Monetary disequilibrium theorists must face this question: "If this recession was caused by an excess demand for money, how come interest rates are so low? Doesn't an excess demand for money mean an excess supply of bonds and rise in interest rates?"
[Warning: this post is long, rambling, and unclear. I ought to tear it up and write a couple of shorter and clearer ones. I may do that later, when (if) I get my head clearer. Read at your own risk. Maybe skim it first.]
Despite all my brilliant theoretical proofs of the metaphysical necessity of monetarism — how a general glut can only be caused by an excess demand for the medium of exchange — Brad DeLong has got the perfect comeback: "OK, the 1982 recession was caused by an excess demand for money, as shown by the very high interest rates. But the recent recession must have been caused by an excess demand for safe assets in general, otherwise we wouldn't be seeing interest rates on safe assets near zero." (He didn't actually say those words, but he might have done.)
So I'm going to sketch a simple model where an excess demand for money causes a recession but no rise in (real or nominal) interest rates.
The basic idea is simple. An excess demand for money causes unemployment for the "unlucky". The unemployed can't borrow (nobody will lend to the unemployed), and can only spend down their money balances by buying from those who are "lucky" and remain fully employed. The lucky employed get the money that used to be held by the unlucky unemployed. So nothing changes for the employed. And the unemployed are shut out of all markets, so can't affect the equilibrium. And if unemployment causes expected deflation, nominal interest rates will fall via the Fisher effect.
[Warning to lefties: before you get too excited by this model, remember that the distribution of money is not the same as the distribution of wealth. This model is much closer to Milton Friedman than to Karl Marx.]
Before getting started on building the model, I need to talk about loanable funds vs liquidity preference, the ISLM model, and take a sort of cheap shot at Paul Krugman to illustrate my point. (It would be a cheap shot if I didn't admit it were a cheap shot, if you can handle the Liar Paradox.)
Digression on loanable funds vs liquidity preference and ISLM and Paul Krugman and stuff.
The price of apples (if it is flexible) is set in the apple market to equilibrate the demand and supply of apples. If there's an excess demand for apples it rises; if there's an excess supply of apples it falls.
OK. So where is the rate of interest set? What market? What are the demands and supplies it is supposed to equilibrate?
Liquidity Preference says it is set in the money market, to equilibrate the demand and supply of money. Which is totally stupid, because there is no money market. Or rather, in a monetary exchange economy every market is a market for money plus one other good. When finance guys talk about the "money market" they are really talking about the market for short-term loans. When you lend someone money, you get an IOU in return. That IOU is a bond. So when we talk about "the money market" we are really talking about the bond market. So let's call the thing by its proper name. The "apple market" is the market where money is exchanged for apples; the "bond market" is the market where money is exchanged for bonds.
But isn't the "bond market" just another name for the market in "loanable funds"? If so, what the hell is the difference between the liquidity preference and loanable funds theories of the rate of interest?
Another way of describing the loanable funds theory is to say that the rate of interest adjusts to equilibrate desired savings and desired investment. OK. But since (closed economy) national savings is defined as Y-C-G, we can do some trivial math and re-write S=I as C+I+G=Y. So loanable funds says that the rate of interest adjusts to equilibrate desired consumption plus desired investment plus desired government spending to desired sales of newly-produced goods? In other words, loanable funds says that the rate of interest adjusts to equilibrate the output market?
Which is weird. Sure, the demand for output may depend on the rate of interest. But can we jump from that to saying that the rate of interest is set in the output market? Can we say that an excess demand for output will put upward pressure on the rate of interest? The rate of interest is the (reciprocal of) the price of bonds, not the price of output. The demand for apples may depend on the price of pears. But we don't say that the price of pears is determined in the apple market.
The ISLM model was supposed to reconcile the liquidity preference and loanable funds theories of the rate of interest. IS shows the loanable funds answer, as a function of Y; LM shows the liquidity preference answer, as a function of Y. In the short run, with M/P fixed, Y adjusts until both curves give you the same answer. In the long run with P and hence M/P flexible, and Y fixed by the LRAS curve, M/P adjusts until the two curves give the same answer.
But the ISLM is trying to reconcile two opposing theories of the rate of interest, neither of which make any sense.
Here's my cheap shot at Paul Krugman:
Paul says (H/T Brad De Long): "Now equilibrium in a three-good model can be represented by drawing curves that indicate combinations of prices for which each of the three markets is in equilibrium."
No it can't. At least, not if one of the three goods is called "money". In a barter economy, with n goods, there are n(n-1)/2 markets. So if n=3 that means three markets. But in a monetary exchange economy with n goods (including money) there are (n-1) markets. So if n=3 that means two markets.
Paul also says: "Although there are three curves, Walras' Law (if all markets but one are in equilibrium, that market is in equilibrium too) tells us that they have a common intersection, which defines equilibrium prices for the economy as a whole."
But Walras' Law is wrong in a monetary exchange economy. It only works in a Walrasian General Equilibrium model with a single market in which all n goods can be traded for each other and no agent is ever unable to buy or sell as much as he wishes. That's very different from a model of a monetary exchange economy used to explain excess supply recessions where people can't sell as much labour as they want.
This is a cheap shot because lower down Paul says: "Sixty years on, the intellectual problems with doing macro this way are well known. First of all, the idea of treating money as an ordinary good begs many questions: surely money plays a special sort of role in the economy."
Yes. Money does play a special role. For one thing, money does not have a market of its own. It is traded in every market against every other good; and all the other goods are traded only against money. For a second thing, if we lump all output into one good, we have to recognise that every agent is both a buyer and a seller of that good. We sell our own output for money; and use money to buy others' output. We don't barter our own output for others' output.
So let's start from scratch.
A sketch of my model.
There are three goods: backscratches; bonds; and money. There are two markets: the output market, where backscratches are traded for money; and the bond market, where bonds are traded for money. The rate of interest (aka the price of bonds) is perfectly flexible. It adjusts instantly to excess demand or supply for bonds, so the bond market always clears. The price of backscratches is sticky, or fixed if you like, in terms of money. So the market for backscratches may not clear.
People must trade, because you can't scratch your own back. And you can't barter backscratches, or trade them for bonds (promise to pay later) because you can't see a person's face when you are scratching his back. (OK, so cook up your own silly story for the microfoundations of monetary exchange).
That makes the output market very different from the bond market. Each agent is either a buyer of bonds or a seller of bonds. But each agent is both a buyer and seller of output.
All agents are identical, except: agents differ by "luck". Luck is distributed along a continuum. In the event of an excess supply of backscratches, where demand is only 60% of supply, the luckiest 60% of agents will be able to sell as many backscratches as they want, and the unluckiest 40% will be able to sell none.
Unlucky agents, who are unemployed, are unable to access the bond market. Everyone knows they are unemployed, and therefore unlucky, so they cannot borrow money from lucky agents because they might stay unemployed and not be able to repay the loan. (OK, this assumption could be relaxed a bit, but shouldn't affect the results too much).
In advance of a recession, agents don't observe their own luck, so all agents are identical ex ante, and the unlucky won't save more than the lucky.
Start in full-employment equilibrium. All agents are buying and selling backscratches for money. But no bonds are traded, because all agents are ex ante identical. In full-employment equilibrium, it's a representative agent model.
Now let's shock the model.
Shock 1. This example is very contrived, but is also the simplest. Assume that a fire destroys all the stock of money held by the unluckiest half of the population. In this example, the unlucky are doubly unlucky. They are unlucky in the market for backscratches, and they are unlucky in the fire too. What happens?
In the new equilibrium the economy carries on exactly the same as before for the lucky half of the population, while the unlucky half of the population is shut out of all markets, and so has no effect on the equilibrium. The rate of interest initially stays the same.
The unlucky unemployed have no money, so can't buy backscratches. They will want to borrow money, but nobody will lend to them, because they are unemployed. They want to sell backscratches, but the lucky employed are already buying as many backscratches as they want from each other, and the unlucky are at the end of the queue supplying backscratches, so the demand runs out at the halfway point. The fire that destroyed their money might as well have destroyed them too, in terms of how it affects the equilibrium in the luckier half of the economy. Except:
Of course, the excess supply of backscratches will slowly cause the price of backscratches to fall (assuming it's sticky but not stuck). Given long enough, this fall in the price level will increase the real money supply by enough to restore full employment. But in the meantime the expected deflation will lower the equilibrium nominal rate of interest.
Shock 2. Now assume half of each agent's stock of money gets destroyed by fire. (So, unlike Shock 1, the unlucky agents are only unlucky in the market for backscratches, not in the fire.) What happens?
Initially, each agent will respond in three ways. He will supply more bonds. He will supply more backscratches. He will demand fewer backscratches. All to try to rebuild his stock of money. But since the stock of money is fixed, they must collectively fail.
Since there is an excess supply of backscratches, some agents near the unluckiest end of the spectrum will be unemployed. They cannot sell backscratches to earn income. They cannot sell bonds to tide them over till the recession ends. They can slowly run down their stocks of money, which is earning 0% interest. Or they can sell some of that money to buy bonds, to earn positive interest, then slowly run down that stock of bonds. Either way, the money once held by the unemployed will, immediately or over time, end up in the pockets of the employed.
What does the new equilibrium look like?
To a first approximation, the equilibrium in Shock 2 will look exactly the same as the equilibrium in Shock 1. The only difference between the two shocks is a small change in the distribution of wealth. The unlucky half of the population is slightly better off, and the lucky half of the population slightly worse off, in Shock 2 than in Shock 1. At the previous equilibrium rate of interest, the unlucky unemployed will want to buy bonds with money, and the lucky employed will want to sell bonds for money. So both the demand and the supply of bonds will increase relative to Shock 1. This small change in the distribution of wealth will have an ambiguous effect on the rate of interest, compared to the equilibrium in Shock 1. And that effect will be small anyway, since stocks of money are such a small part of total wealth.
So, in Shock 2 as in Shock 1, the initial impact of the excess demand for money will be to leave the rate of interest (approximately) unchanged. And since the excess supply of backscratches will eventually cause expected deflation, the nominal rate of interest will fall.
Relaxing the key assumption.
What happens when we relax the assumption that the unemployed cannot borrow by issuing and selling bonds? The unemployed would want to borrow to smooth their consumption stream over time, and will be able to pay back the loan if the recession is short-lived. But the IOUs (bonds) they issue will be riskier, and will have to pay a higher rate of interest, than the safe bonds issued by employed agents. That might be a way to reconcile my model with Brad DeLong's theory that the recession was caused by an excess demand for safe assets. But, I would want to insist that it was the recession that caused previously safe bonds to become unsafe, and the recession was caused by an excess demand for money.
Worthwhile Canadian Initiative 
I’ve read this more than once but I keep ending up thinking you’re almost losing an argument with your own straw man, so I know must be missing something. I don’t understand why the awkward permanent-luck and borrowing constraint features are necessary. Why can’t you just have uncertainty or pessimism about future recovery? Maybe because people aren’t sure if the problem is nominal or structural or because they are worried that future, unaccommodated surprise fires might render the recession a recurring nominal problem. Then with some kind of stochastic unemployment churn you have the unemployed’s demand for bonds offset by the employed’s supply of bonds given their greater risk of future unemployment. Then low interest rates reflect the possibility that things might get even worse or the fear of more future deflationary fires.
Moreover, I don’t see why this sheds any light on whether it was the chicken (flight to safe future consumption) or the egg (demand for the money as a MOE). The low rate question might be quandary in either case. Say there is a flight to safety because someone makes a scary movie called Future Backscratch Volatility. The supply of bonds increases as future consumption becomes more uncertain, which spills into an excess demand for money. Disequilibrium ensues, production declines and unemployment emerges. If recovery is expected, you still might wonder why bond prices don’t decline. Sure, there might be an offset to some extent relative to moment zero if the movie-volatility fear was still around, but not necessarily such that it would actually mean high bond prices net of the expected future recovery. In fact, the economic shock might be seen as a realization of the volatility everyone was fearing (now they are in volatility and expecting recover), so and might go away entirely from expectations. You’d still wonder why rates were low. In other words, the flight to quality leading to an excess demand for money might not be expected to remain a flight to quality once you hit quasi-equilibrium, just as an excess demand for money leading to a flight to quality during disequilibrium might not expected to remain a flight to quality if a recovery is expected. I don’t see how you can tease out the prime mover by going around counting the chickens and eggs after-the-fact. Both explanations have some explaining to do if there are too many eggs and not enough chickens.
“A cash advance on your Visa is a bond market. You get cash, and you give an (electronic) IOU (bond) in exchange. You just sold a bond.”
As Andy Harless says – but I can buy apples with said (electronic) IOU (bonds), so this bond functions as a medium of exchange, in the same way actual cash money does. The set of goods which are “medium of exchange” includes both these particular bonds, and cash.
The point here is that there is a money market, i.e. a market in which the unit of account is traded with the medium of exchange, at a price which may be flexible or fixed.
“Imagine a world with apples, bananas, and money. If the prices of apples and bananas were both fixed, we would be forced to see recessions as being due to an excess demand for money. But if the price of bananas were flexible, recessions would look like being caused by the wrong relative price of apples and bananas. For “bananas” read “bonds”. We mistake the symptom for the cause.”
Well, validating that wrong price by creating more bananas solves the issue…
Besides which, the price of bonds should be lower before the shock and higher after the shock, right? Even if the unlucky are shut out of the market, thus preventing the price of bonds from skyrocketing immediately, at least the price of bonds should not fall, right? How does this square with interest rates actively decreasing during a recession?
I think the original post makes too much of the loanable funds – liquidity preference issue (even if this section is only a digression).
In a Walrasian framework it is the vector of prices announced by the auctioneer which causes the vector of market excess demands to reach zero (ie clear all markets). It is not the case that the apple price clears the apple market, the pear price clears the pear market, and so on.
The reason why it is easy to slip into the argument that the interest rate is set in the loanable funds market, and that the money market doesn’t matter, is that money enters excess demand functions in an atypical way. Excess demands for goods are homogenous of degree zero in nominal money and absolute prices, and the demand for money is homogenous of degree 1 in prices. As a result, if the quanity of money doubles, and we focus exclusively on comparison of the final equilibrium states, we find that real variables (including the interest rate) are unchanged, and all that has changed is the absolute level of prices. In contrast a doubling of apple supply would in general affect the relative price of pears and oranges. This special property of money makes it easy to argue that liquidity preference cannot affect the interest rate, which is determined solely in the market for loanable funds. Nevertheless the increased supply of money has affected excess demands for other commodities: it’s just that the auctioneer ensures that trade does not take place until these effects have worked themselves out, so they are effectively invisible.
Keynes abolished the auctioneer, and this changes the model fundamentally. If prices are sticky (and it is sensible to assume that goods prices and the wage are stickier than bond prices) the initial impact of a monetary expansion will be on interest rates and possibly (if we believe in direct real balance effects) on the demand for goods. In the short-run this means that investment and output will change. In the long-run (particularly if we start from full-employment) we may get back to the original real equilibrium, but with a higher price level. But the liquidity preference schedule is crucial in explaining the size of the inital effect on interest rates, and the circumstances (the liquidity trap) in which a simple-minded quantity theory breaks down.
On a different issue in the post, surely we should effect the relationship between interest rates and recession to depend on the interest rate policy rule (or alternatively the money supply rule) in effect. many postwar recessions in both the US and UK were set off by official action to tighten monetary policy, either because (in the US) the Fed was ‘taking the punchbowl away from the party’ or (in the UK) because there was a perceived need to protect the currency from depreciation. This is not the same type of initial shock as 1931, or 2008. In addition, it’s important to distinguish beteween the crisis and the subsequent recession. In 19th century Britiah banking crises standard policy was for the central bank to supply liquidity at a high (‘penal’) interest rate, both because high rates were necessary at such times in a gold-based system (and the high rate could attract additional gold from abroad, and thus help to maintain convertibility) and because the knowledge that a penal rate would be charged reduced moral hazard and punished improvident bank executives. In 2008 (the first major and prolonged banking crisis since 1931) policy-makers opted instead to provide unlimited liquidity at very low interest rates, in the hope that this would be less disruptive to economic activity than the traditional policy. It’s not clear we have done much better as a result.
[Edited very slightly to fix formatting. NR]
dlr, david, William: those are 3 good comments. I’m not ignoring them. I’m thinking about them. Will respond if and when I think of a good response. My mind is currently wandering off onto upward-sloping IS curves.
dlr: “I don’t understand why the awkward permanent-luck and borrowing constraint features are necessary. Why can’t you just have uncertainty or pessimism about future recovery?”
You might be uncertain about recovery, but the standard idea of a recession is that output is below trend, so in that sense you must be optimistic in a recession. (Which, I admit, is strange). People (unless they don’t realise they are in a recession) believe that things will get better, even if they don’t know how soon or how much things will get better.
Still thinking about your chickens and eggs.
david: I might pay for something by Visa, but what really happens is that the credit card company pays for it on my behalf, and I then owe the credit card company. It is as if I sold a bond to the credit card company, and the credit card company pays the merchant in money. My bond does not circulate, so isn’t a medium of exchange.
“Even if the unlucky are shut out of the market, thus preventing the price of bonds from skyrocketing immediately, at least the price of bonds should not fall, right? How does this square with interest rates actively decreasing during a recession?”
In my simple model, real interest rates will not rise. But yes, they won’t fall either. I reckon I’m halfway there 😉
William: “In a Walrasian framework it is the vector of prices announced by the auctioneer which causes the vector of market excess demands to reach zero (ie clear all markets). It is not the case that the apple price clears the apple market, the pear price clears the pear market, and so on.”
Understood. I wasn’t meaning to contradict that. In my example, I said that the demand for apples may depend on the price of pears, but that doesn’t mean the price of pears is set in the apple market. What I am talking about is disequilibrium price adjustment. An excess demand for apples means the price of apples will rise. It does not (necessarily) mean the price of pears will fall.
Fully agree with your next paragraph, down to:
“Keynes abolished the auctioneer, and this changes the model fundamentally. If prices are sticky (and it is sensible to assume that goods prices and the wage are stickier than bond prices) the initial impact of a monetary expansion will be on interest rates and possibly (if we believe in direct real balance effects) on the demand for goods…..But the liquidity preference schedule is crucial in explaining the size of the inital effect on interest rates,…”
But is an excess demand for money really the same thing as an excess supply of bonds? In equilibrium we may say something like the demand for money is Md/P=L(Y,i). But once we are out of equilibrium, there are potentially n-1 different excess demands for money, and the price of bonds will be responding to only one of those n-1 excess demands for money.
“Even if the unlucky are shut out of the market, thus preventing the price of bonds from skyrocketing immediately, at least the price of bonds should not fall, right? How does this square with interest rates actively decreasing during a recession?”
What is important is the rental rate of capital, not the bond yield per se. If capital values are expected to decline, households will demand a capital rental rate that compensates them for this decline.
The arbitrage free condition would be that buying capital, renting out for one year, and then selling it should give the same return as buying a 1 year bond, plus a risk term.
Even ignoring possible increases in this risk term during a downturn, if capital goods prices are expected to decline faster than bond yields, then the capital rental rate will increase even if bond yields are falling.
Nick
“You might be uncertain about recovery, but the standard idea of a recession is that output is below trend, so in that sense you must be optimistic in a recession. (Which, I admit, is strange).”
You may not be optimistic but you behave opimistically at the very through when for some reason ,maybe merely stochastic, you do something stimulative. Recovery, like depression, might have a Minsky moment.
RSJ: “If capital values are expected to decline, households will demand a capital rental rate that compensates them for this decline.”
What you are talking about there is akin to the distinction between real and nominal interest rates.
Jacques Rene: what you say sounds plausible, but can it be squared with expectations that would be even vaguely rational, i.e. model consistent? If the model says that output is temporarily below trend in a recession, people should expect higher than normal growth.
Nick,
In the spread
(nominal) bond yield – % change in (nominal) capital goods prices
.. the inflation term cancels out, so the spread is a real quantity. It measures the return from supplying a capital good in terms of consumer goods.
I wasn’t arguing that the excess supply of money was equal (one-for-one) to the excess demand for bonds. My argument was that in a Keynesian framework (no auctioneer, so sticky prices) an excess supply of money would have more effect on bond prices (which adjust quickly) than goods prices (which are sticky). Any impact on goods markets would be more likely to affect quantities, and this does not send any signal to agents other than the firms for which demand has risen. I’m not sure what is meant by ‘potentially n-1 different excess demands for money’ – since money is (in my view) the one commodity where rationing cannot occur, the excess demand for money can always be aggregated across markets.
Here is a quick and dirty regression that takes capital values as well as bond yields into account (assuming risk-premia are fixed). The implied capital rental rate increased during the recession even though the CB was cutting yields. Increases to the implied rental rate correspond to negative shocks to real GDP growth.
By way of reporting on Academe, what William Peterson said about planned vs. unplanned recessions was also said to me by Torben Drewes of Trent University in a lecture on the financial crisis (it wasn’t a recession yet) in 2008.
Part of the problem with the debate over this recession is that many economists can’t get their mind around the fact that there is more than one type of recession. 1982 is less relevant to today than 1931 which has more commonalities with what actually transpired.