If Banks Don't Lend Reserves, What Do They Lend?

There have been a few more blog posts of late on the old issue of whether banks lend reserves, including from Nick Rowe and Brian Romanchuk.

I prefer to say that banks do not lend reserves.  However, this raises another question.  If they do not lend reserves, what is it that they do lend?  This question is a useful one to ask, because it forces us to think a bit deeper about loans and money.

The essential features of any loan from A to B, is that A agrees to transfer value in some form to B, and B agrees to transfer value back to A, in equivalent form, at a later date.  There may be other terms, most notably payment of interest, but these are incidental.  A simple example is a loan of A's car.  A agrees to transfer possession of the car to B and B agrees to transfer back possession of the same car at some point in the future.

It's similar with a loan of securities.  A transfers securities to B, and B agrees to transfer back equivalent securities on the later date.  One difference compared with the car loan, is that we are not generally specifying exactly the same securities.  The securities merely need to be of the same type.  It's usually not possible to separately identify securities anyway, but the point is that if B chooses to sell the borrowed securities to C, he does not need to get them back from C in order to meet his obligation to A - he can instead buy equivalent securities from D, say.

In both these cases, the transfer of value involves an actual transfer of property from A to B: either physical possession of the car, or title to the security (a chose in possession in the first case and a chose in action in the second).  Loans of dollars are different.  With a dollar loan, A agrees to transfer value by making a payment to the order of B (or procuring that such a payment is made) and B agrees to reciprocate at a later date.  In this case, there is no actual transfer of property.

To some extent, however, we can interpret this as being a loan of money from A to be B.  After all, in many cases, the result of the initial transfer will be that A's money balances will be reduced and B's will be correspondingly increased, with the position reversed at the end.  This interpretation is popular with economists as many of them like to think of money as an enduring thing that is passed from hand to hand.  And in some cases, such as if we lend cash, there is an actual identifiable transfer of property.

However, this is only an interpretation and it does not work well in all circumstances, particularly if we have certain contracts between A and B where we want to treat the actual contracts themselves as being money, such as if either A or B is a bank.

The fact is that a loan does not have to a loan of anything.  Some loans can easily be treated as being a loan of something, such as a loan of a car.  But dollar loans are not in general a loan of something, even if we feel a desperate urge to think of them as such.  They are in fact just bilateral agreements to procure accounting entries.

For many purposes, it is fine to think of dollar loans as being loans of money.  But we should be careful not to fool ourselves into thinking that is what they actually are, because we need to understand how things work when that interpretation no longer fits.

Money with Passive Banks

It is sometimes said that banks are more than just financial intermediaries, because they can sort of create their own funds to lend.  In this analysis, a simple financial intermediary would be one that takes in funds that the public place with it and then allocates these between different assets.  Banks, in contrast, don't need to wait for funds to come in, because the very act of their lending creates the deposits that fund them.  It might therefore be concluded that bank decisions matter, whereas those of other financial institutions do not (or at least they matter less).

Whatever the merits of this analysis, I always find it useful to look at things from different angles, so I thought I'd describe a slightly different approach.  What I want to do is start from the concept of a completely passive bank, which makes no decisions of any importance, and then see what we need to change to get a more realistic picture.

So, this starts with a bank that does no more than record transactions.  Everyone has an account with the bank and each account carries a balance measured in units of the currency.  Whenever any one person wants to make a payment to another, they inform the bank of the payment and the bank debits one account and credits the other.  The bank only acts on instructions; it makes no decisions for itself.

Account balances can be negative.  If a person makes a payment in excess of the positive balance in their account, then the bank simply records that excess as a negative balance.  The total balance of all accounts in the black is equal to the total negative balance on all accounts in the red.  So whilst individual balances will change from time to time, the total net balance is always zero.

In this model, the balances in these accounts form the medium of account.  We could measure the aggregate positive balance of accounts and call this the quantity of the medium of exchange, but this might be misleading as the negative balances are also a part of that medium.

This concept of money might be said to fit with the basic New Keynesian model.  Agents can transact and make payments to one another, but the aggregate balance is zero.  The total amount of positive balances (which we might want to call the money supply) doesn't really matter.  It would be easy to add an interest rate to this picture - with interest charged on negative balances and credited to positive balances.  However, there are no credit constraints.  Everyone spends purely based on how they wish to spread their expenditure.

So the natural extension is to include credit limits.  With what we have here, individuals hold claims on the bank; the exposure to those with negative balances is pooled.  So it may be impractical for the individuals to decide on how credit is allocated.  This creates a natural role for the bank.  For an individual to make any payment which would result in its balance going negative, it has to be first approved by the bank.  Furthermore, the bank will then also require that the account is made positive again within a specified length of time.

What we have now is more like our normal simple model of a bank.  We have positive balances which we call deposits and negative balances which we call loans.  The bank takes an active role in deciding how much is lent, to whom and for how long.  Individual depositors have no role in this decision.  We could say that loans create deposits, although an alternative would be to say that the two are actually created in parallel by spending decisions.

So we have arrived at the same concept of what a bank does, but via a different route.  I think the alternative perspective is useful for a couple of reasons.  First, it highlights that the credit rationing aspect of bank activity is critical to why they matter.  This helps us frame questions about how credit decisions outside banking might matter as well.   Secondly, it gives a different and, I would say, more realistic concept of how the medium of exchange operates in modern economy.  Rather than the MOE being a quantifiable thing which is created by lending and then circulated until it is extinguished, it is just a system of payments and balances, both positive and negative.

Bank Lending and the Value of Money

There is a theory of money that says that the value of private bank money derives from the need of debtors to obtain money in order to settle their debts.  Some people see this as being rather circular.  If the value arises from the existence of debts, how do those debts come about?  How can you borrow or lend something without knowing its value in the first place?

This is where it is useful to see that debts created through bank lending do something special, that other debts on their own do not.  A useful way to see this is by looking at the monetary circuit described by circuit theorists.  In this scheme, banks lend to firms to enable them to pay their workers.  The bank loans create the money, which the firms then hand over as wages.  At a later stage, workers spend their money, buying goods from the firms.  The firms use the money they receive as sale proceeds to repay their loans.

It is assumed here that firms need to pay their workers in advance.  Production takes time, so the firms cannot simply exchange goods for labour.  One possibility would be for firms to pay workers with contracts for delivery of goods at a future specified time.  So, for example, a worker might receive coupons entitling him to receive 100 widgets at the end of the month.  This is fine, but if the worker doesn't want widgets himself, he then has to find people to exchange his widgets (or widget coupons) with, in return for the goods he actually wants.

On the face of it, money does not help here.  For at least the widget coupons entitle the holder to a specified amount of something.  Money doesn't entitle the holder to anything.  If firms just printed up some coupons, with no specific entitlement attached to them, and tried to use these to pay their workers, the workers would quite reasonably reject them.  They would have no reason to believe that anyone would subsequently accept them in return for anything of value.

Involving banks means that the coupons can become tokens for the cancellation of debt.  Each firm can (in fact must) use coupons to repay their loans from the banks.  The coupons thus have value, without the need to be linked to a specific commodity.  Firms need to obtain coupons to meet their contractual obligations to banks.  Workers will accept coupons as wages, because they know firms will be prepared to accept them for goods at a later stage.

So creating value in the coupons specifically requires the involvement of the bank.  In the monetary circuit, the firms are the debtors, the workers the creditors and the banks are just intermediaries.  But although they only intermediate, the role of banks is critical to establishing a money denominated in an abstract unit.  Firms and workers can create contracts through bilateral agreements, but these can only ever be linked to specific commodities.  Credit money requires a tripartite arrangement.

Of course, there is still no guarantee for holders of coupons of what they will be worth in the future.  They know there will be a demand for coupons, but if there is a rush of new lending there may be an excess supply of coupons depressing the value.  People's belief in the intention of banks (specifically the central bank) to control the rate of credit creation is crucial to establishing the value of this money. 

Cochrane, Fiscal Policy, Interest Rates and Neo-Fisherites Again

John Cochrane's latest paper has raised a few eyebrows, due to some of the distinctly post-Keynesian ideas which seem to have crept into his recent thinking.  His focus is specifically on a world with interest on reserves, but his analysis involves a very significant role for fiscal policy.  This comes down to a point Cochrane has made before about the idea that private sector expenditure can be seen as the counterpart of a demand by the private sector for holdings of public debt.  This concept underlies some important ideas in post-Keynesian economics.

Cochrane considers a model with no government expenditure other than transfer payments and, for the most part, with government debt all of single period maturity.  His main equation (slightly restated here) is:

                                P = D_-1 . ( 1 + i_-1 ) / E(s)

where P is the price level, D is the nominal value of government debt* and i is the nominal interest rate on that debt.  So the top part of the expression is the level of debt left at the end of the previous period plus the interest thereon.  E(s) is the expected real value of current and future primary budget surpluses, discounted at the household rate of time preference.  Because Cochrane assumes no government expenditure, this is basically the expected value of the future net tax burden.

So what this equation amounts to is a statement that the current equilibrium price level is the ratio between the nominal amount of government paper outstanding, divided by the total real value of tax liabilities that will have to be paid using that paper.  The idea here is as follows.  Households expect to have to pay a certain amount of taxes in the future and they hold government debt to cover this.  If the value of the debt they hold is greater than the value of the taxes they expect to pay then, even with consumption smoothing, they can afford to increase current consumption.  This would push up prices, reducing the value of the debt, until the equilibrium price is reached.

So we have here the idea that the real value of the overall government debt is being driven by the future tax liability, and that the price level then depends on the nominal amount of debt available to cover this real value.  This might be said to be essentially a Chartalist position.  But it is also possible to analyse this in terms of backing.  In this view, the value of the debt is driven by the backing of the real value of the expected future tax receipts.  Within this model, these could be said to be just alternative ways of looking at the same thing.

Amongst other things, Cochrane uses his model to look at the relationship between the interest rate and inflation.  One feature of this is easy to see from the above equation.  The growth in the nominal value of government debt depends directly on the nominal interest rate.  If the interest rate is increased by 1%, with no change in expected real primary surpluses, the price level needs to grow at an additional 1% to compensate.

This result should not be a great surprise to post-Keynesians.  In a simple Godley & Lavoie model with only short term debt, they also show how an increase in the interest rate would be expansionary (section 4.5.1).  There are certainly differences between the assumptions in these models, but in essence the reason is the same.  An increase in interest rates transfers nominal value from public to private sector.  If this results in the private sector holding more government debt than it wants it will try to spend to get rid of the excess.

We can see that in Cochrane's simple model, when we have perfectly flexible prices, the price level would rise immediately.  So a rise in nominal interest rates would lead to an immediate inflation, with no temporary deflationary period - the neo-Fisherite result.  I have pointed out before that this is a much less likely result with anything other than single period debt.  This model actually provides a good way of showing this again.

Let's assume instead that government debt also includes some longer term bonds, previously issued at a discount and maturing in the current period.  We now need to write our equation as:

                                P = [ DS_-1 . ( 1 + i_-1 ) + DL ] / E(s)

where DS is the nominal value of the short term debt and DL is the redemption value of the longer dated bonds.  Because these longer bonds were already in issue, their payout is not impacted by any change in the current interest rate.  It is immediately obvious that a 1% increase in the current interest rate now translates into a less than 1% increase in the equilibrium price level.  But if inflation rises by less than the nominal interest rate, the real interest rate has risen which will have real effects.  In fact, the only way that flexible prices could eliminate any real effect is if the price level falls in the period in which the increased interest rate is set.  So we get an initial deflationary period, and only thereafter can inflation increase.  This effect with longer dated debt is also illustrated in Godley and Lavoie (section 5.7).

Cochrane's methodology is clearly mainstream and he excludes many things heterodox economists might consider important.  Nevertheless many of the results he derives are consistent with post-Keynesian analysis.  Cochrane tends to be very dismissive of those who consider themselves Keynesians, but I've been increasingly wondering whether he isn't a closet Keynesian himself.  I hope that at some stage he will be able to come to terms with this.

* Cochrane uses the redemption value of discounted debt as his variable.  I have used issue value, because it helps illustrate my point. 

His B = my D. ( 1 + i )