FDI and educational outcomes in developing countries

FDI inflows data are from the UNCTAD FDI database and the annual growth rate is calculated by the authors.

Education or human capital accumulation can affect economic growth through a few different channels. It has long been argued that an educated labor force is able to either imitate the frontier technology and catch up faster or innovate faster, both positively affecting economic growth (Benhabib and Spiegal 1994). As a result, a more educated society grows faster (Gyimah-Brempong et al. 2006). The World Bank (1991) finds that the most successful developing countries invest more in their people through education. On the other hand, the cost of illiteracy is very high—an estimated $1.2 trillion each year globally due to lost earnings, lost business productivity, and lower technology capacity in the longer term (Cree et al. 2012). Further, education supports the development of civil society and promotes political stability, both of which contribute to economic growth too. The second reason we focus on education in this study is that education is in and of itself an aspect of human wellbeing. Individuals with more education face better opportunities for employment and income generation, have higher chances of good health as they are more likely to adopt good nutritional and hygiene practices, and are shown to be more involved in their children’s education (Martinez and Fernandez 2010).

Higher schooling in Egger et al. (2010) is measured by the years of post-secondary education in the total population.

The MDGs include eradicating extreme poverty and hunger, achieving universal primary education, promoting gender equality, reducing child mortality, improving maternal health, combatting HIV/AIDS and other diseases, ensuring environmental sustainability and developing a global partnership for development.

One might be also interested in the effect of FDI on secondary and tertiary completion rates. However, there are no (sufficient) data on secondary and tertiary completion rates across countries. As a result, we focus on primary completion rate and school enrollment rates at three different levels.

Vertical FDI occurs when MNCs locate different stages of production in various host countries, often taking advantage of input cost differentials between home and host countries. For example, a US MNC produces flat screen TVs in a developing host country due to the lower labor cost in the host country and ship the final products back to the USA for domestic consumption is an example of vertical FDI.

Researchers argue that, from a microeconomic point of view, if children can perform specific jobs adults perform but at a lower wage, FDI may increase the demand for child labor in developing countries that do not have stringent labor standards. In this case, two effects may occur. On the one hand, households may replace children’s school activities by child labor due to the higher opportunity cost of these school activities (a substitution effect). Even if the job opportunities go to adults, school-aged children could spend a significant amount of time on house chores when their parents are employed. Webbink et al. (2012) study 178,000 children in 16 developing countries and find that 30% of children in Africa and 11% of children in Asia work over 15 h a week on housework and family business work. Arguably, more time spent on such “hidden labor” means less time spent on school-related activities. On the other hand, when the family income rises above a subsistence threshold (possibly due to employment opportunities associated with FDI), household consumption of education at different levels, which are economic “goods,” may also go up (i.e., the income effect).

The 12 components of the composite political risk index include government stability, socioeconomic conditions, investment profile, internal conflict, external conflict, corruption, military in politics, religious tensions, law and order, ethnic tensions, democratic accountability and bureaucracy quality.

Exclusion restrictions refer to that the instruments do not have direct effects on the dependent variable.

Robust standard errors corrected for serial correlation and heteroskedasticity are reported in Table 2 and other tables thereafter. Following Di Tella and Schargrodsky (2004), Cameron and Miller (2015), and the notations in Stata (2019), the variance formula for clustered standard errors is \(\hat{V} = q_{c} \left( {X^{\prime}X} \right)^{ - 1} \left( {\mathop \sum \limits_{h = 1}^{M} u_{h}^{^{\prime}} u_{h} } \right)\left( {X^{\prime}X} \right)^{ - 1}\); where \(u_{h} = \mathop \sum \limits_{{j \in G_{h} }} u_{j}\); \(G_{1} ,G_{2} , \ldots ,G_{M}\) are the clusters; M is the number of clusters; \(u_{j} = \left( {y_{j} - X_{j} b} \right)X_{j}\) and \(q_{c} = \frac{N - 1}{{N - k}}\frac{M}{M - 1}\), with N representing the number of observations in the sample and k being the number of regressors; \(y_{jt} = X_{jt} \beta + u_{jt}\), j = 1, …., M, and t = 1, …, T; and b is the point estimate of \(\beta\).

To save pace, the results are not reported in the paper but available upon request.

The gross enrollment rate is measured as the ratio of school enrollment to the number of corresponding school-age children. When over-aged children enroll in, for example, primary schools, this measure can exceed 100%.

The two countries are Botswana and Congo, D.R.

We report the instruments and IV-related test statistics for regressions 3.5 and 3.6 in the bottom of Table 3. Again, the null hypothesis of the under-identification test is rejected, while the over-identification test is not rejected, indicating the validity of the instrumental variables. Similarly, the Wald test for weak instruments suggests that the bias in IV fixed-effects estimates does not exceed 5% of the OLS bias. The F-statistics in the first-stage regressions are also greater than 10.

For secondary enrollment, the Chi-square test to test the equality of the FDI coefficient for female and male reveals a p value of 0.4231, indicating the difference is insignificant. Whereas, for tertiary enrollment, the Chi-square test statistic for the coefficient equality shows a p value of 0.0457, significant at the 5% level.

Our results also show that FDI from both OECD and non-OECD countries has a positive and significant effect on primary school enrollment rates for female and male. These are not reported. In addition, the Chi-square test for the coefficient equality between OECD and non-OECD FDIs yields a p-value of 0.087 for secondary female enrollment, 0.1608 for secondary male enrollment, 0.003 for tertiary female enrollment; and 0.096 for male tertiary enrollment. Overall, the test of coefficient equality can be rejected at the 10% level or better except for secondary male enrollment.

The test was performed by the Stata routine xtcdf, written by Wursten (2017).

There are 52 countries included in the primary enrollment rate regressions. They are Albania, Argentina, Bangladesh, Benin, Bolivia, Botswana, Burundi, Cameroon, Central African Republic, China, Congo Rep., Costa Rica, Ecuador, Egypt Arab Rep., El Salvador, Fiji, Gambia, The, Ghana, Indonesia, Iran Islamic Rep., Jordan, Kazakhstan, Kenya, Kyrgyz Republic, Lao PDR, Lesotho, Mali, Mauritania, Mauritius, Mexico, Mongolia, Morocco, Nepal, Nicaragua, Niger, Pakistan, Panama, Paraguay, Peru, Philippines, South Africa, Sri Lanka, Swaziland, Syrian Arab Republic, Tajikistan, Tanzania, Togo, Tunisia, Turkey, Uganda, Venezuela RB, and Zambia. For primary completion rates, secondary and tertiary enrollment rates, the number of countries decreases as missing observations increase. For the purpose of brevity, the country lists for regressions 10.3–10.8 are available upon request.