We always knew that interest rate targeting could never work in theory, because it left the price level indeterminate. But it seemed to work well in practice, and kept inflation close to target, so we eventually learned to overcome our theoretical squeamishness and embrace it as part of the reality of how modern central banks operate. But now interest rate targeting has failed in practice, and failed badly. It cannot keep inflation, and expected inflation, on target. We want to loosen monetary policy. And because monetary policy is interest rate targeting, we can't, because interest rates on safe liquid assets are already at zero.
OK, this is probably the weirdest post I have ever written. I am going to argue that interest rate targeting is not what central banks really do; it's a social construction of what they really do. Interest rate targeting is not reality, it's a way of framing reality.
That was weird enough, but I'm now going to get really weird. The failure of monetary policy is not caused by anything central banks are actually doing; it's caused by central banks' way of framing what they are doing, and by the rest of us accepting that same framing. The current recession was caused by those (and that includes especially central bankers themselves) who think that central banks use an interest rate as the control instrument. It's the framing of what central banks do that caused the mess, not anything central banks are actually doing. The social construction of reality is what dunnit!
Let me give an example of what I mean by "framing". It would be quite conventional to say that the Bank of Canada sets an overnight rate target as a function of various indicators, including inter alia the exchange rate. Write that reaction function as i=F(S,I), where i is the overnight rate, S the exchange rate, and I a vector of other indicators. Now, just invert that reaction function and write it as S=G(i,I). Exactly the same thing mathematically, but see how the framing has changed. Now the Bank of Canada is targeting the exchange rate as a function of various indicators, including inter alia the overnight rate. The Bank of Canada is doing exactly the same thing, but the account it gives of what it is doing, the framing, has changed radically. And the Americans would get mad at us if the Bank of Canada targeted too low an exchange rate!
Now here's a real-life example of how the Bank of Canada has changed the way it frames monetary policy, and why framing matters. The Bank of Canada used to construct and publish a "Monetary Conditions Index" which was a weighted average of the overnight rate and exchange rate. And it would talk about what was happening to the MCI whenever it changed the overnight rate target. This did not mean it kept the MCI constant over time. It would adjust the overnight rate target and so adjust the MCI whenever it felt that the indicators warranted it. But the Bank of Canada eventually stopped publishing the MCI, and stopped talking about the MCI, because it found the MCI interfered with the Bank's "communications strategy". People were thinking of the stance of monetary policy in terms of the MCI, and the Bank wanted people to think about the stance of monetary policy in terms of the overnight rate target.
When it stopped publishing and talking about the MCI, the Bank wanted to change the framing of monetary policy. And it succeeded, even though anyone with a calculator can construct the MCI from public data.
As any philosopher or sociologist can tell you, the way reality is socially constructed can have real effects. "I may lawfully nourish myself from this tree; but the fruit of another of the same species, ten paces off, it is criminal for me to touch. Had I worn this apparel an hour ago, I had merited the severist punishment; but a man, by pronouncing a few magical syllables, has now rendered it fit for my use and service…" says David Hume. (One fruit tree is on his land, the other isn't; the guy said he could buy the clothes). When one man body-slams another, is it a fight or a hockey game? A British visitor can't always tell the difference, and not because his eyesight is any worse than the Canadians who know how to frame what they are seeing. One framing causes a cheer; the other causes a call to the police.
How could a stranger, with good eyesight and a time series graph of all macro variables, really tell which one of those variables the Bank of Canada was targeting? OK, there is a way, if he has a very high-frequency data set. The stranger would notice that the overnight rate tended either not to move, or else move in discrete jumps, about 8 times a year. But if Fixed Announcement Dates were held monthly, or weekly, or daily, or hourly, the stranger would see nothing. In the limit, with continuous FADs, the "fact" that the Bank of Canada targets the overnight rate would be only a socially constructed "fact". Since I don't think hourly FAD's would make much difference to policy, especially under current circumstances, when nobody expects the overnight rate to change anytime soon, I think I'm safe in saying that interest rate targeting is at least 95% a social construction of reality.
A socially constructed reality, like who owns what, is a game-theoretic equilibrium held in place by players' shared expectations of how each would react to an out-of-equilibrium move, something they never of course observe in equilibrium. They follow the rules because of what they think would happen if they didn't.
A given time path of overnight rates would be highly inflationary under one set of expectations about inflation and real output growth, and highly deflationary under another set of expectations about inflation and real output growth. A time path of interest rates cannot measure the stance of monetary policy. A time path of interest rates does not define a coherent monetary order. An interest rate reaction function, with feedback from expectations to the time path of interest rates, might define a coherent monetary order. (Equilibrium would be maintained by players' shared expectations of what would happen out of equilibrium.) We thought it did for a decade or two, but events have proved us wrong. We have hit the zero limit where it can react no further, so all we have left is the time-path itself, rather than a reaction function.
If expectations of monetary policy coalesced around the time path of some nominal variable (one with $ in the units), we could escape the liquidity trap. Just let that time path for the nominal magnitude grow over time at a fast enough rate and the equilibrium overnight rate will rise above zero. If monetary policy were framed as the central bank setting some nominal variable, expectations of monetary policy could then coalesce around that variable, monetary policy would be loosened, and the overnight rate, as some endogenous response to monetary policy, would rise above zero. Under an alternative framing, a loosening of monetary policy would mean an increase in the overnight rate.
But we are stuck with framing monetary policy as setting an interest rate. So there is no way the Bank of Canada can try to say that an increase in the overnight rate counts as a loosening of monetary policy. Because given the way expectations are determined by the framing, it would be a tightening of monetary policy. The Bank of Canada is fettered by its own social construction of what it is doing.
Oh hell, it's late. I'm posting this anyway.
Worthwhile Canadian Initiative 
I should just add that at the height of the crises when the Fed was lending directly in the money markets this was a good idea at the time because otherwise companies like GE might have failed when they should be viable, but the Fed is correctly trying get out of the business of credit intermediation as quickly as it can.
And the Fed buying things like comercial paper was NOT monetary stimulus in the sense of trying to promote investment, it was saving the economy’s life.
just to be extra clear, this statement:
“GE would not be encouraged to invest more because their private cost of capital has gone down. ”
should be read as:
“GE would not infer that their private cost of capital had gone down and thus be encouraged to invest more”.
I guess it anyway should have been clear from the subsequent discussion…
Adam P,
Assuming that the Fed can influence interest rates by marginal manipulation of the stock of base money (as you know I am sceptical about the extent of that influence), the Fed could get quite a long way in conducting its normal OMOs just by trading GE stock. Without looking it up, I believe the stock of non-interest-bearing base money (which I assume is roughly what the stock of base money would be but for QE) is about $900bn, compared with the market cap of GE of about $170bn, so GE would just about be big enough to support the normal size (say about $10bn) of marginal adjustments to the stock of base money (although the Fed would have to continue using some other assets – historically treasuries – to back – if I may use that term on Nick’s blog – the underlying increase of the base money stock). However, I dare say it would be very expensive for the Fed to make such marginal adjustments by buying and selling so much GE stock almost every day. The repo market is larger with smaller transactions costs, and the rapid maturity of repo loans puts the market in a structural short (base money) position.
I believe that Scott Sumner’s NGDP futures targeting scheme is just about determining the size of OMOs rather than how they are conducted.
Rebel no, you can’t seperate size of OMO’s from what get’s bought. Once, say, T-Bill rates are down to zero the fed is supposed to start buying other, positive yielding assets. Scott has said this.
With respect to conducting monetary policy by trading GE stock you’ve completely misssed the point.
Excellent point, Adam P.
This post was about the transmission mechanism (if it exists) between the FFR and anything else. Nick just assumed a mechanism in his model, sidestepping the question altogether.
Brilliant!
You are also completely correct that this current recession is not a failure of interest rate targeting as a monetary tool. It isn’t even a failure of monetary policy, since monetary policy doesn’t work (not transmission mechanism). It’s a failure of economic management by Obama & Co., aided and abetted by a failure of the economics profession to understand how the financial system, and economy, works.
well, you can’t blame it on Obama’s administration, that’s ridiculous. It began before he was even elected.
There will be some relationship between the GE price and the amount of GE stock traded, which will adjust the base money supply by a certain amount, which in turn will be related to the price level. If these linkages are captured by a model, it would be possible to target inflation by setting the price of GE stock. I agree that it’s not very practical though – I dare say that there are many influences on the GE stock price other than the central bank’s purchases, so the degree of control would be poor.
Your not making any sense Rebel.
Go back to first principles, Adam. The basic relationship is between money stock and the price level. Because money is a close substitute for short term debt, the short term interest rate (ie the price of short term debt) is especially closely related to the size of the money stock. This applies however the money stock is adjusted – whether by trading short term debt, gold, foreign currency, labour, or even, as Eddie George once said, Pink Ladies (gin!). However, because the demand for base money depends strongly on short term influences (eg Christmas shopping), central banks found that communicating monetary policy in terms of short term interest rates was less confusing, so that became the conventional approach. The problem is that, since then, the idea of relating interest rates to inflation has become so embedded that people, especially academics, forget the first principles. Indeed, academic monetary economics reached the point that it hardly mentioned money(though that may be one change for the better to emerge from the financial crisis). The first principles still hold, however, so that monetary policy can be set in terms of any asset price, albeit with less reliability the more distant the asset is as a substitute for base money.
If you want my view of how this works in more detail, please read my blog post: http://reservedplace.blogspot.com/2009/04/easing-understanding.html
I do agree though that this recession was not caused by interest rate targeting overlooking a liquidity trap in the brief period of panic around the failure of Lehman, although I would agree with Scott Sumner that policy was inadvertently tight at this time if it could be shown that banks were deterred by expectations of falling prices from borrowing at the already low interest rates prevailing at that time. I believe that the recession was inevitable long before then.
Go back to first principles, that’s what I’m trying to get you to do.
However, now I see the problem. You think that “The basic relationship is between money stock and the price level”. ACTUALLY NO, although I’m beginning to realize that this is the most mis-understood point in economics (already the most mis-understood disipline in history).
I’ll say it again, in caps again, JOINTLY DETERMINED. We’re talking general equilibrium here. With fiat money the relationship between the money stock and price level is determined in a system of equations that jointly determine rates of substitution across goods, time and states of nature. And that’s when we have frictionless markets (costless and perfect intermediation) to facilitate the solution. Right now we far from perfect intermediation.
To quote Krugman, the intellectual man among boys, “if you try to think about either interest rates or the price level in terms of just a single market — interest rates determined by supply and demand for lending, price level by quantity of money, full stop — you get it all wrong, especially in times like the present.” Hence the idiocy of Scott Sumner…
and of course, even when itermediation is not impaired we have all sorts of REAL frictions (search, time to build etc…)
I said basic relationship; I did not say “only” or suggest that it works just one way round. But I think that it is the key relationship for practical purposes.
Krugman, by the way, is one of the worst offenders among academic economists when it comes to drawing conclusions from an inadequate understanding of monetary policy operations (although I suspect that he really does know better, but a more realistic description would make his arguments harder to make). He frequently describes open market operations as trading t-bills when discussing the ineffectiveness of monetary policy.
RebelEconomist said: “Without looking it up, I believe the stock of non-interest-bearing base money (which I assume is roughly what the stock of base money would be but for QE) is about $900bn, …”
Is the $900 billion all currency, or is there some bank reserves not paying interest?
Here is CURRCIR from the St. Louis Fed (FRED).
http://research.stlouisfed.org/fred2/data/CURRCIR.txt
I have a question about it. Does CURRCIR include currency in bank vaults or not?
Nick:
I can see from the comments how strong is this social conception of reality.
There is a market process by which open market interest rates impact short run nominal interest rates, so the Fed must be targeting those interest rates. It must must!
That there could possibly be other parellel market processes working in a differnet fashion–that just can’t be.
I think it is blindly obvious. If there is a falure of policy so that we have expected deflation and a low or negative natural interest rate rate, then if the Fed is seen willing to do what it takes to break out of the problem, say by purchasing long term and risky bonds, or let the short term low risk rates go negative (which would require a suspension of currency payments,) then it is very likely that expecations would shift and there would no longer be expected deflation and there would be an increase in the natural interest rate, so that short term policy rates could rise.
We plan to raise short term policy interset rates is not an effective way to communiate this policy.
We will raise the quantity of base money until nominal expenditure is back on target, interest rates can go anywhere they want, does a much better job.
Mathematical nit:
If you are targeting X, presumably you are predicting X. Let X’ be the prediction of X. In general, if X’ = F(Y,Z) is the equation used to predict X, and Y’ = G(X,Z) is used to predict Y, you cannot find G by transforming F. It is not simply a question of framing.
Bill Woolsey said: “We will raise the quantity of base money until nominal expenditure is back on target, interest rates can go anywhere they want, does a much better job.”
Could you define base money for me? There seems to be more than one definition of that.
Bill: Yes, social constructions do present themselves to us as a concrete reality. There’s whole schools of thought in sociology that devote themselves to exploring this idea, and examining how the social construction is maintained.
But I’m not sure that communicating policy in terms of M can work well enough in practice.
Min: I’m not sure which of two ideas you are referring to:
1. You can’t always invert a function
2. If y=Bx+e, then the forecast of y given x is Bx, but the forecast of x given y is not (1/B)y
Or both.
I think that 1 is more relevant in this case. It’s like if a crank is attached to a wheel. You can’t always predict which direction the crank will move when you turn the wheel clockwise. Similar case for interest rates and tightening/loosening monetary policy.
Adam: I think I had better post my little toy model, and see what you think.
Base money is currency plus reserves.
The 900 billion figure is for currency.
The base adds to that reserves. It is a bit over one trillion.
The base is made up of vault cash, which is currency, and reserve
balances at the Fed, which is deposit accounts held by banks at the Fed.
It bears (low) interest.
You can take currency held by the public and vault cash to get total
currency and add reserve balances. Or, you can add currency held by the
public to reserves.
Anyway, it is a bit less than 2 trillion.
Adam P,
You seem to regularly confuse ideas in your head or else writing on paper with the real world.
Perhaps it is just a matter of confused language.
In the real world, changes in the quantity of money have multiple impacts. Not all of them can be taken into account, but a decent understanding of the economy requires that the important ones be taken into account.
But there are no simultaneous equations that God is solving to solving in order to direct social life (or nature, for that matter.)
If there was lending, no borrowing, and no interest, does that mean there could be no money and no money prices?
Anyway, I think it is obvious that it is important to take into account the fact that when the Fed buys bonds someone else has fewer bonds. I think taht money is best understood as a financial asset and that interest rates play a key role in monetary disequilibrium.
But that doesn’t mean that interest rate targeting is the least bad option for monetary policy.
What a mess. Sorry.
Trying to finish off before I leave.
If there were no credit markets, there could be no money or prices? Of course not. So, interest rates cannot be the only possible transmission mechanism between monetary disequilibrium and the price level.
There are no simultaneous equations being solved by anyone. It is just a way of keeping track of a set of market processes that are a subset of those operating in the real world.
Anyway…
Nick,
I am no fan of targeting the quantity of money. But remember, index futures contracts don’t impact the quantity of money at all.
Great post, I left a link on my blog. I’ll try to think of something more interesting to say tomorrow.
Nick,
If you produce posts as good as this late at night, you should stay up later. Nevertheless:
“We want to loosen monetary policy. And because monetary policy is interest rate targeting, we can’t, … ”
In much of the world, monetary policy is now socially constructed to include Quantitative Variation.
“A time path of interest rates does not define a coherent monetary order”. Not quite. A time path of interest rates only defines a coherent monetary order if it is expected to do so.
I think there is a very general point here. A coherent monetary order focusses on the expected sum of MV. A coherent monetary order is therefore by definition an ordering of expectations. Sticky prices – and a high proportion of prices are observably sticky – are set for the period to come; that is to say they are set in the light of expected monetary conditions. No expectation of monetary conditions is ever set from a comprehensive and complete model – so all expectations of monetary conditions must take account of what other incompletly informed agents are likely to expect. An accepted social construct for that is inherently stabilising. Which I think is roughly where you went to bed.
If you persist in framing your view of the economy with emphasis not on where we were, nor on where we guess we are, but where we might think we are going, this simple, partial conceptual model may help to relate the ‘real’ and monetary aspects:
1. Offer and bid prices for a period are set (in general, in advance of the transaction) as a function of expected supply, expected quantities demanded and expected availability of the means to pay.
2. Quantities offered are set (in general, in advance of the transaction) as a function of expected quantities demanded and expected availability of the means to pay.
3. The expected availability of the means to pay (expected MV) is the quantity weighted sum of the expected propensities of every asset class (not just financial assets – a small change in the propensity of a large, usually illiquid asset class such as housing can mean a large change in expected MV) to be used in settlement of transactions in the forthcoming period. This is conditioned but not determined by interest rate policy.
4. Quantities demanded are set as a function of incomes, actual and expected, adjusted for willingness to use assets in the settlement of transactions.
5. The general form of expectations is expectation that past trends will continue, modified by learning from past experience and theory. E.g., if pig producers learn to expect a pig cycle, it will not exist. This is true for both old pig farmers who have learned from sad experience and for new pig producers who learned pig cycle theory at college.
Bill Woolsey said: “Base money is currency plus reserves.
The 900 billion figure is for currency.
The base adds to that reserves. It is a bit over one trillion.
The base is made up of vault cash, which is currency, and reserve
balances at the Fed, which is deposit accounts held by banks at the Fed.
It bears (low) interest.
You can take currency held by the public and vault cash to get total
currency and add reserve balances. Or, you can add currency held by the
public to reserves.
Anyway, it is a bit less than 2 trillion.”
The way I read that is this.
Currency held by the public plus vault cash (which is also currency) equals total currency.
And, vault cash (which is currency) plus reserve balances equals reserves.
I believe that makes reserve balances a type of debt. I would call it fed debt. These reserve balances used to not pay interest but now they can (but don’t have to).
Are those statements correct in those three(3) paragraphs?
If reserve balances are a type of debt, what are the “repayment” terms? Thanks to anyone who answers.
Bill Woolsey said: “We will raise the quantity of base money until nominal expenditure is back on target, interest rates can go anywhere they want, does a much better job.”
And, “Base money is currency plus reserves”
So, do you want to price inflate with currency or price inflate with reserves?
And yes, I (and I think JKH) believe that it does make a difference.
Bill Woolsey said: “If there was lending, no borrowing, and no interest, does that mean there could be no money and no money prices?”
Do you mean have ONLY currency and no currency denominated debt?
And, “If there were no credit markets, there could be no money or prices? Of course not. So, interest rates cannot be the only possible transmission mechanism between monetary disequilibrium and the price level.”
How about having the no currency denominated debt discussion with Nick? I am real interested to see it.
~sly smile~
If the fed is paying interest on reserve balances that is higher than treasuries, does that mean reserves balances (fed debt) carry a higher risk?
Milton Friedman claimed central banks only said they were targeting interest rates when they were really fiddling with rates to get the right growth in M2:
http://www.econtalk.org/archives/2009/11/sumner_on_monet.html