Algorithm-Driven SVARs: Navigating the Wilderness of Big Data

The paper argues that SVAR results depend not only on shock-identification restrictions but also on which variables enter the system, a choice usually made manually. It proposes a Bayesian method that constructs information sets, uses an out-of-sample criterion, and retains the largest admissible system. Under recursive identification, output responds to housing production rather than household credit alone. A joint Bayesian proxy SVAR with multiple instruments, without an anchor variable, strengthens the credit-spread channel. Adding a corporate spread to a core system identifies expected default risk as an important transmission margin.