This post is the second in a series. To read the first post, click here.
In this addendum, we apply the same analysis we used in “Are Central Bank Financial Stability Reports Useful? An AI-Based Analysis” to the case of the Swiss National Bank (SNB).
SNB has been publishing an annual Financial Stability Report since 2003.[1] The report has important consequences for bank supervision and regulation. Its analysis forms the basis of potential changes to the countercyclical capital buffer (CCyB); based on that analysis, the SNB recommended to the Federal Council that the CCyB be activated in 2013, increased in 2014, suspended in 2020 and reactivated in 2022.[2] In conjunction with the Swiss Financial Market Supervisory Authority, the SNB’s assessment of risks may also result in specific supervisory actions to be undertaken with banks.[3]
How Well Did the Financial Stability Reports Do?
Subjected to the same analysis outlined in our earlier note, the SNB Financial Stability Reports performed no better than the reports produced by the Fed, ECB or BoE. As the table in the appendix in the table shows, the reports uncovered no novel risks that were eventually realized. The reports identified risks to mortgage markets that were used to justify an activation and increase in the CCyB in 2013 and 2014, but those risks did not materialize in the form of any significant drops in real estate prices or significant increases in bank losses. The reports identified risks to mortgage markets that were used to justify an activation and increase in the CCyB in 2013 and 2014, but those risks did not materialize in the form of any significant drops in real estate prices or significant increases in bank losses. The 2016 edition of the report focused on the risks to residential real estate investment, but there were no subsequent significant market stresses or bank losses. Interestingly, the risks to residential real estate investment reported in 2016 by the SNB had already been flagged well before by UBS in 2015.[4] Like the other central banks’ financial stability reports, the SNB financial stability reports failed to anticipate the risks leading up to the financial crisis of 2008-09.
Thus, since 2003, there have been 1) no examples of successful forecasts of novel risks; 2) multiple forecasts of risks that did not materialize and therefore imposed an unnecessary tax on lending through higher capital requirements; and 3) substantial risk identification misses leading up to the financial crisis. Therefore, the reports demonstrate zero or negative regulatory alpha.
As in our earlier note, we emphasize that this is (surprisingly) a first-of-its-kind effort to use AI to assess the historical performance of financial stability reports. We show our work so that others can question the assumptions and perform their own analysis. We hope that others will continue this type of research.
Implications
Article 44 of Switzerland’s Capital Adequacy Ordinance confers broad powers on the SNB to activate the CCyB when the risks of “excessive” credit growth[5] must be managed. The SNB in response has developed a discretionary risk identification framework[6] that depends on its ability to successfully determine risk measures for lending market imbalances that have forecasting ability to predict significant market and bank stresses. The results of that annual analysis are discussed in the financial stability reports.
However, the over-20-year track record of the financial stability reports suggests the SNB has demonstrated no ability to dynamically identify realistic risks. The SNB should therefore reassess its discretionary framework to set the CCyB. It should also refrain from using the financial stability reports as a guide for potential supervisory actions in its discussions and recommendations to FINMA.
Appendix: LLM Assessment of SNB Financial Stability Reports
Categories: 1 = successfully forecast; 2 = forecast as significant but did not subsequently prove significant; 3 = missed, under-ranked, too weakly forecast or identified without forecasting subsequent deterioration.
[1] See the description at https://www.snb.ch/en/the-snb/mandates-goals/financial-stability/swiss-banking-sector
[2] See https://www.snb.ch/en/services-events/digital-services/faq-overview/qas_finstab#t05
[3] Article 44 of Switzerland’s Capital Adequacy Ordinance governs the SNB’s ability to change the CCyB based on its views on systemic risk. See https://justement.ch/doc/act/ch/952_03/tit_3/chap_1/art_44. In addition, the “Memorandum of understanding in the field of financial stability” governs how the SNB and FINMA collaborate and exchange information to regulate banks. See https://www.finma.ch/en/enforcement/recovery-und-resolution-n/nationale-zusammenarbeit/
[4] Le News reported the UBS analysis on August 4, 2015, under “Swiss real estate bubble index rises significantly,” describing an “investment property boom.” See https://lenews.ch/2015/08/04/swiss-real-estate-bubble-index-rises-significantly/
[5] See https://elischolar.library.yale.edu/cgi/viewcontent.cgi?article=1517&context=ypfs-documents
[6] Swiss National Bank, “Implementing the countercyclical capital buffer in Switzerland: concretising the Swiss National Bank’s role,” 2014, available at https://www.snb.ch/dam/jcr:f67f123e-fe4f-4d20-85ea-f06cb5cd3689/ccb__communication.en.pdf
