Older and Slower: The Startup Deficit's Lasting Effects on Aggregate Productivity Growth

The more than thirty-year decline in the rate of new employer business creation, and with it the significant shift in the firm age distribution of U.S. employers have both slowed U.S. labor productivity growth. This gradually accumulating drag on productivity growth was obscured by the surge in TFP in the late 1990s and early 2000s (see Fernald 2014). Earlier work by Baily et al. (1992) and Foster et al. (2006) among others has shown the crucial role of reallocation among establishments, primarily from entry and exit, in explaining aggregate productivity growth within the manufacturing and retail trade sectors. In this paper, we use administrative Census data with near universal coverage of the non farm private sector to assess the effects on aggregate productivity from the direct effects of the decline in the business entry rate and its indirect effect on the age distribution. We follow the methodology of Haltiwanger et al. (2016) and merge the Census Longitudinal Business Database (LBD) with data on annual sales from IRS records in the Census Business Register to construct a revenue-enhanced LBD with annual firm level measures of real sales per worker. Using these measures as rough proxies for labor productivity within narrow NAICS industries, we quantify how the entire distribution of labor productivity evolves by firm age as well as other firm characteristics by extending the dynamic Olley-Pakes decomposition in Melitz and Polanec (2015). Looking across the entire private-sector economy, we have three main findings. First, we document that the age-profile of labor productivity growth of firms between the ages of 1 and 19 is downward sloping and convex. Productivity growth is highest among young firms and declines quickly with firm age. Since startups have higher productivity growth rates, a decline in the startup rate lowers aggregate productivity growth. Second, we document that this age profile of firms aged 1-19 and the pattern of selection is stable across our sample. This means that the slow down in aggregate productivity growth has not come from changes in the age profile of productivity growth for these age cohorts. Third, the contribution of mature firms (age 20+) to aggregate productivity growth has declined significantly since the mid-2000 and the source of this decline has been a slowdown in allocative efficiency (the most productivity mature firms have not gained market share relative to less productive firms). We use these cross sectional results to calibrate an equilibrium growth model with heterogeneous firms and to perform simple counterfactuals. The model allows us to estimate the aggregate consequences of the declines in entry for aggregate productivity in the time series. Using the model we quantify the total effect of the decades long startup deficit on aggregate productivity growth. Simple counterfactuals suggest that the decline in the startup rate can explain a significant fraction, though not all, of the aggregate productivity growth slowdown.