Floor Effects in the Time Series Quasi-Experiment: Tracing the Impacts of Social Policies

The conventional time series quasi-experiment measures the impact of a public policy in terms of a pre- to post- intervention change in the level of a time series. Policies may also affect a post-intervention change in series variance, however. Such a change may be strictly the result of the policy change, or more likely, may be a statistical anomaly. In this latter case, the change in series variance is due to a floor (or ceiling) effect in which the impact of a polity is such that the time series "bottoms out." Most crime time series, for example, have natural floors (zero crimes) below which the series cannot fall. By ignoring a naturally defined floor, researchers misstate the nominal probabilities of Type I and Type II errors. As a solution to this problem, we propose a data transformation. After demonstrating the practical consequences of a "floor effect," we reanalyze two recently published time series quasi-experiments.