He assumes that family and hired labor are perfect substitutes. The paper focuses on the demand for labor. Under separability, household size, the number of children, etc. should not affect the (total) demand for labor on the household farm. Thus such variables should have no explanatory power in labor demand estimates. This all sounds very simple, but there are several complications, which is what you need to get your paper published in Econometrica .
5 1. Labor demand is seasonal. It may be that labor markets in “slack” seasons are not as well developed as they are in “peak” seasons. Thus separability may hold in peak seasons but not in slack seasons. (Of course the opposite may hold – maybe peak seasons have labor “shortages” that distort labor markets, and such distortions do not occur in the slack season). Unfortunately, Benjamin’s labor demand variables are “sums” over all seasons, which may hurt the power of the test to detect violations of separability if such violations occur only during one season. 2. Household composition may not be exogenous. Suppose separability holds but household size is in part determined by land quality. If land quality is not in the data set, household size may appear to have a significant impact on labor demand but the reason is not lack of separability but correlation between household size and an omitted determinant of labor demand. 3. Measurement error in household demographic variables (data were collected in slack season). Consider the following labor demand function of household farms: L D = L D (w I , w O , A, a )
6 where w I is the wage the household pays for hired labor, w O is the wage household members can obtain if they work off the farm, A is land, and a is household composition variables. If separability holds, w O and a should have no effect on the household’s demand for (total) labor. Benjamin estimates the following: log(L D ) ≈ α + β log(w I ) + γ log(A) + G 1 i δ i a i + ε If none of the δ parameters is significantly different from 0, then one cannot reject the null of separability. Question : Note that w O is not in here. This is because he had no data on it. Will this cause an econometric problem that will invalidate his test? Question 2 : Suppose that the household composition variables are endogenous in the sense that they are correlated with ε . Will this invalidate the test? (For a detailed discussion of this problem, see pp. 304-305). The overall result was rather surprising. Benjamin found very little evidence that household demographic variables affected the farm’s demand for labor. For details, see the paper.
7 IV. Some Theories of Unemployment in Developing Countries A. Minimum wages are too high (yet in most countries these are not enforced) B. Unions keep “modern sector” wages high C. Government wages are higher than the market rate D. Large corporations pay wages that are higher than the market rate (theories B, C and D imply job “queuing”) E. Nutritional wage theories – employers do not want to pay lower wages because reduced
You've reached the end of your free preview.
Want to read all 24 pages?
- Spring '14