The March problems of banks in the US and Switzerland for the first time caused a noticeable wave of preprints written by economists “on the topic of the day.” The failures of Silicon Valley Bank (SVB) and Credit Suisse (CS) have already been analyzed in dozens of academic texts written in one to two weeks, with the usual timeframe for such publications being at least several months. If the trend continues, the question of the quality of “instant scientific papers” for economists will very quickly become relevant.
While regulators are still discussing whether the SVB and CS bank crashes in March 2023 can be considered the beginning of a serious banking crisis, this issue has already been resolved for some economists.
What is happening has already led to dozens of publications of preprints of articles by economists (which did not pass the peer review procedure). Including in a series of publications by NBER, the largest American economic association, two texts were published in March that directly dealt with this problem.
The speed of writing “instant scientific papers” increased dramatically during the COVID-19 pandemic, when the first papers by economists on this very topic were released in April-May 2020 (i.e., written in two months). But the history of SVB and CS has reduced this period to one or two weeks.
March paper by UCLA’s Erika Xian, NWU’s Gregor Matvos, Columbia University’s Tomasz Piskorski, and Stanford’s Amit Seru (all from university business schools) “Money squeeze and US banking fragility in 2023: valuation losses and uninsured deposit flight ? looks like an attempt to simulate (on the existing model for 190 US banks, apparently created for other purposes) a repeat of the SVB collapse in other US lending institutions.
The authors come to a rather banal conclusion that the losses of the bank and its affiliated group by themselves, without the flight of uninsured deposits, could not lead to bankruptcy, and SVB was just characterized by a high proportion of such deposits – only 1% of US banks it is higher, with such a rate of “bank flight” as in California, 190 banks in the United States and deposits in the amount of $ 300 billion are at risk of bankruptcy (the “roundness” of the figure indicates the caution of the authors in such a forecast).
Formally, the article states that the current situation with the loss of the value of assets on balance sheets by banks “very significantly increased the fragility of the US banking system in terms of vulnerability to “bank runs” by owners of uninsured deposits.”
The method of Xian, Matvos, Piskorski, and Seru is uncomplicated, the main advantage of the text, which allowed it to appear in one of the most authoritative series of preprint publications, is probably the NBER-encouraged speed of the appearance of a preprint on a hot topic.
The second March text is a preprint by Andrew Metrick of Yale and Paul Schmeltzing of Stanford (both of whom also work at university business schools) titled “March 2023 Banking Sector Interventions in a Long-Term Context – Silicon Valley Bank and Beyond.”
Here, the authors use their own database of banking interventions from 1257 to 2019 to analyze the collapse of the SVB and, for the first time, CS (the existence of “banking interventions” in the 13th century, however, is no less debatable than the fundamental comparability of fractional reserve banking institutions at that time and systems involving central banks since the middle of the 19th century), stating that injections of the Bank of Switzerland into CS sold by UBS are among the top 20 central bank interventions of this type included in the database (293 in total) and twice as large as the average liquidity injection into a troubled bank of 57 major incidents of this kind.
The last document taken into account by Metric and Schmelzing, which is referred to in the text, is dated March 19, 2023, i.e. the article was actually written in no more than a week – the authors managed to read the preprint of Xian, Matvos, Piskorski and Seru and refer to it. Probably, the main value of the text is an indication of the created database and, again, acute relevance.