TIME AGGREGATION AND TIME-LAG IN MACROLEVEL DETERRENCE RESEARCH

Recent arguments in research on individual-level deterrence suggest that the effect of perceived sanction risk on illegal behavior might occur over a shorter time period than the yearly lags typically incorporated in panel studies. This study raises the same issue for macro-level deterrence research by suggesting that panel studies using yearly data might have failed to capture the deterrent effect. The analysis uses ARIMA models with data aggregated monthly, quarterly, and semiannually to estimate the reciprocal effects between arrests and crimes for robbery, burglary, larceny, and auto theft in Oklahoma City from 1967 through 1989. No effect of crimes on arrests is found, but significant effects of arrests on crimes appear for three of the four offenses in the shorter temporal aggregations. The results suggest a need to reconsider conclusions based on panel studies that have used time aggregations and time lags that might be too long to uncover deterrent effects.