When the tempest over Robert McDonnell’s thesis began, I didn’t think it would exceed the proverbial teapot. Now I’m beginning to wonder. In my wonderment is a lesson about how to identify the effects of campaign events.
The first lesson: here we are three weeks after the story broke, and only now is it possible to see a _potential_ effect. The lesson, in other words, is “wait.” Above I’ve taken Pollster’s graph, and made the smoothing more sensitive. You can see an inflection point around September 1, when story broke. Another way to look at this is to compare the margin between the two candidates in two sets of polls: those in August and the first week of September, when McDonnell’s average margin was about 10 points, and the polls taken after the first week of September, when McDonnell’s average margin was 4.5 points. The race does seem to be narrowing.
The narrowing isn’t necessarily due to McDonnell’s thesis, of course. However, some other evidence suggests that it might. This leads to a second lesson: identify _which groups_ are changing their minds.
Who is most likely to be turned off by conservative social attitudes, particularly with regard to gender rights and responsibilities? Probably swing voters who do not share these attitudes. And in today’s Washington Post, Anita Kumar and WaPo pollster Jon Cohen provide some evidence for this hypothesis. Compared to an August poll, the most recent WaPo poll found particularly significant losses for McDonnell among Northern Virginians and independent women.
The third lesson is: pin down the corollary attitudes that should be affected by the event in question. The Post did just that. (See here). Northern Virginians and independent women are more likely than other groups to identify McDonnell as “too conservative,” to say that his thesis makes them less likely to vote for him, and to oppose traditional gender roles.
Taken together, this evidence is still circumstantial, but it’s far better than what often exists amidst cable-newsy speculations about campaign effects.