• Jeff Ryan

Data science and Trump’s travel ban before the Supreme Court

TL;DR: If we believe that this case hinges entirely on Kennedy’s vote, there’s a 61% chance of victory for Trump. If we believe the other justices’ votes are not certain, there’s only a 53% chance of a Trump victory.

Yesterday the US Supreme Court heard oral argument in 17-965 Trump v. Hawaii, wrapping up its term with one of the highest-profile cases of the year. News media and court watchers appear divided in predicting the outcome of the case. For example, Greg Stohr predicts that the travel ban will be upheld, while Noah Feldman would bet that Justice Kagan will be able to sway the vote enough to strike down the ban. Let us see if we can put some numbers to these predictions.

The court is currently composed of nine justices, who may be described roughly as four liberals (Ginsburg, Breyer, Sotomayor, Kagan), four conservatives (Roberts, Thomas, Alito, Gorsuch), and Justice Anthony Kennedy who is somewhere in the middle. Justices do not always vote along ideological lines; it is common to see unanimous votes. In fact, the correlation between the votes of any two justices is positive, so blindly predicting along ideology isn’t going to get us very far.

Is this case likely to be a unanimous decision? We have a predictive model based on speech analysis and legal information that forecasts the votes of each justice; here is what it predicts:

In our model, a prediction of 1 corresponds to a certain victory for Trump, and 0 corresponds to a certain victory for Hawaii. Given these individual predictions, our model predicts just a 53% chance of the final case outcome favoring Trump.

However, we have evidence that this decision will come from a court divided along traditional ideological lines, so perhaps we can do better than 53%. Looking back at court voting patterns since the beginning of the 2010 term, Justice Kennedy has voted with the majority in a striking 91% of cases. It appears that this case may indeed hinge upon his vote.

If we believe that the outcome of the case depends entirely on Justice Kennedy’s vote, then the outcome of the case will be the same as his vote, which is 61% likely to favor Trump. However, if we are not convinced that the other justices’ votes are set, then the chance of a Trump victory falls to 53%.

Taking either view, this is a case with a narrow margin for victory, slightly favoring the Trump camp. Although the justices will vote tomorrow on the outcome, we will likely have to wait until late June for the result to become public.

Our model’s out-of-sample confusion matrix on Justice Kennedy’s most recent 50 votes. A quick p-value calculation shows that our model has positive predictive power with .025 significance.

© 2018 Lambda Technologies