A Bayesian Model for Estimating Spillover Effects in Unconventional Monetary Policy

Organised by

• Department of Economics

Motivated by the assessment of the European Central Bank's Corporate Sector Purchase Program (CSPP), we propose a Bayesian model for estimating spillover effects when they arise from general economic equilibrium mechanisms. The methodological proposal extends some recent causal analysis methods that are based on factor models for panel data, and it relies on the use of a set of units, referred to as "additional controls," for which we hypothesize a zero probability of eligibility for the policy and immunity from the spillover effects. These units (bonds in the specific application) can be used to predict counterfactual outcomes for both eligible and non-eligible units, as if the policy had never been implemented. By doing so, we can evaluate the overall (direct plus spillover) effects for the eligible group and the spillover effects for the non-eligible group. For inference, we rely on a longitudinal Bayesian causal factor model which offers the flexibility to accommodate various functional forms and integrates shrinkage priors for model feature selection. Our results suggest that the CSPP had a negative impact on the yield spreads of euro-denominated bonds eligible for purchase and dollar-denominated non-eligible bonds, with no significant difference between the two effects.

Co-authors: Sharmistha Guha (Texas A&M University), Taneli Mäkinen (European Central Bank)