Combining Proxies and Narrative Sign Restrictions: Revisiting the Effects of Technology Shocks

ABSTRACT This paper proposes a novel integration of the proxy structural vector autoregression approach with narrative sign restrictions to identify multiple shocks with multiple instrumental variables. We show that this combination leads to informative inferences on shock effects through both analytical demonstrations and Monte Carlo simulations. Applying this method to identify anticipated and unanticipated technology shocks, we find that anticipated shocks account for over 45% of stock price variation but only about 13% of total factor productivity (TFP) fluctuations. The majority of TFP variations are driven by unanticipated technology shocks and measurement errors.