Caveat

Several limitations of this study need to be acknowledged. For one, the CHAID method of analysis is somewhat limited due to the fact that only one predictor can be chosen per level, and cell sizes tend to drop rapidly as the analysis runs forward, thereby rapidly decreasing statistical power and perhaps failing to detect relationships that would be evident under other approaches. But the technique handles a mix of categorical and continuous variables with ease and is free of some of the assumptions required by general-linear-model techniques.

It also may be that the study defined ideological constraint and political knowledge too narrowly. The domestic-agenda issues from which the ideological constraint and political knowledge scales were built may have been relatively minor matters in the election compared to the economy, the war in Iraq, and the so-called war on terrorism. Despite these and perhaps other limitations, though, the study provides several useful insights into measuring evangelicalism and understanding its role in presidential election voting behavior.