The middle productivity trap: Dynamics of productivity dispersion

A few months before leaving office President Barack Obama wrote about the challenges that his successor would have to tackle [1]. Recent innovations, he claimed, “have not yet substantially boosted measured productivity growth. “In fact, since 2004, productivity growth slowed across nearly all advanced economies. Productivity being the most important determinant of economic growth, Obama concluded: “Without a faster-growing economy, we will not be able to generate the wage gains people want, regardless of how we divide up the pie.” The challenge ahead of us begs a better understanding of productivity growth dynamics, which is what this paper aims to do. 

An important set of studies have looked at the possible causes of the slowdown in productivity growth, including mis-measurement, the role of recent innovation in boosting (or not) productivity, the existence of increasing market frictions or decreasing dynamism, and of structural vs. cyclical economic factors. Beyond dynamics in average productivity, large productivity dispersion within narrowly defined sectors has been widely documented. Also, Decker et al. have documented increasing productivity dispersion since the 1990s in the U.S. Yet, there is no consensus on what explains increasing productivity dispersion across time.[2] Increasing productivity dispersion could be a result lack of convergence, which could be interpreted as lack of diffusion of innovations across firms. In fact, Comin and Mestieri show that penetration rates of technologies have declined using macro data. 

Consistently, my working paper documents a robust, yet undocumented thus far, stylized fact linking the structural pattern of firm-level productivity growth to within-industry dispersion: convergence-divergence dynamics, where the fastest firm-level TFP growth rate is concentrated at both extremes of the initial productivity distribution, generating a “U-shaped” convergence curve.

end notes


[2] For Hsieh and Klenow (2009) dispersion is a result of misallocation, which could be interpreted as a static increase in dispersion, unless this misallocation worsens over time.