IMPLICATIONS OF MODEL CHOICE ON THE ESTIMATION OF EFFICIENTFRONTIERS: A SYSTEMATIC LITERATURE REVIEW AND META-REGRESSIONANALYSIS

Abstract

The aim of this meta-regression literature review is to examine the effect of modeling choices
on technical efficiency scores within the econometric literature based on fundamental concepts
rooted in the operations and production management sciences. Building on key modeling
frameworks of efficiency analytics and diverse frontier methodologies relevant for use in health
services research, the focus of this paper is on the major considerations following the selection of
nonparametric data envelopment analysis (DEA) variables/hospital indicators and the choices
applied to this estimation technique, accounting for model specification and variables included in
the efficiency analysis. The review concludes with an empirical section containing a statistical
summary of the literature on hospital efficiency frontier modeling, as well as a meta-regression
analysis aimed at identifying the key factors of DEA model specifications or study
characteristics that influence efficiency estimates. This step, undeniably, is vital in understanding
the limitations and drivers of efficiency score estimations and avoiding certain pitfalls that can
reduce the robustness of frontier modeling methods. Based on the meta-regression analysis, it is
clear that as the number of variables included in the frontier model increases, the average
efficiency predictions drop fairly rapidly when the sample size is fairly small. This significant
phenomenon identifies the importance of the sample size effect, indicating that the inclusion of
an extra variable into a model with more than 10 (hospital) observations does not alter the
average efficiency score very much; as long as the hospital sample size is large and homogenous,
the mean technical efficiency shows little change, and the mean efficiency seems to remain
constant after a sliding threshold is reached. Therefore, correcting for sample size has a major
impact on the assessment of average efficiency estimates and aids in robust empirical analysis
that can potentially be deemed scientifically evidence-based to allow for future health policy
reforms and allocation redistribution decisions to be made.