In a paper recently published in this Review, James Bennett and Thomas DiLorenzo (hereafter, B-D) estimate a model designed to test the hypothesis that conservatives, ceteris paribus, do in fact return a larger proportion of their staff allocations unspent than liberals (1982, p. 1160). While they make no claim to be modeling efficient management behavior, they are sufficiently confident in their ability to for what might be variously labeled as state effects, job effects, and political effects, to conclude that who are conservative in making collective decisions are also fiscally conservative in spending public funds under their direct control (p. 1160). Their regressions indicate that several such factors are significantly related to the percentage of staff funds returned by each senator and hence may influence the estimated relationship between measures of political and fiscal conservatism. However, B-D do not effectively for these variables. In this comment we suggest a natural alternative specification of their model which automatically controls for all state effects. If the B-D model were correct, our specification would produce parameter estimates statistically indistinguishable from those obtained by B-D. However, we find that this is not the case and, in particular, under our specification, the relationship between political and fiscal conservatism disappears. To measure senatorial frugality, B-D suggest using the treatment of clerk-hire budgets by the senators of the 95th Congress. These budgets, allocated to staff each senator's offices on the basis of state population, cannot be carried over from one fiscal year to the next. Any money not spent must therefore be returned to the Treasury. Some politically conservative senators have publicized the fact that they return a relatively large percentage of their clerk-hire allowance. While making no judgment on the prudence of such behavior, B-D correctly note that other factors may influence the percentage of the budget which a senator may return. In particular, they suggest variables which may influence the cost of staffing an office, including 1) the number of committees on which the senator serves, 2) the number of chairmanships held by the senator; 3) the number of years that the senator has held office; 4) whether or not the senator is up for reelection; and 5) a large number of variables describing the state which the senator represents (for example, area of state, per capita income, number of families, etc.) In spite of the fact that B-D tried unsuccessfully to include a number of additional state-related effects, it appears that they failed to take into account other such variables which may, in fact, confound the estimated relationship. A noteworthy example, which might be termed an allocation effect, may be seen by examining the clerk-hire allocation schedule (see B-D, Table 1, p. 1153). This schedule reveals that the state per capita senatorial staff allowance falls as state population increases. Although the costs of staffing an office undoubtedly depend, ceteris paribus, upon the ize of the state population served,' it seems likely that allocation errors which vary with state population may exist. Depending upon the relationship between such errors and state population on the one hand, and political conservatism on the other, it is possible to obtain a spurious relationship between fiscal and political conservatism when state effects are not fully taken into account. *Economists, Bureau of Labor Statistics, 600 E Street, NW, Washington, D.C. 20212, and Department of the Treasury, 15th & Pennsylvania Avenue, NW, Washington, D.C. 20220, respectively. We thank Robert Giflingham, John Greenlees, and Edward Murphy for their suggestions. The views expressed are our own and do not represent those of the BLS, the Treasury Department, or other members of their staffs. 'For example, a linear regression of staff expenditures upon state population levels explains 68 percent of the variation in the dependent variable.

The model of short-run residential demand for electricity combines price-schedule information obtained from the Federal Energy Regulatory Commission with household-specific data from the 1972-73 Consumer Expenditure Survey to determine the household's level of electricity consumption and relate it to the household's demographic profile and appliance stock. The overall short-run price-elasticity estimate of -0.550 supports existing findings that short-run residential electricity demand is price inelastic. However, the variation in elasticities across end-use categories suggests that the overall response is made up of a complex set of responses that vary substantially and significantly across appliances. This implies that it may be more efficient to tailor conservation policies to the stock configuration of the population. 17 references, 6 tables.