I have been arguing with John Cochrane and Steve Williamson over whether central banks announcing higher nominal interest rates is inflationary or deflationary. The very fact that economists are arguing about that very basic question tells us something important about central banks' using nominal interest rates as a communications strategy: it sucks. This is a point that economists like Scott Sumner and I have been making for some time. Do low nominal interest rates mean monetary policy is loose or tight? It depends.
(And if the answer to that question is "it depends", it also means that the Zero Lower Bound on nominal interest rates does not mean that central banks cannot loosen monetary policy.)
Suppose the central bank suddenly announces a new time-path for the monetary base. Or for the money supply, or the nominal exchange rate, or the price of gold, or something else with $ in the units. The new announced time-path is different from the one people had previously expected.
[Update: just to keep it simple, I'm assuming this announcement is not a response to some shock. Or, if it is a response to some shock, people already know about that shock, so we can separate out the effects of the shock and the effect of the announcement. The contents of the announcement are news.]
Consider the set of all such possible announcements.
Let us divide that set into those that cause an immediate increase in the nominal interest rate, and those that cause an immediate decrease in the nominal interest rate. (Strictly speaking there will be some members of the set that have exactly zero effect on nominal interest rates, so you could add a third category, if you insist, but let's ignore those borderline cases, to make my life easier.)
Let us also divide that set into those that cause an immediate increase in the inflation rate, and those that cause an immediate decrease in the inflation rate. (Strictly speaking there will be some members of the set that have exactly zero effect on the inflation rate, so you could add a third category, if you insist, but let's ignore those borderline cases, to make my life easier.)
So we have partitioned the set of all announcements into four subsets.
How exactly you allocate members of the total set into those four subsets will depend on your macroeconomic model.
But ask yourself this question: can you say that two of those sets are empty sets? I don't think you can.
Let's start with an extreme macroeconomic model, in which prices are perfectly flexible, and in which money is always neutral and super-neutral. In that model, monetary policy has no effect on real interest rates, so inflation rates and nominal interest rates must always respond in the same direction. That seems to imply that the North-East and South-West sets are empty sets.
But it's not so obvious. Because some monetary policy announcements will cause a downward jump in the price level and an upward jump in the nominal interest rate. For example: "We will cut M in half immediately but will double it again 10 years from now." If by "inflation rate" we mean "percentage change in the current price level relative to just before the announcement" this is an example where the inflation rate falls and nominal interest rate rises. It is only after the announcement that the (forward-looking) inflation rate increases.
Now let's consider a different macroeconomic model, with sticky prices, where monetary policy is not neutral. It is tempting to say that, with such a model, any monetary policy announcement that caused an immediate increase in nominal interest rates would cause a recession and a fall in inflation, so the North-West and South-East sets are empty sets.
But that isn't right either. For example, suppose the announcement is: "M will stay the same today, but will double in 10 years time". Even if the price level is sticky, so cannot jump immediately on the news, people know that the price level will eventually double, so expected inflation, and hence actual inflation, can increase immediately, and this can cause the nominal interest rate to rise immediately.
Suppose you were an economist who knew that the central bank had made an announcement about monetary policy, but you did not hear that announcement, and the only thing you did know was that the announcement had caused nominal interest rates to rise.
Simply knowing that the central bank's announcement caused nominal interest rates to rise does not give you sufficient information to tell you whether inflation (or the price level) will rise or fall.
If the central bank announces a higher nominal interest rate, that does not give you sufficient information to tell you whether inflation (or the price level) will rise or fall.
Central banks' announcing targets for the nominal interest rate is a very bad communications strategy for monetary policy. We do not know what they mean.
2x2 matrices - exploding myths successfully since 1970!
Great post.
Posted by: Akshay | December 22, 2013 at 10:19 AM
I’m not an economist but two things in this post caught my eye: “it depends” and “we do not know what they mean”. When non-economists listen in on conversations between academic economists, two of the most common reactions are: “surely, it depends” and “we do not know what they mean”.
Here is an example. Suppose that one of your representative agents has some savings and suppose that the rate of inflation increases. What will happen to the representative agent’s spending? The non-economist’s answer is ‘it depends’. Here are some scenarios.
If the agent sees his savings being eaten away by inflation, he may decide to spend his savings before they lose most of their value. His spending will increase and his savings decrease.
If the agent is saving for something specific which he can’t currently afford, inflation will mean that he has to save more to be able to make his purchase sometime in the future. His spending will decrease and his savings increase.
If the agent is saving for a house or another asset which is not included in the rate of inflation, he will carry on as before as a change in the rate of inflation has no impact. His spending and savings will stay the same.
The probability of these scenarios is also likely to change depending on a number of factors about the agent’s circumstances such as the total amount of savings involved, the age of the agent and the likelihood of being able to replace the savings in the future.
Economists talk about representative agents in such abstract and ambiguous ways that it is never clear how their agents would react to this type of situation. Nevertheless, economists seem to believe that they can plug a specific behaviour into their models. As a result, "we do not know what they mean" and "surely it depends". This ambiguity also means that economists are never accountable for what they say or what they predict.
You are discussing an example of this where you are the one who is confused, so I’m glad that you understand the problem, but I disagree on its scope.
In your example, the problem could be alleviated if central banks were much more precise in their statements e.g. a central bank might make a decision and state that it expected a certain measurable outcome in a certain number of months time. The bank would then be accountable as the rest of us could see whether the central bank achieved its expected result or not. Economists could argue about the bank’s expected results and success rate, and make their own alternative predictions which could also be judged against the measurable outcome. However, that would almost be science rather than economics, albeit with uncontrolled experiments, so I don’t hold out much hope.
In summary, it’s not just a central bank problem, and it’s not just a communication problem.
Posted by: Jamie | December 22, 2013 at 10:41 AM
Thanks Akshay! BTW, I found your comment caught in the spam filter, so if your comments don't immediately appear, that's what's probably happened.
Posted by: Nick Rowe | December 22, 2013 at 10:45 AM
Jamie: for every creditor there's a debtor. If one gets hurt from a change in the price level the other gains. This does not necessarily mean the effects cancel out in aggregate, but it does suggest we should focus on substitution effects rather than income/wealth effects. The proverbial "representative agent" is neither a creditor nor a debtor.
Posted by: Nick Rowe | December 22, 2013 at 10:52 AM
Jamie: "In your example, the problem could be alleviated if central banks were much more precise in their statements e.g. a central bank might make a decision and state that it expected a certain measurable outcome in a certain number of months time. The bank would then be accountable as the rest of us could see whether the central bank achieved its expected result or not. Economists could argue about the bank’s expected results and success rate, and make their own alternative predictions which could also be judged against the measurable outcome. However, that would almost be science rather than economics, albeit with uncontrolled experiments, so I don’t hold out much hope."
That is exactly what the Bank of Canada does. It's called "inflation targeting". And I'm one of a group of 12 economists who make alternative recommendations. And (given a long enough data set) we can be judged against each other.
Posted by: Nick Rowe | December 22, 2013 at 11:04 AM
Nick Rowe: "for every creditor there's a debtor. If one gets hurt from a change in the price level the other gains. This does not necessarily mean the effects cancel out in aggregate, but it does suggest we should focus on substitution effects rather than income/wealth effects. The proverbial "representative agent" is neither a creditor nor a debtor"
Thanks for the replies. Yes, I understand that for every creditor there is a debtor. However, when I save or dis-save the creditor/debtor relationship is between me and the bank. If I spend $10 in the economy then someone else gets income of $10. However, if I save the $10 then the default situation is surely that the bank has the $10 and the money is not spent in the economy. You seem to be assuming that the bank will always be able to find someone else who will borrow this $10 (and no more and no less than $10) and spend it. My limited understanding of the zero lower bound is that there aren't enough people willing to borrow the $10 even at 0% interest rates so the money just sits with the bank. Hence, the agent's decision to spend or save is important.
I don't want to take up any more of your time in teaching me basic economic jargon. My key point is that economists don't speak a language that is accessible to non-economists, and if we don't understand you then we are unlikely to vote for the policies you recommend even if they are the correct ones. Note that your reply just made me say again that "surely it depends". When economists also write blog posts saying that they don't understand the language used by other economists, it's not clear how you expect the rest of us to understand.
Regarding your second reply to me on inflation targeting and your role in making alternative recommendations, I am now confused on what you are saying in your initial post. We must be talking past each other. I'll read the post again and see if I can understand it based on your reply. Thanks again for your time.
Posted by: Jamie | December 22, 2013 at 12:34 PM
Nick
I'm not sure that the distinction here is between rate and quantity announcements, so much as short-term and long-term guidance. The differences in your two examples both rely on the (believable) announcement about what is going to happen in the future. You finish by stating that an announcement on current interest rates tells us nothing. But how much would it tell us if the central bank announced today's money supply, but gave no indication over their plans for tomorrow? On the other hand, if they were to lay out the course of the nominal rate for the indefinite future, would that not contain as much information as if they were to a similar quantity announcement?
Posted by: Nick Edmonds | December 22, 2013 at 12:57 PM
Nick Edmonds: suppose the announcement is that the money supply will permanently change by x% compared to what had previously been expected. If, for example, we take the perfectly flexible prices model, that announcement will have no effect on nominal interest rates, either now or in the future, regardless of x. But the price level will permanently change by x%. And x could be either positive or negative. So knowing the course of the nominal interest rate for the indefinite future tells us nothing about whether the price level will rise or fall.
If you tell me M for the indefinite future, I can figure out prices and nominal interest rates. If you tell me nominal interest rates for the indefinite future, I can't figure out the reverse.
Posted by: Nick Rowe | December 22, 2013 at 01:22 PM
And, if instead we have a model in which inflation is sticky, if you tell me nominal interest rates for the indefinite future, I will almost certainly tell you that the economy will eventually either explode into hyperinflation or implode into hyperdeflation. So I will not believe your announcement.
Posted by: Nick Rowe | December 22, 2013 at 01:32 PM
I must confess I'm not sure I know exactly what's happening in that money-neutral scenario. If all nominal interest rates stay the same, doesn't that imply that nominal bond prices (including long-dated and perpetuals) stay the same. In which case, if the general price level has increased by x%, hasn't the real stock of bonds fallen by x%? Is that really money neutral? Maybe that's OT, but it makes it very hard to interpret the scenarios you're describing.
Posted by: Nick Edmonds | December 22, 2013 at 01:33 PM
Nick Edmonds: Yes, the real stock of bonds will be lower. And if that has aggregate real effects (it may) then money won't be neutral. I'm ignoring that for simplicity.
Posted by: Nick Rowe | December 22, 2013 at 01:42 PM
OK - but I suspect that may be relevant to the difference between the way I conceptualise this (as a rates person) and the way you do.
Posted by: Nick Edmonds | December 22, 2013 at 01:53 PM
Nick, what's your view on how central banks do or should look at market expectations, in terms of how markets perceive the path of rates/inflation/demand. Specifically:
- government yield curve, TIPs
- fed funds futures
- credit spreads, CDS, etc.
- REITs, housing- and automobile-sensitive (perhaps consumer durables in general) (equity) stock value
I'm asking a slightly different question to your post. Not only how the CB communicates it's actions, but how the market perceives them is just as important? Should CB's have more tools to measure market expectations, or perhaps you think NGDP is enough?
Posted by: jt | December 22, 2013 at 03:19 PM
What's really fascinating here is that we went through a learning period with money aggregates that taught us that variation in velocity was a problem in practice. Gradually the focus moved to interest rates but really for the same reasons: instability of the real rate and inflation expectations we didn't really solve anything.
And as Scott says, we ended up with the ZLB issue.
We did somewhat better by focusing on IT rather than rate setting or money growth as the market accepts that CB will learn the shocks in r and v...
So what's the next step? Stop announcing policy in terms of the interest rate and money growth. Announce your IT target or NGDP target and just keep saying the CB commits to its target and believes with appropriate execution will meet the target.
Posted by: Jon | December 22, 2013 at 03:22 PM
jt: ideally, if the central bank is targeting X(t+h), where h is the horizon, it would like to have direct data on market expectations on X(t+h). I know the Bank of Canada watches surveys of expected inflation, and the spread between nominal and indexed bonds, but perhaps it pays less attention to asset market data than perhaps it should. Dunno. That question needs a whole post.
Jon: love your first paragraph!
On your final paragraph: yes, but there has to be some sort of feedback rule from expectations about the target variable into something the central bank actually does. I keep thinking about Irving Fisher's compensated dollar plan. Adjust the price of gold up and down to keep the price level (or it could be future NGDP) on target. At least we know that raising the price of gold means loosening monetary policy, and there is no ZLB problem. It doesn't have to be gold. Just some real asset (as opposed to a nominal asset like bonds) the CB can buy and sell.
Posted by: Nick Rowe | December 22, 2013 at 05:30 PM
Nick: does the CB need to announce the policy instrument? does the CB need to announce how it observes the inflation-rate? (I have in mind in particular that a market measure such is the TIPS spread could never be identified by the Fed as driving policy because the market is so small it would be pushed around by participants looking to push the Fed around on policy.)
Does piercing the mystery beyond a target based on a very large aggregate (CPI, NGDP, etc) add value?
Keep in mind that the effective fed funds rate or the h41 report would still be visible.
I'm definitely not saying they only need to be a conductor and not have an instrument; but is it meaningful for the market to know which one they are playing?
Posted by: Jon | December 22, 2013 at 06:24 PM
Does a high money supply mean that monetary policy is loose or tight?
Posted by: Nemi | December 23, 2013 at 05:41 AM
Jon: Dunno. If the inflation (or NGDP or whatever) target is fully credible, it might not matter whether anyone knows what instrument the CB is using. Good question.
Nemi: it depends on whether it's high relative to the demand for money. But an announced permanent increase in the supply of money, that is not a response to some shock to money demand, is loose money.
Posted by: Nick Rowe | December 23, 2013 at 07:28 AM
I will try to give a easily visualized example of a north-east quadrant situation:
Let us assume that central bank announces that all future borrowing for government deficits will be from private ownership of capital. This should drive up interest rates.
This announcement should also drive down inflation because the money supply should stabilize.
Having said that, it must be pointed out that if the deficit of government is the result of a NEW tax credit for new car purchase, then the deficit of government is effectively a sales promotion for the auto industry. This could easily result in inflation of prices related to the auto industry and an average increase in inflation rate. This couple of price increase and interest increase pushes the example into the north-west quadrant.
This example flows from the MMT concept that new money supply enters the economy through government spending.
Posted by: Roger Sparks | December 23, 2013 at 10:22 AM
Following Roger Sparks' example,
A minus 2% real interest rates is about negative 22% return on shorter term risk free funds like reserves that the Fed is forcing the private sector to hold. This is if rates lift off in 2017 like Janet wants. If I told you that 3-4 trillion of private sector capital had a -20% real return - what if the stock mkt had a -20% real return in 2014? The Fed governors would be falling over themselves to prevent it and NO ONE would argue that it was not deflationary. There you have it - lower rates = deflation in this case.
Posted by: ReturnFreeRisk | December 23, 2013 at 11:05 AM
Nick,
Our exchange above prompted me to write the following: http://monetaryreflections.blogspot.co.uk/2013/12/determinancy-of-rates-and-quantities.html
Posted by: Nick Edmonds | December 23, 2013 at 11:39 AM
Sumner says that it doesn't much matter: using either interest rates or the monetary base both "suck" as communication strategies:
http://www.themoneyillusion.com/?p=25642#comments
Posted by: Tom Brown | December 23, 2013 at 12:34 PM
Central banks' announcing targets for the nominal interest rate is a very bad communications strategy for monetary policy. We do not know what they mean. Does any central bank do that? Doesn't an inflation targeting central bank target inflation and use cuts or increases in a particular interest rate as a signal? That is why a change in interest rates by a central bank who is only targeting inflation means something different from a central bank that also has an implicit income target.
All central banks operate within a framework of expectations, the question is which framework. So, simple inflation targeting is like the gold standard--the central bank anchors expectations for the value of the currency but provides no anchor for income expectations.
To put it another way; the problem with the ECB or the BoJ (pre Abe) is not that we don't know what they mean but we know all too well what they mean (and don't mean).
Posted by: Lorenzo from Oz | December 23, 2013 at 08:07 PM
Lorenzo: provided we stay off the ZLB, there seems to be little harm in using nominal interest rates if the central bank has a very clear inflation (or, presumably, NGDP) target. We then know how to interpret them. But nominal interest rates alone, without an inflation target (or similar), are the bad signals.
Posted by: Nick Rowe | December 23, 2013 at 08:38 PM
" But an announced permanent increase in the supply of money, that is not a response to some shock to money demand, is loose money."
But that is (even more) true w.r.t. a permanent decrease in the interest rate as well.
Posted by: nemi | December 24, 2013 at 03:30 PM
Nick: right, that makes perfect sense. I was afraid I was missing something obvious.
Posted by: Lorenzo from Oz | December 26, 2013 at 03:55 AM
nemi: if you believe in perfectly flexible prices levels and inflation rates, the only non-explosive or non-implosive equilibrium compatible with a permanent decrease in nominal interest rates is one where the inflation rate falls.
If you do not believe in perfectly flexible inflation rates, the only equilibrium is an explosive one. So an announced permanent decrease in nominal interest rates is not credible, unless you believe the central bank wants to follow Zimbabwe.
Lorenzo: there is probably something non-obvious we are missing.
Posted by: Nick Rowe | December 26, 2013 at 07:10 AM