What the Evidence Says about Strategic Voting in Open Primaries

~by James Hanley

There’s been some interesting discussion here lately about the potential for, and legitimacy of, strategic voting in open primaries, in response to two posts by the surprisingly controversial Tod Kelly Although I lean Tod’s (amoral) direction on the ethical issues, my comment here is directed only at the question of whether there is evidence to support the claim that strategic voting in open primaries happens in large numbers. I begin by defining terms, then lay out the reasons I expect strategic voting will be rare,* then look at the evidence that generally demonstrates that in fact most cross-over voters in open primaries are not voting strategically, but sincerely.

Definitions

• Sincere Voting: Voting for the candidate, of whichever party, that is your first preference among all available candidates.
• Strategic Voting: Voting for anyone other than your first preference in an effort to defeat a less preferred candidate. In this debate, of course, we’re not looking at the Gore voter who preferred Nader, but at someone who crosses over in a primary to vote for the other party’s most extreme, or generally least-electable (in a general election), candidate.
• Non-Partisan: In the technical sense, not someone who is non-ideological or who truly doesn’t prefer one party over the other but anyone who is not registered with any political party, even if they identify more strongly with one than another.

The Logical Argument Against Strategic Voting

Based on what I’d heard casually from other political scientists and from my own rational choice perspective on politics, I expect strategic voting in large elections to be rare. Here are my foundational assumptions.

1. Parties control state legislatures, so they determine whether a state will have open primaries. They will not vote against their own interests, so it is unlikely they would vote in favor of an open primary if they believed there was substantial likelihood of strategic voting that could hurt their party’s interests.
2. Non-partisans are the primary beneficiaries of open parties. While partisans can cross over to vote in the other party’s primaries, non-partisans are more likely to be drawn into a party’s primary through an open rule. Non-partisans consist of those who are truly independent and those who are more ideologically aligned with one party or the other (generally, in approximately equal numbers). Ideologically aligned non-partisans can be expected to participate most often in the primary of the party they align with, while true independents can be expected to participate in the primary of whichever party has a candidate most preferable to them. Each group is less likely to think in strategic terms than partisans, so the most likely effect of open primaries is to expand the size of the electorate casting sincere votes, with that expansion consisting of generally more moderate voters, resulting in more moderate candidates winning the primaries, not the more extreme candidates strategic cross-over voters would support.
3. Strategic voting in large groups presents a difficult collective action problem (so long as ballots are secret). A rational voter knows that if he alone votes strategically his effort will be futile, and if enough others do it his vote is superfluous for achieving the strategy. He will only cast a strategic vote if he believes it could have some meaningful probability of affecting the election. Having no way to determine how many others actually voted strategically few voters will vote strategically. (Note: strategic voting is investment voting, whose value is solely in the outcome; sincere voting is often consumption voting—the value comes in part from the act of expression itself, regardless of outcome, so sincere voting does not pose such a collective action problem since value accrues regardless of how others vote).
4. Collectively, these assumptions suggest the hypothesis that the most likely cross-over voter is a sincere voter who is either an aligned non-partisan, a true independent, or a partisan of the other party who sincerely prefers—in this particular election—one of this party’s candidates to the candidates in his/her own party.

This analysis is supported by political scientist John G. Greer, who argues that;

“[I]f parties want to nominate electable candidates and they use primaries to choose electable candidates, there are good reasons for making certain that the candidates they select are attractive to independents and opposition party voters. It is unlikely that independents would ‘raid’ a party’s primary to vote for their least favorite candidate to undermine that party’s chances in the general election. A more likely reason for independents (or even partisans of the other party) to [cross over] is that they found a candidate they would be willing to support in November. Certainly the support [John] Anderson and [George] Wallace received from outside their own party is consistent with this argument.

Evidence
So what does the evidence say? There’s surprisingly little available, but here’s what I’ve found. (Note: All sources are listed in full at the end of the post.)

A Laboratory Experiment Supports Strategic Voting
Economists Todd Cherry and Stephan Kroll conducted laboratory experiments to investigate the likelihood of strategic voting in primaries. They find that strategic voting does occur, “although at low levels,” but that the levels are sufficient to influence outcomes. Also, the frequency is affected by the rules of the experimental primaries. They argue that semi-closed primaries, in which non-partisans can participate but partisans of the other party cannot, do not promote strategic voting as much as open primaries, so they are actually likely to promote more moderate candidates more than open primaries do.

However evidence from studies of actual primary voters doesn’t support Cherry and Kroll’s laboratory findings.

Exit-Poll Data Contradicts Strategic Voting
Political scientists Karen Kaufmann, James Gimpel and Adam Hoffman used “state-level exit poll data from 1988 through 2000” to compare the demographics of primary and general election electorates to see if “open and modified-open primaries actually attract a more representative electorate than their closed counterparts.” They find an open primary “can result in a more ideologically moderate and a more ideologically convergent electorate.”

Through the adoption of open primaries, Republicans’ primary electorates often wind up less conservative than their party following and Democratic primaries less liberal than theirs.

This does not speak to the ideological positioning of the candidates who actually win the primary, but it does mean the strategic goal of boosting an extreme or generally unelectable candidate is made significantly harder. With the median voter being closer to the ideological center, the weight of aggregate preferences has shifted away from the strategic voters’ target.

The 1976 Wisconsin Primary Contradicts Strategic Voting
Political Scientist Ronald Hedlund sampled voters from Wisconsin’s open primary in the 1976 presidential election. Among his findings were:

• “Independent voters as a group had a distribution of traits on income, education, and liberal conservative ideology than they did the other voters in the Democratic primary.
• “No evidence was found to document the contention that there was a widespread mischief vote which had a critical effect on the primary’s outcome.

The 1977 VA Gubernatorial Primary Contradicts Strategic Voting
Political Scientists Alan Abramowitz, John McGlennon and Ronald Rapoport studied the 1977 Virginia Democratic gubernatorial primary. This was a particularly valuable study because the Republicans used a convention system, rather than a primary, so Republican voters could, if they chose, cross-over and vote strategically without the loss of the opportunity to vote sincerely, since that option wasn’t open to them anyway. If strategic voting is going to occur, this is the perfect setup for it. In fact “Republicans comprised about 20 percent of those voting in the Democratic primary.”

Nevertheless, they did not find much evidence of strategic voting. Instead, the overwhelming majority of those Republican voters supported the Democratic candidate who was closest to their ideological position, even though that candidate was considered to be a stronger opponent to the Republican candidate in the general election.

The most significant finding of this group is that [80%] voted overwhelmingly for the more conservative Democrat…who was regarded by Republican party leaders as the stronger general election opponent…

Candidate preference, rather than any strategic consideration, seems to have been the foremost concern of crossover voters…The correlation (r) between candidate preference and vote in this group was .95, while the candidate between perceived candidate strength and vote was .34 (the opposite to the direction perceived by the strategic voting hypothesis)… When we control for candidate preference, the partial correlation between perceived candidate strength and vote is 0.17

In short cross-over voters were supporting the candidate they most preferred, despite the fact that they expected that candidate to be a stronger opponent against their own party’s candidate in the general election. This can be explained by those voters wanting to ensure that whomever won the general election it would be someone satisfactory to them, but it can’t be explained by a desire to cause the other party to lose.

Conclusion
The evidence, albeit a bit scanty, seems to suggest that open primaries don’t promote strategic voting as much as they expand and moderate the primary electorate. This suggests they are in fact achieving their intended purpose. According to Hedlund, the original motivation for open primaries was to expand democratic participation.

Open primaries were established in Wisconsin during 1903 to respond to pressure from the progressive faction of the Republican party and to public concerns expressed over the rights of nonpartisan voters [citation omitted]. Arguments supporting open primaries are that they protect the rights of voters to keep their party preferences secret, they maximize a voter’s choice in selecting a primary candidate, and they encourage widespread citizen participation in candidate selection [citations omitted].

And according to Kaufmann, et. al., states have adopted open primaries in an effort “to moderate the ideological extremity of primary voters…” It seems to have worked, without resulting in anything more than rumors of substantial strategic voting.

_______________________________________
*Full disclosure: I make no claim to be an expert on voting and elections, since that’s not an area I’ve studied thoroughly. However I have read some of the important literature in the field (some work on Duverger’s Law, Anthony Downs’ Economic Theory of Democracy, William Riker’s Liberalism Against Populism</em<, for example), and I have access to JStor and the ability to read quickly and organize what I’ve read. So I think what I’ve written here is correct, but I’m not trying to pass this off as the authoritative word on the matter.

Sources
Abramowitz, Alan, John McGlennon, Ronald Rapoport. 1981. “A Note on Strategic Voting in a Primary Election,” Alan Abramowitz,” The Journal of Politics 43(3):899-904.

Cherry, Todd L., and Stephan Kroll. 2003. “Crashing the Party: An Experimental Investigation of Strategic Voting in Primary Elections.” Public Choice 114(3/4):387-420.

Greer, John G. 1986. “Rules Governing Presidential Primaries.” The Journal of Politics 48(4):1006-1025.

Kaufman, Karen M., James G. Gimpel, and Adam Hoffman. 2003. “A Promise Fulfilled? Open Primaries and Representation.” The Journal of Politics 65(2):457-476.

Hedlund, Ronald D. 1977-78. “Cross-Over Voting in a 1976 Open Presidential Primary.” The Public Opinion Quarterly 41(4):498-514.