A Bayesian model to assess spillover effects of unconventional monetary policies

Motivated by the evaluation of the European Central Bank’s corporate sector purchase program (CSPP), this paper proposes a model to assess causal effects whenthe Stable Unit Treatment Value Assumption (SUTVA) is not plausible due to interference between units arising from spillover effects. The proposal involves the implementation of a factor model for panel data within a Bayesian framework, where the latent structure is identified through a Lasso shrinkage mechanism. The model is applicable to policies where it is reasonable to assume the absence of global “general” equilibrium effects. Therefore, it is possible to find units not affected by the policy in question, and use them as “additional controls”, specifically as observations used to predict counterfactual outcomes for both the eligible and non-eligible units of interest. Under the proposed assumptions for identifying treatment and spillover effects, simulation studies demonstrate the strong performance of the method. The application to the real dataset reveals significant treatment and spillover effects; the two effects are moreover not statistically different from each other, indicating that the CSPP strongly affected also the prices of non-eligible bonds of euro area issuers.