Whose Education Matters in The Determination of Household Income? Evidence From a Developing Country

This paper attempts to determine the best model for analyzing how education levels affect household income in developing countries. It compares the head-of-household model, that most commonly used by economists, with three other methods of testing the effects of education on household income in Ghana.

Most studies assume that the head of household's educational status determines his or her family's income. This study finds that assumption to be invalid in Ghana and, by implication, in other developing countries, for two main reasons. First, nonwage labor comprises a significant portion of household income in developing countries. And second, most education data are measured at the individual level, while income data are measured at the household level, making it impossible to accurately correlate an individual's educational status with his or her contribution to the family's income. To compensate for these problems, the author regresses the household income level to the household's educational level.

The Study
Data were collected by the Ghana Living Standards Survey (GLSS), a nationwide household survey conducted by the Ghana Statistical Service, with technical assistance from the World Bank in 1988-89. The survey queried 3,200 households about farm and labor activities, expenditures, education levels, and other determinants of household welfare. A supplemental education module then collected additional data from a subsample of 1,547 households.

The study analyzed three separate income functions—total household income, farm income, and off-farm (self-employment) income—for the head-of- household model and three other paradigms.

Head-of-household model
This paradigm assumes that the educational level of the head of household is the primary determinant of household income.

Maximum education or talented tenth model
This model measures the level of the best-educated member of the family. It rests on the principle that one well-educated member of the household makes business decisions based on education and skills.

Median or average model
This paradigm, measuring the average education level of the household, is based on the idea that the skills of all workers are important in determining household income. Both median and average measures were tested.

Minimum education or weak link model
Based on the least-educated member of the house hold, this assumes that the lowest level of education determines a household's income, that a production model is only as good as its weakest link.

These four paradigms were tested among samples of (1) all households, (2) households of two or more adults, and (3) all households with three or more adult members. Farm income and off-farm income was also tested for differences by gender, household size, wages, and relative productivity.

The tests were then extended in two additional ways. First, the education level of the head of household was included along with the minimum, average, and maximum levels in order to test whether it was a significant explanatory variable. Second, all the analyses were done comparing the average and median levels to see how the results differed.

"...the findings clearly call for surveys and policy models to use maximum or average education levels—household measures— rather than individually-based, head-of-household measures."

Findings
The study provides robust evidence that the head-of- household model significantly underestimates the return to schooling. The estimated return to education was most accurate using the maximum and average levels of education within the household. The maximum or talented tenth model provided the best prediction of total household income—27 percent higher than from using head-of-household measures. (The author assumes that higher predictions are more accurate.) Average or median levels (these results were quite similar) best predicted farm and off-farm income, 22 and 29 percent higher than from using head-of-household levels, respectively. The estimates from using head-of-household levels were very similar to those from using the minimum education levels—results that were not surprising, given that a large percentage of heads of households were the oldest and least-educated members.

When tested for the effects of gender across education levels, the results were ambiguous. While gender may not play a significant role in the relationship between education and income, the author surmises that this area probably was not well represented through the GLSS data.

Policy Implications
In rejecting the traditionally used method of measuring the return to education, the study implies several changes for policymakers. First, the findings clearly call for surveys and policy models to use maximum or average education levels—household measures—rather than individually based, head-of-household measures. Theoretical models and policy directives will best be targeted not solely at the head of household, but at other or all members of the family. Second, although the paper shows that maximum or average education levels are more accurate determinants of income than minimum or head-of-household levels, other factors still need to be better understood. Additional study should examine the way the allocation of education across household members of different ages, and genders, and across different activities, affects household income.

To order a discussion paper, please fill out an online order form, email IFPRI-FCN@cgiar.org or send requests to Food Consumption and Nutrition Division, IFPRI, 2033 K Street, N.W., Washington, D.C. 20006, U.S.A.