The Economy and the Campaign

by Larry Bartels on January 14, 2013 · 1 comment

in Campaigns and elections,Political Economy


Income growth was slow through most of 2012, and prospective voters were correspondingly pessimistic about the state of the economy. So how was Barack Obama reelected? An important part of the answer is that perceptions of the economy became significantly less pessimistic in the fall than they had been in the summer—a shift coinciding with the beginning of a rebound in the actual income growth rate in September. This upturn in economic perceptions probably boosted Obama’s popular vote margin by about three percentage points—suggesting that an election held a few months sooner might have been a good deal closer.

While the election outcome was unsurprising in historical perspective, it raises the question of how prospective voters saw and responded to economic trends. Real per capita GDP was almost 5% higher in the fall of 2012 than it had been in the winter of 2009; but real incomes were less than 1% higher, and the official unemployment rate stood at 7.8%, just as it had in January 2009. More importantly from a political perspective, there was little evidence of economic momentum through most of the election year. Month-to-month real income growth—represented by the brown line in the figure above—declined raggedly but significantly through the first eight months of the year. Public perceptions of whether the economy was getting better or worse grew increasingly pessimistic over the same period, and remained firmly in negative territory throughout the year—especially among propsective voters who began the campaign undecided, represented by the purple line in the figure.

The trends in economic perceptions shown in the figure are based on weekly surveys conducted as part of the 2012 Cooperative Campaign Analysis Project. The same survey data can be used to track the political impact of these economic views over the course of the campaign. The statistical estimates of that political impact in the figure below suggest that, as perceptions of the economy became more pessimistic through the spring and summer, they also became more influential in shaping vote intentions—a process referred to in the scholarly literature as “priming.”


Because all of the CCAP respondents completed a baseline survey in December 2011, they can be partitioned into three distinct subsets based on their predispositions at the beginning of the 2012 campaign: those who reported supporting Obama in the baseline interview (43%), those who reported supporting Romney (38%), and those who reported being unsure who they would support (15%). The figure represents the impact of economic perceptions on evolving vote intentions for these three distinct groups.

The results for 2011 undecided voters, represented by the purple bars, provide strong evidence of an increasing effect of economic perceptions on vote intentions over the course of the campaign. In the first four months of the election year—during the competitive phase of the Republican primary campaign—a previously undecided voter who saw the economy as improving was about 14% more likely to have gravitated to Obama than one who thought the economy was getting worse. However, once Romney emerged as the presumptive Republican nominee, the impact of economic perceptions on vote intentions increased markedly, and that impact remained substantially higher through the summer and fall than it had been earlier in the election year. After Labor Day, a previously undecided voter who saw the economy improving was about 42% more likely to have gravitated to Obama than one who thought the economy was getting worse.

Evidence of economic priming also appears among prospective voters who reported supporting Romney or Obama in the December 2011 baseline survey. Although most of them stuck by their original vote intentions through the campaign season, those who defected were disproportionately those whose economic perceptions were incongruent with their original vote intentions. And that was increasingly true as the campaign wore on—though the increasing impact of economic perceptions was more modest and more gradual for Romney supporters (represented by the red bars) and even more modest for Obama supporters (represented by the blue bars) than for those who had begun the election year undecided.

The monthly income growth rates shown here are based on preliminary estimates of real disposable income per capita currently available from the Bureau of Economic Analysis. The data on economic perceptions are from CCAP surveys of 1,000 registered voters per week conducted by YouGov. Respondents were asked, “Overall, do you think the economy is getting better or worse?”

In addition to providing a measure of pre-campaign predispositions, CCAP’s December 2011 baseline survey makes it possible to mitigate potential bias in the statistical analysis stemming from reciprocal causation—shifts in economic perceptions during the campaign that are consequences rather than causes of changing vote intentions. By employing economic perceptions in the baseline survey as instruments for perceptions measured during the campaign in instrumental variables regressions, the analysis reported here plausibly captures the causal impact of variation in economic perceptions at various points in the campaign. I use the estimated causal impact for each of the three subgroups of prospective voters in the last months of the campaign to assess the likely effect on the election outcome of the fall uptick in economic perceptions.

Years ago, I published a chapter on “Priming and Persuasion in Presidential Campaigns” in an undeservedly obscure edited volume entitled Capturing Campaign Effects. That chapter suggested that the main effect of presidential campaigns is to prime prospective voters’ evaluations of the economy. The analysis was based on data from six presidential campaigns (a total of about 7,000 observations) from the American National Election Studies.

I also published a chapter in the same volume on “The Virtues of Panel Data for the Analysis of Campaign Effects.” That chapter started life as part of an effort, back in the mid-1990s, to revise the standard design of ANES surveys to feature three-wave panels beginning with baseline readings of political predispositions before the start of the election year. Due to funding constraints, that effort never took off. However, the same basic idea is alive and well in CCAP—which, thanks to the internet, also includes several times as many observations as the cumulation of six ANES surveys I had available back in the day.

Happily, the analytical power of the panel design I longed for now makes it possible to bolster and elaborate my old finding regarding campaign priming of the economy. For much more evidence of the virtues of the design, stay tuned for Sides and Vavreck’s forthcoming book-length account of the 2012 campaign, The Gamble.

{ 1 comment }

Comments on this entry are closed.

Previous post:

Next post: