The efficiency of human capital allocations in developing countries

a r t i c l e i n f o For a set of 14 developing countries I evaluate whether differences in wage gaps between sectors – estimated from individual-level wage data – have meaningful effects on aggregate productivity. Under the most generous assumptions regarding the homogeneity of human capital, my analysis shows that eliminating wedges between wages in different sectors leads to gains in output of less than 5% for most countries. These estimated gains of reallocation represent an upper bound as some of the observed differences in wages are due to unmeasured human capital. Under reasonable assumptions on the amount of unmeasured human capital the gains from reallocation fall well below 3%. Compared to similar estimates made using data from the U.S., developing countries would gain more from a reallocation of human capital, but the differences are too small to account for a meaningful portion of the gap in income per capita with the United States. A prominent feature of developing economies is the wide variation in labor productivity between different sectors. Gollin, Lagakos, and Waugh (2013) document that agricultural labor productivity is typically about one-half the level of non-agriculture, echoing the work of Lewis (1954) on dual economies. Moreover, this productivity gap appear appears to be most pronounced in the poorest countries, a fact noted by Kuznetz (1971), and explored further by Gollin et al. (2002), Caselli (2005), and Restuccia et al. (2008). Beyond just agriculture and non-agriculture, McMillan and Rodrik (2011) document that large differences in labor productivity exist across ten broad sectors within each country they study. 1 If these disparities reflect real differences in the marginal productivity of labor between sectors, then aggregate productivity is lower than its potential, offering a partial explanation for low measured total factor productivity in developing countries. Several papers have attempted to estimate the loss in aggregate productivity from such misallocations. Fo-cusing on only the distinction between agriculture and non-agriculture, Chanda and Dalgaard (2008), Vollrath (2009a), and Cordoba and Ripoll (2009) all suggest that there are substantial losses due to misallocations, while Caselli (2005) and Graham and Temple (2006) find much smaller effects. 2 The existing literature infers differences in the marginal product of labor and/or human capital between sectors based on aggregate level information on output and labor inputs. Gollin et al. (2013) are the most sophisticated in addressing the measurement issues arising from the use of …