Monday, March 25, 2013

SCOTUS SSM Bracketology

By Mike Dorf

If either team had an equal chance of winning each game in the NCAA basketball tournament, the odds  against filling out a perfect bracket would be one out of 2 to the 63rd power, or roughly speaking, a 9 followed by 18 zeroes. (I'm ignoring the play-in games.  If you count them, discount your odds by a further factor of 16).  What are the odds of filling out a perfect "bracket" for the Supreme Court's decisions in the same-sex marriage cases to be argued this week?  That depends on what we mean by a successful prediction, so let me make some simplifying assumptions.

There are nine justices and two cases--Perry (the Prop 8 case) and Windsor (the DOMA case).  In each case, there are arguably infinitely many possible outcomes, but we can group these outcomes into four basic possibilities: (1) No ruling on the merits because executive non-defense means no case or controversy; (2) Plaintiffs win "big" in the sense that the Court finds that laws barring SSM are unconstitutional; (3) Plaintiffs win "small" in the sense that Prop 8 is invalid in California (perhaps because, following the 9th Circuit, the decision to take away marriage was rooted in animus) but the Court does not say that other states must recognize SSM, and, in Windsor, the Court relies on some DOMA-specific ground like federalism or the animus that motivated Congress; or (4) the challenges are rejected on the merits.  If we assume that each Justice must choose among each of these options for each case, then a single Justice has 16 possibilities (1,2,3 or 4 in each case), and so the total number of possibilities for the Court as a whole is 16 to the 9th power or, equivalently (since 16 is 2 to the 4th power), 2 to the 36th power, or roughly speaking, 1 in 69 billion.  Those seem like slim odds, but not compared with the odds of filling out a perfect NCAA bracket.  You are about 134 million times more likely to get the SCOTUS prediction right than you are to fill out a perfect NCAA bracket.

But wait, there's more.  Your odds of filling out a perfect NCAA bracket are not quite as bad as mere chance would predict because you can predict most games more accurately than a coin toss would by looking at seeding, won-loss record, etc.  Sure, sometimes Florida Gulf Coast and Wichita State unexpectedly make the Sweet 16, but you still have the odds somewhat in your favor on any given pick.  So you might think that the difference isn't so stark.  Until you consider how much less random the Supreme Court is.

For one thing, there are correlations among the options that you don't see in basketball.  Harvard beating New Mexico did not make it any more or less likely that Syracuse would win its opening game.  But if you know that Justice Alito voted to find no justiciability in Perry, that increases the odds that Chief Justice Roberts voted the same way.  More broadly, for any particular Justice, the choice of option (2) (plaintiffs win big) in Perry will be highly correlated with option (2) (plaintiffs win big) in Windsor.

Likewise, just as it was very unlikely that Indiana was going to lose to James Madison, so it is very unlikely that Justice Thomas is going to vote for option (2) in either case.  (Insert James Madison/originalism joke here.)  Indeed, on the whole, the ex ante predictability of the SCOTUS is probably understated by the factor of 134 million times easier than picking a perfect bracket in the NCAA which you get by assuming randomness for both.  Put qualitiatively, NCAA basketball outcomes are more random than how any particular Justice will vote in the SSM case or any other case.

Just how predictable are the outcomes of Supreme Court cases?  Back in 2002, a group of scholars at Washington University Law School compared the results of predictions made by panels of experts with those made by a computer program that used a modified version of the so-called attitudinal model (which codes cases based on ideological factors).  The computer generally beat the experts but not by all that much, and the overall pattern showed a fair degree of predictability, certainly much better than chance for any given Justice in any given case.  (I published my take on the design and results of the forecasting project here.)

I've been asked by a number of reporters and others to predict the outcomes in Perry and Windsor, and so I have done so, but I don't claim to have any special insight.  Nonetheless, it's clear to me that people are interested in such things, so I'll take a crack at it in general terms.  I think that the most likely outcome in each case is a what I've called option (3)--a narrow win for plaintiffs that doesn't address the question of whether all state laws denying the right to SSM are invalid.  If the Court had only Windsor before it, I would predict that outcome with considerable confidence, because the federal interest in banning SSM from a state that has SSM is quite weak (in light of federal law's acceptance of state-by-state variations in the definition of marriage otherwise).

I do not make that prediction with great confidence, however, because I think the argument for option (3) in Perry is also weak.  Notwithstanding Judge Reinhardt's heroic efforts in the 9th Circuit to come up with a California-only rationale for invalidating Prop 8, I think it will be hard to persuade five Justices that it's unconstitutional to recognize then unrecognize SSM but permissible never to recognize SSM in the first place.  I understand and can make the Romer-based argument along these lines, but it strikes me that some number of Justices will think that the 9th Circuit rule discourages states from granting marriage rights to same-sex couples because doing so will be treated as a one-way ratchet.  If option (3) is effectively off the table in Perry, that makes option (2) more likely in Perry, and if the Court goes that route, then that result leads to option (2) in Windsor as well.  And the worry about moving too far too fast (which I don't share but some Justices might) could then lead some Justices to try to duck the case via option (1).  So there are potentially complex interaction effects between the cases.

People also wonder how particular Justices will vote, by which they mostly mean how Justice Kennedy will vote.  Here I'll say that in light of the fact that Justice Kennedy authored both Romer and Lawrence, I find it very hard to believe that if the case breaks ideologically, he would vote for option (4), i.e., against the plaintiffs on the merits.  And at least one case will break ideologically unless the Court votes to reject jurisdiction in both cases.  Accordingly, it strikes me that option (4) is very unlikely in Windsor and pretty unlikely in Perry.  But if you rely on these predictions to make investments or for any other purpose (like wedding plans), you do so at your own risk.