Correct Voting in the 2008 U.S. Presidential Nominating Elections

Abstract

Criticisms of the system by which the American political parties select their candidates focus on issues of representativeness—how choices are dominated by relatively small numbers of ideologically extreme primary voters, or how residents of small states voting early in the process have disproportionate influence. This paper adds a different concern, albeit one that still addresses representativeness. How well do primary and caucus voters represent their own values and interests with their vote choices? Lau and Redlawsk’s notion of “correct voting” is applied to the 2008 U.S. nominating contests. Four reasons to expect levels of correct voting to be lower in caucus and primary elections than in general election campaigns are discussed. Results suggest that voters in U.S. nominating contests do much worse than voters in general election campaigns, often barely doing better than chance in selecting the candidate who best represents their own values and priorities. Discussion focuses on institutional reforms that should improve citizens’ ability to make correct voting choices in caucuses and primaries.

Appendix: Details of Constructing Correct Voting Measures

Four conceptually distinct sets of items go into the calculation of candidate “utility scores” that determine which candidate best represents a respondent’s own political values and priorities, and therefore which candidate a respondent “should” support in the primary election—policy stands, candidate-group connections, “electability” beliefs, and personality traits.¹² In every case we accept survey respondent’s reports of their own values and priorities, but try to find some semi-objective expert judgment about how closely each candidate actually “fits” those same values and priorities.

First, the survey asked respondent’s to place themselves and the major presidential candidates on five policy scales: a question about illegal immigrants, a question about government health care, a question about increasing taxes on the rich, a question about the war in Iraq, and an overall liberal-conservative scale. I relied on the mean ratings of expert respondents (those scoring in the top quarter of the distribution of political knowledge) to determine where the candidate’s stood on these issues. Respondents were also asked their opinion about the circumstances under which abortion should be legal, whether we should take action to slow climate change or protect jobs instead, how strongly they favored capital punishment, whether gay couples should be allowed to have legalized civil unions, and whether gays and lesbians should be allowed to adopt children; but they were never asked to place the major presidential candidates on those same issues. I therefore recruited a separate panel of experts (seven graduate students) to place these candidates on these five additional policy scales after reading an extensive file of information about the candidates’ actual stands on these issues. I used the mean responses of these seven experts as an objective reading of where the candidate’s stood on these five additional issues so that I could calculate agreement with the candidates on these five additional issues. Table 2 reports the “objective” candidate placement scores calculated for each candidate on each issue.

Table 2 Objective measures of candidates actual policy stands and trait qualities

I then calculated each respondent’s proximity to each candidate on these issues.¹³ The policy proximity scores were reversed (so that policy agreement is scored high) and rescaled to vary between −1 and +1, and then added to each candidate’s utility score. These policy agreement scores only factor into the candidate utility ratings for respondents who cared enough about an issue to report a position on it.

Respondents were also asked to indicate how well three positive traits (strong leader, trustworthy, has the right experience) described each candidate. I assume that everyone universally prefers strong, trustworthy, and experienced leaders. Responses to these questions were recoded to range between −1 “Not Well at All” to +1 “Extremely Well.” I again relied on the mean ratings of our expert respondents, and added these three ratings to each candidates utility score, but again only if respondents showed that they cared enough about these attributes that they answered the questions. These “objective” trait scores are reported at the bottom of Table 2.

To estimate candidate-group linkages, I considered whether a majority of (expert) members of 11 easily-identifiable social groups—men, women, blacks, whites, Latinos, the working class, rich people, Southerners, born-again Christians, Mormons, Muslims—voted for any particular candidate in either party’s primary. This procedure assumes that politically expert members of these social groups are able to ascertain important (linked-fate?) candidate-group linkages that are independent of policy stands. Empirically, there were only two such linkages, between Barack Obama and black Democrats, and between Mitt Romney and Mormon Republicans. I therefore created dummy variables representing membership in these two demographic categories, and added them to the utility scores associated with Obama or Romney, respectively.

An important consideration for voters in their party’s nomination contests is how “electable” the different candidates are—that is, how likely they are to win November’s general election campaign. CCAP respondents were asked this question about both parties’ candidates in the baseline and January waves of interviews, and just about Clinton and Obama in the March interview. No consensus had developed among experts in either party about which candidate was the most electable until the March wave, when a large majority of experts believed Obama was much more likely to win in November than Clinton. This did not give me much to work with.¹⁴ Instead, I relied on the mean results of all survey questions (as reported by Real Clear Politics.com) pitting each viable Democrat against each viable Republican. I broke the nomination campaign into ten distinct periods, and calculated the average margin (across all surveys conducted during that period) by which each candidate would win or lose (“if the election were today”) against each remaining candidate from the other party. The mean of these average margins (averaging across all possible opponents) is the “electability” of that candidate during that time period. The data for all 8 candidates are reported in Table 3. I again normalized these electability scores within party so that they had a maximum range of −1 to +1 across the entire nomination process.

Table 3 Candidate electability estimates across the nominating campaign

An unweighted sum measure of candidate utility is computed by simply adding together these 15 different criteria of judgement for each of the major candidates competing for their party’s nomination—Hillary Clinton, John Edwards, and Barack Obama on the Democratic side, Rudy Guiliani, Mike Huckabee, John McCain, Mitt Romney, and Fred Thompson, on the Republican side. Similarly, an unweighted mean measure of candidate utility is computed by averaging together these different criteria of judgement for each of the major candidates. Many respondents do not provide valid responses on all of these criteria of judgment, of course, but we use as many as we can in calculating the candidate utility scores. With the exception of the candidate-group linkages, however, if a criterion of judgment is available for one candidate it is available for all candidates.

The above procedure treats all 15 criteria of judgment as if they are equally important—a good first approximation, but one my intuitions suggest does not match reality for most people. I also devised simple measures of how important each criterion was to each respondent. People were asked to express an opinion on numerous occasions for almost all of these different criteria of judgment, either their own opinion on an issue across multiple waves of the study, or their beliefs about where different candidates stand on an issue or how a particular trait would apply to different candidates, or both. Thus it was almost always possible to compute implicit “importance” weights for each of the criteria of judgment—simply the proportion of relevant questions on which a respondent provided a valid response. A weighted sum measure of candidate utility is computed for each of the eight candidates by multiplying each criterion of judgment by its importance weight and then summing across these 15 products. An weighted mean measure candidate utility is computed for each candidate by averaging together the 15 different (criterion × importance weight) products.

The method described above treats each distinct consideration equally, but as described above I also considered the possibility that electability considerations should be counted more that once (in fact, up to ten times). The analysis in Table 1 utilizes the weighted-sum measure with electability counted three times, but the results are pretty much the same if we use any of the other three operationalizations of correct voting, and count electability less.