very year the Federal Reserve conducts a supervisory stress test to gauge the health of what they call, “systemically important financial institutions”. The stress test became a key component of economic reporting after the financial crisis in 2008, when several banks collapsed after unexpectedly losing billions of dollars on unknown products.
Since then, central bankers adopted the “new Keynesian model.” This “Zero Lower Bound,” creates the conditions for severe balance sheet expansion with very unconventional asset purchase programs through a prolonged period of zero or near-zero interest rates. The ZLB presents a second issue; liquidity. Because the ZLB counterintuitively encourages people and businesses to hold cash when interest rates become lower, liquidity evaporates while a nominal interest rate floor of zero is stable. So why does this hurt banks?
A bank must collateralize its deposit liabilities, though we assume that the bank can commit (say, by putting cash in the ATM) to meeting its promises to satisfy cash withdrawals. For the bank, collateral consists of mortgage loans, government bonds, and reserves. Further, the bank can create counterfeit loans (synthetic products) in its asset portfolio, and in equilibrium, the bank must have the incentive not to do that. This feature is intended to capture a moral hazard problem that exists in the context of limited commitment. The bank must back its deposit liabilities with collateral, but then the bank has an incentive to compromise the quality of the collateral, unbeknownst to the bank’s liability holders if there is a cost advantage.
Banks serve a crucial role in our financial system. In essence, they are capital intermediaries. They collect customer deposits, loan them to other private individuals, and conduct repurchase actions for the Fed. The repo market is a complicated yet important area of the U.S. financial system where firms trade trillions of dollars’ worth of debt for cash each day.
Repo markets are not particularly complex. One firm sells securities to another and agrees to buy back those assets for a higher price by a certain date, typically overnight. The contract those two parties draw up is a “repo”, or repurchase agreement. Essentially, it’s a short-term collateralized loan. And just as most loans come with an interest payment, you can think of the difference between the original price and the second, higher price, as the “interest” paid on that loan. It’s also known as “the repo rate,” usually the Federal Funds Rate.
When the Fed sells a security to a counterparty and then agrees to buy back that security, it’s a transaction known as a “reverse repo.” Financial firms with large pools of cash would prefer to not just let that money sit around since it doesn’t collect interest. In other words, repo markets are like savings accounts for a bank. They allow financial institutions to borrow cheaply to fund short-term needs. There’s also (typically) not much risk involved.
The Fed has used repo transactions for over 100 years to manage liquidity and reserves, but it hasn’t always worked. In 2017 the Fed was shrinking its balance sheet after purchasing huge amounts of mortgage backed securities and treasuries, but corporate tax payments were due at exactly the same time. Firms wanted to purchase the debt the Fed was selling, but didn’t have the cash to meet their demand. In response to less cash supply, the overnight repo rate surged to almost 10%, when the policy goal was 2 - 2.5%, leading to temporary chaos in interest rate markets.
This matters because the purpose of the Stress Test is to gauge which banks are healthy enough to continue holding cash to keep Treasury markets consistently liquid. The Stress Test ignores an important bias in the system: a bank’s first obligation is to its shareholders, not to the government. If a bank has a choice of a slightly higher risk but higher reward investment, it will (in efficient markets) take that over a repurchase agreement. But regulations prevent that from happening. And if a bank isn’t interested in holding quality reserves to fund those obligations, then perhaps the Stress Test isn’t a valuable metric of a bank's health.
JP Morgan CEO Jamie Dimon believes exactly that. Following the latest release, Dimon commented that the stress test is a “terrible way to run the financial system.” “We don’t agree with the stress test,” he said. “It’s inconsistent. It’s not transparent. It’s too volatile. It’s basically capricious, arbitrary.”
And I have to agree. At best, the Stress Test is a summary of current capital conditions among large banks. At worst, it's a scientifically unrigorous examination of arbitrary economic conditions. Data scientists generally avoid fitting a model to their desired outcome, but that is exactly what the Stress Test does.
But, the Stress Test is all we have, so far. And really, all the Fed cares about is assessing the quality and consistent availability of collateral that supports repo operations. And this exact disregard for shareholder value creation (we’ll call it, government selfishness) is what incentivizes poor quality of collateral.
Recessions severely reduce shareholder value, but the Fed relies on banks to play ball. Capital quality naturally decreases in a recession too, because when interest rates are high, people want to borrow less. If the price of housing and the flow of housing services are sufficiently high, then the buyer will not borrow fully against their housing collateral. And by creating a recession to get inflation under control, the Fed clearly shows its interests are misaligned with banks.
The Fed is painfully aware of this problem. Their policy decisions usually reflect the question of how long is the period of low collateral quality relative to the financial sector's longevity in bad markets.
Housing collateral has a liquidity premium, which arises from binding collateral constraints of banks, which interact in an interesting way with the incentive problems faced by buyers and by banks. This liquidity premium must increase during a recession, given that high interest rates reduce liquidity. However, with no extra liquidity premium available (rates are locked in), banks are stuck holding onto assets with immediately distressed values, even if the health of the housing market has not deteriorated significantly, and it hasn’t.
Stress Tests always come with some moral hazard, a kind of bias that occurs when one party has better information than another prior to a transaction. In this case, the banks have much more information than the Fed prior to the Stress Test and may always undergo some kind of window dressing to “juice” the numbers. Trading or lending arms still may have off-balance-sheet (OBS) products, like levered loan books controlled by subsidiary institutions and other kinds of debt without public market quotes.
And finally, liquidity is always a concern. The Stress Test model does account for the liquidity constraints of certain debt products which would decrease their market value beyond their total credit losses. But, these are only estimates, and it can be difficult to gauge the total impact of a severe decrease in liquidity. But, we can extrapolate based on existing trends.
With mortgage rates nearly double their historic lows, almost all existing mortgages have lower rates than those available today. As high mortgage rates disincentivize people from refinancing and moving, mortgage balances are being paid down more slowly and the resulting runoff of the Fed’s MBS balance sheet is happening more slowly as well. In fact, the maximum runoff caps set by the Fed to facilitate a gradual removal of its footprint likely will not even be reached through natural paydowns: the Fed needs to sell.
The Fed announced plans to gradually and predictably reduce its securities holdings by setting caps on the amount of MBS and Treasury paydowns that are reinvested.This process began in mid-June with an initial monthly cap of $17.5B MBS for the first three months, which shifts up to $35B/month in September. However, with prepayments slowing in response to higher primary mortgage rates, the Fed’s paydowns are already running below the planned maximum $35B cap. In other words, organic MBS portfolio reduction will be much slower than the Fed wants.
To make matters worse, the Fed’s Treasury holdings do not have this problem. The cap for Treasury securities was initially set at $3Bn/month and shifts up to $60B/month in September. Because Treasury securities are not affected by slowing prepayment activity like MBS, principal paydowns of Treasuries will be able to hit the full $60B/month cap much more efficiently. This dynamic will cause the share of MBS in the Fed’s portfolio to increase relative to Treasuries rather than decrease.
The MBS purchased by the Fed during the last round of QE was made up of mortgages with much lower rates than today. As a result, if the Fed were to start selling its holdings now, it would impact the supply of very low-coupon MBS. There would be relatively little effect on the price of higher-coupon MBS, which sets the secondary market for mortgages originated today. In other words, Fed sales of MBS would likely have a rather minor impact on going thirty-year mortgage rates.
The broader issue this loss of liquidity presents is about incentives. If there is no liquidity, banks have no incentives to issue mortgages which must ultimately be sold off and repackaged. After all, banks are not in the business of holding mortgages anymore. They are intermediaries between borrowers and investors. With no liquidity, lending will likely dry up, exactly what the Fed wants. The counter-effect is higher immediate mortgage rates, which may push buyers into dire straits, forcing higher down payments or more creative solutions like interest only periods and adjustable rates. Banks are forced to underwrite more synthetic products to gain the artificial compounding exposure a pool of mortgages would normally occupy. And we can all imagine what happens when ARM’s come back in style, and banks get too creative.