his paper presents an empirical example to demonstrate that how profoundly the temporal variation of the policy effect can affect the results of measuring the safety effect of an intervention policy by using different lengths of study periods. The potential fallacies of applying the pair-t test, causal factor analysis model with dummy specification, and causal factor analysis model with time-based specification to measure the policy effect by using different prior and post implementation periods are demonstrated. The policy of criminal sanction for drunk driving in Taipei City is used as an empirical example. The results indicate the different choice of the lengths of the prior and post implementation periods would obtain the different sizes of the policy effect. Finally, the findings provide a set of valuable information for policy evaluation and alert the analysts to interpret their evaluation results carefully to avoid making an inappropriate conclusion. (Author/publisher)
Abstract