the prices of the other variable inputs. This is a "long run"
demand curve, since all inputs vary. Also, it is assumed there
is no output or cost constraint.

Theoretical Considerations Related to the Businesses Studied
 Four model groups were delineated. Attempts at grouping
various business types were unsuccessful, suggesting the level of
aggregation is an important consideration. Reliable equations
were obtained only at the four-digit Standard Industrial Code
(SIC) level, lending credence to our prediction (from theory)
that only businesses with very similar production processes can
be grouped.
 Another consideration in developing derived demand curves
is that shifts in demand caused by different scales of operation
have to be accounted for in the function. The inclusion of fixed
inputs (measured by size of business in this study) satisfied this
consideration.
 Prices of other inputs, as specified in Equation (A.3), were
not included in the equations for several reasons. It was ex-
pected labor costs, for example, were roughly constant across all
firms since most businesses use the same type of employees
(stock men, sales persons, maintenance persons, and managers).
It is unlikely that costs of these personnel (per labor unit) vary
greatly among stores. Thus, there would be no variation in the
variable and a regression coefficient could not be estimated. Also,
the cost of another major variable input, electricity, does not
vary appreciably between firms because all of Dade county is
served by one power company (only the Florida Key's electric
rate was slightly lower during the period of time for which quan-
tity of water figures were collected). Capital costs were accounted
for by the store size (area) variables.
 Another important economic variable, price of output (as
shown in A.3), was also omitted in all but the motel-hotel model.
A market basket approach could possibly have been used to de-
termine the price of output of each business studied, but would
have been costly and time consuming. Furthermore, since the
businesses studied are very competitive, the differences in product
prices between businesses of the same type may be small.

 The Statistical Model
 Ordinary least squares (OLS) regression- procedures were
 2The theory discussed in this section can be reviewed in Johnston [13]
and Kmenta [14].