On February 4, 2012, at a victory rally in Las Vegas, Nevada, Mitt Romney foreshadowed the future of the 2012 general election campaign:

Three years ago, a newly elected President Obama told America that if Congress approved his plan to borrow nearly a trillion dollars, he would hold unemployment below 8%. It hasn’t been below 8% since. This week he’s been trying to take a bow for 8.3% unemployment. Not so fast, Mr. President . . . if you take into account all the people who are struggling for work or who have just stopped looking, the real unemployment rate is over 15%. (Romney Reference Romney2012)

By almost any measure, the topic most discussed by the candidates in the 2012 presidential election was jobs. From the lack of jobs, to the creation of jobs, to the number of jobs shipped overseas, both candidates spent a good deal of time telling voters about the state of the nation’s economy vis-à-vis the number of jobs in America. But as Romney focused on what he defined as the “real” unemployment rate, just a few weeks later at the United Auto Worker’s Convention, President Obama talked about something else:

Today, GM is back on top as the number one automaker in the world… Chrysler is growing faster in America than any other car company. Ford is investing billions in American plants and factories, and plans to bring thousands of jobs back home. All told, the entire industry has added more than 200,000 new jobs over the past two and a half years. 200,000 new jobs. (Obama Reference Obama2012)

Obama’s discussion of jobs was not about the rate of unemployment—whether real or unreal—it was about the change in unemployment; and as we entered 2012 and closed in on Election Day, the unemployment rate was trending downward.

These competing characterizations of the US economy help to illustrate two important elements of election forecasting. First, the change in an economic indicator, relative to its level, better predicts the incumbent party’s share of the two-party vote. And second, presidential candidates make important campaign decisions, like whether to enter the race or what to talk about if they do, based on the state of the nation’s economy at the start of election year.

As scholars of elections we have paid a good deal of attention to the relationship between economic indicators and vote share, but we have paid less attention to how candidates use information about the nation’s economy to make decisions about messaging or about running at all. We have spilt gallons of ink debating and refining models that forecast election outcomes with little more than a cursory nod to the fact that these predictions are made by evaluating elections in which competing candidates typically wage hard-fought campaigns. In reality then, our economic forecasting models tell us what to expect given the typical level and quality of campaigning in presidential races over the last several decades.

Just as variations in the state of the economy can shift forecasts, so too can variations in the level and quality of campaigning, but we know much less about how this translation works. This gap is mainly because data on campaign effort and intensity are difficult to come by. But when we collect these data and use them to model election outcomes alongside economic indicators, we see that what candidates do and what they say, given the state of the economy, plays a significant role in whether economic forecasts prove correct and in how close they come to the eventual two-party vote (Vavreck Reference Vavreck2009, 109).

There is almost no better pair of campaigns with which to illustrate the importance of variations in candidate behavior to predicting election outcomes than the campaigns run in 2012 by Barack Obama and Mitt Romney. A simple economic forecast done months before the election predicted an Obama victory by about 2.4 points (Sides and Vavreck Reference Sides and Vavreck2013, 177–78). If we knew in April 2012 that Mitt Romney would get the nomination and talk mainly about the fact that the economy was not growing fast enough, would we have more or less confidence in the 2.4-point victory the model delivered? Or is this information irrelevant to that prediction entirely? In The Message Matters (2009, 109), I explain why and demonstrate how this knowledge boosts our confidence in our estimate of the outcome. Here are the stylized facts.

In nearly every presidential election since 1948 at least one candidate has focused his campaign predominantly on the state of the nation’s economy (see Vavreck Reference Vavreck2009, 61). Typically, the candidate who focuses a campaign on the nation’s economy is the candidate predicted to win based on a simple economic forecast (of the last 16 candidates in this position, 11 of them did so). But sometimes candidates who are not predicted to win highlight the economy, too (George McGovern, Bob Dole, John McCain, and Mitt Romney). Clearly, candidates understand the powerful role that the state of the nation’s economy plays in American presidential elections and craft their messaging strategies in light of it.

Leading up to 2012, Mitt Romney surely took stock of the national economy and determined that despite high levels of unemployment and low levels of growth, the changes in these indicators leading up to 2012 were moving in directions that benefited the incumbent party (see Sides and Vavreck Reference Sides and Vavreck2013, 178, for the simple scatter plot showing this relationship). Yet, he and his team decided to focus their campaign on jobs and unemployment—a squarely economic message. The decision by Romney to focus his messaging on the economy increases our confidence in the economic model’s prediction because since 1948 no candidate predicted to lose based on the economy, who made the economy the central message of his campaign, has ever won.^{Footnote 1} It was unlikely that Romney would be the first.

Having said that, however, data on people’s perceptions of the nation’s economy and changes to it over the course of 2012 illustrate why Romney may have deemed this strategy a viable one. In partnership with YouGov, Inc., John Sides and I were able to analyze data from a three-wave rolling panel study each week in 2012. YouGov interviewed 45,000 people representative of the population in December of 2011 and then for the Cooperative Campaign Analysis Project (CCAP) and our work on The Gamble (2013), we re-interviewed 1,000 of this initial set each week of 2012. We then interviewed everyone again after the election. Over the course of the year, the unemployment rate in the country shifted by a half a percentage point from the high to the low (roughly 6 percentage points of change) and the change in gross domestic product (GDP) was just above 1 percentage point. Each week of the survey, including the baseline survey in 2011, we asked people whether the nation’s economy had gotten better, stayed the same, or gotten worse over the last year.

In December of 2011, only 20% of the population thought the economy had improved over the last year. The remainder split nearly evenly between thinking it got worse or stayed the same. Given this distribution of opinion it is not hard to see why Romney thought a campaign focused on Obama’s failure to turn the economy around could win him votes. The problem for Romney was the slippage between people’s perceptions of the economy and the actual economy (Bartels Reference Bartels2002; Campbell et al. 1960, 133 ) and the fact that voters were not blaming Obama for the slow recovery—they were blaming George Bush (see Duch and Stevenson Reference Duch and Stevenson2008 for a discussion of attribution effects and economic evaluations).

In an April 2012 YouGov poll, we asked how much blame Obama and Bush each deserved for “the poor economic conditions of the past few years.” Slightly more than half (56%) of respondents gave Bush a great deal or a lot of the blame, while only 41% gave Obama that much blame. A similar gap existed among independents with no leaning toward a political party: 58% blamed Bush, but only 42% blamed Obama. Altogether, 47% of respondents blamed Bush more than Obama, 21% blamed them equally, and 32% blamed Obama more than Bush (Sides and Vavreck Reference Sides and Vavreck2013, 26). Thus, despite what the Romney campaign may have thought was a favorable distribution of opinions about the state of the nation’s economy, tying those mainly negative opinions to Obama’s presidency was not going to be straightforward. They had found an issue on which public opinion was lopsided, they just made the wrong inference about what that meant for their campaign.

Using our panel data we can illustrate how successful Romney was at convincing voters the economy had not improved fast enough during 2012. All told, 30% of the electorate changed their opinion of national economic trends between December 2011 and Election Day 2012—a share that does not vary significantly by party. Although there is roughly equal movement among Democrats, Republicans, and independents, shifts among Democrats were more likely to be in the “got better” direction (22% of Democrats moved in this way) while among Republicans only 13% shifted in this way (and 16% moved in the “got worse” direction). Independents are between the partisans, with 17% improving their evaluations of the nation’s economy over the year. Still, 11% of Democrats and 13% of independents updated their impressions of the nation’s economy in the way that Romney wanted them to during 2012. Where is the slippage between that movement and Romney’s vote share?

Consider registered independents who eventually shifted their assessment of the economy in a negative direction (of the 45,000 respondents, there are 987 of them in the CCAP data). Their initial distribution of vote choice in December of 2011 was predominantly slanted toward Romney, 59% to 36%. This pro-Romney split is going to significantly limit the benefit, in terms of votes, Romney can get when these people change their assessments of the nation’s economy in his favor because most of them were supporting Romney from the beginning. But among the registered independents who were initially supporting Obama and whose evaluations of the economy moved Romney’s way, how many also reported changing their vote choice and casting a ballot for Romney? Only 4.2%.^{Footnote 2} Doing a little math suggests that Romney could have gained almost .4% of the vote from changing the views of independents on the economy.^{Footnote 3} Four tenths of a point is not nothing by any stretch of the imagination, and by comparing this to some of the work in The Gamble about electioneering effectiveness (2013, 220), this shift is worth a little more than having two Romney field offices in a county, on average and all else equal. But the flip side of this is that Obama picks up about .36% of the vote among the independents in this group who initially preferred Romney (although their economic assessments are becoming more negative). Romney is gaining some votes by convincing independents the economy has gotten worse, but Obama is gaining some votes as well—presumably for other reasons. The shifts are in nearly equal and opposite directions. This is another common pattern that John Sides and I illustrate across multiple dimensions of campaigning in The Gamble (2013).

In the last 60 years of presidential campaigns, all of the candidates challenged by national economic tides, who tried to reshape voters’ assessments of the economy, lost their elections. Surely part of the explanation for this is that people’s opinions are hard to move, but perhaps a larger part of the story is that election outcomes are more highly correlated with objective economic conditions than with people’s retrospective assessments of the economy—something that is counterintuitive for most political analysts and difficult for candidates and strategists to believe. The data on this point, however, are clear. Since the American National Election Study started asking its traditional retrospective economic assessment question (1980), the percentage of people who describe the economy as “getting better” over the last year is correlated with incumbent party vote share at .46. But, the percentage saying the economy got better over the year correlates with the actual half-year growth rate at just .57. And here is the critical piece of evidence: the objective change in GDP over the first half of the election year correlates with incumbent party vote share at .75. Even if a candidate is successful at changing people’s minds about the state of the nation’s economy, it’s not clear the payoff is there.

Would knowing, nine months before the election, that Romney planned to center his campaign on the too-slow pace of the economic recovery have affected the forecasts of the 2012 election? Yes. Among candidates disadvantaged by national economic conditions, who tried to refocus the election off of the economy and on to something else, nearly a third went on to win. But among those who tried to reframe the objective economy to benefit their candidacies—none went on to win.

If Romney had tried to reset the 2012 election off of a discussion of the economy and on to something else—something on which he was closer to most voters than Obama and on which Obama was constrained to his unpopular position (Vavreck Reference Vavreck2009, 33)—he would have had a better shot at beating a sitting incumbent president in a slowly growing economy. The effect of the quality of campaigns is “baked in” to our forecasting models, but we must try to systematize and appreciate how campaign effort and quality shape outcomes and, therefore, forecasts. With 2016 right around the corner, we have our work cut out for us.