New method measures credit reallocation in Italy

The paper introduces a novel methodology to directly measure credit reallocation between firms and banks following shocks.

It addresses the "reflection problem" in decentralized markets by leveraging the sparsity of the credit network, where not all firms borrow from all banks.

This allows for the construction of partially overlapping portfolios instrumental variables (OPIVs), which break simultaneity and mitigate bias in treatment and fixed-effects estimates.

The approach generalizes the widely used Khwaja and Mian (2008) model by not constraining cross-elasticities to zero, making it applicable to standard credit register data without relying on specific quasi-experiments.

This context-independence enables repeated measurement over long time-spans.

Applying their estimator to twenty years of data from the Bank of Italy's Credit Register (2002-2022), the authors found that firms' credit cross-elasticities are negative in normal times, indicating substitutability, but often become positive during recessions, suggesting complementarity.

Conversely, banks' cross-elasticities are predominantly positive and do not vary with the business cycle.

This evidence systematically rationalizes the asymmetric real effects of credit expansions and contractions documented in the empirical banking literature, where firms struggle to substitute credit during downturns, leading to more pronounced negative real effects.