Regulation, Tallahassee, Florida. Information was also col-
lected on number of rooms in each motel or hotel.
 The general form of the model for this group was
 W= W(r, NR, PR, DB) (3.0)
where
 NR = number of rooms for rent,
 PR= average maximum price per room, the weighted average
 of the maximum prices of all rooms in a given motel,
 DB=dining room plus bar room area in tens of square feet.
Water price and quantity data were collected from water com-
pany records. Models using both "primary" (from survey) and
"secondary" (from DHR) data were estimated. The DB vari-
able was not included in the secondary data model, as data is
not available on this variable from secondary sources.

Restaurant and Cocktail Lounge Variables
 Restaurants and cocktail lounges were combined into one
 group, as many businesses have both facilities. It was thought
 water consumption in the two facilities was dependent on a
 similar type of production process; thus, it was expected similar
 variables would affect demand.
 The price of output, again, was thought to be a significant
 variable in restaurants and bars. It was expected price of food
 and/or service would be higher in the better quality restaurants.
 Higher priced restaurants may use more utensils in cooking, as
 well as supply more water for customer consumption. Unfor-
 tunately, price of output would be difficult and costly to measure
 directly. A proxy variable was tried [15], but it proved unsuit-
 able.9 The model illustrated for this group herein does not have
 the price of product (or service) included. The general model for
 this group was
 W=W(r, DH, BH) (4.0)

 9Average residential value in the area of the restaurant or bar was
 chosen as the proxy variable to reflect price of the product or service. It
 was reasoned the higher house value in the area would reflect the higher
 cost of food and drink in the area establishments. This proved to be a
 weak assumption. In many cases, some of the most expensive eating-
 drinking establishments are in a part of an urban area where residential
 dwellings are rather old, and thus, of low value. A better proxy would have
 been land value, but even this variable may not be entirely appropriate.
 The only apparent solution would be to actually measure price of output.
 One possibility would be to use the average price of the main dinner items
 on the menu in any future study.