Further thoughts on NeoFisherism

David Beckworth recently interviewed Stephen Williamson, who is an advocate of the NeoFisherian approach to thinking about monetary policy and interest rates.  Williamson argues that a policy of permanently reducing interest rates is disinflationary.  Others think this idea is crazy.  I’m not in either camp, and I keep looking for ways to explain why.  Here are some facts about monetary policy, which seem related to this issue:

1. In the short run, nominal interest rates can be reduced with a tight money policy.

2. In the short run, nominal interest rates are usually reduced with an easy money policy.

3. Because money is neutral in the long run, any monetary policy that permanently reduces nominal interest rates must be disinflationary.

4. A tight money policy reduces the natural rate of interest.

All of these claims are pretty easy to justify, and none seem particularly controversial.  But they raise an interesting puzzle.  Points #1, #3 and #4 all seem sort of NeoFisherian in spirit, consistent with the claims made by Stephen Williamson and John Cochrane.  So why are so many mainstream economists horrified by NeoFisherism?  I think the sticking point is #2.

The vast majority of the time, a reduction in interest rates on any given day represents an easier monetary policy than a counterfactual policy where the central bank doesn’t reduce interest rates.  Not always (in which case #1 and #3 would no longer be true), but the vast majority of the time.  But the NeoFisherian thought experiment requires that the lower rates be achieved via tighter monetary policy.

I think that people are confused about what NeoFisherians are talking about when they discuss policy option number three.  In the minds of most economists, switching to a permanently lower interest rate seems like an expansionary monetary policy, because on any given day cutting interest rates usually is an expansionary monetary policy.  Here’s why they are wrong:

1. If you don’t want the price level to blow up, then any permanent switch to a lower interest rate must be done with a tighter monetary policy.  If the central bank tried to do it with an easier money policy then they’d have to inject larger and larger amounts of liquidity, eventually causing hyperinflation and then complete collapse of the system.  So any sustainable policy of low interest rates must be contractionary.

2.  A contractionary monetary policy lowers the natural rate of interest.  I think many economists picture a world where the natural rate of interest is not affected by monetary policy.  In that world, lowering the policy rate makes policy more expansionary, because the stance of monetary policy is the gap between the policy rate and the natural rate (assumed to be stable).  In fact, any sustainable policy of low rates must be caused by tight money, and any tight money policy will reduce the natural rate of interest so much that monetary policy does not get easier, despite the lower fed funds target.  This is Japan since 1995.

So far I’ve presented a picture that is somewhat sympathetic to the NeoFisherians.  Let me conclude with a discussion of what I don’t like about the way NeoFisherians present their theory.

1.  The listener is led to believe that if you want lower inflation, you need to cut interest rates.  I’d say if you want lower inflation you need to cut interest rates via a tight money policy.  Any attempt to achieve lower inflation via a cut in interest rates achieved through an easier money policy will end in disaster.

2.  Because the vast majority of rate cuts represent easier money than the counterfactual of not cutting rates on that given day, it is not accurate to imply that the first step to lowering inflation is for the central bank to do the sort of rate cut that it often does do–i.e., liquidity injections.  Instead, the NeoFisherians should argue that the first step to lower inflation is for central banks to do the sort of rate cut that the Swiss National Bank did in January 2015, when they simultaneously appreciated their currency and created a credible policy of further currency appreciation going forward.  That credible promise led to lower nominal interest rates via the interest parity condition, and lower inflation expectations via the currency appreciation (combined with PPP.)

PS.  Has anyone commented on the similarity between the NeoFisherian puzzle identified in points #1 – #4 above, and the puzzle that led to the Dornbusch overshooting model?  The overshooting model was an attempt to resolved the following puzzle in a conventional Keynesian fashion:

Puzzle:  Easy money seems to lead to both actual currency depreciation and expected currency appreciation.

Rudi Dornbusch wanted to show how easier money could lead to expected currency appreciation (which is an implication of lower nominal interest rates combined with the interest parity condition.)  His solution was overshooting.

The NeoFisherian model assumes a permanent change in the interest rate, which rules out Dornbusch’s resolution to this puzzle. If you make the rate cut permanent than his solution no longer works; you take overshooting off the table.  In that case, the NeoFisherian result is the only explanation left standing.  Now it is a tighter money policy that reduces interest rates, and that tighter money also makes the currency become expected to appreciate forever, lowering inflation.


Tags:

 
 
 

20 Responses to “Further thoughts on NeoFisherism”

  1. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    20. December 2017 at 08:50

    For the umpteenth time; there’s no puzzle to explain if you stop talking about monetary policy in terms of interest rates. Doing so only confuses the issue. As Milton Friedman well understood.

  2. Gravatar of Ege Erdil Ege Erdil
    20. December 2017 at 09:05

    I think it’s bad to interpret neo-Fisherism (or, indeed, monetary policy in general) through a lens of “what happens if the central bank raises/cuts rates?”. In the models we work with, policy rules and off-equilibrium actions matter, i.e what the central bank would do if X happened matters even if X doesn’t actually happen. For instance, in the three-equation NK model that Cochrane works with, a policy rule which makes the system stable (in the traditional sense, two negative eigenvalues) requires the central bank to respond to large positive output gaps with lower policy rates (so, in a recession you *raise* rates), but also respond more than one-for-one to inflation above target levels a la Taylor.

    That said, I don’t think the neo-Fisherian argument is just about whether low policy rates are achieved through easy or tight monetary policy, it’s about how models where the IS curve (any relation establishing some connection between the interest rate and output) is not forward-looking tend to behave differently than ones where it is forward-looking. For instance, Krugman’s textbook model in http://web.mit.edu/krugman/www/spiral.html has the property that if the monetary authority permamently reduces the money growth rate, nominal interest rates initially go up in response and then fall, while New Keynesian models with a forward-looking IS relation don’t have this property: reducing the money growth rate lowers nominal interest rates, both in the short run and in the long run. I think the interesting theoretical debate is about which kind of IS relation we should be using to model the effects of monetary policy. Some authors working with NK models (this includes Krugman) turn the New Keynesian IS curve into an “old Keynesian” IS curve by holding output at some future date fixed and working back from that, but this sweeps the different implications of the two frameworks under the rug.

    Another interesting point is about how we think of stability of these systems, or “determinacy” in New Keynesian literature, for instance. When Friedman predicted that the monetary authority could not peg nominal interest rates without the economy exploding, he was using the standard way of thinking about stability: positive eigenvalues are unstable, negative eigenvalues are stable. However, the entire New Keynesian literature does the opposite: they introduce a Taylor rule which produces positive eigenvalues all over the place, and then say that the economy can’t explode by assumption, so the Taylor rule stabilizes inflation at the monetary authority’s target (since that’s the only locally bounded equilibrium). This strikes me as a kind of double standard that’s not often elaborated on. This is how Williamson gets neo-Fisherian results out of basically every mainstream model: the positive eigenvalues explode, so he rules those equilibria out as being impossible paths for an actual economy to follow, and all that remains are the equilibria where a nominal interest rate peg is stable. I think most economists working with New Keynesian models actually don’t realize what they are doing with the Taylor rules in those models so what Williamson is doing strikes them as peculiar, even though it’s essentially isomorphic to what they are doing themselves. This Krugman post (https://krugman.blogs.nytimes.com/2013/11/29/on-the-importance-of-little-arrows-wonkish/) is a very good example of this phenomenon: Krugman himself uses this method of obtaining determinacy in his Debt, Deleveraging and the Liquidity Trap paper with Eggerston, but my guess is he isn’t aware of what I pointed out.

    tl;dr: I’m not so sure that “mainstream macro models” as Williamson would describe them actually say that #2 is possible at all, at least not unless we resolve the stability issue I raise in the last paragraph, and perhaps not even then.

  3. Gravatar of jj jj
    20. December 2017 at 12:25

    Could one summarize with “Never reason from an interest rate change”?

  4. Gravatar of Jared Jared
    20. December 2017 at 12:58

    Could you define “tight monetary policy”? What specific actions would a central bank take in the case of lowering interest rates but running a tight monetay policy? I take it you mean something other than the CB signaling future rate increases.

  5. Gravatar of B Cole B Cole
    20. December 2017 at 16:30

    I wonder if monetary policy is neutral in the long run.

    The consensus seems to be that tight money led to 20 years of very low and no growth in Japan, barely concluding in the present day.

    What if the Bank of Japan had not decided, in very recent times. to become more pro-growth?

    Perhaps Japan would molder for another several decades.

    How long is the long run? Do you have a couple lifetimes to spare?

  6. Gravatar of dtoh dtoh
    20. December 2017 at 16:42

    Scott,
    Two quick points.

    1. I think interest rate movements can be confusing and/or contradictory because looking at them in isolation ignores the market’s expecations of Fed competence.

    2. Looking at interest rates is a mistake to start with. All of this discussion would go away if the stance of monetary policy was evaluated in terms of changes in the rate of Fed asset purchases from the non-banking sector (or, in your transmission mechanism worldview, changes in the amount of money the Fed is injecting into the non-banking sector.)

  7. Gravatar of Jon Jon
    20. December 2017 at 18:24

    Of course it is very easy to tighten money while lowering interest-rates. That describes most recessions.

  8. Gravatar of Benjamin Cole Benjamin Cole
    20. December 2017 at 19:10

    OT but funny, and btw no one really knows what the tax bill will do.

    Sen. Bob Corker defended himself against charges of self-dealing on the tax bill by saying he hadn’t even read it.

    I do not blame Corker; the tax bill is a couple thousand pages, and the federal tax code 75,000 pages.

    I have not read the tax bill either.

    Has anyone?

  9. Gravatar of Tf Tf
    20. December 2017 at 21:29

    The problem is economists keep using the phrase “interest rates” as if it is meaningful. The fed funds rate could not be more different from the 30 yr treasury yield, which couldn’t be more different from the 2 year treasury yield. It is analogous to Biologists using the word animal for every animal. Animals love kibble, just look at my dog! Re read or re write your post being specific which rate you are referring to. It will become much simpler. Different rates move in opposite directions for the same reason! Then teach your field and the financial press to do the same. Sorry, I am a bond trader, a few days trading a yield curve and you will laugh when economists and the press use the term “interest rates” generically.

  10. Gravatar of ssumner ssumner
    20. December 2017 at 22:16

    Ege, In my view the indeterminacy problem could be reduced if we stopped talking about monetary policy in terms of interest rates, and we switched to other less ambiguous indicators such as NGDP futures prices, CPI futures prices, foreign exchange prices, gold prices, etc.

    Thus when I say that interest rates can be reduced in the short run with either an easy or a tight monetary policy, I mean that there are monetary actions that simultaneously reduce interest rates and depreciate exchange rates, and other equally plausible monetary actions that simultaneously reduce interest rates and appreciate exchange rates.

    As long as we keep talking about monetary policy in terms on interest rates we’ll never fix this problem. We need an unambiguous way of measuring the stance of monetary policy, and then we can look at how interest rates respond to various changes in those metrics. But to start the conversation with an interest rate change is a huge mistake. It would be like talking about the effect of a change in the dollar price of yen, or the dollar price of gold, without first indicating whether that price change was caused by a shift in supply or demand.

    jj, Yes.

    Jared, A central bank could promise to gradual appreciate its currency against the dollar. That would be a tight money policy that lowers nominal interest rates.

    dtoh, I don’t think asset purchases are a good indicator. I prefer the price of NGDP futures contracts.

  11. Gravatar of Rajat Rajat
    21. December 2017 at 01:46

    Based purely on that interview, I was and am left completely none-the-wiser about what Neo-Fisherists see as the transmission mechanism or concrete steppes from lower interest rates to lower inflation. Williamson said nothing about targeting or making any promises re the exchange rate. I was left with the impression that Neo-Fisherism is nothing more than looking at some data and confusing correlation with causation. As Colonel Jessup might say, “Is that all you have?” A possible clue is (if I recall correctly) that Williamson said he came to economics from mathematics.

  12. Gravatar of Rajat Rajat
    21. December 2017 at 03:10

    Also, Scott, I think you are being way too charitable to the Neo-Fisherists by suggesting – if you are – that they may be referring to lowering nominal interest rates with a tight money policy. Let’s face it – for better-or-worse, apart from you, Friedman and a handful of others, virtually no-one describes monetary policy that way. And even if that’s what they have in mind, such a policy approach would be (according to my understanding) incredibly inexact and dysfunctional – even bloody-minded. It relies on either: (i) waiting for an exogeneous shock (eg a sub-prime or sovereign debt crisis) to increase money demand and push down the Wicksellian rate and then deliberately cutting the FF rate by less than needed, or (ii) talking the economy (and natural rate) down by saying that the central bank won’t cut rates when needed – kind of like a malevolent form of forward guidance. Sure, central banks may do these things by mistake, but it’s hardly a viable policy framework is it?

    My conclusion is they are crazy. I am hearing and reading nothing that redeems them, but happy to be corrected.

  13. Gravatar of Ege Erdil Ege Erdil
    21. December 2017 at 04:46

    Scott,

    I agree that thinking of monetary policy in terms of interest rates alone is misleading, for essentially the same reasons. Neo-Fisherian models don’t have the property that tight money policy leads to high inflation, which I think is what many people are inferring by looking at the claims being made by Williamson and Cochrane. For instance, Cochrane works with the frictionless fiscal theory equation, in which case increasing IOER increases the growth of nominal government liabilities outstanding, so it’s an “easy money” policy. Indeed, adding some kind of nominal anchor (like a cash-in-advance constraint or the government debt valuation equation) to these models resolves the indeterminacy problems.

    Neo-Fisherism has some other aspects, though. For instance, if we’re in frictionless FTPL land, then it says that if the monetary authority is fixing short term nominal interest rates, they can go out and buy all the mortgage-backed securities they want and it may have no effect on inflation or NGDP. The usual comparison with Zimbabwe is misleading because Zimbabwe had an insolvent government. The most reliable way that the monetary authority can ensure higher inflation is to raise IOER in this case, for obvious reasons: high IOER means high money growth. If the monetary authority instead fixes IOER at 50 bps and starts buying private assets and expanding total government liabilities, then it has to convince investors that the expansion is permanent, and this is very tricky because the expansion can be undone by a fiscal policy change as much as it can be by a monetary policy change. (How reliable are the Treasury’s five year deficit forecasts?)

    In short, if we want to keep the monetary authority and the fiscal authority as separate entities, then what the monetary authority with an IOER or an ON-RRP facility should do is to raise these policy rates sharply in a slump. This way, the monetary authority ensures high money (or, in this case, government liability) growth, which raises nominal incomes and cuts the recession short. If the monetary authority tries lowering rates and hits the ZLB, then they will be following tight money policy even if they don’t realize it.

  14. Gravatar of dtoh dtoh
    21. December 2017 at 05:42

    Scott,
    I think we were talking about evaluating the stance of Fed policy, i.e. whether it’expansive or contractionary.

    You said,
    “I don’t think asset purchases are a good indicator. I prefer the price of NGDP futures contracts.”

    NGDP futures contracts would tell you whether policy is on target but it would not necessarily tell you the stance of policy. If the Fed were perfectly competent, you could tell policy was contractionary if and whenever futures were above target and vice versa. But if the Fed were not perfectly competent, futures would not necessarily tell you anything about stance. You could be under target and the Fed could still be pursuing contractionary policy.

    So in sum i) futures are a good indicator or whether or not policy is on target, and ii) changes in assets purchase from (or money injected into) the non-banking sector are a good indicator of stance.

  15. Gravatar of ssumner ssumner
    21. December 2017 at 09:19

    Rajat, I can’t agree, I think the NeoFisherians are claiming that a persistently low interest rate policy is a tight money policy. The models really do support that claim.
    But I agree that Williamson is not doing a good job explaining the theory in a way that will convert skeptics.

    If they would spend more time talking about cases like Switzerland circa January 2015, they’d have more success.

    Ege, You said:

    “Neo-Fisherian models don’t have the property that tight money policy leads to high inflation, which I think is what many people are inferring by looking at the claims being made by Williamson and Cochrane.”

    No, that’s missing the point. They don’t claim that tight money causes inflation, they claim that higher interest rates are EASY MONEY. At least that’s been my assumption. Do you think they’d call hyperinflation “tight money”

    I’m not a fan of the FTPL, which I don’t think applies to countries such as the US.

    dtoh, I don’t agree. Any excessively contractionary or expansionary policy is a sign of incompetence. A neutral stance means hitting your target.

  16. Gravatar of Ege Erdil Ege Erdil
    21. December 2017 at 10:23

    Scott,

    Yes, that’s precisely their claim: they are saying that easy money leads to higher nominal interest rates, in the short run and in the long run. Cochrane works with a model with money in the utility function in a paper, and he finds with some specification of parameters that to produce an initial decline in nominal interest rates from easy money you need a real money supply that’s about four times annual real GDP. He dismisses this scenario as unrealistic, though of course his model is not any kind of serious econometric work.

    If you’re not a fan of the FTPL, then consider some model where you have the usual MV = PY where V is an increasing function of the interest rate spread between money and what you can get elsewhere in the economy. If you add this to a New Keynesian perfect foresight framework, log-linearize and mix, you get a three-dimensional system with one negative eigenvalue in (nominal interest rate, output gap, inflation) space, and the saddle path is in the (+, +, +) direction. Of course, in this model we no longer have the FTPL concerns, so the monetary authority can increase the money supply to deal with a money demand shock without any constraints, which is the optimal policy for dealing with these transient shocks in the model.

    On the other hand, the model is still neo-Fisherian. It follows from the direction of the saddle path that inflation and nominal interest rates *in the broader economy* (not IOER) always rise and fall together, so an unexpected and permanent expansion of the money supply to deal with a money demand shock produces high NGDP and high nominal interest rates at the same time. Raising the IOER without raising money growth, however, only leads to low inflation and low nominal interest rates. (In the FTPL model, it was impossible for the central bank to raise IOER and not raise government liability growth, so we didn’t have this problem.) “Raise interest rates to raise inflation” doesn’t work if you’re raising the IOER and not using easy money policy to generate expectations of inflation.

    The crucial thing leading to these results is really the intertemporal substitution relation, i.e the IS curve looking like dx/dt = k(r – R) instead of x = k(R – r) where x is the output gap, k is some constant, r is the real interest rate and R is the natural interest rate. In my experience, the neo-Fisherian models are precisely the ones with “dx/dt” and not “x” in the IS curve. The FTPL is not essential to producing most of the neo-Fisherian results, they come out of MV = PY just as readily.

  17. Gravatar of Ege Erdil Ege Erdil
    22. December 2017 at 06:47

    This is an addendum to the previous comment: one way to think about the immediate impact of monetary expansion or money demand shocks in the above model is to project the 3D space onto the pi = m plane, where m is the constant money growth rate. Then, what you get is essentially a perverse kind of IS-LM: the LM curve is upward sloping and it’s given by MV = PY with constant P (prices are sticky on impact), where the “IS curve” is *also* upward sloping and is given by the saddle path projected onto x-i space. If you work it out, you find that the LM curve is steeper than the “IS curve” (saddle path), so monetary expansion (shifting LM to the right) raises both nominal interest rates and output.

    The usual way monetarists tend to think about monetary expansion is directly from MV = PY: with sticky prices, some of high M leads to high Y on impact, and some of it is “absorbed” by lower V, i.e lower nominal interest rates. The MV = PY relation is there in the model above, but instead it predicts expanding the money supply actually increases V (because it generates expected inflation). This is essentially why the model is neo-Fisherian.

  18. Gravatar of ssumner ssumner
    22. December 2017 at 08:10

    Ege, You said:

    “Yes, that’s precisely their claim: they are saying that easy money leads to higher nominal interest rates, in the short run and in the long run.”

    Sorry, I misread your earlier comment. My mistake.

    I think it’s a mistake to focus too much on these sort of models, especially IS/LM, or indeed any New Keynesian model. You can write down models to get any result you want. I prefer to start from empirical data. We observe that some expansionary monetary shocks raise nominal rates and other expansionary monetary shocks reduce nominal rates (in the short run.) So right off the bat we know that any models that predicts rates always move in one direction after monetary shocks is wrong.

    The next step is to figure out why some monetary shocks have one effect and other monetary shocks have the opposite effect. We know that if you evaluate monetary shocks in terms of the “price effect” it is easy to explain why rates rise on some cases and fall in others, but if you evaluate monetary policy in terms of the interest rate effect it gets a lot more confusing, as there are lots of indeterminacy problems.

    BTW, since MV=PY is a tautology, any true model must be able to be explained in MV=PY terms.

  19. Gravatar of Ege Erdil Ege Erdil
    22. December 2017 at 10:27

    Scott,

    “I think it’s a mistake to focus too much on these sort of models, especially IS/LM, or indeed any New Keynesian model.”

    I would draw a sharp distinction between those models. Krugman is wrong when he says that you can think in terms of IS-LM and use a Woodford style NK model to “make that intuition precise”. They are very different animals. That said, I agree neither of these models should be taken too seriously.

    “I prefer to start from empirical data. We observe that some expansionary monetary shocks raise nominal rates and other expansionary monetary shocks reduce nominal rates (in the short run.) So right off the bat we know that any models that predicts rates always move in one direction after monetary shocks is wrong.”

    I don’t think the model makes that prediction, though. If you have a money demand shock that the monetary authority does not respond to by adequate monetary expansion, then the money supply could be going up while you see low output, low interest rates and low inflation, i.e the price effect is present in the model. Of course, they would still be higher than they would be in the counterfactual case where the monetary authority doesn’t respond at all to the shock.

    When I talk about the model being neo-Fisherian, I am referring to a scenario in which there’s no money demand shock, but the monetary authority permanently raises the money supply. It’s hard to glean from empirical data what happens if the monetary authority does this, because in the real world the monetary authority will often be responding (sometimes insufficiently) to money demand shocks anyway. There’s no central bank which will say “we raised our total liabilities by %100 forever so Ege could see what happens if we do that”.

    “The next step is to figure out why some monetary shocks have one effect and other monetary shocks have the opposite effect. We know that if you evaluate monetary shocks in terms of the “price effect” it is easy to explain why rates rise on some cases and fall in others, but if you evaluate monetary policy in terms of the interest rate effect it gets a lot more confusing, as there are lots of indeterminacy problems.”

    I agree, but this doesn’t change much, because I’m not suggesting we think of monetary policy using interest rates. I’m simply pointing out that monetary policy does have an impact on interest rates. To me, neo-Fisherism is the claim that easy money policies which are expected to be sustained create enough expectations of inflation so that real interest rates fall while nominal interest rates rise when the policy is announced. It’s not the claim that “high interest rates raise inflation”.

    I don’t want to look at empirical data crudely to test this claim, because if you throw expectations about future policy into the mix, then it becomes increasingly difficult to entangle the effects of current policy vs. market expectations of future policy. If the market expects future policy to be sufficiently tight, then that can offset any inflationary effects of present easy money policies, and leave us with low nominal interest rates and seemingly easy money.

    “BTW, since MV=PY is a tautology, any true model must be able to be explained in MV=PY terms.”

    Sure, but what I mean by “the model has MV = PY in it” is that V is an increasing function of the interest rate spread between bonds and money, which is far from a tautology.

  20. Gravatar of ssumner ssumner
    23. December 2017 at 22:12

    Ege, You said:

    “I don’t think the model makes that prediction, though. If you have a money demand shock that the monetary authority does not respond to by adequate monetary expansion, then the money supply could be going up while you see low output, low interest rates and low inflation, i.e the price effect is present in the model. Of course, they would still be higher than they would be in the counterfactual case where the monetary authority doesn’t respond at all to the shock.”

    To avoid this problem I try to use event studies, wherever possible. Thus I focus on
    the market response to unexpected monetary policy announcements. One market of interest is TIPS spreads. So if a monetary shock is contractionary, then TIPS spreads should decline on the news. Ditto for the price of foreign exchange.

    The empirical evidence of interest looks at events that are clearly some sort of monetary policy surprise, and then looks at co-movements of interest rates, exchange rates, inflation expectations, etc.

Leave a Reply