Identification of technology shocks using misspecified VARs

摘要

Studies that evaluate the effects of technology shocks often employ structural VARs identified with long-run restrictions. In the presence of a mismatch between the lag structures of the true data-generating process and the adopted VAR, estimates based on long-run restrictions can be biased. This paper offers a method that can reduce this bias substantially. Using artificial data, I assess the performance of the proposed method and find that it can outperform a range of alternative procedures. Applying the procedure to the US data, I find that per-capita hours exhibit a positive hump-shaped response profile in response to a technology shock.