Although the media is now focused on Chief Justice Roberts following his surprising health care vote, long-time Supreme Court watchers know that Justice Kennedy is the Court's true swing voter. An analysis of the Supreme Court 2010 and 2011 terms sheds new light on the mystery of when Justice Kennedy will swing left or right. I find that he is more likely to vote with the liberal Justices when the federal government is involved in a case.

In nineteen of the Court's 5-4 decisions from the past two years, Kennedy was the pivot, the one to convince if you wanted to attain a majority. On March 21, in Missouri v. Frye, Kennedy sided with Justices Breyer, Ginsburg, Sotomayor and Kagan. Yet only a day earlier, in Coleman v. Court of Appeals of Maryland, he wrote the majority opinion for the opposite side, Justices Roberts, Scalia, Alito and Thomas. His vote is split almost 50/50 - he agreed with the liberals in nine and with the conservatives in ten close cases over the last two terms. In contrast, the Chief Justice's vote tipped the scales only once over the past two years.

Since Justice Kennedy's vote is so evenly divided, is there a way to predict which side he will choose in a given case? Analyzing the votes in the last 140 Supreme Court decisions from the 2010 and 2011 terms,  I find that Justice Kennedy is more likely to side with the Court's liberal wing when the federal government is involved as a party or an amicus.

About three-quarters of the cases in these terms involved the government.  To determine whether Justice Kennedy's voting pattern differs in those cases, I compute the agreement rate between Kennedy and the other Justices, the traditional metric used by Supreme Court observers. Kennedy's rate of agreement with the liberal Justices is close to 80 percent in the cases with government involvement, but in the 60s in the remaining cases. These results, displayed in the table below, suggest that Justice Kennedy is more likely to form a majority with the liberal Justices when the federal government is involved in a  case. No comparable trend can be observed for any other Justice. The pattern is even clearer when the analysis is restricted to "close cases," i.e. 5-4 or 6-3 decisions. Once again, Kennedy has a much higher rate of agreement with the liberal Justices in cases which involve the federal government than in other cases, as the table below shows.  On the other hand, the rate of agreement is higher with the conservative justices when the government is not involved. For example, his rate of agreement with Justice Sotomayor falls from 47 to 23 percent, but his agreement rate with Justice Alito jumps from 53 to 77 percent.  Focusing on the close decisions where one Justice's vote can shape the majority confirms that Justice Kennedy is more likely to agree with the Justices on the left side of the political spectrum when the government participates in the case.

One possible explanation for this pattern is that Justice Kennedy has a more liberal perspective when the prerogatives and policies of the federal government are at stake. Interestingly, among the cases that did not involve the federal government, Kennedy was particularly likely to vote with the conservative justices when no government - federal, state, or local - was involved, although the small number of cases suggests caution in drawing firm conclusions.

The Court's next term, which starts in October, will feature another round of high-profile cases that may involve close decisions, including a review of affirmative action in Fisher v. University of Texas at Austin or human rights violations in Kiobel v. Royal Dutch Petroleum. The battle over same-sex marriage may even find its way onto the Court's docket. Justice Kennedy is likely to be the swing voter in some of these cases, and the pattern found here offers some clues to how he will vote in these contentious cases.

Veronika Polakova is the economic policy program manager at AEI.