Bandwagon or strategic voting in Israel? Note on Bargsted and Kedar 2009

• Goal: Re-interpret Bargsted and Kedar (AJPS, 2009).
• Results are very sensitive to the construction of the proxy for coalition expectations.
• Conclusion: expected electoral performance seems to have a much larger impact on voting behavior than expected coalition participation.

The goal of this note is to re-interpret and further analyze the results of Bargsted and Kedar (2009). BK use pre-electoral survey data for the 2006 Israeli legislative elections, and argue that a non-negligible set of individuals cast their vote in order to affect government formation and policy outcomes. Strategic considerations affect the likelihood of voting for Kadima, Labour or Likud, but not smaller parties. I contend that (i) what they are capturing is indistinguishable from a bandwagon effect, and (ii) their findings rely on the particular specification of the proxy for ‘expected coalition’ they use. I carry out the same exercise as BK using an extra set of controls for expected number of seats and an alternative specification of the proxy for expected coalition. My results show two interesting patterns. First, expected seats seem to be more important in voters' strategies than coalition considerations. Second, there seems to be a compensatory strategic behavior among voters, as opposed to BK: increased likelihood of a rightist (leftist) coalition induces voters to vote less for rightist (leftist) parties. Finally, this note shows that model performance is significantly increased when using each of these two new variables, independently or together. These findings support the inclusion of such variables in all models which empirically assess strategic coalition voting behavior.