Yesterday the Senate rejected the United Nations’ Convention on the Rights of Persons with Disabilities (CRPD) treaty. As political scientist Jeff Peake noted, this was the first treaty to die on the Senate floor in over a decade (since the Comprehensive Test Ban Treaty in 1999). Among the measure’s proponents, former Majority Leader Bob Dole took to the Senate floor and pressed his Republican colleagues to vote for the treaty. But all was for naught, as 38 Republican senators voted “no.” Reaction to the vote was sharp, with headlines such as “GOP Senators Ensure Spots in Hell.”
A number of journalists have asked: Why did the treaty fail? Salon.com poses the question is a more loaded fashion: What has the Republican Party become? Asked in this manner, the obvious answer is polarization (and for good reason). While both parties have become more ideologically differentiated and internally homogeneous since the 1990s, the Republican Party has drifted furthest toward the extremes. So yes, it’s almost certainly ideological polarization to an extent.
But there are good reasons why the vote contained a non-ideological element as well. People who study Congress know that disagreement is often strategic. In Beyond Ideology, for example, Frances Lee documents disagreement among Democrats and Republicans on issues that have little to no ideological content (on which both sides should presumably agree). John Gilmour paints a similar picture in his fantastic book Strategic Disagreement. For Gilmour, “strategic disagreement” occurs when politicians avoid legislative agreements that would make them better off from a policy standpoint but worse off electorally. Simply put, Lee, Gilmour, and other political scientists have argued that voting behavior can be driven by electoral politics rather than simply ideological or policy disputes. Disagreement is not simply the inability to compromise.
The unique nature of yesterday’s vote gives us a way of examining these two possibilities (although imperfectly, as I point out below). What’s particularly interesting is that the treaty is almost an exact analogue of the vote on the Americans with Disabilities Act (ADA). The ADA was passed by the Senate in 1990 with a margin of 91-6. Now, because both votes are on similar policies, we can reasonably compare the two votes and assess the effect of greater polarization. Perhaps it’s simply that the two votes have the same ideological structure, just more ideological actors today. All the requisite data can be found at Keith Poole’s VoteView.
The figure to the left is a standard logit model of the 1989 Senate vote on the ADA (1=”yes”, 0=“no”). The lone predictor is ideology. The model fits the data very well, as one would expect. Based on the model’s predictions, I simulated yesterday’s vote (using NOMINATE scores from the 112th Senate). What this does, in essence, is take into account changes in the ideologies of senators from 1989 to today. Because the data are not available for the current Senate, this approach does not account for the votes of 13 Senate freshmen. Senators are predicted to vote for the CRPD based on the model’s predicted probabilities (aye>50%; nay<50%). I feel a chart coming on…
The columns in the chart to the right present the model’s “predicted” yeas and nays on the CRPD (67-17) while the rows chart the “actual” vote (57-27). The chart shows that 72 votes (85%) were correctly classified. Only 12 votes were incorrectly classified. There are two principal observations. First, based on the ideological structure of the 1989 vote, yesterday’s treaty is estimated to pass with at least 67 votes (and that’s with 13 freshman senators unaccounted for). Second, the chart reveals why yesterday’s vote did not—in fact—pass. In the upper right column we see that 11 senators voted against the treaty that were predicted to vote “yes.” Stated differently, 11 Republican senators voted against the treaty for reasons that can’t be explained by their ideology. These senators are presented in the chart above along with their estimated probability of voting for the treaty based on the simulation.
So what I’m arguing is that disability treaty vote was not entirely about policy disagreement, though it certainly was to a large extent. Rather, what is sometimes labeled “strategic disagreement” (voting against policy for electoral reasons) played a critical role and was, perhaps, the difference between passage and failure. And though this approach is certainly imperfect (see notes below), the conclusion is supported by the fact that 5 of the 11 incorrect yes votes (or 45%) are up for reelection this cycle (Lindsey Graham, Lamar Alexander, Thad Cochran, Pat Roberts, Mike Johanns). Indeed, a number of the senators listed above have moved to the right in the 112th Senate due to possible primary challenges.
* Now there are some qualifications needed here. First, the two votes are certainly not perfectly comparable. Yesterday’s CRPD vote, unlike the ADA vote, no doubt tapped into the right’s disdain for the U.N. and their fear of global governance. It’s also unclear to me whether the argument for the CRPD is a strong one given that it’s, well, non-binding. Erik Voeten makes this very point. There are also more sophisticated ways of modeling these kinds of decisions. As an aside, I estimated a second model restricted to Senate Republicans where an indicator of whether the senator is up for reelection this cycle was included along with ideology. The coefficient was negative, consistent with the prior arguments, but only “approaching” standard levels of statistical significance (p=.15; with only 34 observations).