The Surprising Robustness of Surprising Events

Paolino presents two core critiques of “Performing on Cue.†First, he suggests that my dichotomous measures of support for the Iraq War bias against finding evidence of reinforcing cues. Second, he suggests that using party identification (ID) as a moderator for the treatment effects biases against finding an impact for elite cues. Unfortunately, Paolino’s statistical modeling choices do not reflect these theoretical concerns. His arguments about attitude strength and source credibility imply that the experimental treatments should have nonlinear effects. Yet Paolino relies on an ordinal logit model with a linear interaction of the treatments with a six-point index of party ID and Bush approval. A more appropriate approach for capturing Paolino’s critiques would estimate a multinomial logit model with categorical interaction effects between the treatments and the source credibility. These more appropriate statistical modeling choices reveal that the findings of “Performing on Cue†are very robust to the concerns raised by Paolino.