and the less-developed countries. Tower's alternate explanation that A does not differ systematically while (er/f) does is not very appealing. The empirical evidence (p. 453) suggests that r is an external cost and does not vary significantly from nation to nation. Given this, the less-developed countries would have to weight e relatively twice as large as the advanced nations to explain the typical disparity in the estimates of A between the two groups.

N this paper, we first re-estimate the structural equations of the Puerto Rican Model (Dutta and Su, 1969) by seven alternative methods of estimation, and obtain predictions on endogenous variables by using seven sets of alternative estimates. We have obtained three different types of predictions -pure ex-post, pure ex-ante and partial ex-ante. We then compare the predictive power of the various estimators in terms of several descriptive criteria. The specification of the model remains as before (Dutta and Su, 1969). The model consists of 36 jointly determined variables and 36 equations. Twenty-three of these equations are stochastic, and the rest are definitional. Of the 23 behavioral equations, equations (1) to (6) explain consumption expenditures on food, services, other nondurable goods, housing, automobiles, and other durable goods; equations (7) and (8) describe private investment expenditures and changes in inventory investment; equations (9) to (16) cover imports including food, nondurable goods, automobiles, other durable goods, capital goods, raw material, payments on service account, and payments on all other service account; equations (17) to (19) determine exports, namely traditional exports, nontraditional exports, and exports of services; equations (20) to (22) deal with the production sector and explain gross product of three sectors, namely: agriculture, manufacturing, and all other industries and finally equation (23) explains wage share. The relations (24) through (36) are definitional. II Estimation