Pakistan Institute of Development Economics

Working Paper 2024:01
QR Code

Impact of Education Mismatch on Earnings: Evidence from Pakistan’s Labor Market

Publication Year : 2024
Author: Henna Ahsan


During the last 20 years developing countries like Pakistan heavily invested in their education sector to increase enrollment at primary, secondary and tertiary levels to boost their human capital. However, in the presence of poor governance institutions, stagnant labor markets and low educational quality, these additional years of schooling do not necessarily translate into enhanced human capital. It has been argued in literature that Human Capital Model based on Mincer Earning model produces biased results as a mismatch of education exists in the labor market. Therefore, present study investigated the impact of education mismatch on earnings by using the methodology of Duncan and Hoffman (1981).  For this I used Pakistan Social Living Measurement PSLM (2019-20) data. Our results indicate that though over education yields positive returns, but these are less than adequate level of education. However, after controlling unobserved heterogeneity bias, over education has no positive value. The returns of over education from OLS model might be overestimated if overeducated workers have lower average ability levels. Moreover, results seem to support the job competition model where the earnings of the individuals are based on job characteristics and not on individual’s education level.
[1] Research Economist, PIDE. Email:[email protected]

 1.      Introduction

To what extent education play’s role in increasing earnings of an individual is an important question for policy makers and researchers. Although this question existed since decades, it got due attention after publication of the book “The Over-Educated American” by Freeman in 1976. The book brought to light the startling findings that average earnings of high school and college graduates in the USA decreased by 16 to 40 percent between 1969 and 1974 in USA (Freeman, 1976).

There has been a consensus since long that education increases productivity by raising earnings of the individuals. The Theory of Human Capital posits the same stance that each individual is paid according to his/her marginal product. Therefore, Schultz (1961), Becker (1964) and Mincer (1974) concluded that additional years of schooling increase earnings of the individual and so returns to education are positive. However, with the rapid expansion of education, it has been observed that there is also over education in some labor markets, that is education of some workers is beyond and above the level which is required to perform a specific task causing the mismatch of education in occupation (Rumberger, 1981; Hartog, 2000). Further, Bird (1975) predicted that opportunities for new college graduates declined in the labor market, especially during the recession years of 1975 and 1976, leading to a more widespread distrust regarding economic payoffs for the college graduates.

As noted by Pritchett (2001), developing countries like Pakistan during the past 20 years invested heavily in their education sector to increase enrollment at primary, secondary and tertiary levels in order to boost their human capital. However, in the presence of poor institutions, stagnant labor markets and low educational quality additional schooling years do not necessarily translate into enhanced human capital. Therefore, the basic earning model developed by Mincer (1974) that suggests a positive relationship between earnings and years of schooling of an individual may not hold true. Education mismatch leads to misallocation of human capital in labor market and penalizes over educated individuals as they get low earnings than workers with similar education but whose education is in accordance with the job requirement. However, these over educated individuals tend to earn higher earnings than their co-workers who are not overeducated. Further, this educational mismatch also affects the undereducated individuals as these individuals receive low earnings when compared with adequately educated individuals (Duncan & Hofman, 1981; Groot & Maassen, 2000; Rubb, 2003 and McGuinness, 2006).  Sial et al., (2019) suggested that an unregulated expansion of education without healthy growth of labor market is a warning sign for the policymakers as this leads to earning differentials, and hence income inequality in the labor market.

To analyze the impact of mismatch of education on earnings most of the studies have used the methodology of Duncan and Hofman (1981) which decomposes actual level of education in Mincer earning model into three components (adequate, over and under level of education) to estimate the returns of education (for example, Kiker et al, 1997; Clark et al., 2017 and Sial et al, 2019).

Despite a large number of empirical studies focusing on the impact of education mismatch in labor market on earnings there persist a number of problems that may cause bias in estimating this earning effect. The first is the sample selection bias. Recent studies (Nicaise, 2001; Cutillo & Di Pietro , 2006; Lee et al, 2016; Caroleo & Pastore, 2018) point out the issue of this bias in estimating over education as there lie clear differences in the attributes of unemployed and employed individuals. These different characteristics may affect individuals’ choices to work and, thus, their outcomes in the labor market. For instance, due to prevailing high unemployment rate, individuals may be compelled to take jobs requiring less schooling, rendering high chance of being overeducated if employed (Quintini, 2011; Lee et al, 2016). On the other hand, Ghignoni & Veraschagina (2014) found that unemployment is a voluntary choice; individuals with higher skills set and academic achievements will remain unemployed until they find a suitable job. Therefore, omitting individuals’ decisions regarding participation in the labor market may cause a bias when estimating through simple OLS method.

Secondly many recent studies (Dolton and Vignoles, 2000; Kleibrink, 2016) point out that using over education in classical wage regressions depends on the assumption that equally educated individuals possess the same innate ability and, therefore, productivity; leading to unobserved ‘heterogeneity bias’. On the other hand, it is possible that ability of individuals may vary even with the same level of educational achievement. Therefore, to ignore the impact of ability while analyzing the earning model may cause bias known in literature as omitted variable bias (Baurer,2002; Kopri and Tahlin, 2009).  It is widely known in the field of education that productivity is reflected not only by attainment but also through unobserved factors like “ability”. Lee et al., (2016) and Bauer (2002) observe that significant increase in education attainment trend may characterize new workers with an increased heterogeneity biased. Unobserved heterogeneity does not only influence the educational attainment of individuals but also the extent they can make use of it in the labor market. Hence, each state — undereducation, overeducation, and being in an educational match — is the result of the decision process to find a suitable job, given the educational attainment. The question in which of these states’ respondents end up is influenced by unobserved heterogeneity.

As pointed out earlier, both heterogeneity bias and sample selection bias occur while analyzing the impact of education mismatch on earnings. However, it is usually difficult to control both of them at the same time so most of the studies either control sample selection bias or heterogeneity bias. Literature regarding control or correction of unobserved heterogeneity with sample selection bias is not inclusive. In Pakistan limited work (Farooq, 2011; Sial et al., 2019; Khan et. 2022) has been done to analyze the impact of education mismatch on earnings. These studies did not incorporate the heterogeneity bias and sample selection bias.

Therefore, the objective of this study is to analyze the impact of education mismatch on earnings from Pakistan perspective by taking into account the biasedness from both sample selection and heterogeneity perspective. To analyze the returns to education mismatch, the study adopts the methodology proposed by Duncan and Hofman (1981) model.  To handle the problem of sample selection bias I adopt the methodology of Heckman (1979) and Genialized Method of Moments (GMM) under instrumental variable (IV) technique in order to address the heterogeneity bias for observable educational variables included in the empirical specification. The study is based on the data of Pakistan Social Living Measurement (PSLM) 2019-20.

The remainder of this study is organized as follow. Section 2 discusses the literature review regrding the impact of education mismatch on earnings, while section 3 explains the methodology in detail. Data and construction of variables are explained in section 4, while the descriptive statistics and results of econometric analysis are presented in section 5 and the study is concluded in section 6.

For Full Text Download PDF