Does sample selection bias affect the effect of family background on the returns to schooling? Evidence from Korea

Abstract

This article examines the effect of family background on the returns to schooling in Korea. In doing so, sample selection bias is controlled, as the probability of being employed is affected by the schooling of family members. The results of this study are different from previous studies, which did not consider sample selection bias. While the effect of family background on the returns to schooling was obvious in previous studies, it is weakened with the control of sample selection bias. This phenomenon is particularly strong for the wife. Moreover, controlling sample selection bias increases the downward bias due to measurement error. Thus it is concluded from these findings that sample selection bias should be controlled to accurately estimate the effect of family background on the returns to schooling.

Acknowledgements

This work was supported by the special research grant of Sogang University.

Notes

¹ There have been lots of studies on Latin America except these countries. For example, see Behrman and Wolfe (1984) for Nicaragua, and Stelener et al. (1987) for Peru.

² It might cause selection bias. Fortunately, I cannot find any difference in education and income of workers between the group living with parents and the group living without parents.

³ Earnings of each year is adjusted by the year 2005 consumer price index.

⁴ More specifically, an illiterate person is given the number 0 and the person with a graduate school degree is given the number 6.

⁵ According to Lam and Schoeni (1993), men with university-educated fathers have a mean wage 10 times greater than men with illiterate fathers.

⁶ Parental education can reflect family connection (or nepotism).

⁷ In addition to this, wife's schooling can affect the worker's earnings through human capital and the family connection between parent-in-law and son-in-law.

⁸ The content about measurement error hereafter came from Lam and Schoeni (1993).

⁹ In fact, this was originally shown by Welch (1975) and Griliches (1977).

¹⁰ For more details, see Liu et al. (2000).

¹¹ To the extent that wife's schooling is a proxy variable to reflect parents-in-law, wife's schooling can also be used as a variable to reflect ‘family connection’ through father-in-law and mother-in-law.

¹² This method has been widely used as a way to correct sample selection bias. For example, Nayga (1996) used this method to control sample selection bias, which existed in home expenditure on wine and beer.