The relationship between value added per worker (Y) and the wage (W) is repre¬sented by the coefficient a (the elasticity of substitution) which should come close to unity.
Griliches [8, pp. 962-64] and Bautista [2, pp. 76-77] have shown a way to reduce a possible source of bias in the elasticity estimates by introducing a mea¬sure of education into equation (2) on the assumption that education takes into account qualitative differences in the labor input across regions. Our study goes further from their respective approaches by differentiating the education variable into secondary and higher education, respectively.
The data used for agriculture are taken mainly from the 1971 Census of Agri-culture [12], a nationwide sample survey which provides the relevant data for our purposes. It provides cross section observations for sixty-six Philippine prov¬inces—our units of observations—and refers to crop-year 1970/71.
For manufacturing, the data employed are derived mainly from the 1968 Annual Survey of Manufactures [11], a nationwide sample survey covering around 4,000 manufacturing establishments with five or more workers. It contains cross section data for fifty-seven reporting provinces, Manila and suburbs being included as the fifty-seventh “province.”
Considering that the agricultural census data have no observations pertaining to a measure of education variables, such information is obtained from the 1970 Philippine Census of Population and Housing [13].
Table I shows the results of fitting equation (2) with some modifications to manufacturing data. Most of the coefficients are statistically significant at the 5 per cent level, and have the predicted signs. The crucial point in Table I is related to the estimates of the elasticity of substitution as indicated by the co¬efficient of the logarithm of the wage rate (PAYR). By statistically testing (at the 5 per cent level) we found that these estimates support the findings of previous studies [2] [15] that there exists a unitary elasticity of substitution in Philippine manufacturing. Thus, the stage is set for applying the Cobb-Douglas form in the following discussion