Bank Reserves

Paul Krugman responds to a Cullen Roche post at Pragmatic Capitalism:

The Krugman post is reproduced below.

Professor Krugman is a great writer and communicator of economics. Cullen Roche has raised an issue about the relationship between bank reserves and bank lending. And the professor has responded to that. He concludes with a take on “word games” that he feels may be at play here. We think there’s more to it than that.

Professor Krugman approaches this issue indirectly at first:

“Similarly, if we ask, “Is the volume of bank lending determined by the amount the public chooses to deposit in banks, or is the amount deposited in banks determined by the amount banks choose to lend?”, the answer is once again “Yes”; financial prices adjust to make those choices consistent.”

He refers to ISLM in analogous fashion, suggesting that in both cases: “The correct answer is “Yes” — it’s a simultaneous system.”

In the case of the view that “loans create deposits”, this is not necessarily a conflicting statement. A borrower in seeking a loan can obviously create a new deposit with the proceeds. The negotiation between a lender willing to lend and a borrower willing to borrow must be an agreement on that. So “loans create deposits” translates to a prospective borrower who is in a position to create or “issue” a liability to the bank, which is the bank’s loan asset, in exchange for a deposit asset. Indeed, the arrangement can be viewed as an asset swap – loan for deposit – according to mutually agreed pricing, in the sense of the Diamond/Dybvig liquidity transformation that the professor uses in his model.

But the simple observation “loans create deposits” reflects an origination flow of funds – lender to borrower – that is more directly informative as a contrast to the misguided money multiplier theory. In any case, the idea that loans and deposits appear simultaneously informs nothing beyond that which is conveyed by “loans create deposits”. Moreover, “it’s a simultaneous system” doesn’t directly deal with the fact that banks don’t lend reserves, which is the issue that Cullen has raised. In this sense, simultaneity is arguably a straw man point with reference to the relationship between broad money creation and central bank reserves and the validity of the money multiplier theory.

Another economist has embraced Professor Krugman’s “simultaneous system” point as a pat answer to those pesky commenters who keep raising this issue about lending and reserves:

It is a curious point. Any outcome which includes the joint appearance of a loan and a deposit requires a preceding “adjustment” to get to that point. That must be reflected in the path of corresponding accounting entries and the pricing associated with those. But this alone doesn’t say a lot about loans, deposits and reserves.

The more relevant paragraph in Professor Krugman’s post follows next as it pertains to Roche’s point:

“Now, think about what happens when the Fed makes an open-market purchase of securities from banks. This unbalances the banks’ portfolio — they’re holding fewer securities and more reserve — and they will proceed to try to rebalance, buying more securities, and in the process will induce the public to hold both more currency and more deposits. That’s all that I mean when I say that the banks lend out the newly created reserves; you may consider this shorthand way of describing the process misleading, but I at least am not confused about the nature of the adjustment.”

The professor has previously made the same point about using “lend reserves” as a type of “shorthand”. But shorthand for what exactly? That would make an interesting post. And why shorthand? For one thing, a more detailed alternative might not fit easily in the NYT column format. Let’s acknowledge the shorthand, but for purposes here, this will be a temporary pass.

That said, there may still be some potential confusion arising from that second quoted paragraph. That is the focus of the rest of this post.

Consider a simple model of a commercial bank:

There are two asset portfolios – liquid assets and non-liquid loans. On the other side of the balance sheet, we have deposits and capital.

Banks maintain liquid asset portfolios as part of the process of managing the reserve account at the central bank. This area is sometime referred to as the bank’s money market operations. It includes both liquid assets and short term short term wholesale liabilities. These are the portfolios that help the bank to adjust its reserve position each day.

It is instructive in this analysis to consider the difference between the pre-2008 Fed environment and the situation since then with the creation of massive excess reserves through quantitative easing (QE).

First we consider the pre-2008 environment. The bank’s money market operation takes into account daily Federal Reserve policy as expressed in the policy target interest rate and implemented as necessary through the Fed’s open market operations. All interbank transaction activity ends up being settled in the reserve account via the interbank payment and settlement system. This includes everything on behalf of customers who transact with customers of other banks. The bank wants to manage the volatility of the resulting net reserve position to the best economic result. Too much net inflow results in unwanted reserves earning an inadequate rate of interest (zero interest on a pre-2008 basis). Too little means borrowing from the Fed, which is not encouraged and is the reason for liquidity management in the first place. At the same time, management has the job of trying to anticipate Federal Reserve policy as it affects interest rates. The reserve management operation looks to optimize the cost effectiveness of the net interest margin effect and the liquidity volatility protection offered by its liquid asset portfolio, given everything else that’s going on within the bank and outside it.

The key point in all of this is that the main lending operation, which accounts for the balance sheet expansion most relevant to the Fed mandate, does not respond to Federal Reserve policy at the level of reserve operations. That distinction explains the nature of the dilemma that is inherent in Dr. Krugman’s model and perhaps in the academic papers that back it up.

Let’s return momentarily to the issue of “lending reserves”. Dr. Krugman’s simplification is arguably more digestible in the case of the bank’s money market operation than it is with respect to the main lending operations. Bank reserve managers do respond to changes in their reserve positions that might be induced by proactive Fed reserve activity. A reserve manager will take steps to acquire money market assets if his reserve balance is increased because of Fed activity. Conversely, bank lending officers in making their credit decisions pay absolutely no attention to the bank’s daily reserve operations.

The use of the shorthand of “lend reserves” may be comparatively harmless when applied to the money market operation (although this does not fully validate it), but it is misleading when transferred to the area which most matters for the discussion, which is the bank’s main lending portfolio. Bank lending officers do not operate even according to the professor’s simplification. And the lending portfolio, not the liquid asset portfolio, is where the relevant balance sheet expansion take place over time.

Lending managers make decisions on the basis of credit risk assessment and capital allocation. And when they lend, deposits are created as a result. This has nothing to do with the reserve management of the day.

Even in the case of money market operations, and even allowing for the “lend reserves” simplification, the meaning of the adjustment in question must be qualified and limited in a very strict way in order for reserve operations to be depicted correctly. The Federal Reserve adjusts the system reserve setting in conjunction with targeting the Federal Funds rate. The Fed sets the policy rate. That’s what they do. They have tools in achieving that policy rate target. They adjust reserve settings when necessary in order to hit their target rate. (Still pre-2008 here.) But the degree to which they have to make such adjustment is miniscule in its order of magnitude compared to the size of the US banking system and its aggregate loan and deposit portfolios. One merely need look at the statistics going back over decades, pre-2008, to see that excess reserve settings have been tiny in this framing (very low single digit $ billions), and very stable on average over time, even on a high frequency weekly time series basis. The explanation for this is that when reserves pay no interest, the Fed need supply or withdraw very little of them in order to get the interest rate response it’s looking for. The banks’ demand for uncompensated reserves (pre-2008) is so interest inelastic, that a small change in the system position can move the funds rate significantly. Bank reserve managers have to deal with this. And they do so by attempting to get rid of excess reserves and to cover reserve deficiencies through money market operations and to do it very quickly based on perceived and actual Fed actions.

Thus, the Fed’s activity in achieving its target rate (when the actual rate is off course) tends to be quick – because the bank response is quick. And over a longer time frame, the Fed’s activity tends to be divided between steering the trading rate for fed funds higher or steering it lower, depending on the direction of the deviation from target, in order to get the trading rate back on target. That’s how open market operations work when used by the Fed in managing the fed funds rate. And so what happens is that any excess reserve position that is similarly higher or lower than a normal level of excess reserves, as a result of the Fed’s course correcting action relative to an existing target rate, tends to revert back to the normal level in short order – and we’re talking here about a matter of several days. And the result of that is that the Fed’s excess reserve setting has been very low and stable on average over time (pre-2008). And that means in particular that excess reserves which exceed this low average level are short lived. And those short lived episodes in particular never reach the attention of bank lending officers. The lending officer works in the dark as far as reserves are concerned, because there’s no functional reason for him to be interested in that activity. He lends based on risk assessment and capital requirements – not based on the Fed’s manipulation of the excess reserve position. Those short lived bursts of Fed activity in managing excess reserve expansion or contraction have absolutely no bearing on the lending decision through the bank’s main credit portfolio. And it is in this sense that the second paragraph quoted from the professor’s post is potentially misleading.

Two more points warrant comment – regarding required reserves and banknotes.

It is a fact that required reserves are calculated after the deposits that give rise to them are created. This is well documented in various Federal Reserve publications. And the Fed immediately supplies these required reserves to the system at the point they come into effect. It must do this in order to control the fed funds rate relative to target. So, given that time sequence, there is no way that the Fed’s necessary supply of required reserves can influence bank balance sheet activity in any meaningful way beyond that. Everything relevant to this issue is done in the money markets at the margin by way of the daily excess reserve setting as described above.

Next, bank customers decide when they want to exchange bank deposits for banknotes. The Federal Reserve has no operational control over such a decision process. If customers want banknotes, and if the banks’ corresponding purchases of those notes from the Fed reduces the reserve balances available to the banks for interbank payment and settlement, then the Fed will replace those lost reserves with a new injection through open market operations or the like. Just as in the case of required reserves, this is necessary in order to control the fed funds rate relative to target. Again, the active Fed management of the system reserve setting is done at the margin of both required reserves and banknote issuance, by way of the strategy for the excess reserve setting as described above.

It is puzzling how some treat the aspect of banknote issuance. If the banking system and its deposit base grow over time, it seems natural that the private sector will increase its holdings of banknotes on trend and in conjunction with such expansion, as a matter of asset mix (between deposits and banknotes) and portfolio rebalancing. And interest rates including deposit rates may well influence a liquidity preference for banknotes relative to deposits. This doesn’t seem complicated. What is mysterious is why this aspect needs to play a central role in understanding how the reserve system operates. That approach seems to place the supply of banknotes at the epicenter of monetary transmission, dismissing the importance of a more detailed and separate understanding of bank reserve operations.

That should cover the necessary issues regarding Fed operations for the most part, pre-2008.

Returning to the Krugman model of banking, which combines portfolio equilibrium and liquidity services according to the academic papers he cites, this seems like a solid perspective on role of banking at a high level. But the referenced papers, as famous as they are, do not seem to account for the adjustment mechanism in question as it occurs in the modern system.

Now let’s go back to the issue of Dr. Krugman’s “lend reserves” shorthand. Having accommodated this temporarily, we must now say that this is wrong at a technical level, whether we are talking about money market operations or lending operations.

Banks do not lend reserves. More precisely, banks only lend reserves to other banks, which again is only a liquidity management operation, separate from the main lending portfolio. And loans do create deposits. This is a factual matter of double entry bookkeeping. Banks record new loans to non-bank customers, creating corresponding new deposit entries at loan origination. When a brand new deposit departs the lending bank, and traverses to a competing bank, the lending bank will literally pay for that transaction through its reserve account. That is not lending reserves. That’s using reserves as the medium of exchange to pay for an interbank transaction. That’s the functional role of reserves – to pay for interbank settlement. It’s got nothing to do with “lending reserves”.

And this point applies just as accurately to money market operations. When I noted earlier that it might be more right (less wrong) to use the shorthand of “lend reserves” in money markets operations, it was an observation relating to the reserve manager’s way of responding to an excess reserve position. But the manager does not respond by lending reserves in any technical way, unless he is lending them to another bank through a special market for reserves that is directly available only to the banks (i.e. the fed funds market). The reserve manager is acquiring liquid assets and using reserves to pay the bank whose customer sold the securities. The counterparty bank does not owe reserves back to the acquiring bank and certainly no non-bank customer owes reserves to anybody at any time.

It is understandable that Dr. Krugman has used this as shorthand. But it is dangerous in the final analysis to do so – because it misleads on the nature of bank balance sheet expansion, which after all is the main subject beyond this technical issue of reserves. A fundamental distinction in banking is violated by the idea of “lending reserves”. And that is the difference between liquidity management and capital management. Bank reserve managers are liquidity managers. Bank lenders are capital managers, in effect.

(Capital management is a centralized bank function. Lending operations need to pay for their capital funding as supplied by that central function. Lenders determine if a prospective loan is financially viable by plugging in their cost of capital to the relevant profit calculation. Reserves have nothing to do with it other than in the case of a relatively minor cost calculation associated with the net interest margin tax effect of required reserves on prospective deposit funding.)

Bank lending and associated deposit creation – the kind that matters over time – the kind that the Federal Reserve is interested in influencing through monetary policy – requires capital underpinning. Banks make lending decisions based on risk assessment and the cost of capital. This has virtually nothing to do with bank reserves (other than the marginal tax-like cost calculation noted above in the case of required reserves). The lending officer must be connected by policy and operations to the capital account manager – not the reserve account manager. Lending uses up credit and other risk capital. And this difference in connection is evident in the institutional structure that functionally separates money market operations from portfolio lending in banks. (In the case of very large loan commitments, the lending division may well advise the reserve management function on the actual timing of the expected funds drawdown. This information can be useful in the planning of that day’s reserve management operations.)

Conversely, the nature of the money market operation is such that the related liquid assets tend to be high quality, low risk and short duration – meaning that the capital allocation for risk in liquidity management operations is typically not significant when compared to the main lending book. Thus, the money market operation is a relatively low user of risk capital and for good reason according to the nature of the liquidity management function.

The Fed’s monetary policy works through interest rate targeting. Fed reserve operations are a means of achieving the operational objectives of interest rate targeting. This principle holds for QE as well, notwithstanding the popular focus on the enormous level of excess reserves currently in the system. QE in its various forms is an extension of Fed interest rate targeting. The parabolic increase in excess reserves under QE is a byproduct of interest rate targeting of a wider scale.  In the initial credit easing phase, the Fed lent to certain non-bank entities, alleviating money market rate pressures, and creating bank reserves as a by-product in the process. In parallel, it provided reserves directly to banks in order to alleviate interest rate pressures in the inter-bank market. The funds rate would have skyrocketed under these conditions otherwise, without this extraordinary action by the Fed. The interbank lending system had seized up because credit risk had penetrated deeply into money market and interbank operations. The Fed injected reserves so that interbank settlement could take place at the target funds rate. Excess reserves were required to an extraordinary degree to alleviate extraordinary interest rate pressures. But this demand for reserves had nothing to do with banks’ desire to “lend reserves” to non-bank customers (i.e. outside of the fed funds market). And it had nothing to do with banks’ desire to stockpile reserves for future lending. It was about hoarding a risk free asset (bank reserves) in a market environment that had suddenly become very risky – i.e. hoarding the asset that is used for payment and settlement of transactions in a normal market. The Fed provided reserves to satisfy that demand in an attempt to restore normal interest rate conditions, and then started to pay interest on reserves to establish an interest rate floor (albeit an imperfect floor) once rates became better behaved and the outsized supply of excess reserves became redundant to its usual purpose of payment and settlement. In the subsequent stages of QE, with mortgage and Treasury bond purchases, the Fed created reserves again as a byproduct of its policy to extend its direct interest rate influence further out the yield curve. Again, the supply of reserves increased, not for the purpose of “lending reserves”, but this time as a byproduct of bond purchases, most of which were sourced from the portfolios of bank customers rather than the portfolios of the banks themselves. Given the banks’ normal role in servicing customer bond transactions, they received more excess reserves as a result of that activity, when non-bank bond sellers deposited their new money back into the banking system. These interest bearing excess reserves resemble treasury bills captured inside the banking system. There is no need for them for purposes of lending activity.

Cullen Roche was basically correct in saying:

“Banks don’t lend their reserves out so expanding the monetary base was never going to result in consumers getting “the money for deposits”.

Moreover, he would have been correct even if accommodating “lending reserves” as shorthand for actual operations, the way Dr. Krugman suggests. Because the relevant bank reaction function that both of them should be referring to (and which Cullen was referring to by implication) is the main lending portfolio. But the reaction function that Dr. Krugman actually describes is the money market operation and liquidity management function, on a pre-2008 basis for the most part. That is not the same as the bank reaction function that the Fed is interested in from the perspective of its mandate for economic expansion under price stability.

Regarding Dr. Krugman’s citing of several classic academic papers, surely it must be possible that such papers, being great works in economics, might be supplemented effectively by modern updates in some way. Is it possible that today’s monetary system works in such a way that these papers alone no longer offer an encompassing explanation? Specifically, the academic papers cited by the professor don’t seem to say much about bank capital. Portfolio decisions on risky lending require an allocation of capital – not an allocation of central bank reserves. The bank liquidity process described by the Diamond/Dybvig paper involves a risk transformation function that requires an allocation of capital across the organization’s asset portfolios as a necessary condition for the associated creation of liquid, low risk bank deposit liabilities. Virtually all bank liquidity crises start with some element of a perceived risk to bank capital, and capital adequacy plays a central role in the liquidity management framework for any banking institution. The Tobin and Brainard paper examines a range of banking system equilibria that include combinations of reserve requirements and fixed bank deposit rates. That doesn’t seem to model the modern system in which deposit rates float competitively, and where by analogy the only relevant fixed interest rate is the target fed funds rate, which is fixed in the sense of being administered by the Fed until the Fed wants to change it. But this is the anchor rate that propagates influence throughout the system as the basis for a galaxy of differentiated rates with different risk premiums and term structures, including the required return on bank equity capital. Rates of return are not determined by the public demand for banknotes, or by a model that assumes the Fed can determine that for them. The paper seems incomplete relative to an understanding of today’s system.

So there is a great difference between using “lend reserves” as convenient shorthand for operational dynamics in bank money market operations (which is relatively OK) and the material distortion of economic substance that this phrase conveys if entirely overlooked when applied more broadly to lending operations (which is not OK). And that is what deserves a bit more attention, given that this is not a minor issue in understanding the financial crisis and the response of the Federal Reserve to it.

These are not word games. It’s about how the internal institutional organization of banking interfaces with the Federal Reserve in functionally differentiated ways. This multi-layered institutional response provides clues as to how monetary policy actually affects banking. It is about clarifying distinctions between liquidity risk and credit risk and capital, and how banks make portfolio decisions based on these different risk factors and exactly where in the bank they make those decisions and how those decisions relate in their own different ways to Federal Reserve operations. And at the end of the day, the Fed transmission of monetary policy occurs through interest rates being set and not through reserves being provided. The excess reserve setting on a pre-2008 template was a tiny but fairly stable pool of special funds attached to an ocean of financial flows, a pool for the sole purpose of providing interbank settlement liquidity at the prevailing fed funds target rate. The QE reserve setting post-2008 has resulted in a larger lake of the same type of special funds, due to a Federal Reserve interest rate policy that has been extended to include credit risk premiums and the yield curve. These funds still have nothing directly to do with commercial bank lending operations in the main.



August 16, 2013, 11:39 am

Banks and the Monetary Base (Wonkish)

Cullen Roche is unhappy with the way I treat monetary expansion in my old Japan paper (pdf); or actually he’s unhappy with the way I talk about it. I’m actually kind of reluctant to even get into this, because any discussion of these issue brings out the people who believe that they have discovered the hidden secrets of the monetary universe, somehow missed by generations of economists. But here goes anyway.

When I think about the role of banks in the economy, I generally rely on two models, which are both partial pictures but add up to a reasonable overall approach. One is Diamond-Dybvig (pdf), which portrays banks as providers of liquidity services, and also shows how bank runs can happen. The other is Tobin-Brainard (pdf), which portrays the financial system in terms of a portfolio equilibrium, in which each sector — households, banks, firms, etc. — choose the mixes of assets and liabilities they want to hold, and asset prices adjust to make these choices consistent.

What I did in that old Brookings Paper was a quick-and-dirty merger of these two approaches, in which I got the monetary base into the story in the form of required reserves held by banks. That was a strategic simplification, and an unrealistic one — almost all of the monetary base is actually held in the form of currency, not bank reserves. But it was obvious to me that it didn’t really make any difference for the question at hand.

And how did I know that? Basically from Tobin-Brainard, who showed that whether banks hold monetary base in the form of reserves makes no fundamental difference to the monetary mechanism, as long as somebody wants to hold base money — and the public does, in the form of currency.

Actually, Tobin-Brainard is to many of the controversies that swirl around banks and money as IS-LM is to controversies about interest-rate determination. When we ask, “Are interest rates determined by the supply and demand of loanable funds, or are they determined by the tradeoff between liquidity and return?”, the correct answer is “Yes” — it’s a simultaneous system.

Similarly, if we ask, “Is the volume of bank lending determined by the amount the public chooses to deposit in banks, or is the amount deposited in banks determined by the amount banks choose to lend?”, the answer is once again “Yes”; financial prices adjust to make those choices consistent.

Now, think about what happens when the Fed makes an open-market purchase of securities from banks. This unbalances the banks’ portfolio — they’re holding fewer securities and more reserves — and they will proceed to try to rebalance, buying more securities, and in the process will induce the public to hold both more currency and more deposits. That’s all that I mean when I say that the banks lend out the newly created reserves; you may consider this shorthand way of describing the process misleading, but I at least am not confused about the nature of the adjustment.

And the crucial thing is that there are no puzzles or misunderstandings here. Tobin and Brainard got it all straight half a century ago, and anyone who thinks that there’s a big flaw in their reasoning is almost surely just getting caught up in his own word games.



View all posts by

Leave a Reply

124 Comments on "Bank Reserves"

Notify of

3 years 9 months ago


Regarding this statement: “A reserve manager will take steps to acquire money market assets if his reserve balance is increased because of Fed activity. Conversely, bank lending officers in making their credit decisions pay absolutely no attention to the bank’s daily reserve operations.”

I met a former treasurer of regional bank (branches in several states) and he said that he did try to get rid of excess reserves by acquiring money market assets AND by issuing more loans. Now this was pre-2008 and his decisions was based on a number of other factors (capital position of bank, credit risk of borrowers, risk-return tradeoffs, etc), but the point is his story contradicts what you are saying.

Maybe his loan officers, as you note above, were not aware of how the loans got funded but from his big-picture perspective it was clear.

3 years 9 months ago


It is possible for a single bank to be like that. Imagine a bank which is growing fast in the sense that its market share of loans issued in total is increasing. It can find itself constrained by its funding. Perhaps its deposit raising ability is not as good as its capturing the loan market.

Also while this bank’s market share is increasing, others’ is decreasing. In the aggregate lending is demand-led and on top of it also depends on bankers’ own animal spirits so in the whole (unless in crisis, banks as a whole aren’t reserve-constrained).

I think the PKE punchline “banks lend first and look for deposits later” is a bit of an overkill. In fact in the book Horizontalists and Verticalists Moore himself states that he does some overkill.

I think here or any one of the reports near that publication it states how Australian banks first look for deposits and then lend (my words not that of the RBA – something of that sort)

Read it a while back but can’t find where so if you have the time and can find out let me know.

So when a bank raises deposits it will see an inflow of central bank settlement balances or reserves. Perhaps the person you asked is more right talking about deposit inflow rather than reserves.

3 years 9 months ago
You raise an interesting point, so I’ll try and be reasonably thorough in addressing it. Treasurers have centralized jobs with an all-bank scope. But treasurers don’t normally have responsibility over the full lending function of a bank (note the last 3 words of the post). I suppose that would be possible in a small bank, although it would be an unusual organization structure. And I suppose it might have been the case with the fellow you spoke to. But it is certainly not the case in larger institutions. And it fair to say that my description of things may apply more accurately to larger institutions (especially the money center banks, which account for most of the $ trillions of the banking system balance sheet) than necessarily to the smaller ones. In the case of the treasurer you spoke with, it’s definitely the case that his responsibilities would reflect a “big-picture perspective”, which is in the nature of the job, but less certain that everything he was describing (as you describe it) correlated with activity that he was directly responsible for, unless the organization structure was quite unusual, as noted above. In any event, all that said, there is a distinction between the scope of a Treasurer’s responsibilities and the mode in which reserve account management is done on a daily basis, which is what I attempted to highlight in the post. Every cash flow that goes through the bank that goes to or comes from another bank goes through the reserve account, and reserve account management normally is a treasury responsibility. So the treasurer sees everything in that sense. But he doesn’t cause everything to happen that gets reflected in that reserve account. A big part of reserve account management is responding to stuff that’s been caused elsewhere in the… Read more »