Fig. 1 illustrates the impact of scale on costs for non-local variations in scale. It
is derived by calculating the administrative cost of an FHSA from the fixed and
random effects models when the primary care practitioner variables are increased
or decreased proportionately from their mean 1994r1995 values and the values of
all other explanatory variables are held constant at their 1994r1995 means. The
curves plot the resulting total costs divided by the scale factor for the primary care
practitioner variables and can be regarded as the multi-output analogue of a single
output average cost function.