Last month, Arnold Kling caught my attention with his profoundly pessimistic and, I thought, reasonably plausible argument that the Fed could go bankrupt. But I don’t follow Fed balance sheets the way my friend Jeff Hummel, an outstanding monetary economist, does. So I asked him what he thought.

Jeff gave me a detailed answer, which I post below. I made a number of edits and he approved them.

As I wrote before, I don’t believe Arnold Kling’s scenario is quite as scary and imminent as he thinks. To begin with, Kling is making a much stronger claim than the Neil Irwin article he cites. Irwin’s only concern is that rising mortgage rates, by causing a fall in the market prices of Fed held mortgage backed securities (MBS), will prevent the Fed from unloading them. Kling goes further and claims the Fed faces insolvency even if it hangs on to those assets. I think the latter claim is by far the weaker, not that the Fed isn’t facing some problems. But Irwin’s argument is more credible.

It is true that currently the Fed’s overall balance sheet has a significant maturity mismatch, borrowing short (bank reserves, currency, Treasury deposits, and reverse repos) and lending long (both Treasuries and MBS). But marking an institution’s balance sheet to market, as Kling suggests, is not necessarily the best to way to evaluate an institution’s solvency if it intends to hold all its assets until maturity. Fed assets will continue to pay their fixed returns (or inflation-adjusted returns on the less than 7 percent of long-term Treasuries the Fed holds) as long as they are not sold. In that case, the Fed does not have to worry about changes in market prices, unless there is an outright default on its assets.

Even in the case of a default, the Fed doesn’t hold any mortgages directly. It instead can hold only MBS that are guaranteed by one of three federal government sponsored enterprises. So it is not clear that the Fed itself would take a hit from defaults. Moreover, there does not appear to be a major concern about mortgage defaults as occurred in the 2007-2008 financial crisis. The main worry is rising mortgage interest rates due either to inflation or to the Fed’s dumping its MBS on the market, causing a fall in the value of the Fed’s MBS portfolio.

If the Fed doesn’t sell any MBS, Kling argues, inflation will nonetheless cause rising interest rates on Fed liabilities. To address that issue, we need to take a closer look at the Fed’s current balance sheet. Its total liabilities as of May 11, 2021, were $8.94 trillion. Its total assets and capital, of course, equal that. Of its assets, $2.71 trillion were MBS (30.3 percent) and $5.77 trillion were Treasury securities and a minor amount of Federal agency debt (64.5 percent). The Fed also holds as assets gold certificates of only $11 billion, valued at $42.2 per ounce. Other assets include minimal amounts of Treasury coin, SDRs, foreign assets, special lending facilities, bank premises (Fed buildings and other physical assets), etc.

Fed liabilities include $2.22 trillion of Federal Reserve notes (24.8 percent of its total balance sheet), bank deposits of $3.30 trillion (36.9 percent), and Fed borrowing either in form of Treasury deposits of $0.92 trillion (10.3 percent) or in the form of reverse repos of $2.17 trillion (24.3 percent). The remainder on that side of the balance sheet is either other minor liabilities or the Fed’s capital account ($42 billion). However, the only significant liabilities the Fed pays interest on are bank deposits and reverse repos.

Thus, over a third of the Fed’s liabilities are essentially interest-free loans through Federal Reserve notes and Treasury deposits. And only the interest rates on reverse repos are partly market driven, whereas the interest rate on bank deposits (reserves) is set entirely at the Fed’s discretion. Moreover, Kling’s implication toward the end of his post that the Fed might have to “induce banks to keep reserves by raising the interest rate paid on reserves” is completely mistaken. A single bank can get rid of its deposits at the Fed in basically three ways. First, it can sell reserves to other banks, which has no effect on their total outstanding quantity. Second, the bank can have the Fed convert the Fed deposits into currency, which actually would reduce the amount of interest the Fed has to pay. Normally banks do the second only to satisfy their customers’ demand for currency. Third, a bank can convert reserves into a reverse repo loan to the Fed, but because the Fed sets reverse repo rate below the interest rate on reserves, that also reduces the Fed’s interest costs. In short, no matter how high market interest rates go, the Fed does not have to raise interest on reserves to get banks to hold them.

Indeed, despite the elimination of formal reserve requirements in 2020, banks are still subject to implicit reserve requirements through the back door. Back in 2015 the Fed imposed the Basel Accord’s Liquidity Coverage Ratio (LCR), which I wrote about here and here. Not only does the LCR compel banks to hold more high-quality highly liquid assets, such as reserves, than they otherwise might choose. But also other forms of capital requirements, including stress tests, have similar, non-public, and sometimes ad hoc liquidity requirements, as detailed by Bill Nelson and Francisco Covas.

Of course, the rate of interest on reserves can affect how much money banks create through the money multiplier. But because the Fed can always alter the quantity of reserves, we are now in a complicated mix of policy options the Fed can play with. At the beginning of May the Fed raised the interest rate on reserves from 0.04 percent to 0.09 percent and on reverse repos from 0.03 to 0.08 percent. So for the sake of argument, consider an extreme stylized case in which both rates rise by another 1.0 percentage point and everything else remains the same. The Fed’s annual revenue would then fall by $54.7 billion ([3.30 trillion + 2.17 trillion] x .01]). Initially, all that would do is reduce the amount the Fed regularly pays to the Treasury. Although prior to the financial crisis, the Fed was remitting excess earnings to the Treasury totaling only an average of $30 billion per year, since the enormous increase in the Fed’s balance sheet after the financial crisis, the amount has on average more than doubled, with $109 billion being remitted in 2021. Back in 2019 remittances had temporarily fallen to $56 billion, before the Covid further expansion of the Fed’s balance sheet, but that has been their lowest since at least 2010.

So there seems to be ample room for the income spread between Fed assets and liabilities to fall without serious problems for a couple of years, after which market conditions could have changed. Even if continuing inflation at current high levels (something you and I both doubt) ultimately leads to Fed policies that reduce remittances to zero, and losses begin to impinge on Fed capital, the Treasury before then could increase its interest-free deposits at the Fed, easily and quickly substituting for the Fed’s interest-bearing reverse repos. Treasury deposits were used by the Fed as a major source of funds during the financial crisis, and during Covid they rose to a peak of $1.7 trillion. Only after that did reverse repos begin to replace Treasury deposits, which fell to $550 billion in Sep 2021, but they are again on the rise.

Or, with Congress’s approval, various accounting tricks can maintain the Fed’s nominal solvency. For example, back in 2015, Congress passed a Transportation act that put a cap on the Fed’s surplus capital, which is the amount of capital that exceeds the amount provided by the member banks. So that year, the Fed, in addition to its regular remittances to the Treasury of $97.7 billion, had to turn over $19.3 billion from its capital account. Again, in 2018, Fed turned over $3.2 billion from its capital account on top its regular remittances. There is no reason that Congress could not instead provide subventions to the Fed’s capital account, although doing so would probably gain more public attention. Or, if push came to shove, the Treasury’s gold could simply be revalued from $42.2 per ounce anywhere up to its current market price of around $1,800 per once. At the max, that could raise its capital account by $450 billion. Not that I think this will likely be necessary if the Fed hangs on to its MBS.

I do admit that, if the Fed tries to taper down by selling MBS on the secondary market, then rising mortgage interest rates pose a more serious problem. Any net reduction in Fed earning assets is inevitably going to reduce Fed remittances to the Treasury no matter what else happens, just as past net purchases increased remittances. True, a major part of the current rise in mortgage rates is from the unusually low rates of 2021.The annual average rate for 30-year fixed mortgages has now risen to 5.25 percent. The average for the entire year of 2021 was only 2.96, while back in 2018 that average was as high as 4.55 percent. And in most of the years between 2019 and 2021 it was above 4.00 percent. I think the Fed, based on past performance, could weather a sell off of MBS, but only if mortgage rates don’t soar much higher.

Yet before the Fed began purchasing MBS, the rates on 30-year fixed mortgages were generally around 6.00 percent and, if we go back to the 1990s, even higher. So Fed holdings of MBS may have depressed mortgage rates by as much as two percentage points. If so, the price at which the Fed sells off MBS could fall quite low, eventually wiping out Fed remittances to the Treasury. And in this case, the GSE guarantees won’t make any difference. The options available, however, are the same as those I mentioned above for the Fed hanging on to its MBS. But if Irwin is correct, his prediction is quite ironic: the very high inflation that the Fed itself created undermines its ability to reduce the size of its bloated balance sheet. Still, the Fed could easily reduce its MBS portfolio slowly, as Irwin points out, by letting “its holdings shrink” as they mature, and not rolling over the portfolio with new purchases. Unfortunately, most of its MBS portfolio, an amount of $2.6 trillion, has a remaining maturity of over 10 years, unless the underlying mortgages are paid off earlier.