching diagnostics.

Note: The top plot of Figure 4 shows the model estimates of legislative experience’s impact on the probability of striking down a federal law, with dots representing point estimates of the logit coefficient and lines indicating the 95% confidence interval. The bottom left plot illustrates the reduction in the L1 statistic—a measure of data imbalance—before (triangles) and after (dots) CEM matching. The bottom right plot shows the proportion of cases matched under CEM (with exact numbers of cases retained annotated nearby).

While the findings from the main model hold up in two of the three alternate specifications, the coefficient for legislative experience is smaller and no longer statistically significant when the experience variable only codes for prior Congressional experience. While this may be the result of reduced statistical power, given that the matching process in this model retains fewer cases, one still might have expected either a larger coefficient or a statistically significant relationship when the variable is limited to former members of Congress. Overall, the weight of the evidence points to legislative background having a modest negative effect on the likelihood of striking down federal laws, but the failure of experience to matter in the Congress-only model suggests that some caution is warranted.

As for the other variables—the impact of party orientation remains strong and consistent in situations where a justice must rule on a law enacted by the opposite party, however defined. The finding that laws produced under divided government are less likely to be struck down than even laws passed by a Congress of the same party as the justice’s appointing president, by contrast, only appears in two of four models. All other controls behave similarly across all comparisons.

Finally, I test for interactions between the binary measure of legislative experience and the partisan orientation of the law in both my main model and the three alternate models. It seems plausible that legislative experience and partisan orientation might interact in interesting ways. In tune with the literature on background discussed above, where background or background effects in panel decisions dampen the impact of ideological or partisan variables, one might theorize that legislative experience might lessen the effects of partisanship. In other words, partisanship might have more room to operate if legislative experience is absent. The results, however, do not support this theory. Only one interaction in one of the four models reaches statistical significance: when using the trifecta measure for partisan orientation, judges with legislative experience are more likely to strike down federal laws under divided government. As this is a singular finding with no clear theoretical basis, the stronger inference is that (perhaps surprisingly) there are no interaction effects between background and a judge’s partisan orientation towards the law under review. Supplementary Appendix Table D-2 provides complete results of interactions for each of the four models.

Discussion

Regarding legislative experience and its impact on judicial review, the results of this study have more theoretical than practical significance. In terms of the scholarly literature, this study joins many others that find that judicial background impacts legal policy outcomes in an issue area relevant to that background. However, if politicians were interested in knowing whether returning justices with legislative experience to the Court would substantively change the practice of judicial review, the answer is probably no. While the lack of legislative experience among today’s Supreme Court justices probably does increase the likelihood that its members will strike down federal laws, the effect may be too small to justify a change in nomination patterns.

By contrast, the study provides evidence of a more robust impact of partisanship on judicial review votes, reinforcing the commonplace finding that Supreme Court justices are influenced by such partisanship in their decision-making. Importantly, the vast majority of quantitative studies have little to say about judicial behavior prior to the Warren Court, while this study shows partisanship is a factor in a dataset containing late 19^th and early 20^th century cases. That such effects appear even using simplistic measures of partisanship suggests they are rooted in the American judicial system, rather than being only a consequence of modern political and legal dynamics.

As with most large-scale, quantitative studies of judicial decision-making, this study has its shortcomings. Most importantly, the study focuses on an easily measurable outcome: a vote to strike down a federal law. Background experience might instead affect judicial decision-making in more subtle ways, impacting agenda control, opinion content, or amici cited. Given the difficulties with data collection or lack of variation along the dependent variable, this study also does not examine judicial review of state laws or voting decisions by lower federal courts.¹³ Finally, it may be that the mechanisms hypothesized as potentially driving differential behavior for judges with legislative backgrounds—such as empathy, information, or socialization—might manifest themselves more clearly in a small set of cases that directly implicate the institutional power of Congress itself, such as Chadha or Powell v. McCormack (1969). This case set, however, might be too small to assess using standard statistical methods.

Appointing judges with legislative or political background may have other impacts on the Court beyond those examined here. Supreme Court justices are generalists often facing specialized issues. A more diverse court may be better equipped to handle such specialization (Epstein, Knight, and Martin 2003). A Court with a greater diversity in backgrounds might also be less likely to heavily weigh the preferences of judges from their former circuits (Epstein et al. 2009) or judges, in general, relative to other political actors (Hinkle and Nelson 2018), which in turn might improve the Court’s ability to act strategically or properly assess legislative responses to its decisions. When it comes to judicial review, however, returning legislators to the Court seems unlikely to significantly change judicial behavior