Abstract

We document that sectoral TFP growth, despite being largely idiosyncratic and transitory, comoves with real GDP growth two to four years ahead in a way that varies notably across sectors. To explain this finding, we derive a dynamic generalization of Hulten's theorem, decomposing each sector's Domar weight into an `impact' component, capturing the contemporaneous response of aggregate consumption to TFP shocks, and a `propagation' component, capturing the response of future GDP by way of investment and capital accumulation. Focusing on a tractable multisector growth model, we show that for each sector, the effects of  TFP shocks on future GDP growth involve a weighted sum of all Domar weights in the economy. The weights in this representation reflect sectors' positions in the intermediate-input and investment networks and, importantly, the capital intensity of downstream users.  When informed by U.S. industry data, the model reproduces the fact that productivity growth in Construction leads GDP growth up to two years ahead, while shocks in large sectors such as Services have large contemporaneous effects but negligible persistence. Thus, the medium-term effects of sectoral TFP shocks on aggregate GDP can differ substantially from their contemporaneous effect. Over a four-year horizon, the cumulative aggregate effects of shocks originating in goods-producing sectors are more than twice as large as those in service-producing sectors, despite goods producers representing less than 1/3 of GDP.

Download

• Paper