Tuesday, December 13, 2011

The Effects of Early Ads

Political scientist Danny Hayes asks over at WaPo's Behind the Numbers whether candidates have struck too soon in Iowa, based on recent political science research that finds big effects from advertising -- but also alarmingly quick decay rates. That is, TV ads apparently can really matter, but the effects wear off in a matter of days. It's a good item, and I recommend that everyone keep an eye on that space for more posts from Hayes and other political scientists.

So: are the candidates wasting their time by going up in Iowa now and over the last couple weeks, given that there's still three weeks before the caucuses?

I don't think so. The studies that Hayes cites are interesting, but they're still really preliminary. In particular, they are all studies of effects during general elections, when we would expect relatively small campaign effects of any kind (because of the power of party identification and because of the other information available outside of ads).

But in primary elections, most voters have very weak attachments to their preferred candidate. It may be that preferences are even easier to influence, and once formed stay in place longer. Or not -- after all, we obviously have many Republicans who have changed their minds multiple times over the course of 2011.

The other key reason that early advertising might matter in primary elections is that in these types of high intensity, multicandidate contests, such things as "stories" and "narratives" are far more likely to matter than they are in low-intensity elections. Remember the example I gave recently: if Gary Johnson had done better than Herman Cain in the very first debate, it's possible that Johnson could have wound up in every debate and Cain would have been a fringe candidate. Similarly, a TV ad that pushes the polls for even a short time can then affect which candidates get coverage, which in turn can affect future polling.

So while it's certainly possible that early advertising is mostly wasted in normal general election campaigns, I wouldn't put to much weight on those findings -- so far -- for primary elections.


  1. I'm not comfortable with saying that "many Republicans changed their minds mulitple times in 2011." It could be a definitional problem, as I'm reading "many" to be something like most, or at least a plurality, and "multiple" to read 4+ times (as "a few" would be 2-3).

    It could be the case that around, oh 25-35% of Republicans have changed their minds A LOT. Or, it could be the case that 80-90% have changed their minds once or twice.

    Until we get the panel data (Annenberg, others), it's tough to say. I could also see running some kind of multinomial logit on the raw poll data from every week or two, once all that is out. But, I tried to put together data on how much shared variance there's been amongst the non-Romney or Paul candidates shares, and I didn't see a ton (but my method was pretty klugey, and I just tossed all the toplines from RCP in there, so anyone who had dropped out wasn't in there. Palin, Trump, Pawlenty, and Christie, I think, were all missing, IIRC.

  2. The Texas study happened in January 2006, during the primary, though it occurred immediately after Perry's only credible primary opponent declared as an independent in stead. Just sayin


Note: Only a member of this blog may post a comment.

Who links to my website?