Estimating dynamic consumption of antibiotics using panel data: the shadow effect of bacterial resistance

To some extent, antibiotics are similar to addictive goods since current consumption is reinforced by past use because of bacterial resistance, which represents a growing concern in many countries. The purpose of this paper is to explore how consumers adjust their current level of antibiotic consumption towards desired levels over time. We construct a balanced panel dataset (2000-2007) for 20 Italian regions and estimate a dynamic model where antibiotic consumption depends upon demographic and socioeconomic characteristics of the population, the supply of health care in the community, antibiotic price, and the "capital stock" of endogenous bacterial resistance measured by past consumption. We apply alternative dynamic estimators for short panels: the bias-corrected least squares dummy variable (LSDVC) and the system Blundell-Bond GMM estimator (GMM-BB). The estimation results are stable across different model specifications and show that antibiotic use in previous periods has a positive impact on current antimicrobial consumption (between 0.14 and 0.39). This indicates that the process of adjustment to desired levels of consumption is relatively fast (approximately 1.2-1.6 years). Weak persistence in consumption may suggest that individuals are responsive to changes in antibiotic effectiveness.