I’ve said this before, but I think it’s worth adding a few caveats to Dylan Matthews’s roundup of political science literature on the Supreme Courts. Particularly when the attitudinal model is used to defend very strong forms of legal realism, it’s worth noting that the model only applies to Supreme Court votes on the merits. Given that virtually by definition cases in the modern era only reach the Supreme Court when there is substantial legal ambiguity involved, it’s not surprising that politics generally drives votes in politically ambiguous cases. But the law matters significantly more in determining what kind of cases the Supreme Court hears and which it doesn’t, and politics alone also can’t explain why votes on the merits are structured around some issues than others. (Although Scalia and Thomas presumably believe that abortion should be illegal, their jurisprudence doesn’t require states to make it illegal.) There are also times when legal policy preferences aren’t the same as policy preferences per se. If you look at it as a case about federalism, the attitudinal model gets 7 of the 9 votes on Raich right; if you look it as a case of whether the federal government should strictly enforce anti-marijuana statutes, it arguably gets 7 of 9 votes wrong. Bush v. Gore is the absolute poster child for purely political jurisprudence, and yet if you look at it as an equal protection case it gets all 9 votes wrong.
For those interested, I’d strongly recommend Mark Graber’s brilliant and very readable essay on models of decision-making and judicial deference during wartime. One crucial limitation of political science modeling is that there’s no objective way of determining what constitutes sound legal practice or of effectively isolating “law” from “politics.” Supreme Court justices always have some discretion, and they’ve always been constrained in many ways from voting pure political preferences.