On the relationship between university education and economic growth: the role of professors’ publication

Abstract

This paper examines the effect of university education on economic growth across 34 developed countries. Professors’ research output is used as a proxy for the quality of education at the university level. To allow for some degree of difficulties in learning English across countries, an English weight has been constructed. We found that disparities in English proficiency across countries no longer matter in academic publications. The research outputs in science and engineering appear to have a positive and significant effect on economic growth. Economics and business researches also have immediate growth effects, although these effects are a bit smaller. The results are, in general, consistent with the findings of Aghion et al. (2005) and Vandenbussche et al. (2006), although their quality measure of tertiary education is different from the one used here.

Keywords:

• university education
• economic growth
• research publications
• English proficiency
• educational quality

JEL Classification::

• O40
• C20

Notes

1. Theoretical foundation of the role of education is discussed in Nelson and Phelps (1966), and further discussions of the educational role in the model of endogenous growth can be found in Lucas (1988), Romer (1990), Becker et al. (1990), and Aghion and Howitt (1998). Theoretical agreement on the role of education, however, has not been well supported by empirical evidence. Earlier studies in Barro (1991), Barro and Lee (1993, 2001, 2010), and Mankiw, Romer, and Weil (1992), among others, used the quantity measures of schooling and found significant and positive effects on economic growth. However, Levine and Renelt (1992) provided evidence that earlier findings were sensitive to the inclusion of other relevant variables. Bils and Klenow (2000) also found a possibility of reverse causation from growth to education. In addition, Hanushek and Kimko (2000) and Hanushek and Woessmann (2009, 2010), among others, provided evidence that rather than Barro-type schooling the quality of secondary education was a more important aspect of education for economic growth.

2. From a practical point of view, it may seem a bit one-sided to interpret research outputs solely as a measure of the quality of university education. One may argue that more productive researchers are not necessarily better teachers. Liberal art colleges in the USA, for example, produce a smaller output in research, but most of them have a very good undergraduate program which prepares them for the graduate-level studies at prestigious universities worldwide. The number of such decent liberal art colleges is, however, relatively small. Thus, professors’ research publications approximate the quality of university education.

3. This phenomenon is not peculiar to such large economies; small open economies such as Hong Kong and Singapore also benefit from this type of education-led growth, although many overseas talents are attracted to Hong Kong and Singapore after graduating from decent US and UK universities.

4. The convergence hypothesis is based upon diminishing returns to reproducible capital. In other words, poor countries with low capital–labor ratios are able to grow faster than rich countries because the marginal productivity of capital is relatively high in lower income economies. The higher productivity of capital induces the low-income countries to grow faster than rich countries. Therefore, a country's growth rate of real per capita GDP has a negative relationship with its initial level of income per person.

5. It should be noted that our quality measure uses the flow of research publications in recent years. Since educational effects are accomplished over longer horizons, our use of recent publications may not distort the growth effect of using the stock of publications in earlier years.

6. See Appendix 1 for the definitions of variables and the sources of data. Appendix 2 provides the list of sample countries and the publication data used. Although 55 countries were initially used to construct EP in Table 1, many developing countries were not included here due to their few SCI publications. For economics and business, more countries produced papers.

7. The growth literature is often faced with the potential problem of endogeneity. In particular, the number of publications might be endogenous to growth. A simple linear regression model is specified below and it shows that economic growth has a negative but insignificant effect on research productivity, that is, The insignificant parameter estimate together with a small R² suggests that the likelihood of the reverse causality from growth to publications is negligible. It would be desirable to include more relevant variables to control, such as tenure requirements, research grants, openness of universities, and salary of professors, but these are left for further study. In addition, our measure of PUB might be highly correlated with the omitted variables. To test for this potential problem, residuals in Equation (5) were regressed on PUB, and its parameter estimate appeared to be very small. A simple correlation coefficient between them was also close to zero. This indicates that any potential problem of endogeneity appears not to be serious.

8. In our model (4), β₂ = ∂GRGDP/∂ln GDP75, where ∂ln GDP75 = ∂GDP75/GDP75 = relative changes in GDP75, holding other variables constant. Thus, the parameter estimate of −2.228 implies that the growth rate of GDP falls by 2.228 percentage points, on average, if relative changes in GDP75 are increased by one unit (i.e. 100%). If GDP75 rises by 10% only, the growth rate of GDP will fall by 0.22 percentage points on average.

9. In addition to research publications, other types of research activities may also translate into higher economic growth. For instance, R&D would lead to innovations that directly facilitate the growth of a nation (Mansfield 1972). More recent analyses have estimated the direct effects of R&D expenditures on the growth of productivities (e.g. Griffith, Redding, and Reenen 2004). Therefore, we also included R&D expenditures as a percent of GDP in our basic model (4). The result shows that Surprisingly, the estimated coefficient of R&D was found to be negative and statistically insignificant, while other parameter estimates remained significant and almost intact. Multicollinearity may not have caused a problem in this case since ln PUB_SCI and R&D were not highly correlated (r = 0.645). More importantly, however, even after controlling for the growth effects of R&D, research publications were found to have a positive and significant effect on GDP growth.

10. The percentage change is computed as follows: 0.067 (parameter estimate in Equation (3)) * 4.87 (a 10% increase in the mean exam score in our sample) Δ0.3%. Note that Hanushek and Kimko (2000) provided 1.46 percentage points as the growth effect of one standard deviation (s.d.) change of exam scores. Since the distribution of their quality measure (mean = 46.61, s.d. = 10.86) is different from a more dispersed distribution of ours (mean = 17.2, s.d. = 13.67), a direct comparison of the two, using a s.d. change, is difficult to get a good approximation.

11. When the world economics ranking conducted by Kalaitzidakis, Mamuneas, and Stengos (2003) was used to measure the quality of economics education, the growth effect was also found to be positive and significant. However, the data were available only for the top 200 universities worldwide, and the ranking did not sufficiently reflect the publications of Asian universities that appeared, in most cases, below the rank 200. Coupe's (2003) citation rank included only 19 countries, and hence the degrees-of-freedom problem could not be avoided. To reduce this type of potential problems, Jin's (2006) world ranking data that included economics departments in 46 countries were employed here.