Catalog

Education and economic growth

Models of education and growth can be divided roughly into two (Aghion and Howitt 1998). The first type of model considers education to be an input into production, much like equipment or labor (e.g., Mankiw, Romer, and Weil 1992). In this view, economies with greater numbers of educated workers should produce more output. Subject to certain technical assumptions, economies that accumulated more education should have grown faster and obtained higher income levels, other things being equal. Treating education as an input, and by introducing various market failures that could lead to underinvestment in education, numerous growth theorists have attempted to explain divergences in the growth paths of economies in terms of the growth of their education stock. Such market failures derive from spillovers of productivity between workers (Lucas 1988) and the difficulties with financing education given that it cannot be used as collateral for borrowing (Ljungqvist 1993; Galor and Zeira 1993; Azariadis and Drazen 1990).

None of these education-as-input models pay much attention to why education influences productivity, what workers might produce, who should be educated, or what types of education to invest in. The central issues are the amount of human capital and output, not their composition or application. They are therefore fairly blunt in their policy implications. Most of them imply that subsidizing education can stimulate growth.

The second type of model considers education to be integral to an economy's capacity for technological innovation and adaptation. Thus, an economy that is far from some global technological frontier but that has a reasonable supply of educated scientists and managers will be able to catch up more quickly in technological terms, generating higher growth en route.

Nelson and Phelps (1966) are explicit that what matters for growth is not a high level of universal education, but having crucial personnel with the necessary education. According to these writers, to be crucial to transformation, a worker must be engaged in a nonroutine task, face new technological choices, and be in an organizational position to innovate. Presumably this implies the ability to redirect capital to new activities. In this view, productivity increases because education enables well-placed personnel to introduce new technologies, activities, and outputs.

Romer (1990) takes a more inclusive view of the role of education in transformation. In his model, the more education that is applied to research and development (R&D), the faster new activities are generated, and the higher the rate of growth. As educated labor could be attracted to pursuits other than R&D, countries with higher levels of universal education can engage in more R&D and grow faster. Romer's work on R&D is widely thought to describe conditions in advanced economies relatively well, while Nelson and Phelps' model of technology adoption and adaptation is a more apt description of developing economies' experiences.

Turning from models to data, there is substantial microeconomic evidence in favor of the view that the value of education in development depends on the scope for technology adoption. Studies of technology adoption are littered with evidence that more-educated workers have adopted new profitable technologies more readily. One particularly arresting and relevant example concerns the Green Revolution period in India. Foster and Rosenzweig (1996) observed that: more-educated households turned to high-yielding crop varieties (HYVs) more rapidly; those states that adopted HYVs experienced faster agricultural growth; returns to primary education expanded significantly during this time; and these returns increased faster in areas that grew faster. These results, and others like them (see also evidence presented in Rosenzweig 1995), point to a two-way causal relationship between education and growth, conditional on the availability of new and better technology.

As argued above, most of the education-as-input models predict that, other things being equal, output growth rates should correlate positively with the human capital growth rate. In contrast, Nelson and Phelps, and Romer predict that higher initial levels of education capital would drive subsequent output growth. So what do the data suggest? Notwithstanding some serious econometric problems with cross-country growth regressions, it is worth reviewing the evidence-limited and hotly debated as it is.

First, surprisingly, growth rates of education attainment are often found to be negatively correlated with growth in GDP per worker. This regularity was reported by Pritchett (1996), Benhabib and Spiegel (1994), and Islam (1995). Given that models linking education to productivity growth are motivated by microeconomic evidence that employers are willing to pay more for educated workers, and that this willingness has been shown to reflect the greater cognitive skills of the educated (Glewwe 2002) this result appears paradoxical. For if education renders individual workers more productive, then surely across-the-board increases in education should render the aggregate labor force more productive as well? As these expected aggregate productivity improvements did not materialize, then "Where has all the education gone?" asks Pritchett. This extraordinarily important and startling paradox has sparked intense debate on how to measure education-growth relationships.

Nelson and Phelps, and Romer's views appear to survive empirical scrutiny. Several studies have found positive relationships between initial education levels and subsequent output growth. Benhabib and Spiegel (1994) show that even after correcting for income levels and education levels, those countries that were further from the world technological frontier, grew faster. Moreover, this rate of arguably technology-driven convergence is more rapid in countries with higher education levels. Bils and Klenow (2000) calibrate a growth model to see if it can explain the education–growth linkage. Critically, they find that most of the causality from education to growth must be explained by education's influence on technology.

Nevertheless, econometric problems preclude a neat conclusion of this debate. For example, Krueger and Lindahl (2001) have argued that the lack of a measurable relationship between education expansion and growth may simply reflect a failure to measure human capital stocks accurately. Hanushek and Woessmann (2007) review studies with more refined data, and conclude that both the level and growth rate of education attainment matter for growth. Moreover, they develop a dataset drawn by pooling the results of several international standardized tests of skills, and using it, find that growth is robustly related to the quality of education.

One important caveat on these results, entered quite convincingly by Bils and Klenow, is that the association between education and growth could be explained by reverse causality, as richer countries-or those anticipating more investment, higher returns to education, and faster growth-invest in more schooling. To date though, no microeconomic evidence on this question of reverse causality has been drawn from the developing world.

A rather different view of the role of education in growth comes from Lewis (2004), who argues that "public debate on education is confused" (p. 243), essentially because the role of education in development is misunderstood. He defines education as "the means through which societies acquire political philosophies based on individual rights. "Any impact of such education on growth is likely to be long term. On the other hand, trainability, Lewis feels, or the capacity to learn to use new production technologies, is what matters for rapid labor productivity growth. In short, therefore, he argues, education is not a constraint on the ability of current workforces to be trained in operations with much higher productivity levels.