The Monkey Cage

It's not just you Lee. According to a survey conducted by Polimetrix for Tali Mendelberg, Nick Valentino, Vince Hutchings, and myself, as of last month about 25% of respondents claim to have seen the Obama Girl video

Does Adam's post highlight a potential problem with polimetrix? We know from Youtube that the video has been viewed 5 million times. Presumably, some people watch it more than once. Admittedly, some people watch their computer screen with a friend. But, 25% of the polimetrix respondents have seen it. I would think that such a high percentage should result in a lot more YouTube views. When it comes to Youtube viewership of political material, polimetrix respondents seem to be unrepresentative of the nation as a whole. Presumably, this could reflect the fact that the respondents are online and check out the video after you survey them. If polimetrix can show you a time "stamp," it would be interesting to see if those who took longer to complete the survey were more likely to see the video.

I think it depends on how you define problem, Forrest.

Unlike Lee, I had heard about Obama Girl over the summer, but maybe that's because I read the Washington Post while Lee reads the Washington Times. Or maybe Lee skips the Style page. Further unlike Lee, I have still not viewed the video. But I've heard so much about it, it's almost like I've seen it. So how would I respond if faced with a survey question on it? From the media accounts, I know what happens in the video, just like I know what happens to Anna at the end of *Anna Karenina* without having read the book. As a respondent, I'd probably say no, but it's easy to imagine others in similar situations *claiming* yes. We need video validation a la the traditional vote validation to get the true sample%.

Also, measuring the population views is more complicated than using the YouTube count. If you search for "Obama Girl" on YouTube, you get 546 hits. Only the first 3 videos have over a million views, but it's possible to view the video not through YouTube, where it won't show up in the counts.

So while Adam et al.'s 25% is likely a little too high, it's far from obvious how much of an overestimate 25% is. Certainly it would be interesting to compare that estimate to a traditional RDD survey estimate. A comparison like that would be necessary to determine the scope of the problem.

OK, you've piqued my curiousity. I guess I better check this video out.

I had a similar concern when I first ran the numbers, but I think that the data is valid. The exact question we asked was:

"Have you seen the video of “I got a crush on Obama” by Obama girl?"

We're almost certainly capturing people who say bits of the clip on the news. I saw the video for the first time on CNN. I'm sure lots of other folks are in the same boat.

I am sure a lot of people would claim to have seen it if they thought a clip (or even a newspaper story) constituted seeing it. This would be a reasonable way to interpret the question.

Nevertheless, it would still be interesting to see if those who took longer to answer the question were more likely to have seen it. I could imagine that some respondents were prompted to see it by the survey and I suspect Polimetrix respondents are particularly internet savvy. Likewise, if the respondents are part of a panel it would be interesting to ask them where they saw the video.

The video has inevitably been a boost to the Obama campaign. The Polimetrix results make clear that the boost might be a lot bigger than we imagined.

Lee, time to check out "Hillary Boy".