Strategic Retirement in Supreme Courts

The most controversial hypothesis about vacancies in the US Supreme Court states that justices time their retirement to allow presidents to nominate like-minded successors. Strategic retirements can be crucial in a divided court where a single vote can be decisive.Footnote ¹ However, scholars disagree regarding the role of political motivations. Some studies assert that the decision to retire is contingent on the ideological compatibility between judges and the incumbent president (King Reference King1987; Hagle Reference Hagle1993; Ward Reference Ward2003; Calabresi and Lindgren Reference Calabresi2006), while others find no evidence to claim that justices retire for political reasons (e.g., Squire Reference Squire1988; Brenner Reference Brenner1999; Yoon Reference Yoon2006).

Such inconsistent findings may result from two different problems. First, the empirical implications of this argument are not always straightforward, and different authors invoke unique results as evidence in support of the theory. For example, Calabresi and Lindgren (Reference Calabresi2006) find that judges are more likely to retire—rather than die in office—when the incumbent president belongs to the party that nominated the departing judge. Hagle (Reference Hagle1993, 36–37), by contrast, discovers that justices are more likely to retire during the president’s second term in office and are less likely to do so when they are former politicians. Zorn and Van Winkle (Reference Zorn2000) find that justices tend to retire at the beginning of the president’s second term, inferring that justices thus allow the president to fill the position with a like-minded successor. Kerby and Banfield (Reference Kerby2014) find that justices in Canada, Australia, and New Zealand are more likely to retire in electoral years, inferring from this fact that judges expect their parties to lose the general election.

These examples suggest that quite disparate empirical findings are offered as evidence of the theory. In the following pages, we employ a simple model to derive crisp empirical implications from the theory’s core assumptions. We identify the fundamental implication of the theory, plus three secondary corollaries that are often overlooked in the literature.

Second, empirical studies may offer weak or conflicting results because the number of justices serving in the US Supreme Court has restricted the size of the sample available to test models of strategic retirement. Scholars have confronted this challenge by analyzing state supreme courts or lower courts (Nixon and Haskin Reference Nixon2000; Hall Reference Hall2001), but the political issues at stake—and thus the incentives to retire—may be quite different across levels of the judiciary. Other scholars have tested the theory with evidence from parliamentary systems (Maitra and Smyth Reference Maitra2005; Kerby and Banfield Reference Kerby2014; Massie et al. Reference Massie2014). These cross-national studies have found very limited evidence in support of political resignations, but the nomination process tends to be less politicized in most parliamentary regimes.

Besides those two problems, weak findings could also indicate that most judges retire for “sincere” reasons, including bad health and retirement benefits. As Stras (Reference Stras2006, 1444) notes, “Infirmity, both mental and physical, is the single greatest disadvantage associated with life tenure” (Squire Reference Squire1988; Atkinson Reference Atkinson1999; Garrow Reference Garrow2000; Zorn and Van Winkle Reference Zorn2000; Ward Reference Ward2003). Studies have shown that retirement provisions with generous benefits provide strong incentives to retire in the United States (Squire Reference Squire1988; Zorn and Van Winkle Reference Zorn2000; Ward Reference Ward2003; Yoon Reference Yoon2006) and elsewhere (Maitra and Smyth Reference Maitra2005).

The Logic of Strategic Retirements

Although the idea of strategic retirement is straightforward, few studies derive precise behavioral predictions from the theory. As a result, authors invoke disparate findings as confirmatory evidence. To address this problem, we formalize the argument and identify the fundamental empirical implication of the theory, plus a few additional corollaries.

Strategic retirement involves a decision in which a judge compares the utility from giving up her position today against the expected utility of a lottery played if she stays in office.Footnote ² The theory assumes (at least) two moments, which we can generally call present and future.Footnote ³

If the judge retires today, she relinquishes the benefits of office (call them s), in exchange for retirement benefits, r.Footnote ⁴ In addition, the judge anticipates that the partisan composition of the court will remain equal to P/N if the president is a co-partisan or it will become $\left(P-1\right)/N$ if the president is not a co-partisan, where P is the number of justices who belong to the judge’s party and N is the size of the court. For notation purposes, we use $c∊\left\{0,1\right\}$ to represent the president’s expected nomination, with $c=1$ indicating that the president would nominate an acceptable co-partisan replacement and $c=0$ ndicating that he would nominate a distant replacement.Footnote ⁵ Thus, a retiring judge expects to derive utility $u\left[\left(P-1+c\right)/N\right]$ from the new Court, where u is a monotonically increasingly but possibly nonlinear function.

If, on the other hand, the justice stays in office, there is always the possibility that aging, bad health, or other factors will compel her exit in the future. The judge anticipates the risk of involuntary departure with probability ρ. If the justice completes her second period, she receives payoff s plus whatever utility is derived from the existing court, $u\left[P/N\right]$ . If instead nature forces the justice to retire, the judge receives r and the president is allowed to nominate a replacement. Because the executive may have changed by the time the judge is compelled to retire, assume that nature will reset the location of the executive by selecting a president from the opposite party with probability ω. The judge thus anticipates a future court P/N with probability $\left(1-\rho \omega \right)$ and $\left(P-1\right)/N$ with probability ρω.

Figure 1 summarizes the structure of the argument. This sequence represents the core assumptions shared by all versions of strategic retirement theory.

Figure 1.

A model of strategic retirement

Given this setup, a judge retires when

(1) $\begin{array}{}\left(c-1+\rho \omega \right)[u\left(\frac{P}{N}\right)-u\left(\frac{P-1}{N}\right)]>\left(1-\rho \right)\left(s-r\right).\end{array}$

Retirement is rational when the advantage offered by a co-partisan president $c=1$ for the nomination of a successor is greater than the benefits of office (after considering retirement benefits as well as the risk of an involuntary exit in the future). Crucially, the advantage offered by a co-partisan president becomes more relevant as the risks of involuntary exit (ρ) and of party turnover (ω) loom larger in the near future.

Some corollaries of this model simply capture conventional wisdom. For example, justices are unlikely to retire when they are not co-partisans with the president (the left-hand side of eq. [1] becomes negative when $c=0$ ), unless considerable retirement benefits are offered. Similarly, judges will retire when they approach an old age (as ρ approximates one, the right-hand side of the equation approaches zero).

More important, however, equation (1) also provides the logical foundation for empirical implications overlooked in the literature. Prior studies have anticipated a greater rate of retirements among judges who have convergent views with the president, without specifying scope conditions for this hypothesis. The model shows that four factors should moderate the effect of partisan convergence on strategic retirements.

(1) If the judge and the president are co-partisans $c=1$ , the ideological benefits of strategic retirement, represented by $u\left(P/N\right)-u\left[(P-1)/N\right]$ , are conditional on the risk of involuntary exit, represented by ρ. Age and infirmity are not alternative explanations for retirement; they are part of the motivation for strategic behavior. Judges are unlikely to entertain their departure when they are young, but they are forced to consider the partisan consequences of their choices as they approach retirement age.Footnote ⁶ This is the fundamental empirical implication of strategic retirement theory:

The fundamental empirical implication is usually implied in the literature, but it is rarely formulated with precision and tested appropriately. The model also allows us to derive three additional empirical corollaries:

(2) A similar conditional effect is created by the likelihood of party turnover, represented by ω. A judge would have few incentives to strategize the timing of her retirement in an authoritarian system with no party turnover, because the same incumbents would nominate any replacement. Strategic retirements are driven by a comparison between current political circumstances and future political conditions that become uncertain only in contexts of effective party competition. Therefore:

(3) Incentives to retire depend on utility function u. Consider a judge with weak partisan ties, such that $u\to 0$ : her incentives for strategic retirement would be almost null. In contrast, judges involved in politics are more sensitive to partisan outcomes, such that $u>0$ . In line with this claim, Hagle (Reference Hagle1993) argued that justices with prior political experience are more inclined to time their retirement for ideological reasons.Footnote ⁷ It follows that:

(4) It is fair to assume that u is a monotone but nonlinear utility function with respect to (P/N).Footnote ⁸ If a judge retires from a nine-member body, leaving it with eight co-partisan judges would not present a dilemma, but leaving it with only four co-partisans would alter the balance of power in the Court. For example, in their study of the Canadian Supreme Court, Kerby and Banfield (Reference Kerby2014) noted that Liberal Party–appointed judges often retired when Conservatives were in power. The authors concluded that, given their dominant position, Liberal judges could retire “without dramatically changing the ideological balance of the bench” (350). Thus:

The four conditional hypotheses follow from the core model of strategic retirement, without introducing ad hoc assumptions about delays in nominations, the electoral calendar, or other factors. If judicial retirements happen in line with the theory, we should observe the predicted patterns of behavior under a wide range of circumstances. Failure to document such patterns would cast doubts on common assumptions underpinning all strategic retirement arguments.

Case Selection

We test the hypotheses using data for the United States and five Latin American countries. The analysis follows a most-different systems design (Przeworki and Teune Reference Przeworski1970; Berg-Schlosser and De Meur Reference Berg-Schlosser2009). Verification of similar results across disparate historical cases should strengthen the external validity of our findings.

Studies of strategic retirement have focused on the United States and a few industrial democracies with parliamentary systems, such as Australia, Canada, New Zealand, and the United Kingdom (Kerby and Banfield Reference Kerby2014; Massie et al. Reference Massie2014). We complement those studies by comparing the United States with Argentina, Brazil, Chile, Mexico, and Panama. This strategy allows us to focus on presidential systems where the executive plays a central role in the nomination of Supreme Court justices (a crucial fact for the theory), while it maximizes variance in mediating factors and potential confounders.Footnote ⁹

Three concerns affect the use of Latin American cases to assess strategic retirement theory. The most obvious one is the history of regime instability that plagued the region for most of the 20th century. We leverage this experience with authoritarian rule in favor of our empirical design because variance in party competition is necessary to test hypothesis 2. In authoritarian regimes without an expectation of party turnover, strategic retirement should not take place.

Table 1 shows that our cases differ considerably in their historical experience with democracy and thus in levels of party competition and judicial independence. Between 1900 and 2004, the United States was always coded as a full democracy by the Polity IV index, which ranges from −10 (institutionalized autocracy) to +10 (institutionalized democracy). Other countries in the sample, by contrast, varied considerably during this period. Similarly, the United States ranks above all other cases in Linzer and Staton’s (Reference Linzer2015) measure of judicial independence, available since 1948. To account for those differences, we estimate models for each country, pool all cases while controlling for levels of democracy, and test the hypotheses with a subset of democratic regimes. The results are consistently similar across all tests.

Table 1.

Country Comparison, 1900–2004

A second concern, related to the previous issue, refers to the problem of judicial purges. Several studies have noted that judicial tenures in Latin America are often uncertain, as political leaders force the resignation of hostile justices (Helmke Reference Helmke2005, Reference Helmke2016; Lara Borges, Castagnola, and Pérez-Liñán Reference Lara Borges2012; Pérez-Liñán and Castagnola Reference Pérez-Liñán2016). If judicial turnover in Latin America is essentially the result of unconstitutional mechanisms, high courts in the region may not provide useful information to test theories of voluntary retirement.

Fortunately, the proportion of forced departures in most Latin American cases selected for comparison is relatively small. Only 2% of all judicial exits in Chile (3 of 131), 5% in Brazil (5 of 108), and 6% in Mexico (13 of 204) were identified as forced removals—as opposed to voluntary retirements or deaths in office. In contrast, Panama, with 32% (8 of 25), and Argentina, with 58% (41 of 71), present a more dramatic pattern of judicial ousters. The overall proportion of removals for the Latin American sample is approximately 13%.

To account for this problem, we treat all involuntary retirements as right-censored observations; that is, we assume that judges would have stayed on the job if they had been allowed to. Retirements are coded as involuntary when justices were impeached, removed after a military coup, or forced to resign in the midst of a revolution or a regime transition. Observations are also treated as censored when justices died in office, if they reached a mandatory retirement age, and if they departed after completing a constitutionally fixed term in office. Those individuals contribute to the estimation of the hazard function, but their departures do not alter the likelihood of voluntary retirements.

It is possible that subtle political pressures have induced some exits that appear to be “voluntary” in the historical record (Castagnola Reference Castagnola2012). However, such exits should stand out in the empirical analysis, presenting a stark contrast with the expected pattern for voluntary retirements. Incentives for strategic retirement, as indicated by hypothesis 1, should prevail among judges aligned with the president who are close to retirement age. Pressures for departure, by contrast, should target judges who are not aligned with the president and who are far from retirement, in order to preempt their future influence in the Court. We will analyze such “hidden” involuntary retirements in our discussion of empirical findings below.

The third concern is the dearth of historical sources. Although most Latin American countries have few reliable records about past court members, we have selected five cases that have sufficient biographic data for our study. A list of primary sources is provided in the appendix. We conducted archival work with an international team of research assistants, limiting the analysis to justices in office between 1900 and 2004 because sources are scarcer for earlier periods. Our data set constitutes the most extensive biographic file available for justices in the Western Hemisphere.

Despite these challenges, our comparative strategy presents important advantages. The six countries have presidential constitutions that empower executives to influence high court appointments. Table 1 shows that, in all cases, judges need to consider the preferences of the executive when contemplating retirement. Although the specifics of the nomination process may differ slightly (e.g., the president nominates from a restrictive list offered by the Supreme Court in Chile), nomination procedures employed in these legal systems resemble the US case more than the procedures employed in other countries covered by the literature, such as Australia, Canada, New Zealand, or the United Kingdom.

Legal regulations in all six countries give judges the possibility of retiring voluntarily after many years of service. They grant judges life tenure or impose mandatory retirement only at an elderly age. The only case with fixed terms and no reappointment, Mexico since 1994, provides for very long terms in office (15 years). Most justices in our sample have been entitled to receive pension benefits, and conditions for retirement are similar across cases.

While these similarities in the nomination procedure and judicial terms are key to the scope conditions of the theory, differences across countries allow us to test important empirical implications of the model, verifying if predicted patterns of retirement take place under the conditions advanced by the theory.