Wednesday, March 8, 2017

Why the American taxpayer might prefer a large Fed balance sheet

David Andolfatto and Larry White have been having an interesting debate on the public finance case for having a large (or small) Federal Reserve balance sheet. In this post I'll make the case that American taxpayers are better off having a large Fed balance sheet, perhaps not as big as it is now, but certainly larger than in 2008.

To explain why, we're going to have to go into more detail on some central banky stuff.

The chart below illustrates the growth of the Fed's balance sheet. Prior to the 2008 credit crisis, the Fed owned around $900 billion worth of assets (green line), these being funded on the liability side by $800 billion worth of banknotes (red line), a slender $10-15 billion layer of reserves (blue line), and a hodgepodge of other liabilities. The Fed now owns an impressive $4.5 trillion in assets. These are funded by around $1.5 trillion worth of banknotes and $2.3 trillion worth of reserves. So the lion's share of the increase in the Fed's assets is linked to the expansion in reserves, which have ballooned by around 25,000%.

There's a problem with the above chart. It shows reserves clocking in at just $10 billion prior to 2008, but it's important to keep in mind that this *understates* the quantity of reserves issued by the Fed. Prior to 2008, the Fed would typically lend out tens of billions worth of reserves to banks during the course of the day, these amounts being paid back before evening. These loans are referred to as "daylight overdrafts." Because the above chart uses end-of-day data, it omits daylight overdrafts, thus making the balance sheet look smaller than it actually was.

How big did the Fed's balance sheet actually get during the course of a day thanks to overdrafts? Prior to the 2008 credit crisis, daylight overdrafts typically peaked at around $150 billion. So if we recreate the chart using intraday Fed data, the pre-2008 balance sheet would be around $800 billion + $150 billion, or 20% larger than if we use end-of-day data. And rather than a relatively flat pattern, we see a pulsing pattern. I've drawn out the chart by hand to give a sense for how the balance sheet would have looked, although its not to scale and doesn't use real data.

So why does the Fed offer daylight overdrafts? One of the business lines in which a commercial bank participates is the processing of payments on behalf of its clients to other banks, these recipient banks in turn crediting sent funds to their clients. To make these interbank payments, banks use deposit accounts at the central bank, or reserves.

In the U.S., legally-stipulated reserve requirements force banks to hold small quantities of central bank reserves overnight. So when the U.S. banking system opens in the morning for business, a bank will typically already have some funds in their reserve accounts that can be used to make client payments. However, the ability of this small layer of required reserves to carry out the nation's payments will soon be swamped—after all, the quantity of transactions conducted on a single day using reserves is massive, currently clocking in at $3 trillion.

In theory, banks might choose to hold an excess quantity of reserves overnight (i.e. more than the legally mandated minimum) in preparation for the next day's payment cycle. However, the Fed has historically kept the overnight interest rate on reserves at 0%, far below the market overnight interest rate. So no bank wants to hold reserve overnight if they can avoid it. If they did, their profits would suffer.

To ensure that banks have the ability to carry out the nation's business come morning, the Fed has typically provided the necessary reserves via daylight overdrafts. When the banks close for business in the evening, the Fed then sucks the reserves it has lent to banks back in. Alex Tabarrok once fittingly described banks as inhaling credit during the day, "puffing up like a bullfrog" —only to exhale at night.

As I mentioned earlier, before the credit crisis hit Fed-granted daylight overdrafts used to rise as high as $150 billion over the course of the day. Since 2008, the quantity of daylight overdrafts has declined quite dramatically. See the chart below:


Why have banks stopped applying for overdrafts? In 2008 the Fed began to pay interest to any bank that held reserves overnight. Rather than "exhaling at night," it suddenly made sense for banks to hold reserves till the next morning. This new demand for overnight balances was not met by daylight overdrafts, which must be paid back by the end of the day. Rather, a new permanent supply of reserves began to emerge thanks to the Fed's policy of quantitative easing. Under QE, the Fed created reserves and spent them to purchase bonds, these reserves staying outstanding as long as the Fed did not repurchase them, potentially for decades. The upshot is that banks are now quite happy to hold huge amounts of Fed-issued reserves on a permanent basis. As such, they no longer need to make use of daylight overdrafts to carry out the nation's payments. 

So let's bring the conversation back to the taxpayer. As you should hopefully see by now, the debate between keeping a big balance sheet and returning to its pre-2008 size is closely intertwined with the following question: do we want our central bank to provide daylight overdrafts or not? Because if we are to go back to 2008—i.e. to a period when overnight reserves were "scarce," as Larry White describes it—then by definition we are advocating daylight overdrafts.  

I'd argue that taxpayers might prefer that the Fed not provide daylight overdrafts. To begin with there is the question of credit risk. If a bank that has been granted an overdraft were to fail during the course of business, the Fed would be out of pocket. Since the central bank is ultimately owned by the taxpayer, that means taxpayers could take a big hit when a bank fails.

The Fed could protect itself by requiring banks provide collateral as security for access to Fed overdrafts. Now when the offending bank goes under, the Fed has a compensatory asset in its possession that it can use to make good on the loan, thus sparing the taxpayer. However, the protection afforded the Fed by collateralized daylight overdrafts comes at the expense of the nation's deposit insurance scheme, the Federal Deposit Insurance Corporation, or FDIC. To ensure that depositors of a failed bank are made whole, FDIC typically sells off the bank's assets. If the Fed has taken one of those assets for itself as collateral for a daylight overdraft, FDIC will have one less bank asset at its disposal and may have to dip into taxpayer funds to make up the difference. The overall risk faced by the taxpayer has not been reduced.

By maintaining the status quo—i.e. a large quantity of reserves—the taxpayer gets more protection from bank failures. Banks must buy reserves, or tokens, ahead of time to ensure that they can meet the payments needs of their clients. So the Fed acts as a seller, not a creditor, and therefore does not expose taxpayers to risk of bank failures. At the same time, FDIC does not face the prospect of having risk shifted onto it should the Fed seize collateral from a failed bank with unpaid daylight overdrafts.

Now the preceding discussion might seem to tilt me towards David Andolfatto's position of keeping a large balance sheet, albeit for different reasons than him. Not entirely. While a large quantity of reserves will be sufficient to insulate the taxpayer from bank failures, it needn't be as large as the current $2.5 trillion in outstanding reserves. As I pointed out earlier, prior to the 2008 credit crisis the Fed would typically grant around $150 billion in daylight overdrafts. This was sufficient to facilitate ~$2.7 trillion in payments (see data here). So each dollar in reserves was able to support 18x that value in payments. The Fed currently processes around $3.1 trillion in payments, a task that could probably be discharged with ~$200 billion in daylight overdrafts, assuming that the 18x ratio still prevails. So as long as the Fed were to keep at least $200 billion of the $2.5 trillion in reserves outstanding, that amount should be sufficient to replace the need for daylight overdrafts.


1. How the High Level of Reserves Benefits the Payment System and Settlement Liquidity and Monetary Policy Implementation, both by Bech, Martin, and McAndrews
2. Divorcing Money from Monetary Policy by Keister, McAndrews and Martin
3. Turnover in Fedwire Funds Has Dropped Considerably since the Crisis, but It’s Okay by Garratt, McAndrews, and Martin


  1. Interesting post, but I'm not sure about the point that providing collateral might enhance risk to depositors.

    Compare a position where a bank holds either $100 in reserves or $100 in other high-quality liquid assets. In the first instance it covers a $10 payment by simply reducing its reserves to $90. In the latter, it goes into intra-day overdraft for $10 of reserves and places $10 of other liquid assets as collateral. Either way its stock of unencumbered HQLA is still $90.

    I think the argument that there is a material difference in credit risk holds only if the unavailability of intra-day overdraft forces banks to hold higher overall levels of unencumbered HQLA. This might be the case if the higher levels of reserves banks needed to hold were additional to other liquid assets. But, I don't think that's the case. On the whole, those reserves simply displace things like T-bills in the holdings banks carry to meet liquidity requirements.

    1. Yes, good point.

      Whether reserves are built up by reducing HQLA or by reducing risky illiquid loans has a big influence on whether taxpayers face additional risk.

  2. Why do you choose to refer to assets being funded by liabilities, surely it is the case that liabilities are funded by assets.

    1. I can go either way. Isn't it just semantics?

    2. Semantics is the meaning of words.

      There is no alternative words to the ones that describe a process correctly.

      A liability is an obligation that requires funding. A flow from the entity.

      An asset is a source of funds. A flow of funds to the entity.

      Bank liabilities are funded from its assets.

  3. The problem I see is, what the Fed gains in reduced risk from not having to do daylight overdrafts, it loses in increasing its risk of having to pay more in interest on reserves than it receives from its balance sheet portfolio. This could easily start being the case if the Fed should need to increase its interest on reserves higher than 2% in order to keep inflation under control.

    So, I see this as a wash overall. The Fed has traded one type of risk for another.

    1. Good point. In theory, if the Fed holds 3-month t-bills, the return on its portfolio should be nearly perfectly matched to the cost of paying interest. If it needs to suddenly pay 2% IOR, the 3-month t-bill rate will also jump to around 2%. Of course in practice the Fed doesn't match the duration of its liabilities and assets, so it takes on duration risk, as you say.

      And if Nick E. is right, then the gains to not doing daylight overdrafts may be less than I imagined.

  4. 1. Central banks are simply arms of the state. Thus whether state liabilities are classified as liabilities of the CB or of “government” (i.e. the rest of the state) is simply a technicality. If the first option is chosen, the CB has a “big balance sheet”, but that is of no economic significance.

    2. State liabilities can come in interest yielding form (traditional government debt) or non-interest yielding form (the traditional form that reserves / base money has normally taken). Milton Friedman and Warren Mosler suggested there was no case for the former, i.e. interest yielding form. I’m inclined to agree, and for the three reasons I gave at the link below under the heading “What’s the optimum stock of PSNFA?”

  5. Somewhat further to Nick's point (I think), a regulatory system that provides oversight in the form of liquid asset policy guidelines is quite capable of setting up system liquidity to operate in such a way as to make a requirement for excess reserves as the specified form somewhat redundant.

    1. e.g. $ 200 billion in either reserves or treasury bills held by the banks for liquidity purposes both serve to fund the cumulative budget deficit. And for purposes of active liquid asset management, I might argue that the reserve alternative if anything makes for a sloppy, less efficient system.

    2. "...the reserve alternative if anything makes for a sloppy, less efficient system."

      What about the argument that it improves the timing of settlement?

    3. I find it interesting that the 60 and 90 percent time distributions didn't move very much.

    4. Also, the Fed has had decades to observe the operation of the daylight overdraft system prior to QE. I wonder if this type of argument has been considered before simply by proposing an increased in required reserves - e.g. through non-arbitragable capital based central bank reserve requirements. In fact, the world moved in the opposite direction with reduced reserve requirements in countries like Canada. Why should this thinking be undone because of QE ?

    5. "In fact, the world moved in the opposite direction..."

      As a counter-example, there's the New Zealanders who in 2006 injected huge amounts of permanent reserves into the system and stopped providing daylight credit.

    6. If the primary motivation is to reduce settlement time lags along with the Fed's credit risk exposure, I would at least favor a system in which $ 200 billion in excess reserves become required reserves as I suggested above. A penalty rate would be charged on a deficient required average balance over an averaging period - basically the same as the existing Fed system when required reserves were more binding relative to the supply of excess.

      I think having $ 200 billion in excess reserves goes to the point on sloppiness I made earlier. I think it makes for inefficient price transmission and reserve management.

      Do you know if New Zealand did this with their higher level of total reserves?

  6. Your point is that with excess reserve the CB takes less credit risk? I'm not sure the pre-2008 daylight overdrafts are making that point. Yes, daylight overdrafts are unsecured. So are Fed Funds (overnight overdrafts). If I can't get Fed funds (because I'm seen as credit risk by others) I get the penalty rate at the Fed, and someone come knocking to look at my books. The credit risk bit is really happening in the term lending markets, which is where the fed is doing repo away, (collateralised lending). Thats where your credit risk action is, before 2008 and now. And that risk depends on, (1) where the Fed sets the outside spread in open markets and (2) the schedule it posts at its discount window. Consequently, if you look at the big balance sheet at the top of your post, you can see where the credit risk is now (for tax payers). Its in all those asset programmes, especially the ABMS (QE 2 i think?). Just because there are excess reserves doesn't mean we, the public are protected by some kind of pre-funding logic (which doesn't work anyway, as theses balances are still residuals). Look at the remuneration statement to treasury and you can see in numbers, whether the Fed is actually getting paid for doing liquidity ops, or paying to bail people out.

    1. " can see where the credit risk is now (for tax payers). Its in all those asset programmes, especially the ABMS (QE 2 i think?)"

      I don't see much credit risk there. The debt that the Fed holds in its portfolio either issued by the Treasury or is guaranteed by the Treasury. So if the Fed issues some debt to by some Treasury debt, its a wash. That being said, Larry White's point about duration risk makes sense.