where TE VRS is the estimated efficiency score over each observation, i.e., for county council i
(=1,...,19) in year t (=2001,...,2004). The first four independent variables all represent
features of the local county council. ALTER is our main variable of interest, and it is defined
as the number of physician visits at facilities run by others than the local county council
divided by the total number of physician visits in the local county council. Previous studies
have shown that competition from private providers has increased efficiency in other sectors
in the economy (see e.g. Millward and Parker, 1983; Sandstrom and Bergstrom, 2005). In
addition, even if competition does not increase efficiency in the local county councils, an
increase in the amount of health care being performed by alternative producers could have a
positive effect on efficiency. This would, for example, be the case if these production units
owned and operated by alternative producers of health care are more effective than the
production units operated by the local county councils. As such, our hypothesis is that a large
share of health service being produced by others than the local county council will have a
positive impact on efficiency.
FSTATUS represents the surplus or deficit for the local county council in millions of SEK
and our hypothesis is that a high surplus in the local county council will have a negative
impact on efficiency. This has been found in previous studies, see e.g. Gerdtham et al (1999),
and their explanation of the result is that deficits force local county councils to become more
efficient. REGION is a dummy variable equal to one if the local county council in question
has a region hospital. Region hospitals are large hospitals that in many cases are also
responsible for educating physicians and other health care personnel, as well as doing medical
research. As these activities do not produce the types of outputs included in this study, we
expect that having a regional hospital will have a negative impact on efficiency. Finally,
QUALITY is our measure of the quality of care given in the different local county councils,
and it is measured as the 28-day mortality rates among treated stroke patients in the different
county councils.
The two following variables, POPDENS and OLD, have been included to control for
differences in demographics between county councils. POPDENS measures the population
density (inhabitants per square kilometre) in the local county council. One might expect that
more densely populated county councils will be more effective. Such county councils should,
for example, use fewer resources in transports of patients, and these resources could then be