This would would have not been possible without my great coauthors Andrea Orame (andreaorame.github.io) and Rodney Ramcharan (sites.google.com/site/rodneyr...)

@fuster.bsky.social, Teodora Paligorova, and @vickeryjames.bsky.social study how banks manage their security portfolios, and their results confirm that historical cost accounting reduces the incentive to trade securities with unrealized losses (they have a lot of other interesting results too!)

Dan Greenwald, John Krainer, and Pascal Paul find the same result about the impact of security valuation on bank lending (we study Europe, they look at the US). They do much more than we do, such as looking at hedging and building a structural model to estimate aggregate effects

It's also great to see other papers that are looking into similar issues.

This is likely because selling an "underwater" security with a market value less than historical cost implies recognizing the loss into regulatory capital, whereas keeping it in the balance sheet has essentially no impact on regulatory capital.

We also show that accounting rules affect banks' trading behavior. When available-for-sale securities are valued at historical cost, banks do not sell them in response to the QE announcement, but they do when they are marked-to-market.

... but there is a large pass-through when mark-to-market accounting is used. The channel works through capital gains that enter into regulatory capital. This is true not only for QE (in Italy/Europe, our focus) but also for any shock to the long-end of the yield curve.

When monetary policy changes (long-term) yields, it affects the market value of securities. The main result is that the pass-through on bank lending (through the bank lending channel) is essentially zero when securities are evaluated using historical cost accounting...