Abramson, Aldrich, Paolino, and Rohde: 'Sophisticated' voting in the 1988 presidential primaries

Abramson, Aldrich, Paolino, and Rohde. 1992. 'Sophisticated' voting in the 1988 presidential primaries. APSR 86:55-69.

The Argument

Main Point

Voters are strategic ("sophisticated") more than they are sincere.

• Punchline: "U.S. citizens appear to avoid wasting their votes, when they are confronted with a multicandidate campaign in which the wasted vote logic applies."
• In final analysis, both Dems and Repubs have vote preferences consistent with sophisticated voting. Furthermore, voters typically include utility and viability in their calculations.
• Strangely, 15% vote irrationally. They don't vote for their preferred candidate OR the candidate they think is most viable.

Definitions

Sincere voters consider only their preferences among candidates. Sophisticated voters consider both preferences and viability (probability).

The Logic

If a voter prefers 'A' to 'B' to 'C', but 'B' and 'C' are the leaders, he might vote strategically for 'B' in order not to waste his vote on a hopeless candidate ('A'). Voting for 'B' becomes more likely as

1. 'A' and 'B' become more similar (relative to 'C')
2. 'B' becomes more strongly preferred to 'C' (i.e. 'B' and 'C' become more different)
3. 'A' becomes less viable, and 'B' and 'C' get into a closer race

Place in the Literature

See Cox's (1997) formalization of Duverger's law, which develops the logic of strategic voting considerably.

Empirics

Data

ANES Super Tuesday surveys from the 1988 presidential primaries. Measure preferences about candidates with "feeling thermometers." Also ask respondents who they plan to vote for and which candidates are most likely to win the nomination. Examine three time periods: Pre-Iowa, between Iowa and New Hampshire, and between New Hampshire and Super Tuesday.

Advantages of this data:

• No incumbent president running
• Both parties have more than two candidates
• There are no party labels within a primary, so it's a pure open election.

Five Different Tests

• First and second tests: voters vote for the candidate they like the most (they voted for the candidate they rated highest on the thermometer). Distribution of voting types (whether sincere, strategic, etc.) does not vary with political involvement, voting types do vary with preferred candidates, types vary over time in ways consistent with campaign (information). Thus, new information on candidate viability affects candidate preference.
• Third test: assessments of viability will help a voter decide between the first and second preferred candidate. Thus, support given to both inclusion of viability and preference in voting.
• Fourth test: does the name of the candidate determine variation in type of voting? Do people vote for Jackson for different reasons than they vote for Dukakis? Authors find that even among the two least viable candidates, Robertson and Jackson, viability still matters although less so than with the more popular candidates.
• Fifth test: Do people believe their preferred candidate is just more likely to win ("wish fulfillment")? Viability assessments have a small impact on candidate evaluations but evaluations have large impact on viability assessments. So, if I think a candidate is likely to win, that does not make me like her more. But if I already like her, it makes me think she will win. Even with projections, sophisticated voting takes place.

Findings

• Although there are a few sincere voters, there are many sophisticated voters.
• Voters tend to overestimate their preferred candidates' viability.