ANALYSIS OF HOSPITAL PRODUCTION: AN OUTPUT INDEX APPROACH MARTIN S. GAYNOR a,b,c,f , SAMUEL A. KLEINER c,d * AND WILLIAM B. VOGT e a Heinz College, Carnegie Mellon University, Pittsburgh, PA, USA b Centre for Market and Public Organisation, University of Bristol, UK c NBER, Cambridge, MA, USA d Department of Policy Analysis and Management, Cornell University, Ithaca, NY, USA e Department of Economics, Terry College of Business, University of Georgia, Athens, GA, USA f Bureau of Economics, Federal Trade Commission, Washington, DC, USA SUMMARY In this study, we develop and implement an output index approach to the estimation of hospital cost functions that reflects the differentiated nature of hospital care. The approach combines the estimation of an output index within a flexible func- tional form. We find, in an application to California hospitals, evidence of scope economies across specialties within primary care, and diseconomies of scope within secondary and tertiary care. Minimum efficient scale is reached at larger levels of output than would be estimated by conventional techniques. These results indicate the importance of accounting for firm output heterogeneity when estimating cost functions. Copyright © 2013 John Wiley & Sons, Ltd. Received 23 April 2012; Revised 23 July 2013 Supporting information may be found in the online version of this article. 1. INTRODUCTION The estimation of multiproduct cost functions is one of the most important tools for applied production analysis. Cost function estimation has been used to inform aspects of antitrust and regulatory policy in a wide variety of industries, including banking (Pulley and Braunstein, 1992; Huang and Wang, 2004; Humphrey and Vale, 2004), public utilities (Kim, 1987; Garcia and Thomas, 2001), telecommunications (Röller, 1990; Bloch et al., 2001), transportation (Caves et al., 1980; Pels et al., 2003), education (Andrews et al., 2002) and natural resources (Toft and Bjørndal, 1997). Within this literature, the use of flexible form cost functions is widespread across both diversified and specialized firms, as they impose few a priori restrictions on the underlying structure of production. However, due to data limitations, the adoption of these functional forms often necessitates the simplification of the output space, thereby imposing restrictions on the production properties that can be analyzed. Such restrictions are particularly pertinent for industries producing highly heterogeneous goods where the nature of output is differentiated not only by the number of outputs produced by a firm, but also along dimensions within each unit of output (e.g. law firms, car repair, home construction). Hospitals are a noteworthy example of such an industry, as they produce hundreds or even thou- sands of outputs. For example, hospital output is classified using diagnosis related groups (DRGs), of which there are more than 500 categories, and each output differs by individual patient as to the degree of disease severity and complexity required for treatment. As Breyer (1987) notes, when analyzing hospital production, the curse of dimensionality precludes estimation of a production function which adequately accounts for case heterogeneity while also maintaining a sufficiently flexible functional form to allow for theoretically sound estimates of scale and scope economies. The difficulty in analyzing * Correspondence to: Samuel A. Kleiner, Department of Policy Analysis and Management, Cornell University, Ithaca, NY 14853, USA. E-mail: skleiner@cornell.edu Copyright © 2013 John Wiley & Sons, Ltd. JOURNAL OF APPLIED ECONOMETRICS J. Appl. Econ. (2013) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/jae.2371