Nationalizing the Races

One way to look at the 2014 midterms--any midterm, really--is to consider them to be a referendum on President Obama. That is, voters will vote mostly based on their perception of President Obama rather than on their perceptions of the candidates running in their district or state. In such a world, where local races are completely nationalized, Democrats are doomed. As of December 2013 (which is a long time ago according to campaign time, but recent enough to make some judgments), in only one state--his native Hawaii--is his approval rating above 60%; moreover, of the 12 most disapproving states, two, Alaska and Arkansas, are home to two of the most vulnerable Democratic senators in the country. 

Of course, we don't live in that world. Regardless of whatever pundits might say about how the president's favorability impacts the races in the states, the fact is that the elections this year don't have President Obama vs. a generic Republican on the ballot. They're about two candidates that have their own records and their own personal appeal on which to run.

If, however, say, a Senate race would be completely "nationalized" (that is, completely dependent on the state's opinions of national politicians and not local ones), we'd see correlations between the Democratic Senate candidate's standing in the polls and, say, President Obama's approval rating in the state (something like statewide approval of the Affordable Care Act would work just as well). With this as a rough idea, I created moving averages for Sen. Kay Hagan's Senate race in North Carolina and President Obama's net approval rating in the state. (North Carolina happens to have a lot of polling for both the Senate race and for President Obama's approval rating, since the highly prolific pollster Public Policy Polling is headquartered in Raleigh.) I then calculated the correlation between how much Sen. Hagan's margin over Speaker Thom Tillis changed in a four-day period and how much President Obama's net approval rating changed in the same period:

The P-value isn't terribly great. The relationship between the changes in Obama's net approval and the changes in Sen. Hagan's lead (or lag, as the case often was) was significant at a P = 0.10 level (also known as the confidence level social scientists use when they can't get their variables to be significant at the usual 0.05 level). Moreover, the fit is terrible: R-squared = 0.0637, suggesting basically no correlation between how Sen. Hagan's lead changed and how President Obama's approval rating was changing.

A scatterplot doesn't tell a much better story for the relationship:

*One caveat: many polls taken between January 2014 and May 2014 (the points plotted on this graph) asked respondents either who they'd vote for in the Senate race or what they thought of President Obama, but not both. As a result, there are a lot of points on the axes, where either x = 0 or y = 0 because one of the variables didn't change at all while the other did.

This scatterplot should be read as "when President Obama's net approval changed by [x-axis value], Sen. Hagan's margin over Speaker Tillis changed by [y-axis] value." Therefore, if Sen. Hagan's performance is highly dependent on President Obama's approval ratings, we'd expect most of the points to be in the first (both positive) or third (both negative) quadrants. But that's not the case; in fact, it seems that most of the points are in the second (x negative, y positive) quadrant. If anything, this implies that as President Obama's approval rating goes down, Sen. Hagan's poll standing improves. 

That's obviously not the case, because there is a very simple and convincing theory behind tying Sen. Hagan's poll standing to President Obama's popularity: if people don't approve of President Obama's policies, why should they vote for someone who will presumably vote for those policies more often than not? 

But there are two things we should be getting from all of this. First, while President Obama's approval rating may not be tied that closely to Sen. Hagan's poll performance, it is pretty safe to say that she can't benefit that much from President Obama's unpopularity. And second, it may just be showing off the perks of incumbency--Sen. Hagan, precisely because she has a record, can define herself by something other than her relationship to the national Democratic Party. She may find that perk very useful should the president's approval start tanking again.