Max-Share Misidentification

Max-share identification relies on a decomposition of the forecast error variance (FEV) over a target horizon. Consequently, it often conflates multiple shocks because the contribution to the FEV depends on the impulse responses at untargeted horizons and the shapes of the responses to untargeted shocks. We alleviate the issues using a socalled “single horizon” alternative that focuses narrowly on the actual target horizon. We characterize the identified shock in terms of true structural shocks in the single horizon problem and show that this typically bounds results in the literature’s usual implementation. Using a numerical demand and supply example and an empirical news shock application, we show that the traditional max-share approach inadvertently places weight on untargeted transitory shocks, a problem that the single horizon approach avoids.