The bulk of efficiency studies in healthcare literature that use longitudinal data assume that inefficiency is independent over time. However, if the operating environment imposes significant adjustment costs that prevent healthcare providers from operating at optimum levels, inefficiencies are likely to follow a dynamic process. This implies that inefficiencies may persist from one period to the next, thereby violating the widely held assumption of inefficiency independence. In such a dynamic decision-making environment, a healthcare provider may opt to remain partly inefficient over time and aim for a long-run level of equilibrium efficiency that corresponds to the degree of adjustment costs they face. While the measurement of healthcare providers’ total factor productivity (TFP) and its components has gained considerable traction, existing studies exclusively assume either primal or dual specification, without any comparisons between them. The analysis of the productivity using both specifications can act as a robustness check and, most significantly, assist in highlighting the source of discrepancies in the results.
This thesis aims to address some of these gaps in healthcare literature by applying Bayesian estimation techniques on quarterly data on New Zealand District Health Boards (DHBs) for 2011-2018. The first empirical chapter investigates the technical efficiency of New Zealand’s DHBs in providing hospital services by using a dynamic stochastic frontier specification. The empirical results identify an excessive level of persistence in technical inefficiency in the sector due to high adjustment costs. While high adjustment costs resulted in a low national long-run technical efficiency level, on average, DHBs performed close to the long-run level of technical efficiency. The second empirical chapter also used a dynamic stochastic frontier specification to estimate cost efficiency while controlling for unobserved and observed DHB specific heterogeneity. The results show that although cost inefficiency persistence is still relatively high in the sector, DHBs that provide hospital services to rural communities tend to have a higher long-run level of cost inefficiency. The chapter also provides evidence that ignoring heterogeneity in stochastic frontier models can lead to biased estimates of model parameters and efficiency estimates. The final empirical chapter undertakes the TFP change decomposition under both primal and dual specifications. The results from both specifications are consistent and show that the TFP decreased between 2011 and 2018, primarily due to the declining technical change component. However, the scale component remained positive throughout the study period and dampened some of the negative influences of the technical change component on the TFP. Additionally, the results indicated that in 2016, both the scale and efficiency components posted their highest positive growth in response to the introduction of a performance-based ‘elective initiative’ programme, which briefly raised the TFP in 2016.