Judicial–legislative dynamics in separation of powers

Arising out of a separation of powers framework, judicial independence is a fundamental consideration in determining, first, whether judges are free to decide cases according to their sincere beliefs or evaluations of the law – without fear of punishment or retribution. And second, judges must also have their decisions implemented by the governing regime. A wide array of scholarship gives focus to how prospective political competition during constitution-making might yield formalized protections and insulation for judicial institutions – that is, de jure independence (e.g., Ferejohn Reference Ferejohn1998; Ginsburg Reference Ginsburg2003; Melton and Ginsburg Reference Melton and Ginsburg2014).Footnote ⁷ But political competition can likewise engender de facto judicial independence, which allows judges to accumulate power informally to achieve compliance with their rulings, especially among the other branches of government (Linzer and Staton Reference Linzer and Staton2015).

Examining de facto judicial independence within a separation of powers context necessarily considers the intersectionality of judicial policymaking, legislative preferences, and public support. The alignment of preferences between a court and a legislature is a strong consideration in the level of constraint imposed on judicial actors. In the U.S. context, when the Supreme Court is ideologically disproximate from Congress, the justices are hesitant to engage in judicial review (e.g., Segal, Westerland, and Lindquist Reference Segal, Westerland and Lindquist2011), which has a direct impact on independence (i.e., a “court’s ability to make decisions unaffected by political pressure from outside the judiciary” (Clark Reference Clark2010, 5)). While pressure on courts can emanate from a variety of external actors who can affect the implementation of judicial policy (Hall Reference Hall2014), members of Congress have a broad capacity to threaten federal judicial institutions with court-curbing measures – even if such measures are never enacted. Indeed, court-curbing threats can impact the Supreme Court’s institutional support and therefore diminish its capacity to induce compliance among the American public. As Clark (Reference Clark2009) documents, a strategic court will exercise judicial review less frequently when a legislature engages in position-taking on court-curbing measures. Recent research bolsters this line of inquiry; Bartels and Johnston (Reference Bartels and Johnston2020) show that the American public is more willing to curb the Court’s independence when policy emanating from the bench does not align with an individual’s preferences.

Much of the comparative literature on de facto independence has centered on “insurance” theories; anticipating power transitions can give rise to greater power and autonomy for courts. Under those circumstances, legislators – hedging their bets on retaining office – will allow courts to increase independence, which then provides policy insurance in the face of the opposition gaining power. The empirical findings with regard to insurance hypotheses vary, with studies disparately suggesting that independent judges strategically defect when incumbent regimes are failing (Helmke Reference Helmke2002, Reference Helmke2004), impose higher costs on policy changes after party transition (Vanberg Reference Vanberg, Whittington, Kelemen and Caldeira2008), and allow competing parties to impose some mutual restraint (Stephenson Reference Stephenson2003).

Competition can also arise to a significant degree outside of insurance considerations. In the separation of powers game, judges face constraints in exercising de facto independence, which primarily hinges on a judge’s evaluation of the level of political competition within the system. For instance, Randazzo, Gibler, and Reid (Reference Randazzo, Gibler and Reid2016) utilize various electoral indicators of competition to better understand the impact on courts. Among democratic regimes, a party’s electoral performance is a strong signal for when legislators in government need “to develop judicial institutions as protection following a change in leadership” (586). As a more latent process, high levels of ethnic fractionalization lead to an increased likelihood of out-party challenges to the sitting government, which then suggests that courts would have more de facto independence. Courts – when exercising discretion in statutory interpretation – also consider the likelihood of legislative response (Cooter and Ginsburg Reference Cooter and Ginsburg1996). Strategic judges seek to avoid spurring new legislation that would counter the court’s interpretations. Therefore, “prudent judges pursue their own preferences to a point that stops short of provoking legislative repeal” (296). The risk of legislative backlash is increasing as there are fewer veto points to enact new policy. Additionally, legislators – in considering statutory revisions or court curbing measures – must consider public attention to court-curbing attempts (Vanberg Reference Vanberg2001). Judicial actors likewise consider the possibility of backlash based on (1) the expected duration of the current government and (2) the concentration of power (Leiras, Tuñón, and Giraudy Reference Leiras, Tuñón and Giraudy2015).

In particular, legislative effectiveness – as an indicator of competition in separation of powers systems – can have an impact on de facto independence. In Colombia, Rodrıguez-Raga (Reference Rodriguez-Raga, Helmke and Rios-Figueroa2011) provides evidence that courts are increasingly willing to overturn government action as the costs of legislative court-curbing and other anti-court actions increase. Furthermore, political fragmentation or divided government may produce an environment where courts are less likely to face backlash from other branches of government (Ríos-Figueroa Reference Ríos-Figueroa2007), and they would therefore be willing to diverge from the governing majority’s preferences. Research on the Russian constitutional court suggests that judges have incentives to avoid retaliatory attacks from other governing branches (Epstein, Knight, and Shvetsova Reference Epstein, Knight and Shvetsova2001).

These dynamics suggest that courts could find themselves in positions of constraint due to an oppositional legislature and dissatisfied public. Still, other circumstances offer courts opportunities for policymaking. For instance, Whittington (Reference Whittington2005) points to a variety of political factors that would allow or encourage judicial review, as it can be used to strike down statutes enacted by prior governing coalitions. Legislators who are unable – perhaps due to coalitional fragmentation – to repeal existing policies might instead rely on courts to achieve those goals. Among Whittington’s considerations are the division of political power within and across branches of government and preference cross-pressures among constituents or governing coalitions, as majority parties often suffer from preference heterogeneity. Under these circumstances, controversial policies can be shifted outside the legislative arena.

Our theory and empirical tests build on these overarching questions stemming from the literature we discuss above: when will judges have/take the opportunity to assert their independence? What sort of inter-institutional power dynamics would facilitate de facto court power independent of government transitions? The story we tell below offers a possible answer. If a dominant national party or coalition becomes ineffective or reaches political deadlock, then it will be unable to settle a public dispute. In this way, judicial independence may be necessary for effective governance.

A formal model of de facto judicial independence

Existing scholarly explanations of how political competition affects judicial independence focus on two important considerations: first, political agents’ strategic decisions to impair courts’ formal ability to produce policy impact; and second, how electoral vulnerability may affect that decision. We provide a theoretical perspective that examines both of these while also accounting for the impact of political competition within a legislature on demand for judicial policymaking.

We argue that de facto judicial independence is not only determined by legislators’ conscious decisions to create independent judicial institutions. Rather, legislative ineffectiveness develops an environment ripe for judicial action. In other words, as political competition increases, legislative productivity decreases, which leaves more space for judicial independence. We can imagine political competition taking several forms: (1) partisan fractionalization within a legislature and (2) high rates of turnover within a legislature. A third option – electoral volatility changing the composition of a legislature across elections – is more akin to traditional insurance theory. Comparisons of political competition and legislative productivity can occur across time and space.

To make this argument, we study a model of legislative policymaking with costly voting. The model has four players: three legislators (indexed by $ i\in \left\{\mathrm{1,2,3}\right\} $ ) and a court. Each legislator can be from faction $ A $ or faction $ B $ .Footnote ⁸ First, each legislator decides whether to vote for or against a given policy, which we call a reform to the status quo. Each legislator from faction $ A $ supports reform with probability $ \pi $ ; legislators from faction $ B $ always favor the status quo. Voting for the reform costs $ {\kappa}_i $ . If a majority votes for the reform, it passes and is implemented. If the reform does not pass, the court can decide to implement the reform on its own. After the vote, there is an election, and a member from faction $ A $ wins each seat $ i $ with some probability $ {q}_i $ .

We analyze two scenarios. First, passage of reform by faction $ A $ is permanent with no chance of repeal. Second, passage of reform by faction $ A $ is subject to possible repeal from faction $ B $ if faction $ B $ wins a legislative majority. In both scenarios, there is no possibility of reversing a court-created policy. We use these two scenarios to show conditions under which we expect the legislature to defer to the court instead of passing the reform itself.

Information and Preferences

Each member of faction $ A $ wants to pass the reform with some positive probability, $ \pi $ . Each legislator knows whether they favor reform but only knows the probability that each other legislator favors reform. If a legislator favors reform, their policy utility is as follows:Footnote ⁹

$$ {PU}_{i_R}=\left\{\begin{array}{ll}1& \mathrm{If}\ \mathrm{the}\ \mathrm{reform}\ \mathrm{passes}\ \mathrm{and}\;\mathrm{i}\;\mathrm{wins}\ \mathrm{reelection}\\ {}{u}_r& \mathrm{If}\ \mathrm{the}\ \mathrm{court}\ \mathrm{implements}\ \mathrm{reform}\ \mathrm{or}\ \mathrm{reform}\ \mathrm{passes}\ \mathrm{and}\;\mathrm{i}\;\mathrm{loses}\ \mathrm{reelection}\\ {}0& \mathrm{If}\ \mathrm{the}\mathrm{re}\ \mathrm{is}\;\mathrm{no}\;\mathrm{re}\mathrm{form}\end{array}\right. $$

Voting for reform costs $ \kappa $ regardless of whether it passes. Total utility for reform-favoring legislator is

$$ {U}_{i_R}={PU}_{i_R}-\unicode{x1D7D9}{\kappa}_i $$

where $ \unicode{x1D7D9}=1 $ is an indicator function that equals $ 1 $ if the legislator voted for reform, and $ 0 $ otherwise.

A legislator who favors the status quo has a similar utility function:

$$ {PU}_{i_S}=\left\{\begin{array}{ll}0& \mathrm{If}\ \mathrm{the}\ \mathrm{reform}\ \mathrm{passes}\\ {}{u}_s& \mathrm{If}\ \mathrm{the}\ \mathrm{court}\ \mathrm{implements}\ \mathrm{reform}\ \mathrm{or}\ \mathrm{reform}\ \mathrm{passes}\ \mathrm{and}\;\mathrm{i}\;\mathrm{loses}\ \mathrm{reelection}\\ {}1& \mathrm{If}\ \mathrm{the}\mathrm{re}\ \mathrm{is}\;\mathrm{no}\;\mathrm{re}\mathrm{form}\end{array}\right. $$

and

$$ {U}_{i_S}={PU}_{i_S}-\unicode{x1D7D9}\mathrm{l}{\kappa}_i $$

Finally, the court favors reform: $ {U}_C=\left(\begin{array}{cc}1\hskip0.24em & \mathrm{If}\ \mathrm{there}\ \mathrm{is}\ \mathrm{reform}\hskip0.48em \\ {}0\hskip0.24em & \mathrm{If}\ \mathrm{there}\ \mathrm{is}\;\mathrm{no}\;\mathrm{reform}\end{array}\right. $

Analysis

We use Subgame Perfect Nash Equilibrium in Pure Strategies as our solution concept (hereafter, just equilibrium).

Baseline

We begin with the baseline model where faction $ B $ cannot overturn a reform after the election. If the legislature does not pass the reform, the court will implement the reform, and all pro-reform politicians will receive a payoff of $ {u}_R $ . Therefore, a pro-reform politician can guarantee herself at least $ {u}_R $ by voting against reform. To vote for reform, she must receive an expected payoff at least as high as $ {u}_R $ .

However, voting for reform is risky; if at least one other legislator does not also vote for reform, it will fail and she will have paid the voting cost $ \kappa $ without the legislator achieving their sincere policy goals. Further, even if the reform passes, the legislator does not get the full benefit if she loses reelection, which in reality could be more likely with a recorded vote on a controversial issue.

If there are two faction $ A $ legislators, a pro-reform legislator will vote for reform if

$$ \pi \left({q}_i\cdot 1+\left(1-{q}_i\right){u}_R\right)-{\kappa}_i\ge {u}_R. $$

If there are three faction $ A $ legislators, a pro-reform legislator will vote for reform if

$$ \left(2\pi -{\pi}^2\right)\left({q}_i\cdot 1+\left(1-{q}_i\right){u}_R\right)-{\kappa}_i\ge {u}_R. $$

Both inequalities have the same general form. First, the probability that at least one other legislator favors reform is $ \pi $ in the two-legislator majority case and $ 2\pi -{\pi}^2 $ in the three-legislator majority case. Second, the probability of legislator $ i $ ’s reelection is $ {q}_i $ , multiplied by the payoff of passing reform and winning reelection, $ 1 $ . Third, the probability of not winning reelection, multiplied by the payoff of passing reform but not winning reelection is $ {u}_R $ . Fourth and finally, there is the cost of voting, $ {\kappa}_i $ and the payoff from court-implemented reform, $ {u}_R $ . The first proposition characterizes the comparative statics of passing reform:Footnote ¹⁰

Proposition 1. For a pro-reform legislator, the expected utility of voting for reform is

1. Increasing in the number of party $ A $ legislators.

2. Increasing in the legislator’s reelection probability.

The intuition for both comparative statics is simple. First, when faction $ A $ holds three seats instead of two, it is more likely that there will be at least two legislators who favor reform. This makes voting for reform less risky for a pro-reform legislator. Second, as the probability of winning reelection increases, the benefit of passing reform increases. This makes a pro-reform legislator more willing to bear the voting cost.

We refer to the level of partisan division of a legislature as fractionalization. A legislature with two faction $ A $ legislators and one faction $ B $ legislator has greater fractionalization than a legislature with three faction $ A $ legislators. The probability that a single legislator is reelected to a given seat is contestation: the lower the probability a legislator will achieve reelection, the greater the contestation. Proposition 1 states that greater fractionalization and greater contestation lead to lower expected utility from voting for the reform. As we detail in our empirical expectations section, a lower expected utility of voting for reform means that the legislature will be less likely to enact reform on its own and will instead be more likely to turn to the courts.

Repeal

One question that naturally arises relates to faction $ B $ ability to repeal the reform after the election. Would our comparative statics remain as in the baseline model? As the following extension makes clear, even allowing for faction $ B $ to return to the original status quo after an election does not alter the implications of the model.

We now allow the possibility of faction $ B $ – after taking a majority in an intervening election – to repeal a reform passed by faction $ A $ before the election. If anti-reform faction $ B $ legislators repeal a reform, pro-reform legislators receive a payoff of $ 0 $ . The expected utility of voting for a reform now takes into account the possibility of receiving $ 0 $ even if the reform passes before the election:

If there are two faction $ A $ legislators, a pro-reform legislator will vote for reform if

$$ \pi \left[{q}_i\left({q}_j+{q}_k-{q}_j{q}_k\right)\cdot 1+\left(1-{q}_i\right)\left({q}_j{q}_k\right){u}_R\right]-{\kappa}_i\ge {u}_R. $$

If there are three faction $ A $ legislators, a pro-reform legislator will vote for reform if

$$ \left(2\pi -{\pi}^2\right)\left[{q}_i\left({q}_j+{q}_k-{q}_j{q}_k\right)\cdot 1+\left(1-{q}_i\right)\left({q}_j{q}_k\right){u}_R\right]-{\kappa}_i\ge {u}_R $$

where $ {q}_j $ and $ {q}_k $ are the probabilities faction $ A $ wins seats $ j $ and $ k $ in the election.

We refer to the probability of a change in which faction is in the majority as volatility. As volatility increases, reform is more likely to be repealed in the future. This lowers the expected utility of passing the reform today. We would then expect that as volatility increases, legislators would be more likely to let the court implement reform instead of the legislators bearing the cost of voting for reform themselves. This intuition aligns with existing insurance theory; when majorities anticipate transitions in government, those legislators may formally empower or informally tolerate greater judicial autonomy.

A comparison of the model with turnover to the baseline model highlights two important facts. First, fractionalization and contestation are not dependent on volatility – that is, government turnover. They exist both in the first model without the possibility of repeal and in the second model when reform can be repealed. Second, the possibility of repealing the reform does not change the direction of the fractionalization and contestation comparative statics. They point in the same direction regardless of whether policy can be reversed. Therefore, whether we find evidence for insurance theory (by way of volatility) in our empirical exercise should have no bearing on whether we find evidence for fractionalization and contestation on the likelihood of judicial independence.

Expectations for empirical models of judicial independence

Our models above account for political competition generally by allowing for variations in the majority party’s ability to implement its policy. Although research addresses the multitude of ways in which competition can influence a legislature’s policy-making effectiveness, we address three. First, fractionalization is the partisan division within the legislature. In our model, we operationalize fractionalization by changing the size of faction $ A $ ’s legislative majority.

As fractionalization increases, factions need a greater percentage of their members to vote for legislation. Because not every member of a party or faction may support given legislation, needing a greater percentage of a faction makes passing legislation more difficult. In our model, that results in fewer pro-reform legislators actually voting for reform. Therefore, we expect that as legislative fractionalization increases, de facto judicial independence will likewise increase.

Second, legislative contestation conceptualizes a different aspect of political competition – that is, competition occurring through elections, similar to the approach taken by Randazzo, Gibler, and Reid (Reference Randazzo, Gibler and Reid2016). More specifically, contestation conceptualizes seat turnover through elections. This manifests in two ways. First, losing a seat makes the odds of a party maintaining its majority smaller. In turn, the odds of a party maintaining a governing majority decrease, which makes it harder for it to pass its preferred policies. Second, it means the losing candidate cannot take advantage of having passed legislation through fundraising, affecting its implementation, or showing constituents how much she has accomplished.Footnote ¹¹ Both of these effects are captured in our parameter $ {q}_i $ , the reelection probability of a given legislator. A higher probability of reelection translates to a higher probability of voting for legislation and therefore a lower probability of judicial independence:

Finally, we turn to our third measure of political competition – legislative volatility – which captures the degree to which a particular election changes the partisan control of a legislative chamber. Volatility is high when a large majority party loses its control of a chamber, and the incoming majority is likewise large. France’s 2017 elections are a prime example of volatility, as the post-election majority party – the brand-new En Marche! party, which had recently won the French presidency – won 350 seats anew. The pre-election majority party – the French Socialists – lost 286 seats. In the American context, the Tea Party takeover of the U.S. House in 2010 is similar evidence of volatility, as the Republicans took over 63 seats – easily becoming the majority party.

On its face, volatility speaks directly to prior findings regarding insurance theory; a party anticipating losing control of the government would allocate additional power and autonomy to the courts in order to produce policy stability. Indeed, this is what the extension to our baseline model predicts. Still, it is unclear whether volatility should correspond to increased reliance on the courts. A new, incoming majority party is likely to rely on its electorally derived mandate to implement its platform. Our model ends after the election and does not take future policymaking into account. However, if voters have switched their support from the prior governing party to the new majority, that would suggest dissatisfaction with past policies and the prior government (e.g., Helmke Reference Helmke2004) and predict policy overturns.

However, a party that had previously been out of government might need to expand its coalition by moderating its platform. That, in turn, would lead to more heterogeneous policy goals within the new majority party. Given these countervailing considerations, our initial expectations are in line with the existing literature regarding insurance theory, which suggests that the outgoing party anticipated its ouster and therefore allocated greater autonomy and power to courts – whether by informal or formal mechanisms. Although the incoming majority party will rely on its new-found governing status to pass its preferred policies, we would expect the courts to lag in terms of their ideological alignment with the new government. Therefore, we expect that as legislative volatility increases, de facto judicial independence will likewise increase.

Measuring independence and competition

In order to measure political competition and judicial independence, we rely on several different data sources. That being said, it is imperative to make note of how our data are constrained to democratic countries with legislative institutions that have real competition and turnover. We identify democratic states using Polity IV data. We include any state achieving a score of 6 or higher in the polity rating. Additionally, to ensure the effects of regime transitions are not captured by the analysis, we include only democracies that have survived for at least five years.

Measures of judicial independence range widely in how they are constructed. This is of particular importance given the inherent intangibility and difficulty of measuring an idea – such as de facto judicial independence – without a clear and universal definition. The data created by Linzer and Staton (Reference Linzer and Staton2015) combine multiple measures into one continuous proportion measuring de facto independence, with 0 being not independent and 1 being fully independent.Footnote ¹²

They model the measurement of independence for 200 countries from 1960 to 2010 and allow for estimation and a measure of uncertainty.Footnote ¹³ Because the model estimation includes many of the different measurement techniques discussed above, the final product is relatively flexible and includes a great deal of information. As such, this is the most appropriate data to test the theory of this paper.Footnote ¹⁴

We use three different measures to assess the degree of political competition. This helps alleviate some of the potential measurement errors by testing the theory using data generated through different processes. It also allows us to test our theory on a wider range of countries and over a longer period of time because data for each measure is not available for every country-year provided in the Linzer-Staton database.

The first measure of political competition, which we label Fraction (Beck et al. Reference Beck, Clarke, Groff, Keefer and Walsh2001),Footnote ¹⁵ captures fractionalization within a legislature. More specifically, it is the probability that two deputies picked at random from the legislature will be of different parties. Legislators who are of different parties could (1) both be in the governing coalition, (2) both be in opposition parties, or (3) have one in government with the other in the opposition. In this way, fractionalization is meant to capture how many different parties exist within a given legislature, along with the relative size of the parties. A fractionalized governing coalition may have difficulty enacting its platform given variations in the likelihood that some parties fulfill their promises (e.g., Martin and Vanberg Reference Martin and Vanberg2011). This measure speaks directly to the theory of this paper; when there is greater fractionalization, there is greater competition in the sense that one party will have greater difficulty controlling the legislative agenda. These data have 872 observations over 44 countries.

The second variable, Contest, is a principal component factor of many political contestation measures. Teorell et al. (Reference Teorell, Charron, Dahlberg, Holmberg, Rothstein, Sundun and Svensson2013) develop this measure as an indicator of real political competition within a country. Higher values indicate more frequent turnovers through elections and greater power-sharing between parties.Footnote ¹⁶. These data include 574 observations over 37 countries.

The third variable is Volatility, which we utilize to flesh out the causal mechanisms at play (Powell and Tucker Reference Powell and Tucker2014).Footnote ¹⁷ Volatility captures the degree to which an election changed the party composition of a legislature. It is highest when a drastic power shift occurs – that is, when one party loses a very large majority at an election and is replaced by another party gaining a very large majority. It is smallest when no changes occur in party composition within the legislature. The data have 55 observations over 14 countries, typically with between 4 and 6 observations per country.

Unfortunately, the volatility data are only available for European countries in election years, so our empirical models are limited. If competitive parties create independent judiciaries to maintain policy preferences or to create stability in a power-sharing arrangement, then high levels of volatility will be associated with increases in judicial independence. If, however, changes in party control allow a governing majority to consolidate its power and enact its preferred policies, then we would find a null or negative effect on judicial independence. This would further imply that a party would have sufficient power to curb an interventionist judiciary.

Because not all of these measures are available for each country-year in the subsetted Linzer and Staton data, we use multiple imputation to alleviate some of the bias introduced by missingness. As the volatility index is missing a very high proportion of data, we do not impute this variable. However, we repeated the results shown below with case-wise deletion for the volatility variable, and the results are almost identical. Table 1 shows the number of missing observations, and Figure 1 depicts the data availability. We use the mice package (van Buuren and Groothuis-Oudshoorn Reference van Buuren and Groothuis-Oudshoorn2011) in the R software environment (R Core Team 2017) to create 10 data sets.

Table 1. Missingness

Figure 1. A world map showing where the data is available and which countries we analyze.

The first control we use is the Polity IV rating of democratic governance. It is conceivable that high levels of democracy cause both increased measures of judicial independence and an increased degree of political competition. Though all states we consider are democracies, the level of democracy – vis-à-vis democratic consolidation and democratic values – varies across observations. The second control is the length of democratic governance.Footnote ¹⁸ Again, it is conceivable that the length of democratic governance directly affects the level of judicial independence and the degree of political competition.

We also utilize a measurement for presidential systems versus parliamentary systems. The variable takes a two if the system is parliamentary, one if there is an assembly-elected president, and zero if it is a presidential system.Footnote ¹⁹ A presidential system may impact the amount of political competition by unifying strong parties. Furthermore, it may impact the amount of judicial independence by creating an additional check on courts.Footnote ²⁰

Finally, we use a dummy indicating the type of colonial history of a given country. Colonial rule undoubtedly has an impact on the nature of political institutions, and it is possible that a particular colonial history could impact both political competition and judicial independence.Footnote ²¹

Empirically modeling de facto judicial independence

To assess our hypotheses, we utilize time-series data across countries, as political competition and judicial independence can vary across both space and time. Therefore, we employ a multilevel model with varying intercepts for each country. We do not vary the effects of the controls and the independent variables of interest. Each model uses a different independent variable to capture political competition: fractionalization, contestation, and volatility. We logistically transform the outcome variable – the Linzer-Staton measure of de facto judicial independence – as it is a proportional measure.Footnote ²² The first model has Fraction as the variable of interest, the second has Contest, and the third has Volatility. Formally, the three models are specified:

$$ \hskip0.45em {\displaystyle \begin{array}{l}\mathrm{logit}\left({\mathrm{LJI}}_i\right)\sim \mathcal{N}\left({\alpha}_{j\left[i\right]}+{\beta}_0+{\beta}_1\times {\mathrm{interest}}_i+{\beta}_2\times {\mathrm{system}}_i+{\beta}_3\times {\mathrm{polity}}_i+{\beta}_4\times {\mathrm{years}}_i,{\sigma}_y^2\right),\\ {}\hskip9em \mathrm{for}\hskip0.35em i=1,\dots, n,\\ {}\hskip7em {\alpha}_j\sim \mathcal{N}\left({\gamma}_0+{\gamma}_1\times {\mathrm{colonial}}_j,{\sigma}_{\alpha}^2\right),\mathrm{for}\;j=1,\dots, J,\end{array}} $$

where our primary concern lies in the independent variable of interest for the given model, $ n $ is the number of observations across both time and space (936), and $ J $ is the number of countries (44).Footnote ²³ Because the data are not fine-grained with respect to time (i.e., we only have yearly data), we report the instantaneous estimated effects (i.e., no lags are in the explanatory or control variables). This allows us to maintain as much data as possible. However, with the awareness that the effects of these may be lagged, we include in the Online Appendix the results from models, identically specified, with a lag of one year, two years, and three years across all variables but the outcome. The results are very insensitive, and the reliability of the effects is maintained, across lagging schemes.

We model each of the ten data sets separately and pool the results. We accomplish this by averaging the point estimates and calculating the standard errors by taking the square root of the sum of the within variance and the between variance, scaled by $ \left(m+1\right)/m $ , with $ m $ equal to 10 (number of data sets).Footnote ²⁴ The degrees of freedom are estimated as outlined in Barnard and Rubin (Reference Barnard and Rubin1999). The prior distributions are not theoretically motivated and, as such, are conservative. Specifically, the prior distributions for the $ \beta $ ’s and $ \gamma $ ’s are normal distributions with mean 0 and standard deviation 3.

The $ \sigma $ ’s have inverse-gamma prior distributions with shape 2 and scale $ 0.5 $ . The results of all models are shown in Figure 2. All three panels in Figure 2 show the point estimates and 95% credible intervals, with intercepts suppressed.

Figure 2. Multilevel Models Predicting de facto Judicial Independence across Several Explanatory Variables

Note: All model estimates include 95% credible intervals.

Model $ (a) $ in the top-left panel of Figure 2 examines the Fractionalization Hypothesis discussed above. The point estimate for the Fraction variable is both positive and reliable, with 95% credible interval $ \left[.07,.56\right] $ and point estimate $ 0.31 $ . As such, we see that as partisan fractionalization within a legislature increases, de facto judicial independence also increases, suggesting support for our hypothesis. There is also evidence of more variability between countries than within countries, which validates the use of the multilevel model. Though a discussion of the controls is not necessary, it is interesting to note that a country with a colonial past has a negative impact on the intercept, and all but French colonial history estimates are reliable.

In model $ (b) $ in the top-right panel of Figure 2, the variable of interest is Contest, which is both positive and reliable at the 95% level $ \left[.07,.47\right] $ . This is highly suggestive of support for the Contestation Hypothesis above, which states that increases in electoral contestation correspond to increased political competition and, therefore, increased de facto judicial independence. Again, variability is larger between countries than within countries but is slightly lower than in the previous model. There are no noteworthy changes to the controls.

Model $ (c) $ in the bottom panel of Figure 2 includes the volatility measure for our explanatory variable of interest. The estimate for Volatility is very close to zero and slightly negative. The 95% credible interval tightly bounds zero. This indicates that we cannot reject the null hypothesis that judicial independence is not related to volatility. In this case, we have no evidence to support insurance theory. As judicial independence is positively related to competition in our other models, it is somewhat surprising that there is no evidence that it is related to volatility. As we discussed above, although volatility is sometimes linked to particular notions of political competition, it is unlikely that a new majority party has a deficit of political capital. As such, a new majority party likely has the capacity to constrain a court through prospective court curbing. This is in line with the premise that competition is related to independence through its effect to constrain a legislature.

Model implications and limitations

The motivating examples we employ at the start of this paper are worth revisiting in light of our empirical results. The models above reveal relatively broad support for our theory and the corresponding hypotheses across both time and space. As we saw with the controversies surrounding the Defense of Marriage Act (DOMA) in the United States and Pres. Hollande’s “supertax” in France, our empirical results show that higher levels of legislative political competition tend to lead to increases in de facto judicial independence. In the U.S., the Democratic majority was sufficiently heterogeneous to prevent legislative action on DOMA, which activists had ardently demanded. In France, the Sénat was starkly fractionalized between the majority Socialist party, which had won 177 seats in the 2011 elections, and the minority coalition, which had 171 seats. The 60-senator referral of the tax statute to the Conseil constitutionnel, along with the broader partisan dynamics, invited judicial intervention. And Pres. Hollande’s 75-percent tax would ultimately fail by 2015.

In this way, partisan fractionalization and contestation within a legislature speak directly to a majority coalition’s ability to achieve its preferred policies. In our models above, fractionalization and contestation lead to the expected changes in judicial independence. As a governing coalition finds it more difficult to pass legislation itself, the courts are able to step in and make policy independently. As such, we see that courts have greater room to intervene (a) given a more heterogeneous legislature (fractionalization) or (b) through the possibility of a legislator losing their seat (contestation).

Importantly, we do not find evidence that volatility leads to the same changes in judicial independence, and we discussed our reasoning as to why we were uncertain regarding that relationship. Chiefly, large swings in legislative majorities might imply strong public support for an incoming government, and, in turn, an increased capacity for the new legislature to implement its policy platform. Whether or not our null finding in our volatility model poses a direct challenge to insurance theory is an open question and one that we do not address here. However, as our formal model shows, the effects of fractionalization and contestation can occur with or without the effects of volatility.

In short, the mechanisms of political competition that yield higher levels of fractionalization, contestation, and volatility do not necessarily all lead to the same conclusion with regard to court independence. Volatility, in particular, can still allow for a majority party to engage in court curbing, which would decrease perceptions of de facto independence. The difference between volatility – on the one hand – and fractionalization or contestation – on the other hand – shows that competition by itself is too broad a concept for predicting judicial independence. Our model and empirical results show how subtle differences in types of competition can differentially affect whether judiciaries act independently.

Conclusions and future work

Scholars have extensively studied the important relationship between political competition and judicial independence, providing broad empirical support for the proposition that the two are related. While the main thrust of the literature has been to explain the supply-side effect – answering the question of why politicians would provide courts with institutional protections – our theoretical model uncovers and explains important dynamics largely unexplored by the literature so far. Our empirical tests are largely supportive of these conclusions.

In particular, much of the current literature argues that legislators will create an independent judiciary when faced with the probability of losing power as a means to ensure stability in their preferred policies. We, however, argue it may not only be insurance in the face of possible power loss but also a result of a politically constrained legislature. As the legislature becomes less able or willing to deliver policy, the figurative door is left open for the judiciary to exercise independence and provide reform.

Why would the legislature be less willing or less able to deliver policy? Our model gives multiple reasons separate from insurance theory. First, the possibility of losing an individual seat means a legislator will not enjoy the benefits of passing legislation, which will make her less likely to support costly legislation. Second, when a legislator is uncertain over the preferences of her co-partisans, she is less likely to support new policy. Co-partisans from districts that do not support reform may not be willing to bear the cost of voting for reform. This means that even if some faction member votes for reform, it still may not pass. As such, voting for reform is a risky endeavor even when a member supports the policy. Further, these two dynamics interact, making a legislator less willing to vote for reform when her party holds fewer seats.

To test our hypotheses, we estimate three multilevel models across time and countries – each model with a different independent variable of interest. The results strongly suggest that legislative political competition is linked to judicial independence. There is also weak evidence that the driving mechanism is not turnover as would be suggested by insurance theory, but instead fractionalization and competition. High levels of volatility were actually (insignificantly) negatively related to judicial independence, and there is suggestive, but very weak, evidence that there might be an inverted-U relationship between volatility and independence. To further test this, future work should gather more data on volatility to cover a wider range of countries over longer periods of time. Another potentially fruitful avenue would be to gather data on legislative productivity. If, within a country, legislative productivity is negatively related to judicial independence, there would be further evidence that the driving mechanism is a constrained legislature. Furthermore, prior literature suggests that indicators of legislative effectiveness may be interrelated – chiefly, that there are interactive effects between fragmentation and turnover (Leiras, Tuñón, and Giraudy Reference Leiras, Tuñón and Giraudy2015). We believe future studies of de facto independence should examine the conditional relationships between measures of effectiveness and their impact on court power.

Our model provides more predictions than what we empirically test in the current article. In particular, it predicts that as political parties themselves become more diverse – and therefore have more heterogeneous preferences – passing legislation will be more difficult. Therefore, they should rely more often on the courts. We do not directly test this prediction, and we believe it is a fruitful area for future research. Other related empirical tests in line with our theoretical explanation that we do not undertake include (1) analyzing actual court decisions and their relationship to legislative policies and (2) considering the role of (a) party discipline and (b) intra-party preference heterogeneity as dimensions our concept of fractionalization.

If the link between political competition and judicial independence is a politically constrained legislature, we note potential alternative explanations. Specifically, this paper argues that de facto judicial independence may not be a conscious decision made by legislators. Nothing in this paper tests for this, and deriving a test would be challenging. Callander and Krehbiel (Reference Callander and Krehbiel2014) argue that a gridlocked legislature will delegate to other actors to adapt to changing environments. In this context, the courts could be viewed as this actor, suggesting the legislature is consciously granting or allowing the courts to be more independent. Perhaps the best way to test this in future work would be to look at de jure judicial independence. If the legislature is consciously promoting an independent judiciary, legislation should be passed to this end. Studying the timing of independence-promoting legislation might shed some light on the mindset of the legislators. Although this paper strongly challenges some of the previously argued causal mechanisms linking judicial independence to political competition, there is still much work to be done to better understand the latent processes leading to de facto court power.