Table 3 presents each institutionalism model, by making the assumption that the effect of the local and regional level predictor is fixed, as described in formula (4), and there is a cross-level interaction, local plan quality/local planning agency commitment by political culture, civic engagement, and regional agency, respectively, as described in formula (7) . We expect the effect of regional level predictors and the interactions to have a positive effect.
Three institutionalism models show that, like Burby and Dalton (1994)’s model, there are powerful effects of local planning institutional variables. In Model A, which is the sociological institutionalism model (SI), we observe a marginal and positive effect of the political culture predictor at the regional level (p < .1). The cross-level interaction predictors, however, are not statistically significant. In Model B, which is the historical institutionalism model (HI), we observe a powerful effect of the civic engagement predictor at the regional level. The effect is positive and statistically significant at the .05 level. Also, the cross-level interaction between local planning agency commitment and civic engagement is negative and statistically very significant at the .01 level. It seems that the effect of local planning commitment is moderated or conditioned by civic engagement. In Model C, which is the rational choice institutionalism model (RCI), we observe a powerful effect of regional agency predictor at the regional level. The regional agency predictor is positive and statistically very significant at the .01 level. Also, the cross-level interaction between local planning agency commitment and regional agency is negative and statistically very significant at the .01 level. The effect of local planning commitment is moderated or conditioned by regional agency.
Considering random effects, the variance component at the regional-level of analysis of Model A, B and C remains significant. It means that potentially explainable residual variation in fixed effects remains. That is, the political culture predictor, the civic engagement predictor, the regional agency predictor, respectively, at the regional-level do not account for all of the variance in local hazard mitigation policy adoption at this level. We conclude that we should explore the effects of other regional-level predictors (like a combination of political culture and regional agency role) because it might help explain some of the regional-level residual variation. In comparing model fit, the difference in deviance statistics in Model C is lower than that of Model A and Model B. Also, AIC and BIC are higher in Model A and Model B than in Model C. Because it is argued generally that the model with the smaller information criteria (either AIC or BIC) fits “better,” Model C (rational choice institutionalism model) is considered a “better model” among three new institutionalism models.