Estimation of policy preference functions: an application to U.S. beef import quotas

Abstract

IOLICY criteron functions provide a basis for evaluating the desirability of alternative economic outcomes or states. Typically public decision makers must choose between alternative policy proposals which influence different sectors of society in various ways and which have different welfare connotations to these segments of society. We take the objective of economic policy analyses to be that of generating information to aid policy makers in the choice among alternative policy programs. Further, we view the formalized approach of economic analyses to policy making as one which supplements rather than supplants contemporary procedures used in formulating and administering economic policy. In the quantitative analysis of economic policy, two approaches have been advanced with respect to the use of a policy criterion function. The first, which we denote the explicit approach, involves a formally stated objective function as an integral component of the policy analysis. These analyses include not only the various optimizing models of decision making, e.g., Holt (1962), Theil (1968), Prescott (1971) and Chow (1972), but also the work of Fromm (1969) and others who have used an objective function in the explicit evaluation of simulation experiments. The second approach generates, for selected values of the instrument variables, the time paths of the endogenous variables. This approach has been advanced principally by Naylor (1970). While this approach does not involve the representation of a criterion function, such a function may be regarded as a concealed component of the analysis. This approach might be referred to as the implicit approach since (implicitly) a criterion function is used in choosing the policy alternatives for experimentation and in choosing the endogenous or performance variables for which these alternatives are to be compared. In contrasting the first approach with the second it may be argued that explicit specification of the criterion function (or set of criterion functions): (i) does not necessarily involve an arbitrary selection of the policy variable levels to be investigated; (ii) allows the investigator to assist public decision makers with the choice of weights across various arguments or goals entering the criterion function, particularly if a set of criterion functions is examined; (iii) provides an initial and formal basis for interaction between the investigator and the public decision maker; (iv) does not involve averaging over many samples generated by random drawings from the distribution of the stochastic disturbances in the system and as a consequence it is difficult to assess errors in estimating the actual mathematical expectation; and (v) does not typically result in a situation in which public decision makers are inundated with so mnuch data that they cannot realistically make choices. Points (iv) and (v) represent two of the principal difficulties confronted in applications of the implicit experimental approach. The explicit approach usually involves some arbitrariness in the specification of trade-off s between different arguments while the implicit experimental approach involves such elements in the selection of the specific policy alternatives investigated. Since the latter approach is not guided by an optimization or policy improvement method, many different policies must be examined which, of course, aggravates point (iv) and also obviously provides no assurance that good policies will be discovered. Neglecting investigator costs of the two approaches, arbitrariness emanating from the former approach appears less objectionable than the degree of arbitrariness present in the second approach. In view of the above position, this paper contains a general framework for the specification and estimation of policy preference funcReceived for publication April 5, 1973. Revision accepted for publication February 5, 1974. * Giannini Foundation Research Paper No. 378.