Measuring human capital using global learning data

A large number of studies in the development accounting literature have explored the relative contribution of human capital to cross-country income differences. However, the results have been inconclusive, in part owing to difficulties in measuring human capital. Although direct measures of years of schooling exist, the quality of schooling has been harder to measure.

Approaches to estimate the quality of schooling across countries have relied on differences in Mincerian wage returns, immigrant returns and cross-country skill premia. However, these approaches face several challenges, including the need to make assumptions about the substitutability of skilled and unskilled workers. The challenges in measuring quality have contributed to substantial variation in estimates of the role of human capital in accounting for cross-country differences in income, ranging from nearly all to potentially none.

In this study, we provide a more direct and reliable measure of the quality of schooling based on our learning outcome data, which we use to construct measures of human capital stock (Methods).

Our results (Table 1) suggest that human capital accounts for between a fifth to around half of cross-country differences in income—which is an intermediate position relative to the estimates found in the literature, which range from zero to nearly all. These results are consistent with models of human capital that capture the role of educated entrepreneurs and more comprehensive measures of human capital that include schooling, learning and health.

The average relationship between learning and income masks significant heterogeneity across countries (Extended Data Table 2). We find that human capital explains less than two-fifths of cross-country income differences among low-income countries, but more than half among high-income countries. We find even larger differences across regions. For example, when measured by schooling, human capital accounts for 54% of cross-country income differences in advanced economies and only 4% in sub-Saharan Africa. When we include learning, this gap widens to 86% in advanced economies but only 10% in sub-Saharan Africa. This substantial heterogeneity reveals the importance of including a global distribution of countries covering multiple stages of economic development when examining the role of human capital.

Finally, we compare our measure of human capital to alternative measures that are based on prominent global databases such as the Penn world tables, the Barro–Lee Educational Attainment dataset and the United Nation’s HDI. In Table 2, we find that our measure of human capital has a stronger association with growth than alternative human capital measures. This is the case in univariate regressions that include each measure on its own (columns 1–4). We observe that a change of 1% in learning is associated with a change of 7.2% in annual growth. By contrast, a change of 1% in the other human capital measures is associated with a change of between 1.6% and 3.3% in annual growth. We further show that when we include all measures in the same multivariate regression, the association between our measure of learning and growth remains high, between 5.9% and 6.9%, and statistically significant (P ≤ 0.01), whereas other human capital variables have a reduced and statistically nonsignificant association with growth. We find that the model fit improves only slightly when all measures are included with an R² value of 0.32 relative to our measure of human capital with an R² value of 0.30.

Therefore, our measure of human capital appears to have a stronger relationship with economic growth, both individually and jointly. This is probably because alternative measures of human capital rely largely on years of schooling and omit learning. However, the use of these alternative measures remains standard practice, in part because these data have the broadest coverage. By constructing learning data across 164 countries, we fill a key gap: broad coverage over nearly two decades and a measure of human capital with strong links to economic development.