We have frequently demonstrated that the focus of the Fed’s stress tests on bank performance under a single apocalyptic recession scenario overstates the risk of assets that perform particularly poorly in severe recessions, while underestimating the risk of less cyclical bank assets. Such a bias creates strong incentives for banks to shift away from cyclically sensitive assets, such as small business lending, and Covas (2017) shows that such a shift is exactly what is happening.
Another way to identify such a bias would be to determine how banks that concentrate in small business lending – namely, smaller banks – would have performed under the CCAR stress test’s severely adverse scenario. However, because banks with less than $10 billion in assets are not subject to CCAR or company-run stress tests we don’t have such data.
But we can make an informed estimate. In our previous work, Covas (2017) was able to estimate the implicit capital requirements of the Fed’s CCAR stress tests using the post-stress capital ratios published by the Federal Reserve from 2014 through 2017. Specifically, that analysis estimated the risk-weights that would best describe banks’ post-stress regulatory capital ratios under the severely adverse scenario, controlling for banks’ own capital actions and their capital levels at the start of the stress tests. Moreover, that analysis estimated the risk-weights for major loan portfolios such as commercial real estate loans, small business loans, residential real estate loans, and securities. Importantly for what follows, the research demonstrated that the stress tests imposed particularly high requirements on lending, especially lending to small businesses, when compared with Basel standardized models or banks’ own models.
To see how small banks would have performed in the quantitative portion of the most recent test, we applied the risk weights obtained using the reverse-engineered model to all banks’ exposure amounts and estimated their post-stress tier 1 capital ratios under the severely adverse scenario applied to large banks. For this analysis, we used those same implicit risk-weights that were estimated using the Federal Reserve’s Dodd-Frank Act Stress Tests (DFAST) results, which allows to more accurately control the impact of banks’ capital actions on the estimated risk-weights under stressed conditions.1 While a bank passes the actual quantitative portion of the stress test if its post-stress ratio is above 6.0 percent, the 2018 test was markedly more difficult than the tests used to estimate the reverse-engineered model. In a previous blog post, BPI estimated that the increased stringency of the 2018 test was approximately equivalent to raising the stress-test quantitative hurdle rate 0.5 percentage points for non-GSIBs subject to CCAR, and over 1 percentage point for GSIBs. To see how small banks would have performed in the 2018 quantitative test, we therefore checked to see if their estimated post-stress Tier 1 capital ratio was above 6.5 percent. The results are shown in Table 1.
Results of stress tests using reverse-engineered model
(data as of 6/30/2018)
|Number of banks
|Number of failures
|Less than $10 billion
In sum, approximately 800 banks with less than $10 billion in assets – nearly one-fifth – would have failed the 2018 stress tests. As shown in table 2, on average, the failed banks have a higher share of loans-to-assets relative to non-failed banks. In addition, failed banks have a higher share of commercial real estate loans and small business loans and a lower share of securities.
(data as of 6/30/2018)
|Tier 1 Capital Ratio
|Below $10 billion, Pass
|Below $10 billion, Fail
In sum, this analysis demonstrates again that the CCAR stress test creates significant incentives for banks to substitute away from small business lending and certain other asset classes whose implicit risk-weights under the test are significantly higher than appears justified. Smaller banks can continue lending in those asset classes because they are exempt from the test. Unfortunately, as noted elsewhere, they cannot substitute for the loss of activity at larger banks.
1In particular, under DFAST share buybacks are assumed to be zero, and dividends on common and preferred stock are known. The estimated implicit risk-weights used in the blog post are: 226% for C&I loans, 190% for CRE loans, 350% for small business loans, 100% for RRE loans, 110% for consumer loans, 240% for trading assets and 80% for securities.
Disclaimer: The views expressed in this post are those of the author(s) and do not necessarily reflect the position of the Bank Policy Institute or its membership, and are not intended to be, and should not be construed as, legal advice of any kind.