Panel data analysis of dentists’ activity under global budgeting in the presence of activityrelated non-response

This paper evaluates the impact of the introduction of global budgeting on dentists’ activity in Taiwan using a unique and rich panel dataset that was created specifically for the task. The panel data for 4424 dentists over 48 months, January 1997 to December 2000, was drawn from the BNHI’s data warehouse. The dataset has approximately 66% of dentists who are not observed over all periods. The paper examines the existence and consequences of unit nonresponse on estimates of the response of dentists’ activity to the introduction of global budgeting. It is based on the framework of selection on observables. We apply three techniques to assess the existence and magnitude of non-response bias: (1) probit models for non-response (2) variable addition tests (3) inverse probability weighting. The results show evidence of activity-related non-response. Non-response is concentrated among dentists with lower and unstable daily numbers of visits and with higher income. Non-response has substantial influence on the estimates of the mix of services provided to patients. The empirical results shed light on the emergence of a two-tier dental care system due to the introduction of global budgeting.