Law Review
Northwestern Law
Northwestern University Law Review : Colloquy : 2007 : Farnsworth

The Use and Limits of Martin-Quinn Scores to Assess Supreme Court Justices, with Special Attention to the Problem of Ideological Drift

By Ward Farnsworth[*] [pdf]

[Editor's Note:  This piece is a response to a forthcoming article in the Northwestern University Law Review by Lee Epstein, Andrew D. Martin, Kevin M. Quinn and Jeffrey A. Segal, titled Ideological Drift on the Supreme Court:  Who, When and How Important?, 101 Nw. U. L. Rev. (forthcoming 2007).  We posted an introduction to the piece by Epstein et al., 101 Nw. U. L. Rev. Colloquy 127 (2007), (link).]

In their new paper, Lee Epstein, Jeffrey Segal, Andrew Martin, and Kevin Quinn investigate changes in behavior by Supreme Court Justices.[1]  They conclude that the policy preferences of most Justices change during their careers, and suggest that this should cause Presidents to reconsider the use of nominations to try to change the direction of the Court.[2]  I find the authors' evidence and analysis interesting, but am not yet convinced that any rethinking is in order by the people who pick Justices or care about their selection.  I will begin with a general discussion of the model—the Martin-Quinn scores—that the authors use to generate their findings.  It is an ingenious method that is attracting some wider interest,[3] but its basis and workings have not yet been presented in a non-technical fashion that is likely to be understood well by a legal audience.  One goal of this Essay is to explain it in lay terms.  Then I will consider the particular claims the authors make and, finally, their more general thesis about the predictability of behavior by Supreme Court Justices.  My two conclusions, in short, are that the authors have not proven that consequential surprises in the Justices' behavior are more common than has been generally supposed; and that the authors' advice to Presidents (and others interested in the selection of Justices) is premature, because the behavior of some Justices is more predictable than that of others, and this itself can often be predicted by asking how firmly the Justices demonstrated their views before joining the Court.


The authors claim to show that the policy preferences of most Supreme Court Justices change while they are on the Court.  The force and importance of this result depend, of course, on the strength of the method used to produce it.  Their article relies on what they call "Martin-Quinn" scores for the Justices, but where those scores come from and just what they mean are not explained; the reader is referred to other articles for discussion of those issues.[4]  Referring the reader elsewhere is perfectly all right in principle, of course, but in this case the explanations in those other articles turn out to be hard to follow—or at least hard for me to follow.  The fault no doubt is mine, as I am all thumbs with mathematics; then again, that fault is widely shared among the apparent audience for the authors' new article:  lawyers and others arguing about who ought to be put onto the Supreme Court.  For them, a clearer account in words of how the authors reached their conclusions is much needed.  Otherwise there is a risk that readers will ignore the authors' findings because they don't understand them, or accept the authors' conclusions on faith because the findings have emerged from a black box that is mysterious but looks impressive.  I mean no disrespect to the authors or to their ability to explain themselves.  I just view it as a case where they speak a slightly different language than some members of their audience.

In hopes of making the operation of the authors' methods more transparent to the non-mathematician, then, I read the prior work of Professors Martin and Quinn in which they present their methods; more importantly, they were then so kind, patient, and collegial as to let me to ask them many questions, for which I hereby record my thanks.  I will now explain my understanding of their method as plainly as I can.  I apologize in advance to anyone who finds that I am belaboring the obvious, and to Martin and Quinn for any inaccuracies I commit in the course of simplifying and explaining their approach.

The Martin-Quinn method keeps track of only one thing:  whether a Justice voted to affirm or reverse in a case.[5]  The method does not pay attention to what the case was about; the method itself has nothing to do with politics or ideology (or, for that matter, law).  All it knows are things like this:  in the first case decided last year, Justices A, B, C, and D voted to affirm and Justices E, F, G, H, and I voted to reverse.  In the second case last year, Justice A voted to affirm and all the others voted to reverse.  And so forth for every case fed into the model, nothing more.  The authors' findings are all derived from analysis of that data.[6]

The model studies all the cases it is fed—all the patterns of voting, all the coalitions in the cases as just described—and then assigns a number to each Justice.  You can plug those numbers into a formula and ask the formula to generate the votes you would expect the numbers to produce.  Essentially you say to the formula:  "Assume that every case the Court hears presents a choice between the same two poles (putting aside the question of what those poles are); and assume that these numbers reflect each Justice's tendency to vote either way—their preferences between the poles.  Now imagine that they all vote on lots of cases, and show me what sorts of voting patterns and coalitions you would expect to emerge from the collisions of those numbers."  The formula then generates a bunch of patterns.  If the numbers for each Justice are well-chosen, those patterns produced by the formula will look like the voting patterns we see in real life.  The formula will predict, based on the numbers it was given for each Justice, that there will be a certain number of cases where these five Justices will vote one way and those four Justices will vote another; a certain number where this one Justice votes alone and the other eight vote the other way; and so on.  We can, of course, compare the patterns produced by the formula to the actual patterns of voting in any given time period; and if we do, we find (or at any rate the authors report) that their numbers are about 75% accurate.  In other words, over a large run of cases, the scores they give to each Justice can be used to generate voting patterns that match the real patterns about 75% to 80% of the time.[7]  Notice that none of this so far need have anything to do with the Justices' policy preferences.  Nor can it be tested by (or used to make) predictions about the outcomes of individual cases.

So we have a set of numbers for the Justices:  numbers that do the best possible job, when put into a formula, of generating the patterns of coalition that the Justices produce when they decide cases.  If you arrange the numbers in a line, you find that Clarence Thomas and Antonin Scalia (who vote together a lot) are toward one end of the line, and that Stephen Breyer and Ruth Bader Ginsburg (who also vote together a lot, but don't vote often with Scalia and Thomas) are toward the other end of the line.[8]  Martin and Quinn (and then, in this most recent paper, their co-authors) then go another step:  they assume that every case before the Court presents a choice between a liberal vote and a conservative vote and that every Justice votes according to his preference for liberal or conservative outcomes.  They then assume that the numbers produced by their system match those political preferences.  So whatever number Justice Scalia gets in their system not only is the best number for helping to produce accurate simulations of the Court's voting patterns, but also is a statement of his preference for conservative outcomes.  That number is called his ideal point:  a number that shows where, on a political scale, a Justice would like policy to be made.[9]

Ideal points have been devised by various researchers using different methods.[10]  Most of the methods involve looking directly at evidence thought to reveal a Justice's preferences, as by examining votes in various kinds of cases.  A novelty of Martin and Quinn's approach is that it tries to generate ideological ideal points—a measure of where a Justice falls on the ideological spectrum—without ever looking at ideology at all.  This is a strength of their method because it avoids certain kinds of circularity that can arise from other methods;[11] it also may be a liability for reasons discussed later.[12]  For now just observe that the authors' interpretation of the formula's results—their reading of it as an array of policy preferences—is not a product of the formula.  It is an assumption by the people who made the formula.  By itself the formula does not prove the assumption is true, or suggest that it is true, or comment on its truth.  The formula just produces numbers—numbers that produce somewhat accurate pictures of how the Justices arrange themselves when they vote.  True, the formula's accuracy shows that each Justice predictably votes with certain colleagues and not others; but everyone already knows that, and there are various ways to explain it besides the "attitudinal" way—the assumption that it's all politics.  For all the model shows, the Justices could have similar politics but different views about interpretation or other such differences.  It also is true that the array of Justices produced by the model resembles—not always precisely, but noticeably—the array you would get if you tried to arrange the Justices according to conventional impressions of their politics or other measures of their ideal points.[13]  But that doesn't necessarily mean that the model is measuring the politics of the Justices or that it isn't measuring other things.  The positions of the Justices on the spectrum the authors present—and the poles at either end of that spectrum—could reflect combinations of policy preferences, interpretive approaches, judicial philosophies, and perhaps other qualities.  My own view, expressed elsewhere and likewise based partly on statistical analysis,[14] is that ideology does play a major role in the decisions of cases; but for now I just want to clarify what the formulas used by the authors can and can't establish.  The attitudinal view is something the authors assume when they use their model, not something they prove with it.

I have tried to clarify what the authors' model knows and doesn't know, and what it therefore shows and doesn't show.  At a couple of points I have stressed the negative side of these points only because that is where I think the greatest danger lies:  it is easy to be confused into thinking the model proves things that it does not.  It is especially easy in this paper because the authors point to their model and then announce their conclusions without always being clear about what contributions are made by the assumptions behind the model and their own assumptions about the meaning of the numbers the model produces.  I do not blame them for the lack of precision; they probably thought it wasn't needed because they were counting on their readers to already understand the model and its limits.  I just think that assumption is less safe than they do.  As some evidence for my view, Linda Greenhouse writes in her comment that "Epstein and her colleagues prove conclusively that preference change, at least among long-serving Justices, is the rule rather than the exception."[15]  But that isn't what Epstein and her colleagues prove.  They prove—or anyway their evidence suggests—that behavioral change of some sort is the rule rather than the exception.  The idea that this happens because the Justices' preferences change is their supposition, not what they prove; the apparatus they use in this project is incapable of proving it.  This distinction isn't crucial for at least one version of their thesis, which just involves change by Justices; for if judicial behavior does change, then whether this results from changed preferences or from changes in anything else does not affect the bottom line:  their behavior does change.  But Ms. Greenhouse's account reinforces my worries about the need for clarity, and helps explain why I am going to some pains to explain the authors' model to the extent that I can.

In any event, to compensate for this emphasis so far on the limitations of the model, let me now pause to congratulate Professors Martin and Quinn on their achievement.  It is an impressive thing—original, elegant, and praiseworthy—to explain a large share of the Justices' voting patterns as outcomes of their arrangement along a single spectrum, whatever the spectrum may mean.  I have doubts about whether the authors prove as much as they think they have in this new paper, but those doubts do not at all stop me from greatly admiring the ingenuity and promise of their model.


We now can turn to implications for the project at hand:  figuring out how much Justices change while they are on the Court.  Epstein et al. find that the ideal points they generate for most Justices—the numbers, in other words, that, when entered into the formulas, best account for the patterns into which they form themselves when they vote—don't stay the same over the years.  This leads the authors to state a finding and an implication.  The finding is that most Justices drift ideologically:  their preferences change.  The implication, as I understand it, is that Presidents have less ability than previously thought to affect the Court in the long run.  Let me offer some reservations about these claims, starting with the finding that preferences change.

First, Epstein et al. assume, as we have seen, (a) that every case before the Court presents a simple choice between a conservative and liberal outcome, and (b) that the Justices vote according to their preferences for those two sorts of outcomes.  As we have seen, the Martin-Quinn model itself doesn't assume that; the model just assumes that the Justices' behavior can be plotted along one dimension, and the one dimension could be anything, or some package of things.  But when they put the model to use in this project, they assume the poles of the spectrum, and the stakes of every case, are liberalism and conservatism.  Of course the authors know that the world isn't quite that simple, but they evidently believe that any complexities beyond that picture aren't important enough to affect their conclusions.  They might be wrong.  For present purposes I will put aside the point discussed a moment ago—that the model the authors use might measure things besides preferences, and that maybe those other things are changing instead.  As already noted, while that may be true, it wouldn't affect the authors' basic claim about the risk that Justices do change (it just questions what about them is changing).  But I do wish to raise a few other problems.

I will start with two concerns of uncertain magnitude.  The first is that if the Justices' preferences are more complicated than just liberal vs. conservative, their behavior might change without changes in their preferences or changes in anything else about themselves.  Their behavior might instead change because the cases change and the Justices don't all view the politics at stake in them in the same way.  Suppose, to take a simplified example, that Justice Kennedy tends to vote for the government in cases involving criminal procedure, but against the government in cases involving free speech, while Justice Rehnquist—a less libertarian sort of conservative—tends to vote for the government in both situations.  (Both assumptions happen to be accurate.)  Imagine that in term T, there are many criminal procedure cases (where the two Justices vote the same way) and few speech cases (where they don't).  Then in term T + 1, there are lots of free speech cases.  Kennedy's preferences may appear to drift to the left relative to Rehnquist's when they haven't really changed at all.

The authors try to account for this by not just measuring each Justice's ideal point term-by-term; for any given term, rather, they also take into account the Justice's behavior in other terms, with decreasing weight as the other terms get farther away.  But whether this successfully erases the effects of the complexity just described is unknown and may be unknowable; for the extent of the complexity itself is not known.  Even if all that the Justices care about is making policy (which is bound to be an overstatement), the content of those policy preferences probably is complicated—somewhat different from one area of law to the next, and different in the way each Justice thinks it applies to given cases within an area.  If those differences are significant, they foul up the authors' model to an unknown degree; they may produce an impression that preferences are moving around when other things are moving around.

A second worry is that a Justice might appear to change because the people around him have changed; a Justice who stands still while the rest of the court moves to the right (either because some of his colleagues drift or, more probably, because they are replaced) may end up with a different ideal point and an appearance of having changed even if he hasn't.  Again, the authors make an effort to compensate for this by assuming that inertia bears on each Justice.  When a new Justice is added to the Court and differs from his predecessor, for example, voting patterns obviously may change as a result.  The authors react by setting their formulas to assume that Justices generally stay the same and find the way to account for the new patterns that is least disruptive to that assumption:  the formula blames the changes on the newcomer.  But if a Justice goes from usually being in the majority to usually being in the minority, it seems to me that this can't help but eventually affect the ideal point that the formulas generate for him.  It will change his relative position among his colleagues; the changes by others may also affect the cases the Court decides to hear, which in turn can have feedback effects of the kind discussed a moment ago (the changed diet of cases may produce different voting behavior without a change in preferences).  Or the rest of the Court may move when replacements arrive, while the views of a sitting Justice also are changing at the same time (John Paul Stevens may be a good example of this),[16] thus making it hard for the model to detect the influence of each force.

I raise the worries in the last two paragraphs with diffidence because I am not competent to assess the authors' equations and to address the degree of the distortion that could be caused by them.  In my discussions with Professor Quinn, however, he has acknowledged that they are real risks.  He doubts that they are serious in practice; he particularly thinks my example involving Kennedy and Rehnquist is unlikely to be a problem, because the ideal points of the Justices will be based on lots of cases where they don't vote together.  In any event, the precise limits of the Martin-Quinn model—the conditions under which the model, or the authors' interpretation of it, has the potential to fail or mislead because the assumptions made by the model or its users turn out to be too strong—have not yet been carefully and clearly explained.  Perhaps the isolation of these risks and a more exact statement of their size and nature can be the subject of future work by the authors.  Until that work is done, I think their readers are entitled—and probably are prudently advised—to show some circumspection before basing practical decisions on the authors' claims.  This caveat can be disregarded, of course, by readers able to derive precise answers to these questions for themselves based on the authors' published equations.  I suspect that such people are few in number, particularly among readers of law journals, but that is merely a guess.  I certainly encourage anyone interested to give it a try.

Meanwhile, a clearer worry about the authors' model involves measurement of the magnitude of a Justice's change.  There is an important difference between a statistically significant change in a Justice's position within the authors' model and a change in a Justice's behavior that is likely to be significant to the people who supported and opposed his nomination.  A first reason is that the model used in the authors' paper treats every case as equally important, but real people don't view the Court's work that way.  This discrepancy may have important practical consequences.  It means that a Justice can appear to move on the Martin-Quinn scale by, for example, starting to vote for the government somewhat more or less often in an area that isn't of much general public interest (a distinct possibility; a substantial majority of the Court's work is of little general public interest).  But that doesn't mean the Justice moved in a sense that would have been important to those considering his appointment in the first place.  So if, as the authors claim, Justice Rehnquist moved to the left during his career, few of those who supported him—and, for that matter, few who opposed him—will care very much.  They know he mostly held the line on the fronts that were of greatest importance to them.  The authors' model nevertheless makes Rehnquist look like a case study in change.  In one sense the model may be right:  perhaps he did change, even if in ways that most people didn't perceive, and this may be of interest to academic students of the Court.  But in another sense the model is wrong:  he didn't change in ways that, if foreseen, would likely have had an effect on anyone's view of his initial appointment.[17]  A problem with the authors' argument is that it moves from a possible demonstration of change in the first sense to claims about change in the second sense; it says that since Justices change in ways their model detects, Presidents and others should change the way they think about whether their expectations will be fulfilled by Supreme Court nominees.  This doesn't follow, because significance within the authors' model is not the same as significance to their intended audience.

Further examples of the same point can be found by considering Justices who become more of whatever they were expected to be—the possible movement of Justice Scalia to the right, for example, or of Justice Ginsburg to the left.  Nobody much cares, as the authors more or less acknowledge;[18] to be more precise, the movements those Justices have undergone would (if foreseen) probably not have had much effect on anyone's opinion at the time they were nominated.  Then there is Chief Justice Burger, whom the authors cite as yet another study in changed preferences.  They say he went from having quite conservative preferences to having slightly less conservative preferences—and then back to having quite conservative preferences again.[19]  It is not impossible that Warren Burger's policy preferences did perform a back-and-forth odyssey of this sort, though I find myself attracted to other explanations for movement in his voting patterns (such as the ones already sketched).  In any event, though, even if his preferences did move around in this strange way, it is hard to imagine that anyone's view of his appointment would have been changed by foreknowledge of it.  When you remove from the authors' results all the changes in judicial behavior that were of little consequence in the senses just described, the findings they announce in their paper become less striking.  Most of the changers who remain are already well-known to have changed in ways that mattered.

Second, the relationship between the spectrum generated by the authors' model and the spectrum of policy decisions in the real world is a matter of guesswork.  There is no inherent relationship between them.  Here, in effect, is the sequence:  the Justices are placed on a scale for the purpose of generating predictions about their voting patterns and the coalitions they will form.  Then along comes an analyst who notices that the scale puts Justices usually thought to be conservatives at one end and liberals at the other; he suggests that we therefore treat this as a liberal-conservative scale of their policy preferences.  Maybe that jump is fair and maybe it isn't, as discussed earlier—but assume that it is.  Still, while it's then possible to say what a statistically significantly change looks like with respect to the Justices' voting behavior (the coalitions they form, etc.), it isn't possible to convert that reasoning into a statement about when changes of that kind turn into policy differences of a significant size.  In other words, there is no mathematical or otherwise scientific way to take the changes the authors are measuring and say how large they are in familiar political terms or how much they are likely to matter to anyone.

The authors perceive these problems, I think; they address them by choosing their language carefully.  In their conclusion, for example, they say this of ideological drift:  "In some instances, the movement may be relatively inconsequential, but in others substantial doctrinal change may result."[20]  Well, yes, of course; and for practical purposes everything depends precisely on whether and how often the changes are consequential, which the authors aren't in a position to say because their model isn't built to address that point.  It can address the general likelihood that a given Justice at a given time will cast tiebreaking votes,[21] but that's a different issue.  I'm using a definition of consequential that might sometimes overlap with that one but is distinct:  whether the surprises in a Justice's behavior would have materially affected anyone's position on his nomination if they had seen them coming.  I select that definition and adhere to it because it is the pertinent one for purposes of the authors' main payoff:  a set of points for political actors to consider ex ante when thinking about who should be put on the Supreme Court and estimating their likelihood of ideological disappointment.


Last, I want to step back and take a more general view of the authors' project, their thesis, and their advice to Presidents and others in the business of picking Justices.  Here I meet a difficulty arising not from the authors' model but from their rhetorical approach.  They initially create the impression (at least they did for me) that they will be challenging a "strong consensus"[22] about the stability of the Justices' preferences, and advancing a claim "contrary to the received wisdom."[23]  But when they state what might be considered a clear thesis, it is, for better or worse, very mild: 

[T]hose believing that they can entrench their views in the Court for the decades to come are occasionally mistaken. . . .  In turn, because these political actors cannot always accurately predict the future, our results may counsel against ideological appointments—at the least, ideological appointments to the neglect of other factors, especially a nominee's qualifications and his or her ability to advance electoral goals.[24]

The hedge words in these claims—"occasionally," "cannot always," "may counsel," "at the least," "to the neglect"—shield them from objection but also drain them of novelty.  I don't know of anyone who would have started reading the article with a view contrary to the one just excerpted; I would have described it as the received wisdom.  This need not be a criticism.  Confirming received wisdom can be important, though I do not think that is what the authors believe they are doing—and perhaps elsewhere it isn't.  But fussing about how novel a claim the authors made, or meant to make, is of no great interest, so let me instead offer a slightly broader view of the current situation and then consider what light the authors' findings shed on it.

Everyone knows that some Justices change significantly during their careers and that this can cause important changes in the law.  The important questions involve (a) how often this happens, (b) whether it can be predicted, and (c) how Presidents should deal with the risk of it.  I have explained my difficulties with the authors' claims about issue (a):  their model is good at showing that judicial behavior changes rather than stays still, but it is not as good at showing whether or how often those changes would have been considered important by those considering the appointment in the first place.  But a different problem appears when they then offer advice to Presidents or others thinking about the selection of Justices.  They skip from (a) to (c) without considering (b); yet (b) is crucial to (c).  In other words, how Presidents think about the possibility of putting an ideological stamp on the Court depends not just on how predictable Justices are generally, but also on how well their unpredictability can be predicted.

The authors speak as though predictability were monolithic:  either the behavior of Justices can be predicted or it can't be.  They don't pay enough attention to whether Justices might differ in how predictable they will be.  My own view is that they obviously do differ in that way.  If a Republican President wants to pick a Justice who will provide many long years of ideological satisfaction, he picks someone who has earned his ideological spurs by doing time in the federal executive branch while it was under control of the Republican party.  That sort of nominee has shown ideological commitment in one of the so-called political branches, and the time spent in Washington will have given him plenty of chances to be scrutinized for reliability by other people with those same ideological commitments.  Rehnquist, Scalia, Thomas, Roberts, and Alito all fit this mold.  The first three have not changed in consequential ways (though as we have seen, two of them are counted as studies in change by Epstein et al.).  The latter two probably won't go through consequential changes, either, though their behavior, too, may slide around enough to move the needle on the authors' seismograph.

A President who can't get an ideologically reliable nominee through the confirmation process because he is politically weak, or who has other priorities, may resort to outsiders who haven't demonstrated their ideology in the same way:  someone like Stevens, Blackmun, O'Connor, Kennedy, or Souter, all of whom have been ideological disappointments to many of their original supporters.  That is simply a cost of departing from the path of choosing someone with a more clearly proven ideology.  Sometimes Presidents choose to incur those costs because they care more about other things; sometimes they have those costs forced upon them by circumstance.

Needless to say, there are no guarantees provided by nominees of either of these types.  Anyone can change, and perhaps it's even likely that everyone will change over the course of an adult lifetime in ways detectable by statistical scrutiny.  But these two types of nominees carry with them very different probabilities of actual disappointment to those who choose and support them.  I think this point has been well-understood for a long time by people involved in the nomination process, and now Professor Dorf has demonstrated it more carefully and at some length.[25]  I do not know whether it has always been true or always will be; possibly a different story could be told about the predictability of Justices at earlier points in the country's history or at early points in the authors' study set, which goes back to 1937.  But political sociology changes just as people do, so trying to get traction on these questions today—perhaps to advise a current President—based on the calculations that worked or didn't work in the 1940s seems unlikely to be very fruitful.

What the new article from Epstein et al. adds is the knowledge that, to a degree of practical importance that is uncertain, the typical Supreme Court Justice won't behave identically throughout a full career.  That's interesting to me as an academic student of the Court; it makes their paper a valuable contribution.  The difficulty arises because the authors want to do more than make this contribution; they want to give advice.  If I were advising a President, however, I would tell him not to let these findings distract him from the points sketched a moment ago.  For that is my greatest concern about the authors' work:  that some officials or members of the public might be unduly distracted by the claims of Epstein et al., or perhaps by a misreading of them.  For despite the hedge words I mentioned a moment ago, it would be easy to take away from their article a conclusion that it is generally hard to predict what any Justice will think and do ten or fifteen years after appointment.  Indeed, an excellent strategy for a Machiavellian right-winger (or, mutatis mutandis, a Machiavellian left-winger) would be to interpret the article as I just suggested, praise it to the skies, and do whatever can be done to get it discussed in The New York Times.  This could have two good consequences for the hypothetical conservative we are imagining.  It might cause Democrats to take less interest in the ideologies of nominees they themselves put forward, and it might cause Democrats to worry less about the ideologies of the nominees that Republicans put forward.  Then next year, if Michael Luttig is nominated to the Supreme Court, our Machiavellian friend can point to the famous findings of Epstein et al. to reassure everyone on the left that they should relax:  you really can't predict with much confidence what a Justice will be like ten years after appointment.  As the authors have shown, "virtually every" Justice changes during his career;[26] there's every chance that Mike Luttig will, too.  (Snicker.)


*. Professor and Nancy Barton Scholar, Boston University School of Law.  Many thanks to Andrew Martin and Kevin Quinn for discussing their work with me at some length, and to Quinn for comments on an earlier version of this paper; thanks as well to Andrew Kull for comments and discussion.

1. Lee Epstein et al., Ideological Drift Among Supreme Court Justices: Who, When, and How Important?, 101 Nw. U. L. Rev. (forthcoming 2007) (link).

2. Id. (manuscript at 46, on file with the Colloquy).

3. Id. (manuscript at n.85).

4. Id.  The references to which Epstein et al. refer the readers are Andrew D. Martin & Kevin M. Quinn, Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953—1999, 10 Pol. Analysis 134 (2002) (link) and Andrew D. Martin et al., The Median Justice on the United States Supreme Court, 83 N.C. L. Rev. 1275 (2005).

5. See Martin & Quinn, supra note 4, at 137; Martin et al., supra note 4, at 1297 & n.75.

6. Epstein et al., supra note 1 (manuscript at 20—21).

7. See Martin & Quinn, supra note 4, at 150.

8. See Epstein et al., supra note 1 (manuscript at 22, 25, 29) (figs. 5 (Ginsburg), 7 (Scalia), 12 (Breyer and Thomas), respectively); see also Martin & Quinn, supra note 4, at 146 (table 1).

9. See Epstein et al., supra note 1 (manuscript at 20).

10. See, e.g., Jeffrey A. Segal & Albert D. Cover, Ideological Values and the Votes of U.S. Supreme Court Justices, 83 Am. Pol. Sci. Rev. 557 (1989); Glendon Schubert, The Judicial Mind Revisited: Psychometric Analysis of Supreme Court Ideology (1974).

11. See Martin et al., supra note 4, at 1307.

12. See discussion infra at Sec. II.

13. See Martin & Quinn, supra note 4, at 145 (noting the "prima facie sensible" plausibility of the sorting produced by their model:  Marshall, Brennan, and Warren to one side; and Burger, Rehnquist, and Scalia to the other); id. at 146 (comparing results with other measures of conservative and liberal voting).

14. See Ward Farnsworth, Signatures of Ideology: The Case of the Supreme Court's Criminal Docket, 104 Mich. L. Rev. 67 (2005) (link); see also Ward Farnsworth, The Role of Law in Close Cases:  Some Evidence from the Federal Courts of Appeals, 86 B. U. L. Rev. 1083 (2007).

15. Linda Greenhouse, Justices Who Change:  A Response to Epstein et al., 101 Nw. U. L. Rev. (forthcoming 2007), 101 Nw. U. L. Rev. Colloquy 132, 133 (2007), /lawreview/colloquy/2007/9/ (link) (citations infra refer to the Colloquy).

16. See Ward Farnsworth, Realism, Pragmatism, and John Paul Stevens, in Rehnquist Justice: Understanding the Court Dynamic 157, 157—58 (Earl M. Maltz, ed., 2003).

17. For an overview of Rehnquist's appointment and performance, see Keith E. Whittington, William H. Rehnquist:  Nixon's Strict Constructionist, Reagan's Chief Justice, in Rehnquist Justice, supra note 16, at 8.

18. See Epstein et al., supra note 1 (manuscript at 35) ("[T]o be sure, some Presidents would not have objected to the drift exhibited by their Justices.  Scalia is a prime example.").

19. Id. (manuscript at 28—29).

20. Id. (manuscript at 46).

21. See Martin et al., supra note 4, at 1304—1307.

22. Epstein et al., supra note 1 (manuscript at 10).

23. Id. (manuscript at 1).

24. Id. (manuscript at 36) (footnotes omitted) (emphasis in original).

25. Michael C. Dorf, Does Federal Executive Branch Experience Explain Why Some Republican Supreme Court Justices 'Evolve' and Others Don't?, 1 Harv. L. & Pol'y Rev. (forthcoming Spring 2007), available at (link).

26. See Epstein et al., supra note 1 (manuscript at 4).


Copyright 2007 Northwestern University

Cite As:  101 Nw. U. L. Rev. Colloquy 143 (2007),

Persistent URL:

(Download PDF)


(back to top)