On Attempting to Rehabilitate the Proximity Model: Sometimes the Patient Just Can't Be Helped

In a recent article, Westholm (1997) critiques the directional theory of issue voting and seeks to rehabilitate the proximity model. He contends that the theories must be assessed by how well they predict choice, and that in choice-based tests the proximity model is superior. Both of these contentions are wrong. Westholm's empirical results are artifacts of his use of an issue index. Once different issues are allowed to have different impact, directional theory outperforms proximity theory in predicting vote (which Westholm does not analyze) and thermometer differences (which he does). Further, the whole premise of choice-based testing is flawed, because the tests can only be performed on underspecified models that lack verisimilitude. Rather than choice, evaluation is the appropriate focus for theory testing. Both theories predict evaluation, and comparisons based on evaluation are analytically robust. We conclude by discussing the power of the theories to explain real politics.