Skip to content

Household composition, labor markets, and labor demand: testing for separation in agricultural household models

Literature › Works

Dwayne Benjamin (1992). Household composition, labor markets, and labor demand: testing for separation in agricultural household models. Econometrica, Vol. 60, No. 2. The Econometric Society.
Type journal article
Scope Comparative and methods
Method econometric
Period studied 1980–1980
Regions Java, Indonesia
Themes agriculture land, labour markets wages, comparative theory, methods data sources
Citation key benjamin1992

Argument

Under complete, competitive labour markets a farm household's production (labour demand) and consumption (labour supply) decisions separate, so household composition should not affect farm labour use. Benjamin tests this exclusion restriction on a sample of Javanese rice farms; nonseparation from off-farm employment or hiring constraints, or from differing efficiency of family and hired labour, would show up as demographic variables affecting farm labour demand. Using OLS, cluster fixed effects and instrumental variables to handle measurement error and simultaneity, he cannot reject separation. Household structure may affect the family/hired mix but not total farm labour, which follows neoclassical labour demand. The result is not consistent with a surplus-labour or rationed-labour-supply view, though limited power and the landless are acknowledged caveats.

Evidence and method

A 30% cluster subsample of the 1980 Indonesian SUSENAS household survey for rural Java: 4117 households, of which 1675 are rice farms and 1443 have complete labour and wage data. Includes actual family and hired labour days by task, task wages, input prices, irrigation, and head's age and education. Kabupaten soil and climate classifications, 1980 population census density and Geertz's sugar-regency classification are added.

Findings

  • Under separation the optimal farm labour input depends only on technology and the wage, not on household preferences or composition. (p. 5)

    The crucial implication of separation for this paper is that the optimal amount of farm labor L* depends only on the production technology and the wage.

  • Rural Java has an active labour market, with most rice farmers hiring labour and a mean of about 221 person-days per hectare. (p. 14)

    95% of rice farmers hire some labor, though not for all tasks. The average level of labor input, at 221 person days per hectare

  • In simple OLS the wage elasticity of labour demand is about -0.3 and the demographic variables are jointly insignificant. (p. 22)

    There is a well determined negative wage elasticity of -0.3. Each demographic coefficient is insignificant and the joint F test for the inclusion of the demographic variables shows that they can be excluded

  • The effect of an extra prime-age adult on labour demand is economically tiny. (p. 22)

    an elasticity of 0.01 implies that adding such a person to the household increases annual labor demand by 0.8%, or about half a day at the mean.

  • The 2SLS correction for wage simultaneity raises the absolute wage elasticity to about -0.94, and instruments pass overidentification tests. (p. 26)

    the new elasticity estimate rising to a well determined - 0.94.

  • Controlling for farm size, there is no evidence of a family-composition effect of the kind the surplus-labour hypothesis predicts. (p. 29)

    The data show that there is no evidence of a correlation between family composition and labor demand that would result if the surplus labor hypothesis were true.

  • Multiple-cropping farms show a larger household-size effect in OLS, which is not robust to fixed effects or 2SLS but may reflect family labour filling peak-season demand. (p. 30)

    the coefficient on log household size rose to 0.25 with a standard error of 0.07. This result was not robust to other specifications

  • The efficiency test cannot reject equal productivity of family and hired labour. (p. 32)

    The OLS estimate of (1 - a) is 0.009 with a standard error of 0.07, implying that we cannot reject the null hypothesis of equal efficiency.

  • Overall conclusion: evidence is inconsistent with surplus labour or constraints on farm labour supply, though power is limited and landless households are unobserved. (p. 33)

    Together, the evidence is not consistent with surplus labor or constraints on farm labor supply, though a few qualifications must be made.