Jump to ContentJump to Main Navigation
A Mere MachineThe Supreme Court, Congress, and American Democracy$

Anna Harvey

Print publication date: 2013

Print ISBN-13: 9780300171112

Published to Yale Scholarship Online: January 2014

DOI: 10.12987/yale/9780300171112.001.0001

Show Summary Details
Page of

PRINTED FROM YALE SCHOLARSHIP ONLINE (www.yale.universitypressscholarship.com). (c) Copyright Yale University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in YSO for personal use (for details see http://www.yale.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 23 February 2018

Explaining The Puzzle of the Two Rehnquist Courts

Explaining The Puzzle of the Two Rehnquist Courts

Chapter:
(p.141) 5 Explaining The Puzzle of the Two Rehnquist Courts
Source:
A Mere Machine
Author(s):

Anna Harvey

Publisher:
Yale University Press
DOI:10.12987/yale/9780300171112.003.0005

Abstract and Keywords

Chapter 3 looked at econometric analysis, Chapter 4 provided narrative accounts, and together they presented different pictures of the Rehnquist Court. This chapter looks at the picture from a wider perspective and examines the data that has been used to study the question of judicial independence. The estimates reported in this chapter suggest that, at least in cases involving constitutional challenges to federal statutes, the justices of the Supreme Court are surprisingly responsive to the preferences of majorities in the House of Representatives.

Keywords:   econometric analysis, narrative accounts, judicial independence, Rehnquist Courts, Supreme Court

The econometric analyses of Chapter Three and the narrative accounts of Chapter Four present two entirely different pictures of the Rehnquist Court. The econometric analyses depict a Court that was consistently and stably conservative between 1986 and 2004, with the predicted probability that the Court would issue a conservative judgment ranging in a narrow band between .62 and .67. This probability remains unchanged by the 1994 elections; in these analyses the Court was as conservative in its pre-1994 terms, when its average predicted probability of a conservative judgment was .64, as it was in its post-1994 terms, when this probability averaged .63.1

But journalists and legal academics portrayed a very different Rehnquist Court (or Courts). Their accounts of the Rehnquist Court's constitutional jurisprudence depict a relatively moderate, even liberal, Court that made a “revolutionary” turn to the right in the 1994 term. In these narrative accounts, the “second” Rehnquist Court then remained significantly more conservative than its first incarnation through the remainder of its terms.

The fact that these close observers of the Court's jurisprudence saw such a different Rehnquist Court should perhaps be a red flag for those who engage in more quantitative analyses. Perhaps we need to think a bit harder about the data we have been using to study the question of judicial independence. (p.142)

Our measures of judicial and elected branch preferences are estimated from hundreds if not thousands of votes cast by legislators and justices; these measures represent the current state of the art in revealed preference estimation.2 We have seen very similar results using two different versions of these measures. But our measure of the direction of the Court's judgments is a different animal altogether. This measure, as we saw in Chapter Three, is reported in the United States Supreme Court Database (Supreme Court Database), financed by the National Science Foundation and publicly available from the Center for Empirical Research in the Law at Washington University in St. Louis. The Supreme Court Database, often referred to as the “Spaeth” database after its originator and coder, Michigan State University Research Professor of Law Harold Spaeth, has been described as the gold standard for data on the Supreme Court by leading practitioners: “Spaeth's products meet all the aspects of the replication standard … The Spaeth databases are so dominating in our discipline that it would certainly be unusual for a refereed journal to publish a manuscript whose data derived from an alternate source. Even in the law reviews, virtually no empirical study of the U.S. Supreme Court produced by political scientists fails to draw on them.”3

Most of the data reported in the Supreme Court Database comprise relatively objectively coded facts about each case docketed by the Court, such as whether a justice was in the majority or the dissent, whether a justice in the majority joined the majority opinion or merely concurred in the judgment, and so on. But some of the information reported in that database is entirely the product of subjective judgments made by the database's coder. The directional measure of the Court's judgments used in Chapter Three, the measure also used in almost every study failing to find evidence of elected branch constraint on those judgments, is one of these subjectively coded measures.4

It is perhaps surprising that so much research on the Supreme Court has been based on a measure so dependent on a coder's discretion, with so little apparent concern for the measure's validity.5 This discretion prominently enters the Supreme Court Database coding process in the assignment of an “issue code” to each case, a code that will then determine the decision rule for coding the judgment in that case.6 The Supreme Court Database codebook lists approximately 260 choices for such (p.143) issue codes, which themselves are grouped into thirteen broader issue areas. There are no rules or even guidelines for this issue coding decision; the Supreme Court Database codebook acknowledges that “criteria for the identification of issues are hard to articulate.”7 Judgments are then coded as “liberal” or “conservative” according to a series of decision rules that are conditional on the prior assignment of an issue code. For example, in cases given antitrust issue codes, judgments that are “procompetition” are to be coded as liberal judgments. In cases given First Amendment issue codes, judgments that are “pro-civil liberties or civil rights claimant[s]” are to be assigned liberal judgment codes.

The troubling aspect of these coding protocols is that different issue codes may lead to different judgment codes for the same case. Since the coder has considerable discretion in the assignment of issue codes, we might worry that the coder's expectations about how judgments “should” be coded affect his issue coding choices, such that these choices lead to confirmatory judgment codes. In particular, a coder may be improperly influenced by the ex ante expectation that independent liberal courts will generally issue liberal judgments, and that independent conservative courts will generally do the reverse.8 The Supreme Court Database judgment codes might then reflect not the actual nature of the Court's decisions, but rather the coder's self-fulfilling expectations about those decisions.

It turns out that, looking at a large number of generally comparable cases, issue and judgment code assignments in the Supreme Court Database are indeed systematically conditional on the known preferences of the deciding court. Otherwise identical dispositions are assigned one set of issue codes under liberal courts, and another set under conservative courts. These issue codes then lead to judgment codes confirmatory of the expectations one might have about the kinds of judgments typically issued by such courts, under the assumption of judicial independence.9

The apparent confirmation bias in the Supreme Court Database has obvious implications for empirical analyses of elected branch constraint on the Court's decisions. Subjectively assigned Supreme Court Database issue and judgment codes confirm expectations about the kinds of judgments typically issued by more liberal and more conservative courts, under the hypothesis of judicial independence from the elected branches. (p.144)

They therefore cannot be used to test the validity of that hypothesis itself. To test the hypothesis of the Supreme Court's independence from the elected branches, we need to use objectively coded judgment measures.

Measuring Judgments Objectively

It may not be possible to come up with a strategy for objectively coding all of the Court's judgments. However, we can devise a number of such strategies at least for the subset of the Court's cases involving constitutional challenges to federal statutes. The key to all these strategies is the leveraging of additional information about the federal statutory provisions at issue in these cases. This additional information allows us to make inferences about the direction and impact of the Court's rulings on the challenged provisions.

The most precise information we have about these statutes is derived from the congressional roll call votes over their final passage, if such roll call votes exist. We can use information from these final passage roll call votes to make inferences about the policy direction of the Court's judgments. To do so we turn to the analyses of congressional roll call data reported by political scientists Keith Poole and Howard Rosenthal.10 Poole and Rosenthal assume a simple model of congressional voting: a bill, located at some point on a left-right continuum of public policy alternatives, proposes a change in some existing status quo policy, located at some other point on that same continuum. Members of Congress each have some most preferred point on that continuum that they would like to see enacted into public policy. Those members whose most preferred public policies lie closer to the proposed bill, than to the status quo, vote for the bill. Those members whose most preferred public policies lie closer to the status quo, than to the proposed bill, vote against the bill.

Along with estimating members' most preferred policies from their roll call votes, Poole and Rosenthal also estimate the cutpoint for each nonunanimous roll call vote, or the estimated point that separates those who voted for and those who voted against a bill, as a function of their policy preferences. These estimated cutpoints are preferable to the actual separation points we observe between those who voted for and those who voted against proposed bills, because the estimated cutpoints reduce the random (p.145) noise there may be in these votes, allowing us to focus on the effect of members' policy preferences. The cutpoints are estimated with a very low degree of error; they predict actual separation points quite well.11

The estimated cutpoints tell us whether a bill moved policy in a liberal or a conservative direction. Essentially, if the estimated most preferred policies of those voting for a bill lie to the left of its estimated cutpoint, then the bill must have been located to the left of the status quo ex ante: those more liberal members whose most preferred policies lay closer to the bill voted “yea,” while those more conservative members whose most preferred policies lay closer to the status quo voted “nay.” Symmetrically, if the estimated most preferred policies of those voting for a bill lie to the right of its estimated cutpoint, the bill must have been located to the right of the status quo ex ante.12 When there is a nonunanimous final passage roll call vote on a federal statute, then, we can identify with relative precision whether the statute in question moved policy in a liberal or a conservative direction.

The Court's rulings on constitutional challenges to federal statutes then preserve or reverse these statutes' liberal or conservative movements in the original status quo ex ante. One way to measure the Court's judgments objectively, then, at least for this subset of cases, is to code as liberal those judgments that preserve or produce liberal movements in the original status quo ex ante, and to code as conservative those judgments that preserve or produce conservative movements in the original status quo ex ante.

The sample of constitutional challenges to federal statutes between the 1953 and 2004 terms already compiled in Chapter Three can be used to construct this roll call–based judgment measure.13 For each case in this sample the enactment date of the statute or statutes at issue was first identified.14 The final passage roll call vote for each statute in the House of Representatives, if it existed, was then located, and the Poole and Rosenthal roll call estimates for that vote, again if they existed, were retrieved.15 Statutes were then coded as moving the status quo ex ante in either a liberal or a conservative direction.

The Court's dispositions of the constitutional challenges to these statutes then either preserved the direction of movement of the original status quo (by upholding a statute), or reversed the direction of that movement (p.146)

(by striking a statute, causing policy to revert to the original status quo ex ante). This information was used to construct a dichotomous measure of the direction of the Court's judgments in these cases between the 1953 and 2004 terms.16

This roll call–based judgment measure has several obvious disadvantages. It captures neither the magnitude of the movement in the status quo ex ante preserved or produced by a judgment, nor the location of that movement on the policy continuum. It does not incorporate any doctrinal nuances of the Court's opinions, as distinct from its judgments. On the other hand, a judgment measure based on the roll call votes over the statutes challenged in these cases has obvious advantages as well. Most important, it is objectively coded across cases, unlike the Supreme Court Database judgment measure. It measures “liberal” and “conservative” as understood by political actors themselves, most particularly the members of Congress who voted on the challenged statutes. It is specific to the statutory provisions challenged in a given case. And although it measures neither the magnitude of the movements in the status quo ex ante preserved or produced by the Court's judgments, nor the locations of those movements, neither does the Supreme Court Database judgment measure.

Still, the roll call–based judgment measure is available for only slightly more than half of the population of cases involving constitutional challenges to federal statutes between 1953 and 2004, because many statutes are passed via voice votes rather than roll call votes. We may then have concerns about whether analyses using this measure will generalize beyond this limited sample. One strategy to increase the number of judgments in our sample is to leverage Congress-specific rather than statute-specific information about the statutes challenged before the Court.

For example, many of the statutes reviewed between the 1953 and 2004 terms, for which final passage roll call votes are not available, were passed during unified Democratic or Republican control of both houses of Congress.17 We know that the partisan composition of the enacting Congress is highly correlated with the direction in which bills move the status quo: unified Democratic Congresses are more likely to pass bills moving the status quo in a liberal direction, while unified Republican Congresses are more likely to pass bills moving the status quo in a conservative (p.147) direction.18 We can thus use the partisan composition of the enacting Congress to construct a second objective judgment measure for those cases involving constitutional challenges to statutes enacted by unified partisan Congresses. These statutes are coded as liberal if passed by unified Democratic Congresses, and conservative if passed by unified Republican Congresses. If the Court then upholds one of these statutes, the statute's coding is preserved as the Court's judgment code; if the Court strikes a statute, the statute's coding is reversed as the Court's judgment code. This partisan Congress-based judgment measure is available for approximately 86 percent of the population of cases involving constitutional challenges to federal statutes between 1953 and 2004. And despite its apparent crudeness, the Congress-based judgment measure is highly correlated with the more sophisticated roll call–based judgment measure: the two measures agree in 74 percent of the cases for which they are both available.19

Estimating Judgment Probabilities

We can now estimate the effects of both judicial and elected branch preferences on these objective measures of the Court's judgments. We essentially want to replicate the analyses performed in Chapter Three, substituting the objective judgment measures for the subjective Supreme Court Database measure used there. Thus we will estimate the effects of the locations of the estimated most preferred rules of the median justice, the median representative, the median senator, and the president on the probability that the Court issues a conservative judgment, using the two samples of cases for which the roll call–based and the Congress-based judgment measures are available. We will also employ the same controls used in Chapter Three: the direction of the lower court judgment, whether the United States was a party to a case arguing for a liberal outcome, and whether the United States was a party to a case arguing for a conservative outcome. In Chapter Three these control variables were constructed relative to the conservative outcome as defined by the Supreme Court Database judgment codes; here they are constructed relative to the conservative outcome as defined by the objectively constructed judgment measures.

(p.148) Tables 5.1 and 5.2 report the results of analyses of the roll call-based judgment measure, for both the Bailey XTI and the Judicial Common Space preference estimates. Consistent with the results reported in Chapter Three, we see large effects on the Court's judgments from changes in the most preferred rule of the median justice. Using the Bailey XTI preference estimates, increasing the preferences of the median justice from their most liberal value (observed during the 1962 and 1963 terms) to their most conservative value (observed during the 1991 term), while holding other variables at their sample values, increases the average predicted probability of a conservative judgment from .05. to .59.20 Using the JCS preference estimates, moving from the most liberal median justice to the most conservative (or from the 1968 to the 1988 term), under the same conditions, increases the average predicted probability of a conservative judgment in these cases from .17 to .53.21 These predicted probabilities are distinct at conventional thresholds.

Also as we saw in Chapter Three, using the Bailey XTI estimates we observe perverse effects from changes in the preferences of the median senator

Table 5.1: Estimating the Probability of a Conservative Judgment: Bailey XTI Judicial and Elected Branch Preferences, 1953–2004, Roll Call–Based Judgment Measure

Coefficient

SE

95% CI

Conservatism of SCOTUS Median

1.71

.51

(.70, 2.71)

Conservatism of House Median

4.93

1.46

(2.07, 7.79)

Conservatism of Senate Median

−3.08

.88

(−4.80, −1.36)

Conservatism of President

−.37

.14

(−.64, −.10)

Liberal Lower Court Ruling

.42

.24

(−.06, .89)

US Liberal Party

−.17

.37

(−.88, .55)

US Conservative Party

.94

.33

(.29, 1.59)

N = 166

Wald chi2 = 36.26

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at the .10 level (two-tailed).

(p.149)

Table 5.2: Estimating the Probability of a Conservative Judgment: Judicial Common Space Judicial and Elected Branch Preferences, 1953–2004, Roll Call–Based Judgment Measure

Coefficient

SE

95% CI

Conservatism of SCOTUS Median

2.18

.78

(.65, 3.70)

Conservatism of House Median

2.41

1.16

(.13, 4.70)

Conservatism of Senate Median

−.06

1.45

(−2.90, 2.79)

Conservatism of President

−.34

.28

(−.89, .21)

Liberal Lower Court Ruling

.36

.23

(−.09, .80)

US Liberal Party

−.16

.35

(−.86, .53)

US Conservative Party

.90

.33

(.25, 1.55)

N = 166

Wald chi2 = 40.60

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at the .10 level (two-tailed).

and the president; increases in the conservatism of these elected officials' most preferred rules are associated with decreases in the probability that the Court issues conservative judgments. However, these perverse effects disappear when we use the JCS preference estimates. As reported in Table 5.2, using the JCS preference estimates there are no estimated effects on the roll call–based measure of the Court's judgments from changes in the most preferred rules of the median senator or the president.

These results are perhaps no surprise. The surprise comes with the predicted effects of the preferences of the House median on the Court's judgments, using either the Bailey XTI or the JCS preference estimates. Using the estimates reported in Table 5.1, setting the most preferred rule of the House median at the most liberal observed value (in the 1974 term) produces an average predicted probability of a conservative judgment of .20. But setting that rule at the most conservative observed value (in the 1995 term) increases that probability to .62.22 Likewise, using the JCS-based estimates reported in Table 5.2, moving from the most liberal to the most conservative observed value of the median representative's preferences (or from the 1974 term to the 2004 term) increases (p.150) the average predicted probability of a conservative judgment from .30 to .58.23 Both sets of predicted probabilities are distinct at conventional thresholds.

Tables 5.3 and 5.4 report the results of analyses of the partisan Congress-based judgment variable; the control variables in these regressions are coded relative to this new dependent variable. Strikingly, in both analyses we now see no effects on the Court's judgments from changes in the preferences of the median justice. We also see no effects on those judgments from changes in the preferences of the median senator or the president. But we see large effects on those judgments from changes in the preferences of the median representative. Using the Bailey XTI estimates, setting the preferences of the median representative to their most liberal value returns an average predicted probability of a conservative judgment of .10, increasing to .53 when the preferences of the median representative are set to their most conservative value.24 Using the JCS preference estimates returns average predicted probabilities of .10 and .74, respectively.25 Both sets of predicted probabilities are again distinct at conventional thresholds.

Table 5.3: Estimating the Probability of a Conservative Judgment: Bailey XTI Judicial and Elected Branch Preferences, 1953–2004, Partisan Congress–Based Judgment Measure

Coefficient

SE

95% CI

Conservatism of SCOTUS Median

−.25

.31

(−.85, .36)

Conservatism of House Median

4.52

1.33

(1.90, 7.13)

Conservatism of Senate Median

.12

.71

(−1.26, 1.50)

Conservatism of President

−.18

.15

(−.48, .11)

Liberal Lower Court Ruling

.08

.20

(−.31, .47)

US Liberal Party

.21

.29

(−.35, .78)

US Conservative Party

.77

.33

(.11, 1.42)

N = 265

Wald chi2 = 46.35

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at the .10 level (two-tailed).

(p.151)

Table 5.4: Estimating the Probability of a Conservative Judgment: Judicial Common Space Judicial and Elected Branch Preferences, 1953–2004, Partisan Congress–Based Judgment Measure

Coefficient

SE

95% CI

Conservatism of SCOTUS Median

−.50

.47

(−1.43, .43)

Conservatism of House Median

5.08

1.01

(3.09, 7.07)

Conservatism of Senate Median

−.74

1.15

(−2.99, 1.51)

Conservatism of President

−.25

.20

(−.63, .14)

Liberal Lower Court Ruling

.02

.21

(−.39, .43)

US Liberal Party

.20

.28

(−.34, .74)

US Conservative Party

.61

.35

(−.07, 1.30)

N = 265

Wald chi2 = 65.10

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at the .10 level (two-tailed).

The effect of the median representative's preferences on the Court's judgments is perhaps easier to interpret by looking at each term between 1953 and 2004. To that end Figures 5.1 through 5.4 display predicted probabilities derived from the estimates reported in Table 5.3. These are the estimates obtained using the larger partisan Congress-based sample of cases and the Bailey XTI preference estimates. Because these estimates generate a smaller predicted effect for the preferences of the median representative, relative to those reported in Table 5.4, they offer perhaps a conservative interpretation of the magnitude of this effect.

Figure 5.1 reports two series of predicted probabilities for each term between 1953 and 2004. One series, the “Actual Constrained Court” series, reports the average predicted probability of a conservative judgment in each term, setting judicial and all elected branch preferences to the correct values for that term. Case-specific control variables retain their sample values for each observation. This series may be interpreted as our best guess for how the justices would have decided cases involving constitutional challenges to federal statutes in a given term, factoring in the effects of both judicial and elected branch preferences. (p.152)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.1. Predicted Judgment Probabilities, 1953–2004

Simulated from the estimates reported in Table 5.3. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term. Case-specific control variables remain at their sample values. The two series are distinct with at least 90 percent confidence in the 1954, 1956–1957, 1962–1971, and 1982–1993 terms.

The second series, labeled the “Counterfactual Unconstrained Court” series, also reports the average predicted probability of a conservative judgment in each term, setting the preferences of the median justice, the median senator, and the president to the correct values for that term, but counterfactually setting the preferences of the median representative to be identical to the preferences of the median justice in that term. Case-specific control variables again retain their sample values for each observation. This series removes any preference divergence between the Court and the House of Representatives. These predicted probabilities may consequently be interpreted as our best guess for how the justices would have decided cases involving constitutional challenges to federal (p.153) statutes in a given term, had the Court been completely unconstrained by the House of Representatives.

The differences between the “Counterfactual Unconstrained Court” and the “Actual Constrained Court” series in Figure 5.1 are then a measure of the magnitude of the effect of the preferences of House majorities on the Court's judgments. Perhaps the first thing to notice about this effect is that, as reported in the note to Figure 5.1, it is present at conventional levels of statistical significance in about half the terms between 1953 and 2004. That is, in about half the terms over this approximately 50-year period, the preferences of majorities on the Court and in the House were not sufficiently divergent for the latter to exert a detectable effect on the Court's judgments. These terms include most of the Burger Court terms during the 1970s, when a relatively liberal majority on the Court faced generally liberal majorities in the House, and all of the Rehnquist Court terms between 1994 and 2004, when a conservative majority on the Court faced conservative majorities in the House.

But in the remaining terms between 1953 and 2004, the preferences of House majorities were sufficiently divergent from those of the justices to exercise discernible effects on the Court's judgments. These terms include those of the later Warren and early Burger Courts, when more conservative House majorities appear to have increased the likelihood that the Court would issue conservative judgments, relative to the unconstrained likelihood, and those of the late Burger and early Rehnquist Courts, when more liberal House majorities appear to have decreased the likelihood that the Court would issue conservative judgments, relative to the unconstrained likelihood. These effects of House majorities on the Court's judgments are significant at conventional thresholds between the 1962 and 1971 terms, and again between the 1982 and 1993 terms.26

Recall that one objection to the hypothesis of elected branch constraint on the Court is that the preferences of the Court and the elected branches are unlikely to be sufficiently divergent for the elected branches to even want to cabin the Court's judgments. But as we saw in Figure 3.1, there in fact appears to have been considerable divergence between judicial and elected branch preferences between 1953 and 2004. And as we can now (p.154) see in Figure 5.1, this divergence was sufficiently large for the preferences of House majorities to exercise a detectable pull on the Court's judgments in about half of the Court's terms over this period.

But this pull was more consequential in some terms than in others. House majorities appear to have made less of a difference to the Court's judgments during the later Warren and early Burger Courts, relative to their impact during the late Burger and early Rehnquist Courts. During the former period, a very liberal majority on the Court faced somewhat less liberal Democratic majorities in the House of Representatives. But at least after the 1964 elections, the Court was still most likely to issue liberal judgments in cases involving constitutional challenges to federal statutes, even after factoring in the effects of slightly less liberal House majorities. That is, the conservative effect of House majorities during the later Warren and early Burger Courts appears to have been largely a moderating effect, moving the Court from being extremely unlikely to issue conservative judgments in these cases, to being simply unlikely to issue such judgments.

During the late Burger and early Rehnquist Courts, however, the effect of House majorities on the Court is sufficiently large so as to actually alter the character of the Court's jurisprudence. By the mid-1980s, as a consequence of an unbroken series of appointments made by Republican presidents, the majority on the Court had become sufficiently conservative that, at least in cases involving constitutional challenges to federal statutes, had this majority been free to decide cases at will, it would have been most likely to issue a conservative ruling in any given case. Left to their own devices, the majority of the justices' decisions would have been in a conservative direction, meaning that most of their rulings in these cases would have struck liberal and upheld conservative federal statutes.

But because the conservative majority on the Court was not free to decide cases at will, we in fact saw a very different Court during these terms. We saw a Court that was by far most likely to issue liberal decisions in cases involving constitutional challenges to federal statutes, and that consequently decided only a small proportion of these cases in a conservative direction. Indeed, Figure 5.1 may reveal the nature of conservatives' frustration with the late Burger and early Rehnquist Courts. As we saw (p.155) in Chapter Four, conservative judicial activists in the Reagan and George H. W. Bush administrations thought that they were engineering a conservative majority on the Court, a majority that would decide cases in a way perhaps well characterized by the “Counterfactual Unconstrained Court” predicted probabilities reported in Figure 5.1. But the House of Representatives remained dominated by liberal Democratic majorities during these administrations. And so conservative judicial activists instead got a Court that decided cases in a way perhaps more accurately characterized by the “Actual Constrained Court” predicted probabilities reported in Figure 5.1.

Overall, then, these estimates suggest not only that the Supreme Court is constrained by the preferences of majorities in the House of Representatives, the institution perhaps most prominently possessed of the power both to punish and reward the justices, but also that between 1953 and 2004 the degree of this constraint was at times sizable enough to effect dramatic changes in the nature of the Court's constitutional jurisprudence. Moreover, as we will see when we examine the Warren, Burger, and Rehnquist Courts in more detail, even during the terms when the magnitude of this constraint was relatively slight, the Court's judgments in cases involving constitutional challenges to federal statutes were still conditional on the presence of friendly House majorities. Had these majorities been less friendly, we could have seen very different decisions in these cases.

The Warren Court, 1953–1968

Although the Warren Court of the 1950s issued a handful of prominent liberal rulings, like the school desegregation decisions in Brown v. Board (1954) and Cooper v. Aaron (1958), it was not until the 1962 term that this Court had a clear majority of liberal justices. In this term Arthur Goldberg replaced Felix Frankfurter, giving the Warren Court five reliably liberal justices: Chief Justice Earl Warren, Justice William Douglas, Justice William Brennan, Justice Hugo Black, and Justice Goldberg. As we saw in Figure 3.1, in this term the preferences of the median justice shifted sharply to the left, and remained there throughout the remaining Warren Court terms.

(p.156) We are accustomed to thinking of this later Warren Court, the Court sitting between the 1962 and 1968 terms, as acting as “an independent and aggressive guarantor of constitutional rights,” in the words of writer Jeffrey Toobin.27 And it does at least appear to be true that the effect of House majorities on the later Warren Court was relatively small. We likely would have seen an even more liberal Warren Court in its later terms than the Court we actually saw, had House majorities had no effect on the Court. But even after factoring in the effects of slightly less liberal House majorities, we still saw a pretty liberal Warren Court between 1962 and 1968.

What if, however, the Warren Court justices had faced conservative Republican majorities in the House of Representatives? What if, for example, they had faced House majorities as conservative as the 104th Republican House majority, the majority elected in the 1994 elections? Figure 5.2 reports for the Warren Court terms these counterfactual predicted probabilities, along with the two series of predicted probabilities already reported in Figure 5.1. This new series of counterfactual probabilities, labeled the “Conservatively Constrained Court” series, simulates the average predicted probability of a conservative judgment in each term, setting the preferences of the median representative to their 1995 value. All other preference variables are set at their correct values for each term, while case-specific control variables again retain their sample values.

Now we see a Warren Court that, in each of its terms, is most likely to issue conservative judgments in cases involving constitutional challenges to federal statutes. Even during the later Warren Court terms between 1962 and 1968, the terms wherein the justices are predicted to have decided virtually none of these cases in a conservative direction, had they been unconstrained by House majorities, the Court is predicted to issue conservative judgments with probabilities ranging from .57 to .71. These predicted probabilities are distinct at conventional thresholds from the unconstrained probabilities in each of these later Warren Court terms.

Under the condition of conservative Republican majorities in the House of Representatives, the Warren Court looks very different from the Warren Court of legend. It decides most of its cases in a conservative direction, striking liberal federal statutes and upholding conservative federal (p.157)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.2. Predicted Judgment Probabilities, 1953–1968

Simulated from the estimates reported in Table 5.3. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term; the “Conservatively Constrained Court” series sets the preferences of the median representative to their 1995 value. Case-specific control variables remain at their sample values. The “Conservatively Constrained Court” series is distinct front the “Counterfactual Unconstrained Court” series with at least 90 percent confidence in the 1954, 1956–1957, and 1962–1968 terms.

statutes. While counterfactual predictions must always be taken with at least a few grains of salt, those reported here can nonetheless perhaps offer some insight into the conditional nature of the Warren Court's judgments. For example, in its 1964 and 1965 terms, the Warren Court upheld key provisions of the Civil Rights Act of 1964 and the Voting Rights Act of 1965. In Heart of Atlanta Motel v. U.S., 379 U.S. 241 (1964), the Court upheld provisions of the Civil Rights Act of 1964 prohibiting racial discrimination in hotels and motels serving interstate travelers, while in Katzenbach v. McClung, 379 U.S. 294 (1964), the Court upheld (p.158) provisions of the same Act prohibiting racial discrimination by restaurants serving food that had moved in interstate commerce. In South Carolina v. Katzenbach, 383 U.S. 301 (1966), the Court rejected a challenge to the Voting Rights Act of 1965 by the state of South Carolina, ruling that the Act was well within congressional power under Section 2 of the Fifteenth Amendment; in Katzenbach v. Morgan, 384 U.S. 641 (1966), the Court upheld that Act's ban on literacy tests as applied to former residents of Puerto Rico, ruling that the ban was within congressional power under Section 5 of the Fourteenth Amendment.

Both the Civil Rights Act of 1964 and the Voting Rights Act of 1965 had been enacted by unified Democratic majority Congresses; both statutes are also estimated to have moved the status quo in a liberal direction. The Court's rulings upholding these statutes then preserved policy movements in a liberal direction. Using the “Counterfactual Unconstrained Court” predicted probabilities reported in Figures 5.1 and 5.2, had the Warren Court justices been free to decide cases independently of the preferences of House majorities, the predicted probabilities that they would strike these liberal federal statutes would have been less than one percent.

In these terms the Court enjoyed large Democratic majorities in the House of Representatives. The preferences of the median representative during these terms are estimated to have been only slightly less liberal than those of the median justice. Factoring in the effect of these slightly less liberal House medians, the Court is predicted to have issued conservative judgments in its 1964 term with a probability of only .24, and in its 1965 term with a probability of only .20. That is, even after adjusting their decisions to take into account the presence of somewhat less liberal majorities in the House of Representatives, the justices were still very likely to have rejected the challenges to the Civil Rights Act of 1964 and the Voting Rights Act of 1965. However, had the Warren Court faced a House majority as conservative as the 104th House majority, the probabilities of conservative judgments in these cases would have increased to .70 in the 1964 term, and to .69 in the 1965 term. In this different political context, the Court would have been very likely to strike these statutes as unconstitutional.

Similarly, in its 1967 term the Warren Court upheld amendments to the Fair Labor Standards Act of 1938, extending that Act's coverage to (p.159) state-owned enterprises like schools and hospitals (Maryland v. Wirtz, 392 U.S. 183 [1968]). The amendments, enacted in 1966, had been challenged by 28 states on the grounds that they exceeded congressional power under the Interstate Commerce Clause. The 1966 amendments had been enacted by a unified Democratic Congress, and are estimated to have moved policy in a liberal direction. The Court's ruling upholding the amendments then preserved this liberal policy movement. Again, had the Court been completely unconstrained by the House of Representatives in the 1967 term, it would have struck liberal federal statutes like the one at issue in Maryland v. Wirtz with a predicted probability of less than one percent. Factoring in the effects of the somewhat less liberal Democratic House, which had moved in a more conservative direction after the 1966 midterm elections, raises this probability to .47. Nonetheless, the Court was still likely to uphold liberal federal statutes challenged during this term. But had the Court faced a House as conservative as the 104th House, the probability of a conservative judgment would have increased to .69, putting such statutes considerably more at risk.

While these counterfactuals should of course be interpreted with caution, they are nonetheless suggestive of how history might have turned out differently. Facing a conservative Republican majority in the House of Representatives, even the famously liberal justices of the later Warren Court are predicted to decide the vast majority of their cases in a conservative direction.

The conventional wisdom is that the later Warren Court terms exemplify the Supreme Court's independence from the elected branches. According to this widely shared belief, during these terms the Warren Court boldly defied the elected branches by protecting individual rights from majoritarian encroachment.28 Although revisionist historians have challenged the accuracy of this story, it still receives widespread deference.29 But this story is belied by the predicted probabilities reported in Figures 5.1 and 5.2. These figures suggest that the late Warren Court was permitted to be “an aggressive guarantor of constitutional rights” only by relatively liberal Democratic House majorities. Had those majorities instead been significantly more conservative, it is unlikely that the Warren Court majority would have been able to sustain its preferred liberal course. (p.160)

The Burger Court, 1969–1985

As we saw in Figure 3.1, during the Burger Court years the most preferred rule of the median justice began to move steadily to the right. Consequently, as reported in Figure 5.1, over these terms the unconstrained predicted probability of a conservative judgment by the Court also grew steadily. While this probability was only .09 in the 1969 term, by the 1985 term it had grown to .45. Meanwhile, the most preferred rule of the median representative was moving steadily to the left. While during the early Burger Court terms the median representative remained to the right of the Burger Court, after the 1974 elections the most preferred rule of the median representative moved to the left of that of the median justice, and remained there throughout the remaining Burger Court terms.

For the first few terms of the Burger Court the slightly less liberal preferences of the median representative exert a detectable conservative constraint on the Court's judgments. However, even after factoring in the effects of this constraint, the predicted probability of a conservative judgment between the 1969 and 1971 terms still ranges only between .29 and .35. The early Burger Court was still very likely to rule in a liberal direction on constitutional challenges to federal statutes, even after factoring in the effects of slightly less liberal House majorities.

As the Court gradually moved to the right during the 1970s, and the House to the left, the preferences of the median justice and the median representative converged sufficiently that the effect of House majorities' preferences on the Court is too small to detect at conventional thresholds. Between the 1972 and 1981 terms there are no effects of House majorities on the Court that are discernable at standard levels of statistical significance. During these terms the unconstrained predicted probability of a conservative judgment grows from .15 to .32; the first slightly more conservative and then slightly more liberal preferences of House majorities have no detectable effects on these predicted probabilities. Both the Burger Court majorities and those in the House of Representatives had moderately liberal preferences, and these predicted probabilities reflect those preferences.

However, had the Burger Court faced a significantly more conservative House, we would likely have seen significantly more conservative (p.161) rulings. Figure 5.3 reports for the Burger Court terms the same three series of predicted probabilities reported in Figure 5.2. As in Figure 5.2, the “Conservatively Constrained Court” series reports the average predicted probability that the Court issues a conservative judgment in each term, under the condition that it faces a House majority as conservative as the 104th House. This conservatively constrained Burger Court would have been much more likely to have issued conservative judgments throughout the 1970s, relative to the unconstrained Burger Court, with predicted probabilities of a conservative judgment ranging from .50 to .59. Between the 1969 and 1980 terms these predicted probabilities are distinct in every term at conventional thresholds from those predicted for an unconstrained Court.

Again, although these counterfactual predictions are admittedly speculative, they perhaps illustrate the conditional nature of many of the Burger Court's judgments. Between its 1969 and 1980 terms the Burger Court upheld numerous federal statutes enacted by unified Democratic Congresses that are estimated to have moved status quos in a liberal direction. These included the Black Lung Benefits Act of 1972 (Usery v. Turner Elkhorn Mining Co., 428 U.S. 1 [1976]), the 1975 amendments to the Voting Rights Act of 1965, extending that Act's protections to language minorities (Briscoe v. Bell, 432 U.S. 404 [1977]), and provisions of the Public Works Employment Act of 1977 requiring that 10 percent of federal funds for public works be given to minority-owned businesses (Fullilove v. Klutznik, 448 U.S. 448 [1980]).

As reported in Figure 5.3, the unconstrained predicted probabilities that the Court would strike liberal federal statutes in these terms ranged from .18 to .33. The slightly more liberal preferences of the sitting Democratic House majorities reduced these probabilities to range between only .09 and .22, although the differences between the two sets of predictions are not significant at conventional thresholds. But had the Court faced instead the conservative median representative of the 104th House, the predicted probabilities of conservative judgments in these terms would have ranged instead between .50 and .58. These counterfactual predictions are distinct at conventional thresholds from both the unconstrained and the actual constrained predicted probabilities in each term.

By the Burger Court's 1982 term the justices had become sufficiently conservative, and House majorities sufficiently liberal, that liberal House (p.162)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.3. Predicted Judgment Probabilities, 1969–1985

Simulated from the estimates reported in Table 5.3. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term; the “Conservatively Constrained Court” series sets the preferences of the median representative to their 1995 value. Case-specific control variables remain at their sample values. The “Conservatively Constrained Court” series is distinct from the “Counterfactual Unconstrained Court” series with at least 90 percent confidence in the 1969–1980 terms.

constraint on the Court had begun to bite. Now the presence of liberal Democratic majorities in the House began to decrease the probabilities of conservative judgments by detectable amounts. As depicted in Figure 5.3, in the Burger Court's last four terms the presence of Democratic House majorities decreased the average predicted probabilities of conservative judgments by magnitudes ranging from 18 to 26 percentage points; these predicted decreases are all significant at conventional thresholds.

During these terms the justices upheld provisions of several federal statutes enacted by unified Democratic Congresses and estimated to have moved the status quo in a liberal direction. These included the 1976 (p.163) amendments to the Federal Election Campaign Act of 1971, prohibiting corporate political action committees from soliciting contributions from nonmembers (FEC v. National Right to Work Committee, 459 U.S. 197 [ 1982]), Title IX of the Education Amendments of 1972, prohibiting sex discrimination in any educational institution receiving federal funds as applied to students' receipt of federal financial aid (Grove City College v. Bell, 465 U.S. 555 [1984]), the “union shop” provisions of the Railway Labor Act of 1951, as applied to union expenditures for conventions, publications, and social activities (Ellis v. Railway Clerks, 466 U.S. 435 [(1984]), the minimum wage and overtime pay provisions of the Fair Labor Standards Act, as applied to nonprofit religious organizations (Tony & Susan Alamo Foundation v. Secretary of Labor, 471 U.S. 290 [(1985]), and the 1977 amendments to the Social Security Act extending survivors' benefits to wage earners' widowed spouses who remarry after age 60 (Bowen v. Owens, 476 U.S. 340 [(1986]).

Had the Court been unconstrained by the Democratic House during these terms, it would have struck liberal federal statutes like these with probabilities ranging from .35 to .45. But because of the presence of liberal House majorities between the 1982 and 1985 terms, the actual probabilities that these statutes would be struck ranged only between .15 and .21. The two sets of predicted probabilities are distinct at conventional thresholds in each term.

The Rehnquist Court, 1986–2004

By the 1986 term, the first term of the Rehnquist Court, the distance between the preferences of the more conservative median justice and those of the more liberal median representative had become substantial. As we already saw in Figure 5.1, between the Rehnquist Court's 1986 and 1993 terms, liberal House majorities decreased the probabilities of conservative judgments by magnitudes ranging from 30 to 71 percentage points; these decreases are all significant at conventional thresholds. Had the Rehnquist Court been unconstrained by the liberal Democratic House between its 1986 and 1993 terms, it would have issued conservative judgments in cases involving constitutional challenges to federal statutes with predicted probabilities ranging from .45 to .84. But the (p.164) presence of liberal House majorities during these eight terms constrained these probabilities to range between only .10 and .17.

The effect of the House median's preferences on the Court then disappears after the 1994 congressional elections, as a result of the median representative's rightward shift toward the most preferred rule of the median justice. Between 1994 and 2004 the constrained Rehnquist Court was as likely to issue conservative judgments in cases involving constitutional challenges to federal statutes as if it had been unconstrained by the preferences of House majorities; the effect of these preferences on the Court's judgments is small and insignificant at conventional thresholds during these terms.

But had the 1994 elections not changed the composition of Congress, we likely would have seen a very different Rehnquist Court in its last eleven terms. Figure 5.4 reports for the Rehnquist Court terms the two predicted probability series displayed in Figure 5.1, along with a predicted probability series for a “Liberally Constrained Court.” This latter series reports the predicted probability of a conservative judgment between the 1994 and 2004 terms, had the House remained as liberal as the 1994 Democratic majority House. In all terms from 1994 through 2004 these predicted probabilities are distinct at conventional thresholds from both the “Counterfactual Unconstrained Court” and the “Actual Constrained Court” series also reported in Figure 5.4. During these terms the unconstrained Rehnquist Court is predicted to issue conservative judgments in cases involving constitutional challenges to federal statutes with probabilities ranging from .58 to .64. Although the House was slightly less conservative than the Court during these terms, the differences between the two institutions were insufficiently large to exert any detectable effect on the Court's decisions. But had the House remained as liberal as the 1994 Democratic majority House, the predicted probability of a conservative judgment in these terms would have ranged between only .12 and .18.

Figures 5.1 and 5.4 may help to explain many of the cases that puzzled observers during both the first and the second Rehnquist Courts. Recall from Chapter Four that the Rehnquist Court's decision in Morrison v. Olson, 487 U.S. 654 (1988), upholding the federal independent counsel statute, was one of the first prominent decisions to call into question the (p.165)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.4. Predicted Judgment Probabilities, 1986–2004

Simulated from the estimates reported in Table 5.3. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term; the “Liberally Constrained Court” series sets the preferences of the median representative to their 1994 value. Case-specific control variables remain at their sample values. The “Actual Constrained Court” and “Counterfactual Unconstrained Court” series are distinct with at least 90 percent confidence between the 1986 and 1993 terms; the “Liberally Constrained Court” series is distinct from the other two series with at least 90 percent confidence between the 1994 and 2004 terms.

Rehnquist majority's commitment to conservative principles. The case had resulted from the efforts of the Democratic majority in the House of Representatives to wield the independent counsel statute against prominent members of the Reagan administration. This statute had originated as a provision in the Ethics in Government Act of 1978, and had most recently been reauthorized in the Independent Counsel Reauthorization Act of 1987 by a unified Democratic Congress; the reauthorization is estimated to have moved the status quo in a liberal direction.30 Court (p.166) watchers assumed that the conservative majority on the Court, four of whose members had been directly appointed or promoted by the Reagan administration, would side with the administration's claim of executive power and strike the statute.31

And had the 1987 term Rehnquist Court been unconstrained by the considerably more liberal House, it is predicted to have struck liberal federal statutes with a probability of .51. But the Court rejected the constitutional challenge to the Independent Counsel Reauthorization Act of 1987 in a “crushing defeat” for the Reagan administration.32 Although the Court's ruling in Morrison was characterized as “astonishing” by Court watchers, it was perhaps predictable. The Court was not unconstrained in its 1987 term; it was facing a considerably more liberal House of Representatives. Moreover, Morrison presented an explicit, overt conflict between the House, the body possessed of the powers to both punish and reward the justices, and the executive branch. Although the conservative justices owed their appointments to the executive branch, and likely shared the Reagan administration's policy preferences for the most part, that administration could neither save them from House impeachment investigations nor initiate appropriations measures increasing their salaries. Because of the presence of the liberal House majority, the actual predicted probability that the Court would strike a liberal federal statute in this term was only .13.

In its 1989 term the Court upheld the minority ownership preferences in awarding and transferring broadcast licenses that had been implemented by the Federal Communications Commission in 1978 (Metro Broadcasting v. FCC, 497 U.S. 547 [1990]). The FCC's regulations had been challenged on Fifth Amendment equal protection grounds. At issue in the case were not only the FCC's regulations, issued during the liberal Jimmy Carter administration, but also the efforts of the Democratic Congresses between 1987 and 1989 to prevent the Reagan-era FCC from revising those regulations. In each of these years the Congress had attached a rider to the FCC's appropriations bills, preventing the agency from altering the preferential treatment given to minority-owned businesses. The last of these, enacted in 1989 by a unified Democratic Congress, had been largely supported by Democrats and opposed by Republicans; the bill is estimated to have moved the status quo in a liberal direction.33 Like Morrison (p.167) v. Olson (1988), then, Metro Broadcasting directly implicated the conservative Rehnquist Court majority in a fight between the conservative Reagan administration and the liberal Democratic House of Representatives.In the same term the Court also upheld Section 1 of the Sherman Antitrust Act, prohibiting anticompetitive practices such as horizontal price fixing, as applied to economic boycotts intended to affect the price paid for legal services (FTC v. Superior Court Trial Lawyers Association (SCTLA), 493 U.S. 411 [1990]). Section 1 had been amended most immediately prior to SCTLA by a provision of the Consumer Goods Pricing Act of 1975, striking exemptions from Section 1's provisions for state fair trade laws; that Act had been passed by unified Democratic majorities and is estimated to have moved the status quo in a liberal direction. By upholding the application of Section 1 in SCTLA, the Court reversed a ruling by a conservative majority panel for the D.C. Circuit that had weakened the per se illegality of price fixing, a foundational principle in the Court's antitrust jurisprudence.34 The appellate panel's opinion had been widely viewed with suspicion by those endorsing a robust application of antitrust statutes; many feared that Judge Douglas Ginsburg's opinion, if allowed to stand, would “open the door to economically-powerful bullies disrupting governments and markets.”35

And had the Court been unconstrained by a liberal House majority, it might very well have invalidated the liberal statutes at issue in both Metro Broadcasting and SCTLA: the unconstrained predicted probability of a conservative judgment in this term was .57. But as a result of the presence of the sitting liberal House majority, the actual predicted probability that the Court would strike liberal statutes in this term was only .14. The Court upheld both statutes.

Only a few terms later, however, the political landscape changed dramatically. As we have seen, the House median shifted significantly to the right in the 1994 elections. Now House majorities no longer constrained the conservative Rehnquist Court majority from issuing conservative judgments.

Many of these conservative judgments involved strikes of liberal federal statutes previously enacted by Democratic Congresses. They included U.S. v. Lopez, 514 U.S. 549 (1995), striking the Gun-Free School Zones Act of 1990, as amended in 1994;36 Seminole Tribe of Florida v. (p.168) Florida, 517 U.S. 44 (1996), striking the Indian Gaming Regulatory Act of 1988;37 Colorado Republican Federal Campaign Committee v. Federal Election Commission, 518 U.S. 604 (1996), striking provisions of the Federal Election Campaign Act of 1971, as amended in 1976, imposing limits on political parties' independent expenditures on behalf of congressional candidates; City of Boerne v. Flores, 521 U.S. 507 (1997), striking the Religious Freedom Restoration Act of 1993; Printz v. United States, 521 U.S. 898 (1997), striking the Brady Handgun Violence Prevention Act of 1993;38 Eastern Enterprises v. Apfel, 524 U.S. 498 (1998), striking the Coal Industry Retiree Health Benefit Act of 1992; Alden v. Maine, 527 U.S. 706 (1999), striking provisions of the Fair Labor Standards Act, as last amended in 1977;39 College Savings Bank v. Florida Prepaid, 527 U.S. 666 (1999), striking provisions of the Trademark Remedy Clarification Act of 1992, making states liable for damages caused by unfair practices such as false advertising; Florida Prepaid v. College Savings Bank, 527 U.S. 627 (1999), striking provisions of the Patent and Plant Variety Protection Remedy Clarification Act of 1992, making states liable for damages caused by unremediated patent infringement; Kimel v. Florida Board of Regents, 528 U.S. 62 (2000), striking provisions of the 1977 amendments to the Fair Labor Standards Act, as applied in the context of the Age Discrimination in Employment Act; U.S. v. Morrison, 529 U.S. 598 (2000), striking provisions of the Violence Against Women Act of 1994;40 and Board of Trustees of the University of Alabama v. Garrett, 531 U.S. 356 (2001), striking provisions of the Americans with Disabilities Act of 1990 (ADA), permitting state employees to sue their employers for disability-based discrimination.41

As we saw in Chapter Four, these rulings were generally characterized by observers as constituting large departures from the Court's prior jurisprudence. The Court's decision in Lopez, for example, was described by professional Court watchers as “extraordinary,”42 “stunning,”43 “radical,”44 and “a distinct surprise.”45 Seminole Tribe was characterized as “shocking”46 and “amazing.”47 Boerne was described as a “new departure” in the Court's Section 5 jurisprudence,48 one that added “an entirely new” and “entirely alien” chapter to the Court's Section 5 precedents.49 Printz was seen as “huge,”50 as having “enormous consequences,”51 and as a decision that “would have been inconceivable to many people only (p.169) a few years ago.”52 The Court's rulings in Alden, College Savings Bank, and Florida Prepaid were characterized as decisions that “few would have predicted.”53 Garrett was described as signaling a conservative “revolution” on the Court.54

But the estimates reported in Table 5.3 suggest that perhaps these rulings should not have come as such a surprise. Given the preferences of the new Republican majority that assumed control of the House in January 1995, and that maintained this control throughout the remainder of the Rehnquist Court's terms, the Court's rulings in these cases were instead relatively predictable. Because these preferences were only slightly less conservative than those of the Rehnquist Court, the House now exercised no appreciable constraint on the Court's judgments. The Court was therefore free to strike liberal federal statutes, which it was predicted to do in these terms with unconstrained probabilities ranging between .58 and .63.

But, as reported in Figure 5.4, had the post-1994 House medians remained as liberal as that of the 1994 House, most of these statutes likely would have been upheld: the predicted probabilities of strikes in these cases would then have ranged only between .12 and .18. The Court appears to have struck these liberal federal statutes not because the justices had suddenly become more conservative, but because the House had suddenly become more conservative.

In short, the estimates reported here would appear to fully explain the puzzle of the break between the “two” Rehnquist Courts, a break characterized by many as of “historic importance.”55 The dramatic changes in the Rehnquist Court's jurisprudence in the mid-1990s, changes seen in several different areas of constitutional law, appear to have been the product not of changes in the preferences of the justices, but rather of changes in the preferences of House majorities.

The estimates reported in Table 5.3 can also be used to interrogate one of the more prominent claims made about the second Rehnquist Court, namely that it was the product of an apparently newfound interest by the conservative majority in federalism or states' rights. Presumably if the federalism interpretation were correct, the Rehnquist Court would have been equally likely to strike both liberal and conservative federal statutes. All federal statutes should have been similarly at risk during the second (p.170) Rehnquist Court, as the justices sought to free state governments from burdensome federal regulations.

But this interpretation is belied by the estimates reported in Table 5.3. The second Rehnquist Court was as likely to uphold conservative federal statutes between its 1994 and 2004 terms, as it was to strike liberal federal statutes. Examples of such conservative upholds included provisions of the Antiterrorism and Effective Death Penalty Act of 1996, restricting state prisoners' ability to file habeas corpus petitions in federal courts (Felker v. Turpin, 518 U.S. 651 [1996]),56 provisions of the Prison Litigation Reform Act of 1996, as amended in 1997, limiting federal civil actions challenging the conditions of confinement in correctional facilities (Miller v. French, 530 U.S. 327 [2000]), the 1996 amendments to the Immigration and Nationality Act, permitting the Attorney General to detain aliens convicted of certain categories of crimes pending deportation hearings (Demore v. Kim, 538 U.S. 510 [2003]), the 2000 Children's Internet Protection Act, prohibiting public libraries from receiving federal funds for internet access unless they restricted access to obscene or pornographic imagery (U.S. v. American Library Association, 539 U.S. 194 [2003]), and the Schedule I classification of marijuana under the Controlled Substances Act, originally enacted in 1970 and most recently amended by the Hillory J. Farias and Samantha Reed Date-Rape Drug Prohibition Act of 2000, adding Gamma-Hydroxybutyric acid (GHB) to the list of Schedule I substances but leaving the Schedule I classification of marijuana untouched (Gonzales v. Raich, 545 U.S. 1 [2005]).

All these statutes had been passed by unified Republican Congresses, and all are estimated to have moved the status quo in a conservative direction. Had the second Rehnquist Court faced the liberal Democratic 1994 House, these statutes would have been upheld with predicted probabilities ranging from only .12 to .17. But because the Court was unconstrained by a liberal House, these conservative statutes were instead upheld with predicted probabilities ranging from .61 to .64.

These predicted probabilities suggest that the federalism explanation for the second Rehnquist Court does not hold much water. That Court had few problems with federal statutes enacted by Republican Congresses. It was only statutes enacted by liberal Democratic Congresses that the Court found objectionable. (p.171)

Estimating Strike Probabilities

Our two strategies for measuring the Court's judgments objectively still have left out many cases involving constitutional challenges to federal statutes for which no roll call votes are available, and which were enacted during periods of divided government. One way to include these statutes in our analyses is to measure the average locations of statutes enacted by these Congresses, as predicted by prominent models of congressional behavior. Three of these models, namely a floor median model, a committee gatekeeping model under an open rule, and a party gatekeeping model under an open rule, specify the chamber medians as the pivotal legislators in determining the policy location of legislation.57 We can then take the midpoint between the enacting chamber medians' estimated most preferred rules as a proxy for the rule embodied by any given statute.

We are still left with the question of how to code the Court's rulings on these statutes, however. Our previous strategies for coding the Court's judgments took advantage of the fact that we were measuring the direction in which a statute moved the status quo ex ante. We measured this movement directly with the roll call–based judgment measure, and indirectly with the partisan Congress-based judgment measure. When the Court upheld a statute's constitutionality it preserved the direction of this movement; when it struck a statute as unconstitutional it reversed the direction of this movement.

However, locating statutes at precise policy points complicates this measurement strategy. When the Court upholds a statute, it preserves a policy located at some precise point; we could take that point as a measure of the policy location of the Court's judgment. But when the Court strikes a statute, what value should be assigned to its judgment then?

One way out of this measurement conundrum is to code the Court's constitutional rulings on these statutes as simple strikes and upholds, and then to estimate the probability that the Court strikes a federal statute as a function of both judicial and elected branch preferences over that statute. Recall our assumption from Chapter Three that federal statutes embody standards of constitutionality endorsed by the enacting elected branches. When a constitutional challenge to a federal statute comes before the Court, the justices articulate the standard they will use to gauge (p.172) the constitutionality of the challenged statute. They may accept the standard endorsed by the enacting elected branch majorities, as embodied in the statute, or they may move that standard in a liberal or conservative direction, as much or as little as they prefer.

The traditional view of an unconstrained, independent Court assumes that the Court is free to locate this standard at will, and ceteris paribus will apply the median justice's most preferred constitutional standard in every case. A model of elected branch constraint on the Court predicts instead that the Court will defer to the most preferred standards of pivotal elected officials.

We can think about the probability that any given statute is struck or upheld by the Court as a function of the distance between the constitutional standard embodied by the statute at issue, and the standard implemented by the Court in its majority opinion. Should the standard endorsed by the Court be identical to that embodied in the challenged statute, the statute will surely be upheld on constitutional grounds. But as the standard applied by the Court to a particular statute moves further away from that embodied by the statute, either to the left or to the right, then the probability that the statute will be upheld must surely decline.

If the constitutional rule or standard announced by the Court in its opinions is a function of judicial preferences, then we should find that a measure of the distance between the most preferred rule of the median justice and that embodied by a challenged statute is a good predictor of the probability that a statute is struck by the Court. If the location of the Court's rule is not responsive to elected branch preferences, then we should find that strike probabilities are not responsive to the inclusion of measures of the distances between the most preferred rules of pivotal elected officials and that embodied by a challenged statute. If, however, we do find that increases in the latter distances are associated with increases in the probability that statutes are struck, then we will have evidence that the Court's legal rules are in fact responsive to elected branch preferences.

Tables 5.5 and 5.6 report the results of these analyses, which include the same controls used in previous probit regressions, albeit coded for this new dependent variable. These controls include the lower court's disposition in a case (that is, whether the lower court struck or upheld (p.173)

Table 5.5: Estimating the Probability of a Strike: Bailey XTI Judicial and Elected Branch Preferences over Challenged Federal Statutes, 1953–2004

Coefficient

SE

95% CI

SCOTUS Median/Statute Distance

1.60

.49

(.65, 2.55)

House Median/Statute Distance

2.50

1.44

(−.32, 5.32)

Senate Median/Statute Distance

−.08

1.35

(−2.72, 2.56)

President/Statute Distance

−.59

.30

(−1.18, −.01)

Lower Court Strike

.06

.20

(−.32, .45)

US Pro-Strike Party

.83

.47

(−.08, 1.75)

US Pro-Uphold Party

.35

.26

(−.17, .87)

Statute Age

−.01

.02

(−.05 .03)

N = 309

Wald chi2 = 27.12

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at the .10 level (two-tailed).

Table 5.6: Estimating the Probability of a Strike: Judicial Common Space Judicial and Elected Branch Preferences over Challenged Federal Statutes, 1953–2004

Coefficient

SE

95% CI

SCOTUS Median/Statute Distance

1.90

.78

(.36, 3.43)

House Median/Statute Distance

3.85

1.50

(.91, 6.79)

Senate Median/Statute Distance

2.24

1.63

(−.96, 5.44)

President/Statute Distance

−1.71

.49

(−2.67, −.75)

Lower Court Strike

.04

.20

(−.34, .43)

US Pro-Strike Party

.99

.49

(.04, 1.95)

US Pro-Uphold Party

.30

.29

(−.27, .87)

Statute Age

−.01

.02

(−.04, .02)

N = 309

Wald chi2 = 57.66

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at die .10 level (two-tailed).

(p.174) the statutory provision at issue), whether the United States was a party to a case arguing against a statute's constitutionality, and whether the United States was a party to a case arguing in favor of the statute's constitutionality. These analyses also control for statute age; a statute may be less (or more) likely to be struck by the Court as time passes.58

These estimates of strike probabilities, using the largest possible sample of cases involving constitutional challenges to federal statutes between 1953 and 2004, are entirely consistent with the estimates of the probabilities of conservative judgments reported in Tables 5.1 through 5.4. In Tables 5.5 and 5.6 there are no effects in the predicted direction on the probability that the Court strikes a federal statute as a function of either the median senator's or the president's preferences over that statute, using either the Bailey XTI or the JCS preference estimates. There are clear effects on that probability from the preferences of the median justice; the Court was more likely to strike a federal statutory provision between 1953 and 2004 when those statutes were more distant from the preferences of the median justice. Using the Bailey XTI preference estimates, increasing the distance between the most preferred rule of the median justice and that embodied by a challenged statute from the smallest to the largest observed distance between 1953 and 2004 increases the probability of a strike from .13 to .81; using the Judicial Common Space estimates this probability increases from .16 to .45.59 These predicted probabilities are distinct at conventional thresholds.

However, there are also large effects on strike probabilities from the preferences of the median representative over challenged statutes. Using the Bailey XTI preference estimates, the smallest distance between an enacted statute and the most preferred rule of the median representative is observed for statutes enacted in 1987 and reviewed during the 1988 term; the largest observed distance is for statutes enacted in 1976 and reviewed in the 1995 term. Using the JCS preference estimates, the smallest such distance is observed for statutes enacted in 1970 and reviewed in the Court's 1980 term; the largest such distance is observed for statutes enacted in 1976 and reviewed in the 1995 term. Moving from the smallest to the largest observed distance between the most preferred rule of the median representative and that embodied by a challenged statute increases the probability that the Court will strike a statute from .16 to .47 (p.175) using the Bailey XTI preference estimates, and from .15 to .59 using the Judicial Common Space estimates. Again, these predicted probabilities are distinct at conventional thresholds.

These estimated effects of changes in the preferences of the median representative become even larger in magnitude when we isolate the Rehnquist Court terms, terms of particular interest for estimating the Court's responsiveness to changes in elected branch preferences. As reported in Tables 5.7 and 5.8, during these terms changes in the median justice's preferences over challenged statutes appear to have no effect on the predicted probability that a statute will be struck, regardless of the preference estimates used. There are also no effects on this probability from changes in the preferences of the median senator or the president. But moving from the smallest to the largest distance between the most preferred rule of the median representative and that embodied by a challenged statute increases the predicted probability of a strike from .06 to .96 using the Bailey XTI preference estimates, and from .15 to .68 using the Judicial Common Space preference estimates; these predictions are distinct at conventional thresholds.60

Table 5.7: Estimating the Probability of a Strike: Bailey XTI Judicial and Elected Branch Preferences over Challenged Federal Statutes, 1986–2004

Coefficient

SE

95% CI

SCOTUS Median/Statute Distance

−1.74

1.40

(−4.49, 1.00)

House Median/Statute Distance

8.88

2.37

(4.23, 13.52)

Senate Median/Statute Distance

1.82

1.86

(−1.81, 5.46)

President/Statute Distance

−.57

.49

(−1.54, .40)

Lower Court Strike

.92

.28

(.37, 1.48)

US Pro-Strike Party

1.09

.77

(−.41, 2.60)

US Pro-Uphold Party

.21

.47

(−.72, 1.14)

Statute Age

−.05

.04

(−.14, .04)

N = 107

Wald chi2 = 61.67

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at die .10 level (two-tailed).

(p.176)

Table 5.8: Estimating the Probability of a Strike: Judicial Common Space Judicial and Elected Branch Preferences over Challenged Federal Statutes, 1986–2004

Coefficient

SE

95% CI

SCOTUS Median/Statute Distance

.15

2.03

(−3.82, 4.12)

House Median/Statute Distance

4.47

1.74

(1.06, 7.88)

Senate Median/Statute Distance

4.74

3.13

(−1.41, 10.88)

President/Statute Distance

−1.92

1.47

(−4.80, .95)

Lower Court Strike

.81

.29

(.24, 1.37)

US Pro-Strike Party

1.28

.78

(−.25, 2.80)

US Pro-Uphold Party

.27

.50

(−.72, 1.25)

Statute Age

−.04

.04

(−.11, .04)

N = 107

Wald chi2 = 57.87

Note: Probit regression; robust standard errors reported, clustered by term. Intercept term not reported. Italicized variables are statistically significant at die .10 level (two-tailed).

We can use these estimates to characterize the alternative likelihoods that the Rehnquist Court would strike statutes enacted by a given Congress, as a function of both the actual judicial and elected branch preferences in each term, and the counterfactual preferences generated by assuming that the median representative's preferences were identical to those of the median justice in every term. The difference between these two series of predicted probabilities will reveal the magnitude of the effect of House preferences on the Court's dispositions of these statutes.

Figure 5.5 displays these two probability series for statutes enacted by the liberal 1965 Congress, using the Bailey XTI preference estimates. This is the Congress that enacted a number of landmark liberal statutes, including the Elementary and Secondary Education Act, the Social Security Act of 1965, establishing Medicaid and Medicare, the Voting Rights Act of 1965, and the Housing and Urban Development Act of 1965. Statute age is set at the correct value for each term, while case-specific control variables are held at their sample values. (p.177)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.5. Predicted Strike Probabilities for Statutes Enacted in 1965

Simulated from the estimates reported in Table 5.7. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term; the “Liberally Constrained Court” series sets the preferences of the median representative to their 1994 value. Case-specific control variables remain at their sample values. The “Actual Constrained Court” and “Counterfactual Unconstrained Court” series are distinct with at least 90 percent confidence between the 1986 and 1993 terms; the “Liberally Constrained Court” series is distinct from the other two series with at least 90 percent confidence between the 1994 and 2004 terms.

Had the median representative's preferences been identical to those of the median justice in every term, as reported by the “Counterfactual Unconstrained Court” series, we can see that the probability of a strike of a 1965 statute would have ranged between .71 and .98 between 1986 and 2004. However, Figure 5.5 also reveals the degree to which a liberal House protected these liberal statutes during the 1980s and early 1990s. As reported by the “Actual Constrained Court” series, the presence of liberal (p.178) House majorities held the predicted probability of a strike to between .08 and .17 between the 1986 and 1993 terms. The differences between these two series are distinct at conventional thresholds in every term.

After the 1994 elections, however, this congressional protection vanished. Now the Court was free to strike these statutes as a function of its own preferences. Between the 1994 and 2004 terms, the predicted probabilities reported in the “Actual Constrained Court” series are not significantly different from those reported in the “Counterfactual Unconstrained Court” series. But had the House not moved to the right after the 1994 elections, these liberal 1965 statutes would have remained under its protection. The “Liberally Constrained Court” series reported in Figure 5.5 reports the predicted probability of a strike of a 1965 statute under the counterfactual assumption that the House had remained as liberal as the 1994 Democratic majority House. Under this counterfactual assumption, the predicted probabilities of strikes of 1965 statutes range between only .16 and .32 between 1994 and 2004; these predicted probabilities are distinct from those reported in both the “Counterfactual Unconstrained Court” and the “Actual Constrained Court” series at conventional thresholds.

The estimates reported in Tables 5.7 and 5.8 can also help us to understand some of the cases from the Rehnquist Court involving federal statutes passed during periods of divided Congresses. For example, in its 1988 term the first Rehnquist Court upheld a federal statute on jurisprudential grounds that it would later reject after the 1994 elections. As we saw in Chapter Four, in Pennsylvania v. Union Gas Co., 491 U.S. 1 (1989), the Court upheld the Superfund Amendments and Reauthorization Act of 1986 (SARA) against an Eleventh Amendment challenge. This statute had made explicit the congressional intention to render states liable in federal courts for damages caused by environmental harms. While statutory efforts to address environmental damage arising as a consequence of industrial activity were agreed to be a valid exercise of congressional powers under the Interstate Commerce Clause, the state of Pennsylvania argued that the Eleventh Amendment nonetheless prohibited the Congress from allowing states to be sued in federal courts for violating the provisions of such statutes. (p.179)

Although SARA had been enacted by a divided Congress, we can nonetheless estimate the probability that a statute enacted in 1986 would be struck by the 1988 term Rehnquist Court. Using the estimates reported in Table 5.7, had the conservative majority on this Court been unconstrained by the preferences of the liberal majority in the House of Representatives, it is predicted to have struck statutes enacted in 1986 with a probability of .79. In other words, left to its own devices, the 1988 term Court likely would have accepted the Eleventh Amendment challenge to the Superfund Amendments and Reauthorization Act, and struck the statute. However, because of the presence of that liberal Democratic House majority, the actual predicted probability of such a strike in this term was only .22. The Court upheld the Superfund Amendments, rejecting Pennsylvania's Eleventh Amendment challenge.61

Only seven years later, however, the Court would overrule its Eleventh Amendment decision in Union Gas. As we have seen, in Seminole Tribe of Florida v. Florida, 517 U.S. 44 (1996), the Court reviewed an Eleventh Amendment challenge to the Indian Gaming Regulatory Act, enacted in 1988 by a Democratic-majority Congress. As in Union Gas, the congressional power to regulate Indian gaming under the Indian Commerce Clause was not at issue in the case. The issue in Seminole Tribe was whether, in the course of pursuing valid statutory goals under the Interstate and Indian Commerce Clauses, the Congress could make states liable in federal court for having violated such statutes. In Union Gas the Court had rejected such a challenge to the Superfund Amendments and Reauthorization Act. But in Seminole Tribe the Rehnquist Court overruled its decision in Union Gas, accepting the Eleventh Amendment challenge to the Indian Gaming Regulatory Act and striking the offending statutory provisions.

In Seminole Tribe the justices were not constrained by a liberal House majority. Using the estimates reported in Table 5.7, the unconstrained predicted probability that the 1995 term Rehnquist Court would strike a statute enacted by the 1988 Congress was .77; the preferences of the conservative 1996 Republican House were sufficiently similar to those of the justices to have no detectable impact on this Court's dispositions.62 However, had the 1995 term Court instead reviewed the Indian Gaming (p.180) Regulatory Act while facing a House as liberal as the one faced by the 1988 term Rehnquist Court, the Court that had upheld the Superfund Amendments and Reauthorization Act, the Indian Gaming Regulatory Act would have been struck with a predicted probability of only .09. This probability is distinct at conventional thresholds from the unconstrained predicted probability.63 Conversely, if the 1988 term Rehnquist Court had faced a House as conservative as the House faced by the 1995 term Rehnquist Court, it would no longer have been constrained in its constitutional rulings on federal statutes. Under this condition it would have struck the Superfund Amendments and Reauthorization Act with an unconstrained predicted probability of .79.64

In other words, the difference between the outcomes in Union Gas and Seminole Tribe was not a product of a change in the justices' preferences. Had the Rehnquist Court justices been unconstrained by a liberal House majority in their 1988 term, they likely would have accepted the Eleventh Amendment challenge to the Superfund Amendments and Reauthorization Act, and struck the relevant statutory provisions. Because they were unconstrained by a liberal House in their 1995 term, they were able to endorse the Eleventh Amendment challenge to the Indian Gaming Regulatory Act, and strike the offending statutory provisions, overruling their own recent Eleventh Amendment precedent. However, they most likely would not have taken this jurisprudential step, had they not faced a friendly majority in the House of Representatives. The second Rehnquist Court's dramatic turnabout in its Eleventh Amendment jurisprudence appears to have been a direct result of the similarly dramatic 1994 House elections.

Likewise, in its 1989 term the first Rehnquist Court upheld the National Trails System Act Amendments of 1983 in Preseault v. ICC, 494 U.S. 1 (1990). This statute had authorized the Interstate Commerce Commission to convert unused railroad lines into recreational trails. It had been challenged on the grounds that it exceeded congressional power under the Interstate Commerce Clause, the creation of bike trails not having much to do with the regulation of commerce across state lines. In its ruling, the Court had declared that the congressional power to regulate interstate commerce extended even to such noneconomic aims as the (p.181) promotion of the “enjoyment and appreciation of the open-air, outdoor areas and historic resources of the Nation.”65

Preseault was significant because only a few terms later the Court would reject this interpretation of the Interstate Commerce Clause in its landmark Lopez opinion. Understanding what changed between Preseault and Lopez is thus critical for understanding the Rehnquist Court's about-face on congressional power under this clause.

The National Trails System Act Amendments were enacted during a divided partisan Congress, and we do not have a final passage roll call vote for the statute. But we can estimate the probability that the 1989 term Rehnquist Court would strike statutes enacted by the 1983 Congress. Using the estimates reported in Table 5.7, had the Court been unconstrained by the liberal House in its 1989 term, there was a predicted probability of .73 that it would have struck the statute at issue in Preseault. But given the liberal House median sitting during this term, the actual predicted probability that the Court would strike this statute was only .12.66 The Court upheld the statute.

However, had the 1989 term Court faced a House as conservative as that faced by the 1994 term Court, the Court that issued the ruling in U.S. v. Lopez (1995), the justices likely would not have endorsed such an expansive reading of the Interstate Commerce Clause. Under this condition, the preferences of the majority in the House of Representatives would have been insufficiently distinct from those of the majority on the Court to constrain the latter's decision making. The Court would have struck the National Trails System Act Amendments with an unconstrained predicted probability of .73.

Likewise, using the estimates reported in Table 5.7, the 1994 term Court is predicted to strike statutes enacted in 1994, like the amended Gun-Free School Zones Act, with a probability of .96; the preferences of the conservative Republican majority that assumed control of the House of Representatives in 1995 do not decrease this probability. However, had the 1994 term Court instead faced a House as liberal as the 1990 House, the House faced by the 1989 term Court that had upheld the National Trails System Act Amendments, this Court would have struck the Gun-Free School Zones Act with a predicted probability of only .17.67 (p.182)

In other words, the remarkable change in the second Rehnquist Court's interpretation of the Interstate Commerce Clause appears to have been directly related to the perhaps equally remarkable 1994 House elections. Left to their own devices, the conservative Rehnquist Court justices likely would have struck not only the Gun-Free School Zones Act in their 1994 term, but also the Trails System Act Amendments in their 1989 term. That they did not alter their Interstate Commerce Clause jurisprudence earlier seems largely due to the presence of liberal House majorities until the 1994 term.

We can also use the estimates reported in Tables 5.7 and 5.8 to further consider the “federalism” interpretation of the second Rehnquist Court. This interpretation suggests that the second Rehnquist Court began to strike federal statutes out of a newfound interest in states' rights. But the estimates reported in these tables suggest that the Rehnquist Court had quite different plans for federal statutes enacted by liberal Democratic majority Congresses, and those enacted by conservative Republican majority Congresses.

Figures 5.6 and Figures 5.7 illustrate this differential treatment by using the estimates reported in Table 5.7 to simulate the predicted probabilities that statutes enacted in 1987 and 1995 would be struck by the Court, as a function of both actual and counterfactual values for the House median's preferences. In the 1986 midterm elections the Democrats won majorities in both the House and Senate, holding those majorities until the 1994 midterm elections. In both the Bailey XTI and the Judicial Common Space preference estimates the House and Senate medians shift markedly to the left as a result of the 1986 elections; the statutes enacted by the 1987 Congress thus measure as relatively liberal. By contrast, the statutes enacted by the Republican majority 1995 Congress measure as relatively conservative. In both Figures 5.6 and 5.7 the preferences of the median justice, the median senator, and the president are set at their actual values for each term, while case-specific controls are held at their sample values.

In Figure 5.6, the “Counterfactual Unconstrained Court” series reports predicted strike probabilities for 1987 statutes under the assumption that the House median shared the same most preferred rule as the Court median in every term. The “Actual Constrained Court” series reports (p.183)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.6. Predicted Strike Probabilities for Statutes Enacted in 1987

Simulated from the estimates reported in Table 5.7, with 95 percent confidence intervals. The “Actual Constrained Court” series uses the actual values of judicial and elected branch preferences; the “Counterfactual Unconstrained Court” series sets the preferences of the median representative to those of the median justice in each term. Case-specific control variables remain at their sample values.

predicted strike probabilities for these statutes while setting the median representative's most preferred rule to its actual value in each term. The 95 percent confidence intervals for these predicted probabilities are reported as well. We can see that between 1987 and 1993 these two series are distinct with at least 95 percent confidence in every term. Had the Court been unconstrained by the liberal House of Representatives during these terms, it would have struck a statute enacted in 1987 with predicted probabilities ranging from .79 to .99. But because of the constraint posed by the presence of the liberal House, these predicted strike probabilities ranged from only .08 to .14. After the 1994 elections, however, the Court was free to strike statutes enacted by this liberal Congress. Between its 1994 and 2004 terms an unconstrained Court would have struck statutes enacted in 1987 with probabilities ranging from .79 to .91. The only (p.184)

Explaining The Puzzle of the Two Rehnquist Courts

Figure 5.7. Predicted Strike Probabilities for Statutes Enacted in 1987 and 1995

Simulated from the estimates reported in Table 5.7, with 95 percent confidence intervals. Preference variables are set to their actual values, while case-specific control variables are held to their sample means.

slightly less conservative House median has no effect on these predicted strike probabilities; the 95 percent confidence intervals for the two series overlap in every term between 1994 and 2004.

So far, these results are consistent with the “federalism” story. Starting in the 1994 term, the Rehnquist Court began to take a much harder line on evaluating the constitutionality of federal statutes. But this hard line did not extend to all federal statutes. Figure 5.7 reports the predicted strike probabilities for both 1987 and 1995 statutes, along with the 95 percent confidence intervals, using the actual values of both judicial and elected branch preferences, while holding case-specific control variables at their sample values. The two series are clearly distinct in every term between the 1995 and 2004 terms. While the strike probabilities for 1987 statutes range between .52 and .71 between the 1995 and 2004 terms, the strike probabilities for 1995 statutes range between only .16 and .32 over the same period. (p.185)

In short, the estimates of strike probabilities reported in Tables 5.5 through 5.8 are quite consistent with the estimates of the probabilities of conservative judgments reported in Tables 5.1 through 5.4. Both sets of estimates indicate that the Court's judgments in cases involving constitutional challenges to federal statutes between 1953 and 2004 were rather remarkably responsive to the preferences of majorities in the House of Representatives. We see that responsiveness most clearly during the Rehnquist Court terms, when a consistently conservative majority on the Court faced first a series of liberal Democratic House majorities, and then a series of conservative Republican House majorities. During these terms we appear to have observed two distinct Rehnquist Courts not because the justices' own preferences suddenly became more conservative in the 1994 term, nor because they one day discovered an interest in preserving federalism, nor because they abruptly got fed up with Congress. Instead, we appear to have observed two Rehnquist Courts because the Supreme Court is not, and perhaps was not designed to be, independent of congressional influence.

Constitutional Review of State Action

One limitation of the findings reported here is the narrowly defined sample from which they were estimated. Many prominent decisions made by the Court lie outside even the largest sample of cases used here. We have no estimates of the factors that affected the Court's rulings in these cases. These cases include, perhaps most importantly, those involving constitutional review of state and local action. Several of these cases, particularly those from the Warren Court, are frequently cited as evidence of the Court's independence. Ronald Dworkin, for example, after asserting the important role given to independent federal courts in the U.S. Constitution, cites Brown v. Board of Education of Topeka, 349 U.S. 294 (1954), finding a Fourteenth Amendment equal protection violation in racially segregated public schools, and Griswold v. Connecticut, 381 U.S. 479 (1965), finding a “right to privacy” violation in state statutes restricting access to contraceptives, as “shining examples of our constitutional structure working at its best.”68 Justice Stephen Breyer points to Cooper v. Aaron, 358 U.S. 1 (1958), ordering the immediate desegregation of Little Rock's schools, as an example of an “unpopular” decision (p.186) that compelled the elected branches to respect the Court's desegregation rulings.69 The Constitution Project also cites several prominent Warren Court cases as evidence for its claim that “[a] wide array of constitutional and civil rights have been recognized and upheld only because of an independent judiciary.”70 These cases include Engel v. Vitale, 370 U.S. 421 (1962), prohibiting organized prayer in public schools, Gideon v. Wainwright, 372 U.S. 335 (1963), holding that the constitutional right to counsel applies to state prosecutions, Miranda v. Arizona, 384 U.S. 436 (1966), holding that, prior to questioning, police must clearly advise suspects of their rights respecting custodial interrogation, Loving v. Virginia, 388 U.S. 1 (1967), nullifying state statutes prohibiting interracial marriage, and Tinker v. Des Moines Independent Community School District, 393 U.S. 503 (1969), affirming that symbolic speech is protected by the First Amendment.

In these cases some observers see a defiant and independent Court standing up to conservative elected branch preferences. Is it possible that our estimates of the Court's deference to elected branch preferences do not generalize to such cases?

It is possible, but perhaps not very likely. The powers of the elected branches to both discipline and reward the justices are not limited to cases involving constitutional rulings on federal statutes. We looked only at these cases simply because they presented easily measurable outcomes. But we have no obvious reason to think that the judicial deference to elected branch preferences we have observed is limited only to these cases.

Of course, low salience cases that elected branch majorities don't care about likely won't result in elected branch sanctions, no matter what the Court decides. But the cases frequently cited as evidence of the Court's independence would be hard to characterize in these terms. Instead, those convinced of the Court's independence cite cases that drew significant national attention and that had national policy impact. These cases probably drew more attention than did many cases involving constitutional review of federal statutes. It is difficult to imagine that the Court would have deferred to elected branch preferences in the latter cases, but not in the former.

Moreover, the state action cases typically cited to support the thesis of an independent Court tend to fit the pattern we have observed in cases (p.187) involving federal statutes. The string of widely cited Warren Court state action cases starting with Brown, for example, all resulted in what are thought to be liberal rulings by a Court that, as we have seen, was generally predicted to issue liberal rulings even after factoring in the effects of slightly less liberal House majorities. However, we also saw that the Court's liberal rulings on federal statutes appear to have been contingent on the presence of a moderately liberal House: had the House been sufficiendy more conservative, many of these liberal Warren Court rulings are predicted to have been conservative rulings instead. By extension, the famous liberal Warren Court rulings on state action, starting with Brown, were likely also contingent on the presence of a moderately liberal House.

This inference is strengthened by looking at some of the most prominent Rehnquist Court rulings involving state action. In its first eight terms the Court reaffirmed the fundamental right to an abortion established in Roe v. Wade, 410 U.S. 113 (1973) (Webster v. Reproductive Health Services, 490 U.S. 492 [1989]; Hodgson v. Minnesota, 497 U.S. 417 [1990]; Planned Parenthood of Southeastern Pennsylvania v. Casey, 505 U.S. 833 [1992]); struck state actions on the grounds that they violated the robust interpretation of the Establishment Clause articulated in Lemon v. Kurtzman, 403 U.S. 602 (1971) (Allegheny County v. Greater Pittsburgh ACLU, 492 U.S. 573 [1989]; Lee v. Weisman, 505 U.S. 577 [1992]); and ruled that federal judges could order local governments to increase taxes to remedy constitutional violations like race-segregated schools, “one of the major surprises of the [1989] term” (Missouri v. Jenkins, 495 U.S. 33 [1990]). These liberal rulings suggest that the first Rehnquist Court was pulling its punches not only in cases involving federal statutes, but also in cases involving state action.

But relieved of the need to defer to liberal House majorities after the 1994 congressional elections, the Court's rulings appear to have changed not only in the former cases, but also in the latter. In these post-1994 terms the Court backed off on its prior commitments to the Establishment Clause (Rosenberger v. University of Virginia, 515 U.S. 819 [1995]; Agostini v. Felton, 521 U.S. 203 [1997]; Zelman v. Simmons-Harris, 536 U.S. 639 [2002]); school desegregation (Missouri v. Jenkins, 515 U.S. 70 [1995]); and the role of racial composition in drawing congressional districts (Miller v. Johnson, 515 U.S. 900 [1995]). (p.188)

Perhaps most famously, the second Rehnquist Court also reversed the Florida Supreme Court's order of a recount in the contested 2000 presidential election, effectively declaring George W. Bush to be the winner of that election (Bush v. Gore, 531 U.S. 98 [2000]). Many have criticized this decision as a naked assertion of power by the conservative Rehnquist Court majority.71 But, if the evidence reported in this chapter is any guide, it was likely not only the conservative justices' preferences that were driving the outcome in that case. The five justices in the Bush v. Gore majority knew that they could anticipate an incoming Republican-majority House, a majority that surely would not seek to punish the justices for halting the Florida recount. Indeed, that incoming House majority presumably shared the conservative justices' apparent preference to end the uncertainty over Bush's election. But it is difficult to imagine that the Rehnquist Court would have been as willing to interfere in the Florida recount process had it instead faced an incoming liberal Democratic majority in the House. That counterfactual majority would have possessed not only divergent preferences over the Florida recount, but also the means to punish a Court that did not defer to those preferences. In other words, Bush v. Gore, along with other prominent conservative rulings from the second Rehnquist Court, was perhaps just as much a product of a conservative Republican House of Representatives, as of a conservative Court.

Of course, this narrative account of the Court's likely rulings in cases involving constitutional review of state action should perhaps be given little weight. There are no quantitative results reported here making use of the Court's judgments in such cases. It seems reasonable that the estimates reported here would extend at least to the Court's prominent rulings involving state action, the rulings we would expect members of Congress to care about. But we have no direct evidence to support that inference.

The Deferential Supreme Court

Existing empirical studies are nearly unanimous in supporting the conventional wisdom that the Supreme Court decides cases independently of elected branch preferences. But it looks as if those studies, far from providing rigorous confirmation of the truth of our beliefs, instead (p.189) may have been inappropriately influencedby the power of those beliefs to shape what we see. Existing studies have almost uniformly relied on a subjectively coded measure of the direction of the Court's judgments, a measure that turns out to reflect not only how the justices actually decide cases, but also our expectations about how they decide cases. These expectations have perhaps permitted our belief in the Court's independence to color the very nature of the data we have used to test the veracity of that belief.

The results reported here may be misleading in different ways. While the judgment measures used here have been objectively coded, our samples have of necessity been relatively small, and limited to a particular kind of case. They have also been limited to the modern Supreme Court between 1953 and 2004. It is possible that the results reported in this chapter don't generalize beyond these limited samples, or this somewhat limited time frame.

But even if these results don't generalize, they would still seem to be worthy of note. In cases involving the constitutional review of federal statutes between 1953 and 2004, the justices of the Supreme Court of the United States appear to have been profoundly influenced by the preferences of majorities in the House of Representatives. The Court's deference to House majorities was sufficiently large over this period that, when the preferences of majorities on the Court diverged from those of majorities in the House, the Court was more likely to issue judgments that conformed to the latter than to the former.

These results suggest that the justices respond to the specific institutional incentives created by the U.S. Constitution. The consistent responsiveness of the justices to the preferences of the median representative, and not to the preferences of the median senator or the president, is particularly telling. The Constitution gives to the House the “sole” power to initiate impeachment investigations of the justices. Deterring these investigations may give the justices particularly compelling incentives to defer to the preferences of the median representative. The Constitution's Origination Clause likewise gives to the House a special role in initiating appropriations bills, a category of bills including those increasing judicial salaries. The justices need House majorities regularly to introduce and support judicial salary increases if their salaries are not to decrease every (p.190) year in real terms; they thus have particular incentives to defer to the preferences of those majorities. Like other holders of public office, the justices appear to respond to the constituency with the power both to remove them from office and to financially reward their performance.

We might wonder, however, whether there aren't other factors that might be driving the justices' apparent responsiveness to these constitutionally specified incentives. Neither the justices nor members of the House of Representatives exist in hermetically sealed isolation from economic, social, and political currents. These currents may drive changes in both the preferences of House majorities, and the justices' choices in cases involving constitutional interpretation. Without identifying and controlling for these broader forces, we may be wrongly attributing the justices' choices in these cases to their deference to the preferences of House majorities, when in fact both the justices and members of the House are simply responding to broader trends. This is the problem to which the next chapter turns.

Notes:

(1) . These predicted probabilities are represented graphically in Figures 3.5 and 3.6. The comparable predicted probabilities of a conservative judgment using the Judicial Common Space preference estimates range from .50 to .67 between the 1986 (p.329) and 2004 terms. The average predicted probability of a conservative judgment between the 1986 and 1993 terms using the JCS estimates is .63, and between the 1994 and 2004 terms is .57.

(2) . See Fischman and Law (2009) for a discussion of revealed preference estimation.

(3) . Lee Epstein, Jack Knight, and Andrew D. Martin, “The Childress Lecture Symposium: The Political (Science) Context of Judging,” 47 Saint Louis University Law Journal 783 (2003)L 812. Epstein et al. note that between 1991 and 2000, seven of nine articles on judicial behavior in the leading political science journal used the Supreme Court Database, while it was used in fifteen of seventeen such articles in the number two journal.

(4) . These studies include Segal (1997), Hansford and Damore (2000), Spriggs and Hanford (2001), Segal and Spaeth (2002), Sala and Spriggs (2004), Martin (2006), and Clark (2011).

(5) . Some have raised isolated concerns about the Supreme Court Database judgment measure. Landes and Posner (2009) object to the Supreme Court Database decision rules for coding judgments in several issue areas but do not question the validity of the Supreme Court Database issue or judgment codes more generally; William M. Landes and Richard A. Posner, “Rational Judicial Behavior: A Statistical Study,” Journal of Legal Analysis 1 (2009): 775. Shapiro (2009) suggests that the Supreme Court Database issue codes may fail to adequately reflect the doctrinal issues involved in a case as understood by legal academics, but does not address the more general problems raised by subjective issue coding; Carolyn Shapiro, “Coding Complexity: Bringing Law to the Empirical Analysis of the Supreme Court,” 60 Hastings Law Journal 477 (2009). Gillman (2003) and Young (2005) suggested problems of circularity in the Supreme Court Database judgment codes, but did not pursue these suggestions; Howard Gillman, “Separating the Wheat from the Chaff in the Supreme Court and the Attitudinal Model Revisited,” 13 Law and Courts 12 (2003); Ernest Young, “Just Blowing Smoke? Politics, Doctrine, and the Federalist Revival after Gonzales v. Raich,” 2005 Supreme Court Review 1. Some researchers cite as justification for their lack of concern about the Supreme Court Database codes the reliability tests performed on a random sample of Supreme Court Database cases from the Warren and Burger Courts, which resulted in relatively high rates of inter-coder reliability (Supreme Court Database Documentation, Appendix I). These reliability tests do not, however, address the issue of validity.

(6) . Although it is possible for a case to be assigned more than one issue code, Harvey and Woodruff (2011) calculate that about 93.7 percent of cases in the database are assigned a single issue code. Anna Harvey and Michael Woodruff, “Confirmation Bias in the United States Supreme Court Judicial Database,” Journal of Law, Economics and Organization (2011), doi: 10.1093/jleo/ewr003.

(7) . Supreme Court Database Documentation, 42.

(8) . The primary coder for the Supreme Court Database, Harold Spaeth, has written, for example, “Not only are [the justices] independent of the other branches, but their lifetime appointment also insulates them from factious electoral pressures … (p.330) Members of the Supreme Court can further their policy goals because they lack electoral or political accountability.” Segal and Spaeth (2002), 21, 92.

(9) . Harvey and Woodruff (2011).

(10) . Poole and Rosenthal (1997).

(11) . Ibid.; Keith Krehbiel, Adam Meirowitz, and Jonathan Woon, “Testing Theories of Law-making,” in Social Choice and Strategic Decisions: Essays in Honor of Jeffrey S. Banks, ed. David Austen-Smith and John Duggan (Berlin: Springer, 2005).

(12) . The Poole-Rosenthal estimates also report estimated bill and status quo locations. Although some recent papers have used the estimated bill location as a measure of policy location, e.g., Sala and Spriggs (2004), these estimates are highly dependent upon model specification in the DW-NOMINATE estimation procedure; Poole and Rosenthal (1997), Krehbiel et al. (2005). Here we will use simply the relative positions of these estimated bill and status quo locations, namely whether the estimated bill location lies to the left or the right of the estimated status quo location.

(13) . There are 309 such challenges in the sample.

(14) . These statutes have been frequently amended. The decision rule used was to first identify the specific section or sections of the statute actually being reviewed by the Court, and then to identify both the original enacting date and all reenactments of or amendments to this section or sections. As long as the challenged language of the statute remained substantially intact through all amendments and/or reenactments, the most recent reenacting or amending Congress was adopted as the enacting Congress. Cases involving multiple statutes were divided into separate observations, one for each statute.

(15) . The House of Representatives is used because there are more nonunanimous roll call votes in the House for these statutes, relative to the Senate.

(16) . This measure could be constructed for 166 challenged statutes during these terms.

(17) . There are 265 statutes reviewed by the Court between 1953 and 2004 enacted by unified partisan Congresses sitting between 1950 and 2004.

(18) . Harvey and Woodruff (2011).

(19) . Coding unified Democratic Congresses as liberal and unified Republican Congresses as conservative, this indicator matches the Poole-Rosenthal dichotomous statute measure in 74 percent of the 133 observations for which both measures are available between 1953 and 2004; an exact binomial test returns a .00 probability of a random association between the two measures.

(20) . The 95 percent confidence intervals are (.00, .13) and (.48, .69), respectively. All predicted probabilities reported in this chapter hold all other variables constant at their actual values. Because these predicted probabilities take into account how all variables in the model actually covary in a dataset, they are more representative estimates of the size of a variable's effect than those generated by the approach of holding all variables constant at their means or other values. Gelman and Hill (2007), 101–103. Probabilities are simulated using the margins command in Stata 12.

(21) . The 95 percent confidence intervals are (.03, .31) and (.42, .64).

(22) . The 95 percent confidence intervals are (.10, .29) and (.49, .75).

(23) . The 95 percent confidence intervals are (.19, .41) and (.37, .78).

(24) . The 95 percent confidence intervals are (.02, .19) and (.38, .68).

(25) . The 95 percent confidence intervals are (.05, .16) and (.58, .90).

(26) . These effects are also significant during the 1954, 1956, and 1957 terms, when more conservative House majorities appear to have increased the likelihood of conservative judgments, relative to the unconstrained likelihood, with at least 90 percent confidence.

(27) . Toobin (2007), 2.

(28) . See, e.g., Morton J. Horwitz, The Warren Court and the Pursuit of Justice (New York: Hill and Wang, 1998), 3; and Cass Sunstein, Radicals in Robes (New York: Basic Books, 2005), 36.

(29) . Revisionist accounts of the Warren Court include Gerald Rosenberg, The Hollow Hope (Chicago: University of Chicago Press, 1991); Lucas A. Powe Jr., The Warren Court and American Politics (Cambridge: Harvard University Press, 2000); Mary Dudziak, Cold War Civil Rights (Princeton: Princeton University Press, 2002); Michael Klarman, From Jim Crow to Civil Rights (New York: Oxford University Press, 2004); and Corinna Barrett Lain, “Countermajoritarian Hero or Zero? Rethinking the Warren Court's Role in the Criminal Procedure Revolution,” 152 University of Pennsylvania Law Review 1451 (2004).

(30) . Seventy-four percent of the votes for passage in the House came from Democrats; 97 percent of the votes opposed came from Republicans.

(31) . New York Times, July 1, 3, September 11, 1988.

(32) . New York Times, September 11, 1988.

(33) . Eighty percent of those supporting the bill in the House were Democrats, while 73 percent of those opposing the bill were Republicans.

(34) . The panel was composed of Reagan appointees Douglas Ginsburg and Lawrence Silberman and Johnson appointee Spottswood Robinson III.

(35) . Donald I. Baker, “The Superior Court Trial Lawyers Case—A Battle on the Frontier Between Politics and Antitrust,” in Antitrust Stories, ed. Eleanor M. Fox and Daniel A. Crane (New York Foundation Press, 2007), 257–286, 258.

(36) . Eighty percent of the votes in the House in support of the 1994 amendments came from Democrats; 67 percent of the votes opposed came from Republicans.

(37) . Sixty-two percent of the votes in favor of the statute came from Democrats in the House, while 51 percent of opposing votes came from Republicans.

(38) . Seventy-seven percent of the bill's support in the House came from Democrats; 62 percent of the votes in opposition came from Republicans.

(39) . Ninety-two percent of the Act's votes in support came from Democrats; 66 percent of the votes opposed came from Republicans.

(40) . Eighty percent of the votes in favor of the Violence Against Women Act came from Democrats; 67 percent of the votes in opposition came from Republicans.

(41) . Sixty-two percent of the votes in support of the Americans With Disabilities Act came from Democrats; 82 percent of the votes in opposition came from Republicans.

(42) . Justice John Paul Stevens, in New York Times, April 27, 1995.

(43) . New York Times, April 28, 1995.

(44) . Jeffrey Toobin, “Chicken Supreme: The Rehnquist Court Is Political in Every Way,” New Yorker, August 14, 1995.

(45) . Jonathan L. Entin, “The New Federalism After United States vs. Lopez: Introduction,” 46 Case Western Reserve Law Review 635 (1995–1996): 636.

(46) . Justice John Paul Stevens, in New York Times, March 28, 1996.

(47) . Justice David Souter, in ibid.

(48) . New York Times, June 26, 1997.

(49) . Robert C. Post and Reva B. Siegel, “Legislative Constitutionalism and Section Five Power: Policentric Interpretation of the Family and Medical Leave Act,” 112 Yale Law Journal 1943 (2003): 1952, 1964.

(50) . Senator Charles Schumer (D-NY), in New York Times, June 28, 1997.

(51) . Yale Professor of Law Paul Gewirtz, in New York Times, July 1, 1997.

(52) . New York Times, July 1, 1997.

(53) . Schroeder (2001), 317.

(54) . New York Times, July 15, 2001.

(55) . Gerhardt (2000), 10980.

(56) . House Republicans had contributed 64 percent of the votes in favor of the Act, while House Democrats had contributed 65 percent of the votes in opposition.

(57) . See Krehbiel (1998).

(58) . Testing for nonlinear relationships between statute age and the probability of a strike did not produce significant improvements over the linear measure reported in Tables 5.5 and 5.6.

(59) . Using the Bailey XTI preference estimates, the smallest distance between an enacted statute and the most preferred rule of the median justice is observed for statutes enacted in 1970 and reviewed during the 1971 term; the largest observed distance is for statutes enacted in 1952 and reviewed in the 1963 term. Using the JCS preference estimates, the smallest such distance is observed for statutes enacted in 1999 and reviewed in the Court's 1999 term; the largest such distance is observed for statutes enacted in 1954 and reviewed in the 1967 term.

(60) . Using the Bailey XTI preference estimates, the smallest distance between an enacted statute and the most preferred rule of the median representative is observed for statutes enacted in 1987 and reviewed during the 1988 term; the largest observed distance is for statutes enacted in 1976 and reviewed in the 1995 term. Using the JCS preference estimates, the smallest such distance is observed for statutes enacted in 1989 and reviewed in the Court's 1989 term; the largest such distance is observed for statutes enacted in 1976 and reviewed in the 1995 term.

(61) . These simulations set statute age at two years and hold all other variables at sample values. The two predictions are distinct with at least 90 percent confidence.

(62) . Statute age is held at seven years and all other variables are held at sample values.

(63) . Statute age is held at seven years and all other variables are held at sample values.

(64) . Statute age is held at two years and all other variables are held at sample values.

(65) . Preseault v. ICC 494 U.S. 1, 17–18 (quoting the statute).

(66) . These simulated probabilities hold the statute's age at six years and all other variables at their sample values. The predictions are distinct with at least 90 percent confidence.

(67) . Statute age is held at zero years; all other variables are held at sample values.

(68) . Freedom's Law: The Moral Reading of the American Constitution (Cambridge: Harvard University Press, 1996), 12–13. The justices in Griswold were not united in identifying the constitutional source of the privacy right endorsed in that case.

(69) . “Serving America's Best Interests,” Daedalus (Fall 2008): 139–143, 142. Cooper v. Aaron is also cited as a prominent example of the Court's independence by Saikrishna Prakash, “America's Aristocracy,” 109 Yale Law Journal 541 (1999): 575.

(70) . “The Newsroom Guide to Judicial Independence,” The Constitution Project, http://www.constitutionproject.org/pdf/37.pdf.

(71) . See the comments of legal academics in New York Times, December 13, 2000. Also see Alan Dershowitz, Supreme Injustice (New York: Oxford University Press, 2001).