Fatas and Hunt on Reserves and Quantitative Easing

By Dan Kervick

Lacy Hunt reports on three recent academic studies indicating that the Fed’s unconventional asset purchasing programs have failed. Antonio Fatas is “sympathetic to the argument that Quantitative Easing has had a limited effect on GDP growth”, but takes issue with some parts of Hunt’s analysis, and argues that the way Hunt analyzes the relationship between reserves and the money multiplier “is not consistent with the conclusions reached about the lack of effectiveness of monetary policy actions.” I believe there are problems with both Hunt’s analysis and Fatas’s analysis of that analysis. My best guess is that QE has had negligible macroeconomic effects. But some of the considerations Hunt and Fatas adduce in attempting to evaluate that question are red herrings, and don’t get us closer to an answer.

Hunt claims:

If reserves created by LSAP (Large Scale Asset Purchases) were spreading throughout the economy in the traditional manner, the money multiplier should be more stable. However, if those reserves were essentially funding speculative activity, the money would remain with the large banks and the money multiplier would fall. This is the current condition.

To which Fatas responds:

How can reserves be funding speculative activities if they remain in the balance sheet of the banks? Reserves represent an asset in the balance sheet of commercial banks. They have increased by having commercial banks selling other assets to the central bank. So the amount of “riskier” or “less liquid” assets must have decreased.

But contrary to what both Hunt and Fatas seem to imply, whether or not banks are using reserves for riskier speculative activities cannot be determined by looking at changes in aggregate bank reserve balances.

Most businesses in the modern world do not do business with piles of physical cash. The money they hold, and use for day-to-day business operations, is held on account as a deposit at some bank. People or firms who are selling stocks or commodities obviously have bank accounts. If some speculating bank purchases assets of those kinds from the seller, then a payment will be made by the bank to the seller. In order for the payment to be settled and cleared, a transfer will have been made from the reserve account of the bank purchasing the assets to the reserve account of the seller’s bank. Total bank reserves will be unaffected. (If the seller happens to have a deposit account at the purchasing bank itself, then the bank will just credit more money to the seller’s deposit account, and both the bank’s own reserves and aggregate banking system reserves will be unaffected.)

Aggregate bank reserves do not drop as a result of an increase in economic activity unless people for some reason show an increased desire to withdraw physical cash and hold onto it. Note that when people withdraw cash for ordinary, temporary commercial purposes, the cash rapidly cycles back into banks where it again forms part of banks’ total reserves. I personally withdraw some cash each week so I can use it in my company’s vending machines or to make very small, over-the-counter purchases at local stores. Once I have made those purchases, the mangers of the businesses who have received the cash from me collect it together with other daily cash receipts and deposit all of that cash at their bank. Cash is constantly flowing out of bank vaults and back into bank vaults, but the impact on aggregate reserve levels is negligible.

Hunt seems to recognize part of this process in stating that if banks choose to make speculative asset purchases rather than ordinary loans, then “money would remain with the large banks.” But Hunt also seems to suggests that if banks were doing more ordinary lending instead, the money would not remain with the banks. But bank lending does not result in reserves leaving the banking system. When the bank makes a loan, a borrower is given money on account at the bank. As that borrower then draws on the account to make payments of various kinds, the bank will either be required to settle those payments by (i) participating in transfers of reserve balances from the bank’s own reserve account to reserve accounts of other banks, (ii) debiting and crediting money from and to its own depositors’s accounts without drawing on its reserves, or (ii) dispensing cash reserves to the borrower which, as before, are quickly recycled back into the banking system’s total reserves. Bank lending does not result in bank reserves going out of the banking system.

Hunt’s analysis also relies on claims about the so-called “money multiplier”.  This troublesome and much-misused term can be defined in different ways. Sometimes the term just refers to the ex post ratio of commercial bank deposit balances to total reserves, i.e. the ratio of commercial bank money to central bank money. Call this the empirical money multiplier. Others use the term to refer to the reciprocal of a central bank’s formal reserve requirement. Call this the policy money multiplier. Many analysts err greatly in assuming that there is some stable connection between the empirical money multiplier and the positive money multiplier, so that deliberate central bank changes in central bank money will generate predictable changes in the quantity of commercial bank money. This assumption never had much to be said for it, and has been decisively crushed by the events post 2008.

But let’s focus entirely on the empirical money multiplier, which is clearly the quantity to which Hunt is referring. Is Hunt correct in claiming that if banks start using reserves to engage in more speculative asset purchases, then the empirical money multiplier will fall?  That’s possible.  But the empirical money multiplier can also fall as a straightforward consequence of QE without any changes in bank speculation or lending behavior. If, starting with some fixed and unchanging quantity of aggregate reserves, banks increase the ratio of loans to other assets in their portfolios, the empirical money multiplier will grow; and if they instead decrease the ratio of loans to other assets, the empirical money multiplier will fall. But reserve balances are actually fluid, especially given the Fed’s asset-purchasing programs, and so the fact that there has been a fall in the empirical money multiplier cannot be used to conclude banks either are or are not increasing speculative activity or decreasing lending.

Suppose a bank owns $1 billion dollars in reserves and $9 billion in other assets, including $2 billion in treasury securities. It’s total assets are thus $10 billion. Suppose it has $8 billion in deposit liabilities, so that the ratio of deposits to reserves is 8/1. Now suppose the bank sells $1 billion in treasuries to the Fed.  As a result, the bank’s total assets will still be $10 billion, but its reserves will now be $2 billion. Suppose the bank’s rate of lending and loan collections remains unchanged, so that deposits do not change and remain at $8 billion. Then even if the bank makes no other changes in its portfolio whatsoever, its deposit-to-reserve ratio will have been slashed in half to 4/1. Lending has continued unchanged; and the bank has not increased its speculative investments. It has just swapped one kind of low-risk asset for another. But since the assets for which it swapped the treasuries – dollars on account at the Fed – comprise a part of its reserve balance, the deposit-to-reserve ratio is affected drastically. So clearly, attempts to draw conclusions about bank behavior from changes in either reserve balances or the empirical money multiplier are fraught with peril.

Hunt also claims:

(Banks) can allocate resources to their proprietary trading desks to engage in leveraged financial or commodity market speculation. By their very nature, these activities are potentially far more profitable but also much riskier. Therefore, when money is allocated to the riskier alternative in the face of limited bank capital, less money is available for traditional lending. This deprives the economy of the funds needed for economic growth, even though the banks may be able to temporarily improve their earnings by aggressive risk taking.

To which Fatas responds:

Yes, banks can trade their Reserves for either bank loans to the private sector or by purchases of risky assets (stocks). But in both cases the amount of reserves has to go down. One can make the argument that the injection of liquidity is triggering one type of lending more than another but this is inconsistent with the view that the problem with QE is that banks are simply sitting on reserves without doing anything with them.

Fatas makes a good point here: if banks are using reserves to buy other assets, then that is quite a different matter than “sitting on them”. But he is quite wrong, as we have already seen, in asserting that if banks were using reserves to purchase risky assets, the quantity of bank reserves would be going down. And he also seems to accept Hunt’s assertion that if banks are using reserves to speculate in stocks and suchlike assets, then that activity is putting financial constraints on their ability to generate loans. Again, bank reserves do not go out of the banking system when banks make loans, and aggregate reserves do not represent a relatively fixed quantity of loanable or available funds. Bank liquidity managers will generally respond to any need that added lending creates for additional reserves by acquiring those additional reserves, either by borrowing them from other banks or selling treasuries to the Fed. And the Fed will generally accommodate any need for additional reserves by the banking system as a whole by supplying them in a way that maintains the Fed’s target policy interest rate. Right now the cost of acquiring reserves is exceedingly low – approaching 0%.

But in the present circumstances the cost of adding reserves for the banking system as a whole is a moot point, since banks are already carrying abundant excess reserves. So allocating some of those reserves to the purchase of risky assets, if that is indeed what banks are doing, leaves plenty of reserves left to cover any increased volume of interbank payments and depositor cash transactions that would be sparked by expanded lending, with no need for additional liquidity.

The whole idea that one can measure whether either lending or speculating is increasing or decreasing simply by looking at fluctuations in bank reserve accounts, or changes in the empirical money multiplier, is deeply flawed. My suggestion for economists is that they give up on the pointless task of analyzing bank behavior through the study of reserves and ratios involving reserves. If you want to see whether banks are increasing or decreasing their lending, study the statistics on bank lending; if you want to see whether they are increasing or decreasing their risky asset purchases, study bank balance sheets and measure changes in those asset classes. If you want to study whether changes in these activities are tied to Fed asset purchasing programs, study whether there is a robust temporal correlation between changes in bank behavior and changes in Fed asset purchases.

Cross-posted from Rugged Egalitarianism

Follow @DanMKervick