A1815
Title: Testing instrument validity in proxy-SVARs
Authors: Luca Fanelli - University of Bologna (Italy) [presenting]
Giuseppe Cavaliere - University of Bologna (Italy)
Giovanni Angelini - University of Bologna (Italy)
Abstract: A pre-test of instrument validity in proxy-SVARs (SVAR-IVs) is designed based on bootstrap resampling, free from the usual pre-testing issues. The null hypothesis is that (i) the proxies used for the instrumented structural shocks identify the proxy-SVAR, hence are `strong' (relevance condition), and that (ii) the proxies are uncorrelated with the non-instrumented structural shocks (exogeneity condition). The tests for (i)-(ii) are run sequentially: a bootstrap test for relevance is computed and, conditional on not rejecting the null, a bootstrap test for the exogeneity condition is computed without relying on additional external information other than the proxies used for the shocks of interest. Notably, the outcome of the bootstrap pre-test of relevance does not affect asymptotically the bootstrap test of exogeneity and both do not affect post-test inference. A test for the exogeneity condition (ii) is also designed for cases where the test of relevance (i) rejects the null and the inference in the proxy-SVAR is carried out by weak-instrument robust methods.