Zusammenfassung:
After the 2008 global financial crisis, U.S. bank holding companies needing to cover larger-than-expected loan losses raised concerns that existing provision accounting may be procyclical. Most related studies have found evidence of procyclicality using either aggregate time-series data or 'as-reported' panel data. We test the null hypothesis that provisions were a constant fraction of nonperforming loans across the economic cycle. We create a 'forced' panel, which incorporates the entities acquired by each holding company in the quarters prior to their mergers. As in the related literature, we fail to reject the null hypothesis with 'as-reported' data; however, we reject the null hypothesis with the 'forced' panel. This finding suggests that holding companies built up provisions to some degree during the pre-crisis period to cover larger future losses. These actions reduced capital and likely depressed lending in the pre-crisis period; such countercyclical impacts are consistent with post-crisis macroprudential policies.