Shabman and Stephenson: A critique of the self-interested voter model

Shabman and Stephenson. 1994. A critique of the self-interested voter model: The case of a local single issue referendum. Journal of Economic Issues 28 (December): 1173-1186.

In Brief

• Who/what are they challenging? Traditional literature that limits an individual's utility function to self-interested behaviour. [cite Mueller 1989 and Mitchell 1990 as examples].
• What are they testing? Self-interested voting behaviour in local referenda.
• Question: Can self-interest explain voting behaviour?
• Answer: No. Shabman and Stephenson argue that we need to take altruism (Ba) and community duty (D) into account when modelling utility functions.
• Data used:
  • Single case study: Roanoke, VA
  • Voters decide on referendum to increase taxes to pay for flood damage prevention.
  • They use this case because it eliminates some of the problems associated with aggregate data over multiple elections and multiple issues. This is one referendum with one issue in one area.
  • Floods only affect 5-10% of residents, so this is a good case to test self-interest hypothesis.
  • Need to predict 2 separate things: 1) whether to vote; 2) how to vote.

Hypothesis	H1: Those who live or work in floodplain (10% of population) are more likely to vote for tax increase than those who don't live/work in floodplain.	H2: Those who live or work in floodplain (10% of population) are more likely to cast a vote than those who don't live/work in floodplain.
IV(s)	protect home protect work knowledge that tax increases income	protect home protect work knowledge that tax increases income retire mobility
DV	How Vote	Cast Vote
Model	How Vote = f(protect home, protect work, increase tax, income)	Vote = f(retire, mobility, income, protect home, protect work, increase tax)
Findings	Only protect home and protect work were statistically significant in determining who voted for the tax increase. Model predicted 42% would vote for project, but 54% actually voted for it. S&S argue that we need to acct for Bn (altruistic behaviour). Therefore: self-interest was important determinant, but was insufficient to predict voting behaviour.	Only retire, mobility, income, and tax increase were statistically significant. Protect home/work were not significant. Model predicted 4% voter turnout. Actual turnout: 18%. Self-interest model = bad predictor: We need to account for P, D (intrinsic value of voting), or Ba (altruistic). If it were Bn, turnout should have been higher than in regular elections (not the case). If it were P, turnout should have been higher b/c more likely to be decisive in small elections (not the case). D (intrinsic value of voting) most likely to be determining factor.

The Bottom Line

These findings present various problems for the public choice model. Although the PC model does not preclude altruism and civic duty, their inclusion means that the model is non-falsifiable, and therefore pretty much useless.