Goldman Sachs Fails Fed Stress Test

Cross-post from The Institutional Risk Analyst

Last week, the Federal Reserve Board released the annual bank stress tests mandated by Dodd Frank. The big headline is that Goldman Sachs (GS) failed the stress test, but there are a number of other banks of concern as well. The Fed's bank stress tests are a bit of a mess in terms of analytical methodology, one reason we rarely look at them. The description of how the Fed performs the stress tests on Page iii of the report illustrates the problem.

The caption at the bottom of the graphic above tells you that the Board staff will "adjust" the already subjective bespoke inputs from the banks and then produce "custom" economic scenarios? The Fed is clearly conflicted by conducting these discussions in private. We don’t need an economist or an economic scenario to construct a bank stress test. Indeed, the focus on speculative economic scenarios as the point of departure for the Dodd-Frank bank tests clouds the analysis and renders the whole exercise irrelevant in terms of understanding the bank.

To conduct a classical stress test for a bank, all you need is a loss rate and a time horizon. The net income, credit provisions and counterparty losses in the Fed results are useful in this regard. But why the losses occur really does not matter. The Fed has chosen to inject a huge degree of subjectivity into a legally mandated public policy analysis of bank soundness. By employing bespoke economic scenarios produced in secret with no transparency or accountability, the whole bank stress test process is compromised.

Given the clear conflict by the Fed staff in Washington, the stress test results for GS are rather alarming. This is the test result for Goldman that the Fed assigned after the private consultations and hand-holding with Board staff? Notice that GS ended the stress test with a Tier 1 Leverage Ratio of 5.3%.

Readers of The IRA know that the public data provided by large US banks provides more than sufficient information to conduct a stress tests w/o any economic inputs. The bank-specific data provided in response to the Fed-generated DFAST economic scenarios are a hodgepodge. There is no comparability from one bank to the next. Unlike the consistent public data reported by all large US banks, the Fed's stress test is a deeply biased chopped salad in analytical terms.

By using economic scenarios and then "adjusting" the particular, non-standard inputs from each bank, the output becomes entirely speculative and is useless from an analytical perspective. Remember, bank stress tests are meant to be a public process and not part of the confidential bank supervisory relationship. Since few members of Congress understand finance or economics, nobody notices the obvious analytical bias shown to each bank, unevenly, and other shortcomings of the bank test process required by Dodd-Frank.

Anther big drawback of the Fed stress tests is that they work with risk-weighted assets, the four-decade old Basel III construct that focuses on credit risk for assets in the event of default. But the loss rate assumptions that are published in the test are instructive, however, both about how the Fed thinks about a specific bank and the US economy more generally.

Below we show the key results from the Fed stress tests for the 24 largest domestic US banks and also the US businesses of seven foreign banks. Memo to Fed Chairman Jerome Powell: Next year, if you want to dispense with the rest of the report and just show us the spreadsheet below, that would be great. First and foremost, notice which banks ended the stress scenario with higher capital and/or net earnings. Earnings get you through periods of loss, not capital.

Federal Reserve Stress Test Results

Source: Board of Governors

As we all learned with Silicon Valley Bank, credit risk is not the only problem facing banks. You can destroy a large regional bank by gambling naively on interest rates via investments in mortgage-backed securities (MBS). The risk weight on Ginnie Mae MBS is zero under Basel. But the market risk of these variable duration securities makes them among the most problematic assets a bank can own. This is especially true if the people running the bank happen to be idiots, to paraphrase Mark Twain. But we digress.

Notice that the Fed stress test is silent on trillions of dollars in mark-to-market losses facing US banks from COVID-era MBS and US Treasury debt. How can the Fed report to Congress on the state of US banks and not mention that many institutions (and the central bank itself) are insolvent? This sad state is due to radical swings in interest rates engineered by the FOMC since Q1 2020. The chart below comes from our earlier comment with the full mark-to-market analysis for all US banks ("Q2 2024 Earnings Setup: JPM, BAC, WFC, C, USB, PNC, TFC").

Source: FDIC/WGA LLC Adjusted capital = tangible capital less 17.5% M2M haircut on total HTM loans/securities in Q1 2024.

Just by coincidence, Stephen Gandel and Joshua Franklin of the FT wrote an important piece last week reporting on the results for Goldman Sachs. They suggest that the bank is attempting to change the Fed stress test results. "Goldman Sachs faces long odds in getting the Federal Reserve to reconsider its disappointing grade in this year’s bank stress tests, according to regulatory experts," they relate.

We reminded readers in our last missive (“Who Leads the Asset Gatherers? | SCHW, MS, GS, AMP, RJF & SF“) that GS continues to have the worst credit performance of the group. We regret to say that the Fed test results confirm our view and more. If you page through the individual bank stress test results, you will see that GS is clearly the outlier in the group in terms of assumed credit and counterparty risk losses through to Q1 2026.

Note that monoline card issuer Discover Financial (DS) has a $21 billion credit loss provision assumed, but the bank's strong profitability renders any concerns moot. The bank's capital increased during the DFAST test simulation, which means that net income exceeded credit costs. There is a reason why DS is consistently among the top bank stocks in the WGA Bank Top Indices.

Federal Reserve Stress Test Results ($B)

Source: Board of Governors

“Goldman in some ways fared better on the Fed’s stress test than it did on its own, which the banks also had to disclose last week,” the FT reports. “Its internally run stress test predicted a bigger drop in revenue and larger trading losses in an economic downturn that the one run by the central bank.”

While other, large banks such as Citigroup (C) were able to negotiate the stressed scenario losses imposed by the Fed’s test, Goldman missed the target completely and ended up below the minimum capital level for the bank. Remember, each test is different for each bank, thus the test results suggest how the Fed views a given bank business model. Just remember that the Fed stress tests are not a fair comparison of the soundness of one bank to another.

The FT reports that GS may not be able to increase dividends or share buybacks as a result of its poor test performance. Perhaps the Fed has finally discovered that GS is one of the most risky banks in the US. Sure, Morgan Stanley (MS) has a slightly bigger derivatives book than does Goldman, but rarely reports a significant credit loss.

Goldman under CEO David Solomon, however, is shoveling money into the furnace in terms of credit losses even with an 11% gross loan spread. Notice, by comparison, that MS barely saw a reduction in capital and reported a small loss relative to GS and other universal banks. Citi compares well with the other top-five moneycenter banks.

Federal Reserve Stress Test Results ($B)

Source: Board of Governors

In the stress test, Citi had a $47 billion slug of income placed into loan loss reserves, $11 billion for counterparty losses, and generated a net loss in the Fed’s two-year scenario. Yet CEO Jane Fraser and her team came through the stress test with flying colors in terms of minimum capital. Capital is what matters at the Fed, even if the rest-of-the-world does not care. Citi’s stressed loss rate on credit cards was inside 20%, but meanwhile the Fed hit Ally Financial (ALLY) with a 40% loss rate on credit cards in the stressed scenario? Hello.

Goldman has almost $20 billion allocated to loss provisions and another $18 billion for trading and counterparty losses in the stress tests, easily the highest in the group vs total loans. The good news with GS and MS is that they don’t have a big securities book with hidden unrealized losses like most of the industry. As we’ve noted for subscribers to our Premium Service though, GS loses more money on credit than its larger peers. Total loans and leases for GS were only $256 billion out of $1.6 trillion in total assets at Q1 2024, yet the bank lost 62 bp vs total assets in Q1 2024? More than JPM?

Of note, the Fed modeled a stressed 16% loss rate on Goldman’s C&I book, just shy of 16% on commercial real estate and 25% on credit cards. The stressed loss rate on cards was only 16% for Citibank. At CapitalOne (COF), the modeled loss rate on credit cards was only 24%. Again, Goldman is the outlier in the group looking at the Fed's loss assumptions. The credit loss rate on the Goldman credit portfolio says a lot about the target customer for that bank.

Goldman said last week that the Fed’s scoring didn’t reflect the evolution of the company, which has been exiting certain businesses and shrinking outside investments. “We will engage with our regulator to better understand their determinations,” said CEO David Solomon.

Q: If Marcus and related retail exposures are gone, why do the credit losses continue? The Fed's loss rate assumptions for C&I loans and total loss for GS suggest that institutional exposures are the culprit. If the folks at GS really, really want to do credit going forward, then David needs to have lunch with Jane Fraser soon. He and the Goldman bankers run the broker-dealer, but Citibank runs credit. Problem solved.

The bank founded in 1869 by Marcus Goldman and Samuel Sachs to discount short-term commercial paper has proven, once again, that trading commodities and extending credit to second-tier corporate credits are two different skill sets. As we noted to our readers last week, GS reported an average gross yield of 11% on its credit book or 2x JPM. Just imagine the credit default profile of the average GS institutional customer given that gross spread. We could name names, but let's not.

The chart below shows actual net losses for the top banks through Q1 2024. No, we don’t make these numbers up. These numbers come from the Fed and the Federal Financial Institutions Examinations Council. The public data from the FFIEC is more than sufficient for stress testing, but instead the economists at the Board demonstrate how not to perform a financial assessment. And remember, the modeled loss rates tell you how the Fed's supervisory staff sees each bank. The Fed stress tests are not about the bank, but rather about the bank stress test model and the legions of Fed economists who love to build models. The purple line is GS.

Source: FFIEC

Goldman Sachs is easily the laggard among the top-10 US banks this year, but they can take some comfort in the fact that they are not the loss-leading appendage of a foreign bank operating in the US. The history of foreign banks investing in the US is abysmal, as the table below suggests. We wrote about Bank of Montreal (BMO) and their purchase of Bank of the West earlier this year. When you see that many of these foreign-owned operations start the stress test with negative income, that tells you what you need to know.

More, look at the size of the provision for credit loss among these banks vs net revenue. Look at BMO, a bank with a US operation that is arguably insolvent if you subtract the intangibles and M2M losses. Like we said. What does this say about how the Fed views credit risk inside these banks? Note that the $165 billion asset US unit of Banco Santander (STD), the largest retail bank in the Eurozone, has the highest assumed loss rate in the group after the monoline consumer lenders.

Federal Reserve Stress Test Results ($B)

Source: Board of Governors * Top tier BHC in the US

The moral of the story, of course, is that income matters more than capital. Think of bank capital as a security deposit, a performance bond with the FDIC. Banks with strong asset returns like DS, American Express (AXP) and Charles Schwab (SCHW) ploughed though the Fed loss scenarios with little trouble. And JPM easily handled some very large loss assumptions in the test. But the projected losses for Wells Fargo & Co (WFC) and Bank of America (BAC) suggest that the Fed expects rough sailing for the banking industry ahead. Buckle up.

Today's Top Stories