Intra- and Inter-sectoral Knowledge Spillovers and TFP Growth Rates

In this paper I estimate unobserved labor-generated knowledge spillovers within and among six large macroeconomic sectors covering the totality of the US civilian economy from 1948 to 1991. Unobserved spillovers are identified by observed TFP changes measured using Dale Jorgenson’s quality-adjusted factor and product panel data. I construct a series of sectoral knowledge spillover matrices that show that changes in the magnitude and direction of spillovers are associated to the productivity slowdown in the US economy of the early seventies. These matrices also allow me to compute the gap between the market and the optimal allocation of labor among sectors. Moreover, I show that sectoral market wages do not capture the totality of spillovers and I measure the difference. My measurement of spillovers shows that from 1948 to 1991 manufacturing generated knowledge for all sectors, being overall the main engine of growth. Using Samuel Kortum’s data on patent production and use in the U.S. I find that sectoral labor generated knowledge flows coincide with the sectoral patterns of other, disembodied information flows. I also find the productivity slowdown coincides with a change in the pattern of generation and diffusion of spillovers. In the mid seventies manufacturing initiates its decline as the engine of growth and trade starts to catch up as the main generator of knowledge spillovers to the economy; simultaneously, the relative weight of all intra-sectoral spillovers diminishes in favor of spillovers between sectors. Finally, I find that the market allocates resources inefficiently, as spillovers are measured to be significant. More resources should go to the main spillover generating sector, that is, manufacturing, so that employment in this sector increases by 32%, and output by 8%; and wages increase in all sectors, except for services and mining.