“The idea that there are no excess reserves stems from the fact that new regulatory rules underwrite the demand for a big Fed balance sheet for a long time to come (possibly forever).”
– Zoltan Pozsar (Credit Suisse), in “Global Money Notes #5,” April 13, 2016
For the last few months we have been working hard on the question of the optimal Fed balance sheet size, based on the demands created by regulatory rules. Zoltan’s writings about the essentiality of excess reserves have helped enormously, as have data from Barclays, Bloomberg, and elsewhere. We have discussed the issue with sitting central bankers (Fed, Euro-system, others) and with former ones. We have canvassed some commercial bankers. We have met with credit rating experts. This commentary is the result of several months of research and discussions.
Our conclusion is that many Wall St. pundits have taken their eyes off the ball. They prefer to harangue about the Fed or issue warnings of doom. They are missing a regime change of significant proportions. It is impacting asset prices and altering global flows. So our purpose today in this lengthy commentary is twofold. First, we wish to advise our existing clients as to why their portfolios look the way they do and why we have not succumbed to eight years of Cassandra-like warnings of gloom about the bond markets. Second, we wish to trigger some debate among policy makers who read our notes and others who may receive them via social media links or by forwarding.
We believe that the Fed’s balance sheet may be too small, not too large. Later in this commentary we will argue for it to be about $5.5 trillion and to grow at about $350 billion per year. That rough estimate reflects a number of assumptions that we will recite below. We believe that the outcome of the Fed’s under-sizing its balance sheet is to distort the pricing structures of certain asset classes. Some of these pricing distortions are visible and measurable. We will note a few of them as the commentary proceeds.
Today’s comments owe a debt to seminal work by Zoltan Pozsar of Credit Suisse. He has published analysis on the evolution of the liquidity coverage ratio (LCR) under Basel III. We read his work carefully and credit him with expanding our thinking about LCR. What follows are our own estimates, but we are compelled to give him the credit for planting the original seed about LCR and for offering his own estimates, which have provided a basis for our thinking.
LCR requires large banks to hold specified assets in the forms needed to meet a forward-looking liquidity test. Very large banks, over $250 billion, operate under one set of rules. Banks between $50 billion and $250 billion are governed by a different set, whereas banks under $50 billion are considered too small to be a systemic threat, so the LCR rules do not apply to them. That said, many smaller banks are following the behavioral changes required of the larger banks, as Andy Peters explains in, “Why Banks Are Dumping Fannie, Freddie Debt.”American Banker, May 27, 2016. The American Banker article (subscription required) demonstrates why and how large American banks have sold their Fannie and Freddie holdings in order to meet LCR. Fannie and Freddie are an 80% qualifier under LCR; US Treasury direct debt, on the other hand, is 100%.
Reserves deposited at the Fed by a bank are the ultimate and lowest-cost qualifier. They are 100% eligible. They require no capital. They meet all LCR tests worldwide. They pay a current interest rate of 50 basis points. So for any American bank or any US subsidiary of a foreign-owned bank, the riskless and costless LCR threshold is the interest rate paid by the Fed on overnight excess reserve deposits (IOER). Every decision a bank makes about its portfolio starts there. While banks under $50 billion in size are currently exempt, some are growing and anticipate crossing that $50 billion threshold soon, so they are complying now. Others feel a “regulatory or surveillance” pressure, so they are moving toward the LCR standard. For the markets, it makes no difference why they are doing so. It is their collective actions that count.
In the case of GSE debt, we have seen some market reaction and expect to see more. Remember that GSE debt is also part of the asset holdings of the Fed. The Fed has held that level constant since the end of QE and has rolled maturities. The Fed ignored the change in GSE debt character with the advent of LCR and did not make it an issue in policy pronouncements. The Fed did discuss the matter internally but didn’t see it as large enough an issue to drive a change in policy. Query for consideration: Would markets have been more stable if the Fed had systematically and transparently acquired additional GSE debt in order to stabilize market transition when federally-backed GSE debt became less than a 100% qualifying item? Now then, we admit we have asked a counterfactual question, and we know there is no answer except to speculate about it. But our purpose is to ask policy makers to think about this issue. As a money manager we were able to take advantage of the change because GSE debt traded cheap to references. So clients saw some federally backed mortgage paper in their portfolios.
A second derivative of LCR rules appeared in the tax-free municipal bond space. State housing agency bonds are not High Quality Liquid Assets under the Fed’s rules. The Fed ignores their AAA credit structure. It is the liquidity situation that is dictating policy, and the Fed believes that these bonds will not be liquid in a crisis. So Fannie and Freddie are at 80%, but a state tax-free bond is at zero and/or it requires capital adjustments for a bank to own it. Add to that the complexity of the income tax code and one sees few state housing bonds being held by banks unless the raw yield is high enough to be attractive. That situation leads to a pricing anomaly.
The state housing bond may be secured by claims on federally guaranteed (GSE) mortgages. Most are partially or wholly secured in this way; thus the security is similar to that of Fannie and Freddie. So the question of creditworthiness is not the issue. Instead, it is the liquidity of trading them in a crisis that is driving the banks and the LCR decisions. Therefore pricing is markedly distorted. While Fannie and Freddie paper was being disgorged by banks because of LCR, state housing bonds were forced to compete with higher yields. They were being marketed at tax-free rates over 100 basis points above comparable and taxable US Treasury securities and also above taxable Fannie and Freddie paper. The market views Fannie and Freddie as US government paper, so it trades with a spread to Treasury debt. We think the Fed ignores this pricing distortion in policy making. We did not ignore it. For Cumberland, the distortion presented opportunity to buy very-high-credit-quality instruments for clients at tax-free rates above 4% when comparable taxable structures were trading below 3%.
A third anomaly results because of the FDIC rules that govern American banks but do not apply to American subsidiaries of foreign banks. Remember that both sets of banks are covered by Basel III rules and the need to meet LCR tests. The Fed is actually becoming more stringent than its national counterparties in other jurisdictions. The Federal Deposit Insurance Corporation (FDIC) charges an American bank 15 bps on total assets. Thus, the American bank receives 50 bps, pays 15 bps, and nets 35 bps.
The large customer of a large American bank knows those details. That customer expects its commercial bank to pass through the net 35 bps as a payment on the company’s cash, with only a slight reduction for costs. If there is an American subsidiary of a foreign commercial bank depositing the same money at the Fed, the FDIC doesn’t charge the 15bps fee. Therefore the customer of the foreign bank’s American subsidiary has a 15 bps advantage over the American-domiciled bank. Were you that customer, which bank would you select to handle the transaction?
We estimate that almost half the excess reserves on deposit at the Fed are placed there by US subsidiaries of foreign commercial banks. About 44% of total reserves are placed at the Fed by foreign-owned subsidiaries. This number allows us to estimate how much of the total reserve is required and then to guess at what is excess. Our guess is that nearly half of the excess reserves deposited at the Fed today originate in the American subsidiaries of foreign banks.
In large banking transactions, 15 bps is a lot of money. Note that this rate difference delivers an advantage to a large commercial banking enterprise housed outside the US versus an American competitor. Also note that 15 bps becomes a pricing factor on the repo structure. Repo is an alternate form of cash management. Remember that the reverse repo (RRP) is a liability of the Fed, just like an excess reserve deposit. Its use is similar, though its settlement timing and pricing are slightly different. And some agents (GSE) cannot legally deposit reserves with the Fed. In the last year, this pricing anomaly caused a significant shift. The big American banks (JPM, BoA, Citi, MS, GS, and Wells) shrunk their money fund Treasury repo by about $20 billion. Meanwhile, Swiss, German, Canadian, Japanese, and French banks grew, along with others. The total shift in one year was about $100 billion (Barclays, Crane’s Data, Federal Reserve).
For details on RRP see the NY Fed website. A $100 billion shift is not a lot in the scheme of a multi-trillion-sized Fed balance sheet. But the directional trend from domestic banks to American subsidiaries of foreign banks suggests that something is changing. Is it the pricing differential we are thinking about?
Let’s move on.
For the purpose of illustrating how LCR impacts a single bank transaction and why the Fed’s balance sheet may be too small, we are going to simplify this discussion. We will examine a single transaction flow between two companies with two different home currencies (euro and dollar) in two central bank jurisdictions (ECB-Bundesbank and Fed).
Here is some background needed to understand the transaction. A reserve deposit at the Fed is a method of meeting the LCR requirement for that depositing bank. Also note that the payment between two banks is nothing more than a transfer of reserves. Whether denominated in euros or in dollars, the reserves never leave the system. The ownership of the reserve changes, but the total reserves do not change. It takes the expansion or shrinkage of the central bank’s balance sheet asset size to impact the total reserves. This is true whether the bank is an American bank or a foreign bank. Remember that a deposit at the Fed qualifies for 100% treatment under LCR and does not require the depositing bank to find any additional capital to support the reserve deposit as an asset of the bank. The same is true for a German bank or a German subsidiary of a US bank when it comes to a reserve deposit at the Bundesbank. For LCR purposes, the Fed and the Bundesbank are both 100% qualified. Eventually, LCR will be currency-specific, but currently there is an arbitrage inducement, as we will see in an example below.
This LCR computation has the function of turning an excess reserve deposit into a new version of an optional required reserve deposit. The LCR requirement is met by the election of the commercial bank. Each bank, pricing its available alternatives, determines how to comply. Our argument here is that the current worldwide pricing of alternatives favors the use of reserve deposits at the Fed. That explains why about half of the excess reserves at the Fed are placed there by US subsidiaries of foreign banks. Those foreign-owned deposits meet LCR. At the same time those banks are earning 50 bps paid in US dollars instead of paying 40 bps in euro. That 90-bps spread is serious money and may be changing agents’ behavior.
Each of those banks has other options to meet its LCR. For example, it could own a German government bond or a US Treasury note. Either of these securities counts as 100% in meeting LCR, and neither requires the bank to raise additional capital to own. Simply put, owning a US Treasury security with a positive interest rate has a higher yield than owning a German security with a negative interest rate. We can quickly see that a foreign-exchange-rate currency risk develops between the euro and the dollar due to this construction. But is it really a risk, or is it driven by commercial transactions? We think the latter is in play now.
Now let’s get to the single transaction to illustrate the decision.
Let’s assume that a German commercial bank is handling the transaction for a German company. In Germany, the large commercial customer knows that excess euro cash deposited with its commercial bank costs 40 basis points (bps) if deposited with the ECB-Bundesbank. The customer, bank, and ECB all know that 40 bps is the threshold of cost. Why? It is the basic negative interest rate in the Eurozone for the marginal deposit.
The same situation in the US correlates with a different rate. When funds are deposited at the Fed, the bank receives 50 bps (remember that an American bank nets 35 bps due to the FDIC). Both the commercial customer and the bank know about the receipt of 50 bps.
Let’s examine the commercial transaction between the German company and an American company. The hypothetical transaction involves a $100 million sale of a widget. The German company sells to the American company, and the final payment from the American company is $100 million. The German company has a choice. It has a payment coming in US dollars. It can convert dollars to euros now by using the facility of its German bank, or it can wait and get the euros in future at the then-current exchange rate. The German company is charged 40 bps if it converts to euros. Instead it wants to hold in dollars and be paid 50 bps. The difference in earnings is 90 bps on $100 million, or $900,000. This figure appears in the reported earnings of the company, less the costs it incurs for any hedging program in the foreign exchange market. Alternatively, the company takes the risk and opts not to hedge. It may do that if it thinks the dollar will strengthen against the euro in the coming year.
The German commercial bank meets its LCR with the deposit at the Fed instead of a deposit at the Bundesbank. The bank has accomplished this banking service for the company without needing more capital and within the LCR requirements.
If the transaction is a daily rolling arrangement, the spread is 90 bps. But if it is extended over a longer period, the spread will widen due to the adjustment for risk. That is why the spread is widening throughout the euro-dollar term structure. The longer the time compared, the wider the spread. At two years, the spread is about 100 bps. At 10 years, the spread is about 160 bps. On our website we have a chart that shows the yield curve in dollars and in euros on December 31, 2013. That 2013 date precedes NIRP. A second set of yield curves is dated May 31, 2016. Note how NIRP widened the spread even as the Fed hiked a quarter point last December. Also note how the spread has widened while both yield curves have flattened.
Many factors go into the determination of these yields and spreads, and LCR is only one of them. But we can quickly see the effects if the expectation in the US is for a higher Fed yield due to a Fed tightening policy while the expectation in the Eurozone is for continuing negative rates. The flows into dollars are not symmetrical between dollar and euro. That means the demand for those dollars to be held at the Fed is rising. Yet there are only so many dollars, so a scarcity may be forming. The Fed can either meet that demand by providing more dollars, or it can refuse to do so. Query: Is this refusal a tightening? We think the answer is yes.
Those flows of dollars are likely to grow, as will the demand for those dollars to figure in the LCR calculations of those banks, whether they are foreign or American banks. Thus dollars at the Fed are not truly excess reserves. They are functioning like required reserves because of LCR. If the Fed does not provide enough of those dollars, the prices of other assets change in response. We see that in the two yield curves presented in the chart.
Note that the Fed could still raise the short-term policy-setting interest rate. It can set the IOER and size of its balance sheet independently. How it does so and what assets it acquires are the important issues. What happens if it ignores the essential role that “excess” reserves play? We don’t know. Query: Could the Fed trigger the next crisis with a dollar shortage?
As we consider that question, let’s turn to another: What should the size of the balance sheet be in order to meet worldwide demands for deposits that can be created by the Fed? Remember, the Fed creates money in the form of electronic cash every time it buys an asset. So it is the liability side of the Fed’s balance sheet that drives its optimal size. This is true regardless of which assets the Fed buys. The Fed traditionally limits itself to assets that are liabilities of the United States government.
Here is our estimate:
About $1.5 trillion is needed to meet the demand for physical cash (currency) worldwide. More than half of that amount circulates outside the US, and the number is growing every year. In the US, some of that cash is actually held in ATM machines. It counts as bank reserves in the computation for American banks. The old notion of vault cash has been replaced.
An additional $1.5 trillion is needed to meet the demands of foreign banks and cash pools that are shadow banks or non-banks, and that number is growing every year. RRP can be used by some since it is an alternate form of Fed liability.
Estimates are that $2 trillion is needed to meet the LCR requirements of American banks by the start of 2017. That number is also expected to grow every year. In the Dodd-Frank Replacement Plan, House Financial Services Committee Chairman Jeb Hensarling wants to replace capital requirements with a higher LCR. That change might, if enacted, add $700–$800 billion to the demand for reserves.
Then, some $300 billion is needed to provide the US Treasury with cash. That number is expected to grow to $500 billion.
Add the numbers and we have a starting estimate of nearly $5.5 trillion as an optimal equilibrium size of the Fed’s balance sheet by the end of 2016. We estimate the growth requirement at about $300–$400 billion each year. But there has been no growth since QE stopped.
Therefore, the Fed’s balance sheet is currently $1 trillion under the equilibrium estimate. The results show up in the pricing of other assets. If reserve assets are scarce, then the prices of substitutes should be bid up, and rates should fall. The shift in the yield curves and the widening spread support this conclusion.
So let’s sum up the case:
- The Fed may be too tight, not too easy.
- The Fed’s current balance sheet size is already too small by $1 trillion or more.
- The impact of a Fed balance sheet too small to meet worldwide demand for dollars is to alter the pricing of alternatives in the worldwide marketplace for all level 0–1 assets that are required under Basel III requirements for LCR.
- The Fed has “piled on” by adding rules to make American banks more restricted than Basel III rules require.
- Banks have to adapt to High Quality Liquid Asset Tests (HQLA) in addition to LCR. That requirement also alters asset prices.
Note that in our example we used a single commercial transaction of $100 million between a German company dealing in euros and an American company dealing in dollars. Now let’s begin to think about the fact that six currencies in 24 countries, together representing one-fourth of the commercial output of the entire world, are using negative interest rates as part of their monetary policy. And one-fourth of the planet is housed in the US or dollarized economies where Fed policy is trying to raise the short-term interest rate. The rest of the world is the other half. In those countries interest rates are headed lower. Nearly all of them are easing.
Is it any wonder that yield curves are flattening? That is happening both in the US and in NIRP countries. And the spreads we talked about are widening between the dollar term structure and the others.
What if the Fed hikes the 50-basis-point IOER rate to 75 basis points? The spread widens more. The curve flattens more.
Our thought is that the Fed needs to expand the size of its asset holdings by several trillions as it goes about raising interest rates in the US. Our projection is that such activity could make the US yield curve very flat. Every time the Fed hikes, the curve will flatten. Every time the spread widens between NIRP countries and the Fed, the distortion intensifies. And the act of the Fed’s buying assets will also flatten the curve; but it will, at least, provide the world with the dollars it seeks in order to meet LCR. We are suggesting that the Fed needs to increase the volume of the scarce LCR-qualified resource every time it raises the policy rate (IOER).
Here are some final things to think about:
- At the end of March, about $7.7 trillion in debt was trading at negative rates. Two months later that number is $8.2 trillion (Bloomberg, Bianco Research).
- There is no reason we can see that alters the rate of change in that trend.
- Zoltan Pozsar notes that, “Under the new rules, interbank deposits do not count as HQLA, and foreign banks are increasingly settling Eurodollar transactions with reserve deposits at the Fed.”
- Had the Fed, using a modest $50 billion a year, started a reserve growth liquidity regime in 1971 at the end of Bretton Woods, the Fed’s balance sheet would be over $5 trillion today after including the expanding demand for currency in circulation.
As Zoltan Pozsar said, “Big ain’t inflationary. Big is necessary. It is the future.”