“Experimental science is the queen of sciences and the goal of all speculation.”

Roger Bacon

Election Dynamics: Santorum vs. Romney

An online message experiment testing the impact of positive and negative clips on Santorum and Romney in the context of positives and negatives on President Obama.

image

2012 General Election Dynamics: Santorum vs. Romney

Evolving Strategies partnered with The Weekly Standard to explore the general-election impact of Rick Santorum’s and Mitt Romney’s most serious vulnerabilities; social issues for the former and the Bain “Vulture” Capital issue for the latter. Some analysts claim that Santorum’s positions on social issues such as abortion and contraception will be devastating to his support among women, making him unelectable. Others argue that Mitt Romney’s business credentials are a double-edged sword, leaving him open to attack for destroying jobs and companies while he got rich; the business experience will be attractive to some voters, but will devastate his support among blue-collar workers in crucial swing states. These are open questions, impossible to answer with standard surveys.

To answer these questions, Evolving Strategies executed an online message experiment testing the impact of positive and negative clips on Santorum and Romney in the context of positives and negatives on President Obama. We utilized a non-probability, opt-in respondent sample of 1,320 registered voters approximating the national population, fielded February 22-23rd. Using well-established methods and analyses from the fields of political behavior, experimental psychology, and behavioral economics, we identified true causal impacts on vote choice. (See the treatment details here.)

Our evidence suggests that conventional wisdom about Santorum’s strengths and weaknesses is incorrect; Santorum does not do worse with women than he does with men. The most significant distinction between Santorum’s and Romney’s performance is not gender, but education.  The Romney and Santorum treatments have a huge impact on those with a 4-year college degree; Romney over-performs and Santorum under-performs to a surprising degree. Santorum doesn’t appear to be seriously harmed by social issues, however there is evidence that he is harmed by a lack of focus on economic issues. If Santorum centers his campaign on the economic debate, there is evidence he could gain ground on Romney’s electoral advantage.

  • Romney does substantially better than Santorum overall, does better with Obama ’08 voters, and does much better among the college-educated.
  • A focus on the economy and criticism of Obama’s economic plans seems to explain much of Romney’s advantage.
  • Santorum seems to do poorly because his treatment focuses far less on the economy.
  • Santorum does not do substantially worse with women than he does with men.
  • Santorum does not do substantially better with voters who do not have college degrees, our rough proxy for “blue-collar” voters.
  • Santorum does substantially worse than Romney with voters who have a college degree, driving them into the Obama column on that basis, but no better with those who do not have a degree.
  • There is some evidence that Romney does substantially worse with voters who have some college education, but not a 4-year degree.

Impact Analysis – All Respondents

Mitt Romney significantly outperforms Rick Santorum in a general-election match-up with President Obama after respondents viewed video treatments. (Romney’s negatives revolve around Bain “Vulture” Capitalism critiques and his positives entail Romney presenting a stark contrast between his and Obama’s economic vision; success through the free-market vs. failed crony capitalism.)

As you can see in the chart, Romney moves from parity with Obama in the Control condition to a 9-point margin of victory when voters watch positive and negative clips about Obama and Romney. Despite his clear advantage over Santorum and improved performance compared to the Control condition, Romney falls short of significantly outperforming the Control in which he is named as the Republican candidate.

Impact Analysis – Gender

Conventional wisdom currently holds that Santorum has a “women” problem, rooted in his aggressive social policy positions, that could be devastating in a general election against Obama. Romney, on the other hand, is generally believed to have an edge with women that will help in the general election.

As you can see in the chart below, however, there is little indication of a serious gender gap, whether comparing Romney and Santorum to Obama, or to one another. Santorum does more poorly than Romney across the board, but does not do relatively worse among women. In fact, although not a significant difference, Santorum actually performs marginally better with women than with men.

When we look to the data, we see that being socially conservative or liberal has a significant impact on a respondent’s vote, but men and women are fairly similar, with women a bit less conservative overall.

Furthermore, there is no significant interaction between being socially conservative or liberal and the Santorum treatment. In other words, there does not seem to be a significant backlash against Santorum in response to his social policy views.

Voters are overwhelmingly concerned with the state of the economy; the Santorum treatment focus on social issues might indicate to voters that he is not focused enough on economic issues in general. In other words, Santorum’s social policy positions per se don’t seem to be a serious drag, but a heavy focus on social issues do seem to explain much of his poor performance. In the absence of a strong economic argument and the heavy presence of social issues, however, marginal voters appear to rely more heavily on social considerations for their vote choice.

Impact Analysis – Education

Santorum’s vulnerability in the general election seems to center on education (potential proxy for class), with college-educated voters responding disproportionately negatively to the social-issue focus of the Santorum treatment.

The predicted MoneyVote probabilities for specific demographics, such as gender and a college degree, are predictions for the vote probability in each condition, based on that single characteristic while holding other factors in the model constant.

As you can see in the chart below, there is a dramatic difference in how the college-educated respond to the Santorum and Romney treatments compared to less educated voters. Predicting the Obama-Republican vote split based on whether or not a voter has a college education, we find that Santorum is left with a stunning 27-point deficit compared to the 4-point advantage for a “generic” Mitt Romney in the Control. Romney, in contrast, holds a 16-point advantage.

Other voter characteristics moderate the vote-choice, and the raw vote averages show a much smaller deficit for Santorum and moderately smaller advantage for Romney. These predictions highlight the relative importance (or, in the case of gender, unimportance) of that voter characteristic in predicting the vote for each candidate (condition) and the whether the interaction between the treatment and voter characteristic is positive or negative.

What this analysis shows is that the Santorum treatment caused very serious problems for him among the college-educated, even if other characteristics (e.g. income) ideology might hold their vote.

The gender analysis demonstrates that gender is not a significant issue in how voters respond to the Santorum and Romney treatments, whereas the education analysis demonstrates that a having a college degree is of great importance in explaining whether a voter responds positively or negatively to the treatments.

Obama Economic Plan Impact – Explanation for the Education Results?

We asked respondents about their attitude toward the economic plans of President Obama and the Republicans in order to gauge expectations on the most salient election issue; jobs and the economy. Did respondents believe that these plans would significantly help?

You can see that respondents in the Control mostly believe the President’s plans will not help the economy. In the Santorum condition the balance of support for the President’s plans moderates slightly to Santorum’s disadvantage, but remains negative. In the Romney condition, however, skepticism towards the President’s plan increases significantly, from a -12 point margin of opposition to the President’s plan in the Control to a -22-point deficit in the Romney condition. Interestingly, neither treatment seems to have a significant impact on voter perception of the Republican economic plans.

Statistical analyses confirm that a voter’s attitude toward Obama’s economic plans has a huge impact on vote-choice, and that the Romney treatment in turn had a significant impact on this attitude. Indeed, we found that there is a significant interaction between the Romney treatment and the impact of a voter’s attitude on the Obama MoneyVote. In other words, it seems Romney’s advantage among voters is due in significant part to his impact on how respondents view Obama’s economic plans and how that attitude impacts the vote choice.

The Romney treatment appears to have an outsized effect on the attitude of voters with a college degree. College-educated voters are on average more favorable toward Obama’s plans than those without a degree in the Control condition, but they come to parity with non-degree voters in the Romney treatment. Using the same model to produce predicted probabilities, their margin of opposition to Obama’s economic plans increases from a deficit of around 4 to over 20 points while non-college voters shift far less.

Regression analyses indicate that the Romney treatment has a very strong and significant impact on the economic views all respondents on average, and a particularly large impact on those with a college degree, but this impact fades to insignificance when looking at those without a college degree. Romney’s economic arguments, in other words, are disproportionally impacting the views of more educated voters, and this impact appears to translate into Romney’s big gains in the MoneyVote with college respondents.

In the Santorum condition, we see a slight, but insignificant, decrease in skepticism about Obama’s economic plans and no significant difference between college and non-college educated voters.

Santorum’s treatments do not effectively address the economic issue and do focus heavily on social issues. Although there is little evidence of a net drag due specifically to Santorum’s social issue positions, it seems voters rely more heavily on their social views in the absence of countervailing economic considerations. College voters are less socially conservative than those without a degree, and the Romney treatment impact (coupled with statistical analyses) suggests they respond disproportionally to economic arguments. Santorum, in other words, finds a particularly serious problem with the college educated when marginal voters rely more heavily on social, rather than economic, cues to make their vote choice.

A Note on the Statistical Analysis

The analysis in this report uses statistical models to predict the probability of a MoneyVote for President Obama and the Republican candidate (compared to choosing the default/NS option). The model was based on these respondent characteristics; 1) Santorum/Romney treatment, 2) 2008 vote for Obama/McCain, 3) gender, and 4) race (black/not black).

The predicted MoneyVote probabilities for specific demographics, such as gender and a college degree, are predictions for the vote probability in each condition, based on that single characteristic while holding other factors in the model constant.

These predictions highlight the relative importance (or, in the case of gender, unimportance) of that voter characteristic in predicting the vote for each candidate (condition) and the whether the interaction between the treatment and voter characteristic is positive or negative.

In other words, the gender analysis demonstrates that gender is not a significant issue in how voters respond to the Santorum and Romney treatments, whereas the education analysis demonstrates that a having a college degree is of great importance in explaining whether a voter responds positively or negatively to the treatments.