It has now been four years of contested elections for the Council of the Association. In this note we ask: What can we learn about these elections from a political science perspective?1
The only study we are aware of is an APSA report, entitled “2004 APSA Election Data Analysis Summary,” which breaks down the participation rate of various subgroups of the Association, and provides a summary of the comments left by voters after casting their votes. There is an explicit statement that no information on votes was gathered. The report can be obtained at the following URL: http://apsanet.org/imgtest/electiondataforweb2004.pdf (accessed November 15, 2006).
In each election, there was a slate of eight candidates nominated by the APSA Nominating Committee to fill eight vacancies. There were also one or two candidates nominated by a petition of APSA members. These candidates nominated by petition emerged from the “Perestroika” movement, which has advocated opening up the APSA Council by filling its vacancies through competitive elections. Eligible members of the APSA were provided access to a web site where they could cast up to eight votes, with no more than one vote per candidate. Each candidate's vote totals were then tallied and the top eight finishers declared elected to the Council. The descriptive evidence displayed in Table 1 makes the point that there is considerable variation in vote totals, which implies that a sizable number of members are making their own, individual choices for whom to vote, rather than just following the recommended slate put forth by the Nominating Committee or another entity.2
These figures were obtained from the Association's web site: http://apsanet.org/section_462.cfm (accessed November 15, 2006).
Vote Totals of APSA Council Candidates, 2003–2006

This variability in vote totals across candidates suggests that more members pick and choose from the list of candidates than one might expect a priori. Some of this picking and choosing might take the form of “bullet voting” or “plumping.” The aggregate evidence suggests that many members are not simply “voting the APSA slate.”
Of course, without seeing the actual ballots cast, we can only speculate as to such member-specific voting patterns. Yet, even absent the individual ballots, we can use the vote totals to ask the kinds of questions standard in the study of elections: What factors contribute to the relative success of some candidates over others? How do voters (in this case the academics who largely make up the discipline of political science) decide for whom to vote? If we consider APSA elections as simply “elections,” what can we say about the vote patterns we see? Can we predict those patterns as voters responding to ballot cues? The way of approaching this question empirically is quite straightforward: we seek to predict a candidate's electoral success from the range of cues provided to voters by information on the ballot itself.
One obvious ballot cue is whether the candidate was nominated by the Perestroika group, a rough equivalent to an endorsement by local notables or the editorial pages of the local newspaper. Given the heated debates within the discipline, such a cue could either have increased or decreased the likelihood of a member voting for the candidate.
But the ballot contains other cues as well, some of which have clearer expectations. First and foremost, is the name of the individual nominated. Some of these scholars are, if not household names, at least known to a large number of academics either within the various sub-fields or, for the better-known scholars, across them. One might expect widely known scholars to receive more votes.
Second, there is the institutional affiliation of the candidates. We would expect that candidates from the better-known institutions have more support in the profession at large. Part of the reason for this might be reputational: “good” schools are perceived to hire “good” scholars. Alternatively, better-known institutions provide the profession large numbers of graduates, whom we might expect to vote along some version of a “home state” or “friends and neighbors” line.
The most recent election also provided a new institutional cue: two candidates taught at non-American institutions (in both cases Canadian schools) and it may be that being “foreign” was also a cue acted upon by Association voters. With 1,845 APSA members—or about 13% of the total membership3
Source: APSA web site www.apsanet.org/section_170.cfm.
A third cue is the gender of the candidate and fourth is the ethnicity of the candidate, i.e., whether the candidate belongs to a visible minority group. Given the emphasis on diversity within the profession, we might expect both attributes to be positively related to vote share. Also of note, the 2004 APSA report on the election states that a large number of negative comments left by voters pertained to a lack of diversity in the candidates on offer. If enough members are concerned with this lack of diversity, presumably this would translate into higher vote totals for those candidates who are, in fact, contributing to increasing diversity within the discipline.
In order to assess the impact of these various cues, we estimate a series of linear regression models. In the first set of models, the dependent variable is each candidate's share of the vote, while in the second set of models the dependent variable is each candidate's rank position (first, second, third, and so on).
Most of our independent variables are simple dummy variables capturing the race and gender of the candidate (and their combined effect) as well as the institution. We take two different approaches in estimating the effects of race and gender. One approach is to include a dummy variable for gender and a dummy variable for race. The second approach is to include dummy variables for non-White men, non-White women, and White women (thus making White men the omitted category). We show the results of these two different specifications, partly to show that the effects we describe are robust across various specifications, and also to get a better sense of the interplay of race and gender.
With regard to institutional affiliation, there are several different measures of prestige, such as the U.S. News & World Report rankings of universities. But U.S. News does not cover overseas institutions, and it can be difficult to put teaching and research institutions on the same metric. Similarly, rankings of graduate programs would mean no ranking at all for teaching institutions, or for those departments that do not make it onto the rankings. We therefore choose a simpler dummy variable approach where “1” denotes an affiliation with an Ivy League institution, and “0” does not. We also include a dummy variable which equals “1” for candidates teaching at foreign institutions. We finally include a proxy for name recognition, namely citation count found for each scholar on “Google scholar.” While this is admittedly a crude measure, it is a simple one that has face validity in terms of the underlying concept of name recognition. Table 2 presents the results from these regression models.
Correlates of Candidate Success in APSA Council Elections, 2003–2006

Even with the small sample size (37 cases), a number of patterns stand out. First, the name recognition of candidates does matter and in the expected direction: more famous scholars (as measured by Google cite count) receive more votes. The difference between the 25th percentile in citations (46) and the 75th percentile (322) is about 2 percentage points, or nearly 1 full position in the rank position. Not only does this finding make sense, it is also one that does seem appropriate for an academic professional organization.
Second, women candidates do quite well in APSA elections. They consistently outperform men on the ballot. On average, women get about 3.5 percentage points more than men, or rank about 1.4 places higher in the results. When we interact race and gender, we see that White women do much better than White men, to the effect of close to 6 percentage points, or 2 spots in the ranking. Given the range of our dependent variable in Eqs. 1–4, (from 43 to 67), this is a substantial shift in vote totals.
These results also show some other patterns that might give pause, as the voting patterns of members seem to run counter to some publicly stated goals. First, the flipside of the result showing the importance of name recognition is that this vote pattern does undercut attempts to make the Association reflect the interests of teaching colleges. Those members of our profession who work and teach in junior colleges and liberal arts schools are simply not as likely to accumulate the sort of citation count as those who work at R1 schools. Ivy League affiliation does seem to give a bonus of between 3 and 4 points and a couple of positions to the candidate with such an affiliation on the ballot.4
Experimentation with alternative measures, e.g., “Big 10,” did not produce any similar kind of effect nor did a dummy for Western-based schools (a literal “west of the Mississippi” category that encompassed institutions ranging from Washington University in St. Louis and the University of Washington to those on the Pacific). Both of these dummy measures were negatively signed, suggesting that—for voters—there may be something more to having an Ivy League affiliation than simply having produced lots of grads who now vote in APSA elections.
Individual voted ballots might help answer questions related to the way in which the votes cast followed these sorts of patterns.
Second, and to some extent a little worrying, is the negative sign attached to the dummy variables measuring foreign members and minority status. While the p-values do not always meet conventional thresholds used to reject the null hypothesis, they generally approach standard levels of two-tailed significance. However, the parameter estimates are consistently signed across the various specifications. Indeed, in some of the models these parameters do meet the standard benchmarks. Furthermore, the size of the parameters is substantively consequential, ranging between 2–5%. Again, given both the small N and, also, the limited range of the dependent variable, these estimates do give pause. It would seem that candidates are systematically disadvantaged if they are not at American institutions or, perhaps less understandably, if they are not White.
Third, the interaction of race and gender shows that while White women do better on average than White men, they also do better than non-White women (and men). However, among non-Whites, it doesn't appear that gender matters. Thus, we are left with a portrait that looks like this: race matters only among women (Whites do better than non-Whites), and gender matters only among Whites (women do better than men). So, while these results suggest that efforts at diversity might be paying off, there would still seem to be room for improvement.
Finally, Perestroika candidates do not seem to outperform APSA candidates on average. However, a look at the showings of Perestroika candidates shows a clear trend toward success: from dead last in 2003 (with 50% of the vote), candidates nominated by petition finished 5th and 9th in 2004 (with 50% and 49% of the vote, respectively), 3rd in 2005 (with 57% of the vote), and 1st in the most recent election in 2006 (with over 62% of the vote). What this suggests is that mobilization efforts among those supporting the Perestroika cause, as well as perhaps an increased propensity to bullet vote—note the decline in the average vote per ballot cast reported in Table 1 between the first two elections and the two most recent ones—have increasingly benefited the candidates nominated by petition. To capture this effect (in an admittedly inductive fashion), we include a time trend and an interaction of the time trend and Perestroika candidacies. A different picture emerges from previous models: while a liability at first, it appears that, controlling for other factors, “outside” candidacies now receive more votes than their APSA-nominated counterparts. Mr. Perestroika's call for openness in our Association's elections seems to have been heeded.
Discussion
After making a series of caveats that can only be described as “completely expected” (the low N, the fact that we only have four elections, the roughness of some of the measures, the preliminary nature of these results, and so forth), there are two points to be made on the basis of our results that might be worth thinking about as we move forward.
First, a cue-taking framework does seem to help understand some attributes of APSA elections. These may very well be contests conducted by and for professional political scientists—but they are, at heart, elections and are understandable within a standard framework of cue-taking in low-information elections. So that's nice.
Second, and less nice perhaps, there are indications of some patterns that would seem to require a little more explaining: most especially the disadvantage “enjoyed” by foreigners and non-White women and men (compared to White women). In light of the small number of cases in the analysis, it may be a mistake to make too much of these patterns. Moreover, these results are also based on aggregate vote counts. Seeing the actual voted ballots would provide the opportunity for more interesting insights both on these elections and, in passing, on the profession. For instance, if voters bullet vote, do they tend to disproportionately support certain types of candidates? Are there other forms of structuring of preferences (by ideology? by location?). Nevertheless, these patterns do give us something to watch for as we move into 2007 and the possibility of a fifth contested election in as many years.
Author Bios
Shaun Bowler is professor of political science at the University of California, Riverside.
Antoine Yoshinaka is assistant professor of political science at the University of California, Riverside. He received his Ph.D. from the University of Rochester in 2005. He has authored a book and several articles on elections, institutions, and parties in both American politics and comparative politics.
