Over the last few decades, the opening up of business micro data by national statistical agencies, and the vast improvement in computer power to store and analyze very large and complex datasets have facilitated the careful documentation of the enormous variation in productivity across countries, firms, and time. Figure 1 shows the correlation between GDP per capita and Total Factor Productivity for a large number of countries (Jones and Romer, 2010). It is clear that those countries with high TFP are also the countries with high GDP per capita, suggesting that TFP is important for understanding cross-country success. Development accounting (e.g., Caselli, 2005) focuses on how to account for these large cross-sectional differences across countries, but a puzzle remains: observables such as human and nonhuman capital seem unable to account for the large GDP per capita differences observed across countries.
Total factor productivityIn economics, TFP is a fraction of total output, such as GDP, that is not explained by the aggregate in-puts of production. It is particularly studied in macroeconomics as it highly affects economic growth. TFP growth is usually measured by the Solow residual.
Aggregate TFP differences across countries are also influenced by how different economies allocate output to plants of heterogeneous productivity levels. For example, Figure 2 shows the estimated productivity distribution of the manufacturing sectors in the United States and India (Hsieh and Klenow 2009). Compared to the United States, India appears to have a much longer left tail of low-productivity plants. This suggests that something about the structure of the Indian economy allows less-productive plants to survive more easily than they do in the United States.
At the micro level, a substantial body of evidence shows persistent heterogeneity in firm productivity (and other dimensions of performance) in narrowly defined industries in many countries and time periods (e.g. Foster, Haltiwanger, and Syverson 2008; Bartelsman and Dhrymes 1998). Differential observable inputs, heterogeneous prices and idiosyncratic stochastic shocks are not able to adequately account for the remarkable dispersion of productivity. So, what else could account for these persistent productivity differences? One of the possible causes of productivity differences that has been the focus of much of the growth literature is “hard” technologies. This refers to the generation of new technologies, as proxied by measures of R&D or citation-weighted patents, or the adoption of technologies, as proxied by use of, for example, hybrid corn, new drugs, or information and communication technologies (ICT).
Differences in hard technologies, however, are not able to fully account for productivity spreads for at least two reasons. First, even after controlling for a host of observable technology measures, a very large TFP residual remains. Second, the impact of observable technologies seems to vary systematically with the management and organization of the firm. This has been seen most clearly in studies of the effect of ICT on productivity (e.g., Bresnahan, Brynjolfsson, and Hitt 2002). The effects of ICT on productivity range widely, and the impact seems to be much higher when firms are more decentralized and have stronger “people management” practices—structured hiring policies; and a strong emphasis on ability and effort when determining promotion, and dealing with underperformance and pay (Bloom, Sadun, and Van Reenen, 2012). The generation and diffusion of hard technological innovations are therefore unlikely to be the only drivers of the productivity dispersion observed across firms and countries. Another important factor could be “soft” technologies such as management practices.
