continued with this line of inquiry using data on a sample of California hospitals for five outputs. The results
indicated ray scale diseconomies. Breyer (1987) was critical of rudimentary classification of patients in most
output measures. He proposed a specification in which the flexibility of the functional form is applied to the
three global output categories of cases, patient days, and staffed beds. This approach offers potential for
reducing the number of parameters to be estimated without sacrificing detailed measurement of case mix. Thus
far a consensus has not been reached on the appropriate form of the hospital cost function.
The existing cross-sectional studies are not equipped to control for individual hospital differences that cause
variation in costs. One exception is the study of hospital competition by Melnick and Zwanziger (1988), that
draws on California data for the years of 1980-1985. These authors tested for the presence of hospital - specific
effects in the residual. Finding a very high degree of intra-hospital correlation, they used a variance components
model in their study. Gaynorand and Anderson (1995), in an analysis of the cost of hospital beds, also used
panel data in a fixed-effects model, which they applied to American Hospital Association (AHA) data for 1983-
87.
With the growing movement toward hospital cost containment, considerable concern has arisen over the
relationship between cost and quality of care. While there is some evidence that quality improvement is
consistent with lower costs (Fleming (1991), Binns (1991)), the underlying relationship between the quality and
the cost factor is unclear. Despite the interest in the quality variable, very few cost studies include measures of
quality as an explanatory variable. Inclusion of a quality variable has been attempted by Fleming (1991), who
found mortality and re-admission indexes to be significant determinants of cost. Gertler and Waldman (1992)
developed an empirical model in which costs are adjusted for unobserved endogenous quality and applied it to a
sample of long term care facilities. The parameter estimates differed significantly from those in which quality
was treated as an unobserved factor subsumed in the error term. If quality is an important determinant of
hospital costs, routinely ignoring it in cost function estimation can be a serious problem. Yet reliable measures
of quality may be generally unavailable. Incorporating an individual hospital effect through estimation of a
panel data model is an alternative approach that may capture variation in cost due to quality or other
unobservable differences among hospitals.