Omitted variable bias and hospital costs

Abstract

This research examines the impact of omitted variables on the accuracy of parametric hospital cost function estimations based on Québec hospital level data. We assess the effect of omitted variables resulting from incomplete data on technology and performance measurement and on tests of the cost minimizing behavior of the institution. Our results show that important characteristics of hospital technology, such as returns to scale, are extremely sensitive to omitted variable bias. Similarly, estimates of hospital performance are poor indicators of actual performance when data are incomplete.