From Jan. 1 to Jan. 25, Mitt Romney and the “super PAC” supporting his campaign broadcast 12,768 advertisements in media markets that include Florida voters. Newt Gingrich and his super PAC ran only 210.
But were the ads responsible for Mr. Romney’s resurgence in Florida? Certainly the timing seems right: Mr. Romney spent a great deal of money, perhaps five times more than Mr. Gingrich, just as his poll numbers rebounded in Florida.
We can shed some light on the question by looking at variation in advertising within Florida. The Wesleyan Media Project reports the number of ads shown in various Florida media markets. We can then compare the advertising in these markets to the opinions of voters who live in them. Were voters in markets with heavier Romney advertising more likely to support him?
Thanks to the generosity of SurveyUSA’s Jay Leve, I obtained the raw data for their Jan. 8 poll and Jan. 27-29 poll, which bracket the period in which most of the ads reported in the Wesleyan data ran. With this raw data, I was able to place each respondent into their Florida media market based on their county of residence. (See this document for information on Florida counties and media markets.)
Were respondents in the poll more likely to vote for Mr. Romney if they lived in markets where larger numbers of Romney advertisements ran?
A first cut at the data, looking at the Jan. 27-29 poll alone, suggests no. There is no significant relationship between the number of Romney ads in the respondent’s media market and either the likelihood of supporting Mr. Romney or Mr. Romney’s favorability rating. This is true regardless of whether the ads are simply summed up or whether that sum is logged to take account of diminishing returns. It is also true regardless of whether I control for other factors, like respondents’ ideology, views of abortion, identification with the Tea Party, and other demographics.
A second cut at the data compares the Jan. 8 poll to the Jan 27-29 poll. Perhaps the effects of Mr. Romney’s ads were invisible in the later poll because I had not taken into account where opinion had been before they ran? It may also be important to compare support for Mr. Romney to that for Mr. Gingrich, since so many of Romney’s ads attacked Mr. Gingrich. Thus, I looked at changes in Mr. Romney’s support relative to Mr. Gingrich’s in terms of both vote intention and favorability ratings.
One important caveat: sample size is a major constraint. There are only 10 media markets to compare across time, and relatively small numbers of respondents in each one. I’ll therefore look at a simplified comparison: voters in markets or close to Florida’s Panhandle (the media markets of Gainesville, Mobile, Ala., Panama City and Tallahassee) versus voters in the other markets (Fort Myers, Jacksonville, Miami, Orlando, Tampa and West Palm Beach). This is a useful comparison because Mr. Romney and his super PAC advertised much less in the Panhandle (2,657 ads) than elsewhere (10,111).
Within the Panhandle markets, Mr. Romney’s vote share actually decreased slightly, although insignificantly, between the two polls: it was 37 percent in the Jan. 8 poll, but 34 percent in the Jan. 27-29 poll. Outside of the Panhandle, where he advertised much more, his vote share increased to 42 percent from 37 percent. This increase is also statistically insignificant given the sample size.
But other evidence suggests that Mr. Romney’s ads helped him. In the Panhandle the percent rating him favorably dropped to 41 percent from 51 percent over this period — a trend that’s consonant with the negative tenor of news coverage about him during much of this time. But outside the Panhandle, his favorable rating was stable (58 percent vs. 56 percent).
Meanwhile, Mr. Gingrich’s favorability ratings show the opposite pattern. Within the Panhandle area, they were 53 percent favorable to 25 percent unfavorable in the Jan. 8 poll, and a similar 50 percent favorable to 24 percent unfavorable in the later poll. But outside of the Panhandle area, they slipped from 50 percent to 22 percent, favorable/unfavorable, to 43 to 36 percent. That’s a 14 point increase in unfavorablity in the area where Mr. Romney was advertising most heavily. And since a large share of Mr. Romney’s ads were negative and attacked Mr. Gingrich, it isn’t surprising that we might see these effects.
Let me reiterate again that caveats abound in such a simple analysis. These caveats pertain first to the challenges of using survey data to isolate the effects of news media. But even if we take the survey results at face value, it’s also important to remember that advertising is not the only element of the “information environment” surrounding voters. As research by the political scientists Kevin Arceneaux and Gregory Huber has noted, media markets that witness a lot of television advertising usually experience other kinds of campaign activity — like voter contact — making it difficult to attribute any shifts in the polls solely to advertising.
At this point, I would say there is suggestive evidence that Mr. Romney’s advantages in advertising helped him win in Florida – but it qualifies as circumstantial. The longer the campaign goes on, and the longer Mr. Romney’s advertising advantage persists, the more data we will accumulate to test these effects.