This is a response to an article by Nick Rowe, "Repeat after me: people cannot and do not 'spend' money", in which he states that banks lend reserves. As can be guessed from the title of my article, I disagree. But the difference in view is more nuanced than is suggested by the title. There is a good deal of disinformation spread about banking on the internet, so I think this is an important subject. Although this is fairly theoretical, it touches on the topic of the effectiveness of Quantitative Easing (spoiler: it isn't effective).
Justifying The "Banks Lend Reserves" View
In his article, Nick Rowe appears (justifiably) impatient with some of the discussion of banking that floats around on the internet. Although some of this is based on disinformation that is spread by the "Fractional Reserve Lending is Fraud" crowd, some of it is based on exaggerated misreading of "endogenous money" theory (which is used by Modern Monetary Theory). Paul Krugman's outburst about endogenous money appears to be another example of this frustration.
The key to understanding this view is that we cannot just look at aggregate banking system behaviour, we need to look at individual banks. If there was only a single bank, it would face almost no liquidity constraints. This is similar to just looking at the banking system in aggregate. But there are constraints on individual banks, which should show up in aggregate behaviour.
It makes absolutely no difference whether banks make loans in the form of currency or in the form of creating demand deposits. An individual bank that makes a loan of $100 by creating a deposit of $100 will lose $100 of reserves to a second bank when the borrower spends that $100 on a bike, and the bike seller deposits the cheque in that second bank. If the bike seller will only accept currency, so the first bank swaps $100 in reserves for $100 in currency, then lends $100 currency to the borrower, the loss in reserves is immediate, rather than delayed by a day or two. But the end result is exactly the same.
Let's cut to the [mild profanity deleted] chase: banks lend reserves.
In other words, as banks lend, they should expect the money to be transferred to other banks or to cash, which implies a loss of reserves. They cannot completely ignore their liquidity position, unlike some more extreme positions you read on the internet, in which banks can just "print" money out of thin air to cover liquidity losses.
Why This Is Literally Incorrect
Since I am in broad agreement with Modern Modern Theory, it is no surprise that I disagree with his statement ("banks lend reserves"), when taken extremely literally. And this is not just a question of doctrine, we can see this in the data.
Firstly, I will ignore the issue of people withdrawing notes and coins ("currency") from banks. Under normal circumstances, changes in currency outstanding is not a major issue (although there are seasonal effects, such as a lot of people taking out cash in order to go partying between Christmas and New Year's). If we follow Minsky and analyse everything as a bank (and one should follow Minsky), these transactions can be interpreted as depositors transferring their cash to the central bank - notes and coins can be viewed as a deposit at the central bank in bearer form. (I am consolidating the central bank with the central government, which is the correct way to analyse the economy, even if it runs afoul of legalistic analysis of the central bank.) The only time such withdrawals would matter is during a bank run. But in the modern era, generalised bank runs towards currency in sensibly run economies do not happen, and so it does not affect observed behaviour.
For my example, I will look at the "modern" Canadian banking system, which abolished the archaic required reserve system in the 1990s. Thus, it makes no sense to discuss "reserves" in Canada, rather one can speak of deposits at the central bank (the Bank of Canada).
As shown above, members of the Canadian Payments Association (which includes the banking system) have dropped their deposits at the Bank of Canada to inconsequential amounts in recent years. (There was a spike during the financial crisis, when nobody trusted private sector short-term debts.) Although such institutions appear to prefer to keep a small positive balance at the Bank of Canada, the target is to have a $0 balance at the end of the day. (Note that if there is some miscalculation by a bank's treasury department, it can borrow from the Bank of Canada to make up for any settlement imbalance. Since such borrowing is at a small penalty rate - and it is discouraged by regulators - this is generally avoided.)
Therefore, the statement that "banks lend reserves" has to be wrong - Canadian banks have lots of loans outstanding, but they have no deposits at the Bank of Canada ("reserves").
(An additional point is that modern central banks normally target* interbank short-term interest rates. If a bank is short reserves, it can always borrow them in the market at that rate, and the central bank has no choice but to create them if there is a shortage. This is what happened in the pre-QE era in the United States.)
Banks Lend Against Liquid Assets
A more accurate characterisation of the situation is to replace "banks lend reserves" with "banks lend against liquid assets". If we look at Canadian Chartered Banks, we see that government securities (at all levels of government, not just Federal government) are typically about 10% of their Canadian dollar assets. (Note that the Canadian banks now have large U.S. dollar subsidiaries, but those operations are run on a matched-currency basis. In other words, we can look at the Canadian-dollar denominated assets and liabilities in isolation.) These holdings provide the liquidity buffer that is used to insure against the loss of deposits. For example, if there was a withdrawal wave leaving Bank A going to Bank B, this could be met by Bank A selling government bonds and Bank B buying them.
If we amend Nick Rowe's text in this fashion, there does not appear to be much of a disagreement between us. If there is, it is probably the result of the fact that I do not attach particular significance to "money" versus other governmental liabilities (treasury bonds and bills). Correspondingly, I attach a greater weight to fiscal policy (which determines the amount of government liabilities outstanding) than to monetary policy (which is the split between the monetary base versus other governmental liabilities).
What Are The Behavioural Constraints?
It is clear that a bank cannot run down its holdings of governmental securities forever. As a result, I will now outline what is the practical limit for bank lending based on liquidity considerations (that is, ignoring bank capital constraints and borrower demand, both of which matter).
The situation in the United States is more complicated by the split between large money centre banks and the smaller banks. This creates structural imbalances between sub-sectors of the banking system. So I will discuss a situation similar to that in Canada, where the banking scene is dominated by a small number of players.
Imagine that there are five banks, and each has a 20% market share in both bank deposits and lending (and for simplicity, there is no shadow banking system). The size of each bank's balance sheet is a nice round $100.
If one bank increases its loan book by $10, we expect that:
- $2 would remain as a deposit at the same bank; and
- $2 would be transferred to each of the other banks.
Therefore, if you hold all else equal, it would need to raise $8 to cover the $10 in new loans.
Of course, not everything else is equal. It will receive deposits from new loans made by other banks. Since the other banks have 80% of the market share of total loans, that can be a lot.
If every bank increases its loan book by $10, then each bank would "lose" $8 of the new deposits, but it would gain $2 from each of the other banks, netting out to no net transfers between the banks.
Therefore, all balance sheets expand equally, and there is no need to find "reserves" to finance that expansion. But the ratio of liquid asset holdings relative to the total size of the balance sheet would drop, and so it is likely that the banks would need to raise their liquidity (somehow) to keep that ratio near target levels. But this is relatively small; if liquid assets are 10% of the balance sheet, the banks would only need to raise about $1 (1% of the balance sheet) in order to keep the liquid asset ratio constant. If we are in a country with required reserves, the central bank would have to create those reserves, which is a small fraction of the increase in loans.
The key for a bank's liquidity position is its deposit market share versus its peers, as well as its rate of growth relative to the average. A bank that is growing faster than average, or has a smaller market share of deposits than average, will end up with a need to raise money in order to finance its balance sheet expansion. (Losses to "shadow banks" is an important effect in recent decades.) This financing can either be done via equity or bond (term deposit) issuance, or in the money markets. But as we saw in the last cycle, financial institutions that financed their expansion using the money markets were vulnerable to runs by institutional investors. As a result, regulators frown upon expansion strategies that are not funded with deposits. (And if you are a credit analyst, this is why you should dislike rapidly growing financial companies, unlike equity analysts.)
In summary, liquidity considerations pose only a limited constraint on aggregate credit growth; rather the constraint is on relative balance sheet expansion within the banking system.
Relationship To QE
The practical conclusion of this analysis is that Quantitative Easing (QE) will have no impact on the banking system. From the point of view of banks, it is just a change of allocation within liquid assets, and they have no additional capacity to lend.
If there is an impact from QE, it results from supply and demand factors in the yield curve. I am extremely skeptical that this matters much, at least at the front end of the curve. People price the short end off of expectations, and it is nearly impossible to detect supply and demand dynamics. For example, when a central bank announces a surprise rate hike, market makers change prices immediately, without any need for securities to change hands.
* Many economists would object to my phrasing here; what I am referring to is the fact that interest rates are the target control variable for monetary policy for modern central banks (excluding those pursuing QE policies). The objective is to use this policy variable to guide the economy so that the inflation target is hit. Since there is no guarantee that the central bank will be able to force the interbank rate to a desired level, I prefer to write that they target the interest rate.