Saturday, May 5, 2018

Neo Fisherism: Look, it Works!

We know that Neo-Fisherism works in our models. In baseline macroeconomic models in which money is neutral, increases in inflation and increases in nominal interest rates go hand-in-hand. If we incorporate standard types of frictions that give rise to monetary non-neutralities in our models, for example New Keynesian sticky prices or segmented markets, an increase in the central bank's nominal interest rate target will in general raise inflation - in the short run and in the long run. See for example Cochrane (2016), Rupert and Sustek (2016), and Williamson (2018). And note that these theoretical predictions don't come from some freakish concoction of a demented neo-Fisherian, but from the mainstream modern macroeconomic models widely used by academics and central bankers.

But these predictions run afoul of standard central banking practice. I have yet to meet a policymaking central banker who thinks that inflation goes up if the central bank "tightens," in the usual sense in which that word is used. Standard central banking practice is of course enshrined in basic Taylor rules, which dictate increases in the central bank's nominal interest rate target when the inflation rate increases. In fact, the "Taylor principle" implies a more-than-one-for-one increase in the nominal interest rate target in response to an increase in the inflation rate.

Standard theory is pessimistic about the ability of a Taylor-rule central banker to successfully control inflation. Given a fixed inflation target, the Taylor-rule central banker can get stuck in a policy trap in which the nominal interest rate is as low as the central banker wants it to go, inflation is lower than the inflation target, and low inflation continues in perpetuity. But apparently real-world central bankers don't follow a rigid Taylor rule. For example, central bankers in the United States and Canada have recently raised their interest rate targets in the face of inflation that was falling below the inflation target. The Fed and the Bank of Canada, in their public statements, have both used what an Incipient Inflation Argument (IIA) in order to extract themselves from the low-inflation policy trap. That is - as they argue - inflation may not be high today, but given a tightening labor market, the Phillips curve will surely reassert itself, and we have to get ahead of the curve. If we don't get ahead of the curve then, as the argument goes, we'll have to tighten at a higher rate later on, with dire consequences.

The IIA allows the central banker to do the right thing - the neo-Fisherian thing, basically - while not abandoning the Phillips curve in some obvious way. To quote yours truly from September 2015:
What are we to conclude? Central banks are not forced to adopt ZIRP [zero interest rate policy], or NIRP (negative interest rate policy). ZIRP and NIRP are choices. And, after 20 years of Japanese experience with ZIRP, and/or familiarity with standard monetary models, we should not be surprised when ZIRP produces low inflation. We should also not be surprised that NIRP produces even lower inflation. Further, experience with QE should make us question whether large scale asset purchases, given ZIRP or NIRP, will produce higher inflation. The world's central bankers may eventually try all other possible options and be left with only two: (i) Embrace ZIRP, but recognize that this means a decrease in the inflation target - zero might be about right; (ii) Come to terms with the possibility that the Phillips curve will never re-assert itself, and there is no way to achieve a 2% inflation target other than having a nominal interest rate target well above zero, on average. To get there from here may require "tightening" in the face of low inflation.
At the time, there was no shortage of opinion to the contrary. As Larry Summers wrote in the Washington Post in August 2015:
The Fed, like most central banks, has operationalized price stability in terms of a 2 percent inflation target. The dominant risk of missing this target is to the downside — a risk that would be exacerbated by tightening policy.

In December 2015, as you may recall, the Fed in fact began tightening in the face of low inflation. Here's what happened:
So much for downside risk. It took a while, but raising the Fed's nominal interest rate target coincided with an increase in inflation, to the point where the Fed has now hit its 2% PCE inflation target. So, the Fed's policy moves have been a success, and those who feared that interest rate hikes would take the US economy over a cliff should have calmed down and slept more soundly. After six 25-basis-point increases in the fed funds rate target range, the unemployment rate has fallen from 5.0% to 3.9%.

Should central bankers just declare neo-Fisherism a success, throw away their Phillips curves, and move on? Phillips curves should have of course been thrown out long ago, but we want more evidence to convince people that changing the sign in the policy rule (increasing the nominal interest rate target when inflation is below target) is the right thing to do. The recent US experience is only one episode, and there are many factors other than monetary policy (oil prices, factors affecting the real rate of interest) that affect inflation over the short term and the long term.

What has happened in other countries over this same period (2012 to present)? Here's Canada:
The Bank of Canada has a 2% inflation target, in a range of 1-3%, so as you can see the Bank has not been outside its target range much in the last six years. Average inflation has been below 2%, but the current inflation rate in Canada is currently at 2.3%. In terms of interest rate hikes, the Bank is one behind the Fed, with its target rate at 1.25%, as compared to the ON-RRP rate in the US (the comparable secured overnight rate), which is pegged at 1.5%. The Bank of Canada was one of the first central banks in rich countries to increase its policy rate after the financial crisis, though the interest rate target dropped after the fall in oil prices (associated with a drop in real activity in Canada). Overall, inflation performance in Canada relative to the US is consistent with neo-Fisherism. After the financial crisis, the Bank of Canada didn't set its nominal interest rate target as low for so long as did the Fed, and inflation has been on average somewhat higher in Canada.

In most of the other rich countries in the world, central banks have kept their nominal interest rate targets close to zero or below zero for a considerable time. I've selected five key ones: the European Central Bank, the Bank of England, the Swiss National Bank, the Bank of Japan, and the Swedish Central Bank. Here are the inflation time series (from 2012) in those countries:
In this picture, the UK stands out as having inflation above the 2% Bank of England inflation target for more than a year. Of course the Pound has also depreciated by about 20% against the US dollar since 2014, for reasons having little to do with monetary policy. In Sweden and the Euro Area, inflation has at times been at the 2% target, though inflation is now softening in those jurisdictions, particularly in the Euro area. Inflation has come up in Switzerland and Japan, but is still well short of 2% in both countries. As well, we could look at a scatter plot of inflation vs. the overnight nominal interest rate in each of the seven countries we have been looking at:
So, that's a nice neo-Fisherian picture. If good labor market performance produces high inflation, why is inflation so much lower in Japan and Switzerland than in Canada? If unconventional monetary policies make inflation go up, why are the countries with the most extreme unconventional policies (negative nominal interest rates, quantitative easing) - Japan, Switzerland, and the Euro area - the ones with the lowest inflation in the picture? Canada, which didn't indulge in any unconventional policies, but has a higher short-term nominal interest rate than all these countries but the US, has an inflation rate of 2.3%. How come? Neo-Fisherite policy works, that's why.

But I think (in part because I've been told) that central bankers are skeptical that they could ever sell a neo-Fisherite monetary policy rule to the public. No one gets excited about higher interest rates, and the average layperson has been conditioned by decades of central banker dialogue about heating the economy up, cooling it down, taking away the punch bowl, etc. It's much easier to swallow higher interest rates if your neighborhood central banker is telling you that this keeps the economy from overheating. We know what happens when things overheat. They break down and explode. In some theories (New Keynesian models, old money surprise models) you can have too much output, but in practice I think this is nonsense. The unemployment rate can't be too low, unless there's some long-run inefficiency at work. The overheating economy is simply part of the IIA argument. That is, a tight labor market is excellent cover for interest rate hikes, which are going to bring inflation up to target. Jim Bullard likes to say: "tighten on good news."

So what's the harm in using the IIA argument, if it's just carrying out the neo-Fisherite program in a more palatable way? First, there are circumstances in which it would be optimal to increase inflation, even if real activity is sub-par. The right thing to do could be to raise the interest rate target, but if the economy isn't "heating up," the IIA argument can't be used. Second, central bankers can run through the story so many times that they believe it. In current circumstances, the risk is that central bankers end up exceeding their inflation targets because of a misunderstanding of the effects of their policies. For example, officials at the Bank of Canada and the Fed think of their current policy settings as "accommodative." From the April 18 Bank of Canada statement:
Inflation is on target and the economy is operating close to potential. That statement alone underscores the considerable progress seen in the economy over the past 12 months. That said, interest rates remain very low relative to historical experience. This is because the economy is not yet able to remain at full capacity on its own.
And from the May 2 FOMC statement:
The stance of monetary policy remains accommodative, thereby supporting strong labor market conditions and a sustained return to 2 percent inflation.
Policymakers at the Bank of Canada and the Fed think in very similar ways. In both places, policymakers think of policy in terms of a neutral nominal rate of interest (NNRI) which is thought to be in the range of 2.5-3.5%. According to their thinking, if the central bank's nominal interest rate target is less than the neutral rate, this puts upward pressure on the inflation rate and downward pressure on the unemployment rate. So, the ultimate goal is to raise the nominal interest rate to the NNRI, at which point the economy will be operating at potential and inflation will be at its target.

The key problem with this reasoning is that, if the central bank holds the nominal interest rate constant for a long time, policy ceases to be "tight" or "loose." For example, if the Bank of Canada had pegged its interest rate target to 2% in 2009 and kept it there, real economic activity in Canada today would be indistinguishable from what we're seeing. But inflation would be higher. In Canada and the US, nominal interest rates have been low for going on 10 years, and those interest rates are moving up. So the Bank of Canada and the Fed aren't removing accommodation - they're tightening. And tightening means a negative effect on real economic activity and a positive effect on inflation. The Bank of Canada is achieving its goals. The Fed is achieving its goals. Time to stop tightening.


  1. Learned a lot from your piece again. Thank you. The FOMC's normalization plan has two dimensions: (i) increase in the FFR target and (ii) reduction in balance sheet size. Even though FOMC agrees to adopt (ii) in an addendum in June 2017, it seems that it was hesitant about doing it. Since the size of balance sheet reached its maximum at $4.5 trillion in February 2015, there has been a merely small decline in the balance sheet size as it is currently at $4.3 trillion. Why don't they get rid of excess reserves by increasing the ON-RRP rate equal to the IOER and swicth from floor system to the corridor system? Second, some members of FOMC proposed to employ (ii) before (i). What is your view on that?

  2. Stephen, why is this evidence for the Neo-Fisherian view? Isn't this evidence also consistent with the current Wicksellian view? In fact, if I look at the Canadian picture the average CPI started to increase well before the rates were raised. That would be seem to be inconsistent with this Neo-Fisherian view.

    I am not convinced by this evidence because this is a low hurdle for the Neo-Fisherian hypothesis to clear. This could easily have happened even if the Neo-Fisherian hypothesis is false. It's therefore a poor test of the Neo-Fisherian hypothesis.

    What would convince me more is the following. When the interest rate is raised, despite a plausible Taylor Rule calling for lowering or keeping the interest rate the same, inflation as measured by the CPI increases. This would be a test that could distinguish the traditional view from this view. We, however, seemed to have run such a test already. During the crisis the Taylor rule called for a lower interest rate and inflation went down. This is consistent with the traditional view and not the Neo Fisherian view.

    1. What's are the alternatives? Raising the nominal interest rate makes inflation go down? The Phillips curve explains the data? Certainly neither of those works, right?

      "the average CPI started to increase well before the rates were raised"

      Monetary policy isn't the only factor driving the inflation rate, as I mentioned above.

      "When the interest rate is raised ... inflation as measured by the CPI increases."

      Exactly. That's what I'm showing you.

      "During the crisis the Taylor rule called for a lower interest rate and inflation went down."

      But in that case there were other factors at work - principally the drop in asset prices, generally, and the corresponding drop in the prices of commodities. That's not a clean monetary policy experiment.

    2. "What's are the alternatives? Raising the nominal interest rate makes inflation go down? The Phillips curve explains the data? Certainly neither of those works, right?"

      Why doesn't it work? The evidence seems to be perfectly consistent with it. E.g. the Volcker disinflation. Interest rates were raised. Inflation rate dropped.

      "Monetary policy isn't the only factor driving the inflation rate, as I mentioned above."

      Sure, but that argument can always be used no? c.p. is never c.p.

      "Exactly. That's what I'm showing you."

      What I see is that higher interest rates are associated with CPI increases. But this is consistent with both views! E.g. the central bank expects CPI increases and starts to raise rates as these materialize. Or the CPI increases and the central bank starts to raise rates. How do you distinguish these from interest rates are raised to increase the CPI? They seem to be observationally equivalent. And no wonder, clever people before us interpreted the evidence differently from you.

      It reminds me a bit of the following conversation:

      A1: People believed the sun to go around the earth because it looked that way.
      A2: But how would the earth moving around the sun look then?

      "But in that case there were other factors at work - principally the drop in asset prices, generally, and the corresponding drop in the prices of commodities. That's not a clean monetary policy experiment."

      How would a clean monetary experiment look like? Or what would be as close to a clean monetary experiment as possible?

      I am thinking here of the central bank that errs or sets policy somewhat randomly. That would count as good evidence right?

      Alternatively, I am thinking here of the political business cycle literature or some other public choice literature on monetary policy. The presumption in that literature is that the actors are self - interested and know how the world works (or at least monetary policy). The fact that they are able to achieve their political objectives suggests that their picture of the world is the right one as far as monetary policy is concerned. In that literature, as far as I am aware, there are no Neo Fisherian central bankers.

      Or perhaps I don't understand your view well enough. I'll admit that it is counter-intuitive to someone who is mainly aware of the macro models taught to undergraduates. Perhaps if I was taught the Neo Fisherian view I'd find the evidence for the traditional view unconvincing. But I have a hard time imagining such a counterfactual world. Mostly because of how monetary history is told.

    3. "the Volcker disinflation. Interest rates were raised. Inflation rate dropped."

      Go back and look at the data. Volcker actually reduced money growth, and ultimately nominal interest rates and inflation fell.

      "I'll admit that it is counter-intuitive to someone who is mainly aware of the macro models taught to undergraduates."

      That's not intuition. That's using the wrong model.

    4. Why does it matter how rates were raised? We see there a pattern of high interest rates and falling inflation.

    5. It's a very different policy experiment. In some models (e.g. segmented markets), a permanent decrease in the money growth rate leads to an increase in nominal interest rate in the short run, and a decline in the long run, with inflation falling in the short run and the long run. Take the same model, and reduce the nominal interest rate permanently, and inflation falls in the short run and the long run. Thus, Volcker is consistent with neo-Fisher.

    6. OK, let's decompose this. We all agree that lower inflation requires slower growth in the money supply. So for neo-Fisherites to be right, the relationship between money growth and short-term interest rates needs also be positive. However, when I included M1 growth in a regression as a control variable (Alexandrakis, 2014) the coefficient came out negative. But OK, maybe I messed up. However, Barro and Sala-i-Martin (1990), surprisingly to them, got the same result. So if slower money growth results at least initially in higher interest rates, then we would expect a temporary rise in interest rates to be associated with lower, not higher inflation. Of course, since much of this depends on how people set and change their expectations this relationship may be unstable. This means that neo-Fisherites could be right under certain circumstances. But I don't see how the data suggests that they have been right in the past.

    7. "We all agree that lower inflation requires slower growth in the money supply."

      In theory. But in practice, M1 growth is not helpful for thinking about inflation and monetary policy.

    8. Honest question, is there a monetary aggregate that is helpful? I remember discussing a paper a long time ago that was trying to construct a monetary aggregate that was treating the various liquid assets as imperfect substitutes. Everyone was New-Keynesian at the time and thinking along interest rates so I told him that I don't see the paper going very far in terms of publication. I run into him a few years later and he told me I was right. But I feel it was a worthwhile endeavor nevertheless.

    9. 1. William Barnett (Kansas) has been working on divisia aggregates for a very long time, and seems to want to claim success, but I'm not sure we could say these are any more successful than the simple sum aggregates.
      2. Lucas and Nicolini claim to have found a stable money demand function (for M1):

      But due to changing technology and regulation, how assets are used in transactions, and as collateral is an evolving process, and there are new assets being created all the time. So, I don't think we can ever find an asset total for which there is a stable and simple relationship with real economic activity, prices, and interest rates. Given that central banks can peg short-term nominal interest rates successfully, I think it's better to think about policy rules that connect nominal interest rate targets with the ultimate goals of the central bank.

  3. Professor, the time series chart for Canadian inflation rates and overnight (bank) rates for the period from 2015 to mid-2017 do not appear to reflect a neo-Fisherian effect. Although the overnight (bank) rate declined from 1%/yr to 0.5%/yr in two stages during the first half of 2015 and remained constant at 0.5%/yr from mid-2015 to mid-2017 (a period of 2 years), the inflation rate for Canada appears to have been unaffected relative to the inflation rate (average) in the period leading up to 2015. As I understand it from reading your papers on the subject, the neo-Fisherian hypothesis states that the nominal interest rate, d(ln(1+i(t)))/dt, minus the inflation rate, d(ln(P(t)))/dt, equals the real rate of interest, d(ln(1+r(t)))/dt, to use the convention that appears in at least one of your papers on the subject. If r(t) = r, a constant (as often assumed), then the term on the righthand-side of the equation becomes, simply, r. It is difficult to conceive of the real rate of interest bobbing up and down in that two-year period to the extent it would have to in order to satisfy the neo-Fisherian relation between nominal interest, inflation and the real rate.

    I don't disagree that the Taylor Rule approach is misguided. Your paper titled "Inflation Control: Do Central Bankers Have It Right?" (June 26, 2017) lays down a very convincing and cogent argument in that vein. I wonder, however, whether the models which are based on assumptions (necessary as those are for tractability) might not be telling the whole story, that there might be other factors bearing on the determination of inflation rates, especially in an open economy such as Canada's. The chart of Canadian inflation vs. the Bank of Canada overnight interest rates appears to lend itself to an alternative (if unknown by this reader) explanation.

    1. In some of the theory, the long-run real interest rate is a constant, for illustrative purposes. But that's a bad assumption to make when looking at the data. There are long-run and short run factors that matter for the real rate - non-monetary factors in particular, like safe asset shortages, demographic factors, and low productivity growth. So, inflation is affected by factors that affect the real interest rate, and by factors affecting relative prices, like the price of oil, in addition to monetary policy. So, there is a lot going on in the data and, in general, determining how inflation is being affected by monetary policy is difficult. I'm just offering up some crude evidence that might make you want to think twice. But, I think the onus should by on those those advocating conventional wisdom to show why that works.

    2. Thank you for the thoughtful response. It is a very complex and complicated subject, but one which is of high importance given the influence that US monetary policy has on global as well as domestic economic activity. Viewing the time series data of the effective Fed Funds rate vs the occurrence of economic recessions in the US going back as far as 1970 up to 2017, it is striking the correspondence in the rise in the Fed Funds rate has to the officially determined onset of an economic recession in the US. I can think of only one instance in that time span when the Fed Funds rate steadily increased and a recession was not officially noted. One would definitely not welcome a series of interest rate hikes in the current environment, simply because a majority of the FOMC thought it necessary in order to later have some ammunition, as it were, to counter-act the ensuing recession that might result from the anticipatory rate hikes. Such a notion is hardly worthy of the moniker, "sound monetary policy".

      One point that did occur to me as I was working through your June 26, 2017, paper on central bank policy was the observation that according to your deterministic model an inflation rate target of zero leads to a stable equilibrium (output gap = zero) under a modified interest rate control policy. The modified policy includes feedback from the contemporary output gap as well as the observed inflation rate to determine the nominal interest rate (the manipulated variable) in a discrete time model. Complications, such as sticky prices, etc., are ignored but could be incorporated. The policy generator is linear, not relay actuated or limiter constrained. The controller efficacy can be demonstrated by numerical simulation.

      Aside from the details, which are trivial, the interesting insight is the observation that zero as a target for the inflation rate and the output gap yields a stable equilibrium. A question then arises as to why one would as a matter of policy choose some value other than zero, say 2%/yr, as a target given that the theoretical equations comprising the model (cf., eqn. (3) & (4) of the paper) clearly indicate the direct relationship between inflation and the output gap: When pi = 0 and R = rho, the output gap, y, goes to zero in finite time. The simple Taylor Rule (eqn. (5)) with pi-asterisk set to zero will yield the same result, although addition of a term proportional to the output gap lagged by one period ensures the desired stable equilibrium.

    3. The reduced form model in my paper (in St. Louis Fed Review now: is actually a simplified version of reduced form New Keynesian (NK) model. There are sticky prices implicit in what is going on there - that's where the Phillips curve comes from. In some versions of NK models (the one in my paper included), an optimal policy is indeed zero inflation, though in some of those models there are other distortions that make optimal inflation something other than zero (sometimes less than zero, in fact). There's certainly no good reason for a 2% inflation target, rather than a 0%, 10%, or -2% target. That's what the Reserve Bank of New Zealand chose in 1990, and most central banks that chose inflation targets since then simply followed suit.

  4. Hi Stephen. The potential problem I see with the Neo-Fisherian view for the short-run is that it's derived under rational expectations. But what comes out very clearly from your post and the previous comments, is that in this case, under economic agents have a highly misspecified personal model of the monetary policy transmission mechanism under the hypothesis of short-term neo fisherianism, so RE is systematically violated in this case. -> we should formally explore an economic model in which agents use a phillips curve framework to understand inflation, even though under RE the model would produce the neo-fisher effect in the short-term. I'm not sure anyone has done this (this isn't about learning style misperceptions that are usually explored, it's more like a long-term use of a wrong state space representation of the economy in agents' expectation). In any case, I doubt the result would validate the simple neo-fisherian hypothesis for the short-term, since it's so much based on RE (at least the way I understand it).

    1. I have no doubt that, with enough work, one could find arbitrary beliefs that would support orthodox views about how inflation should be controlled. There's a paper by George Evans and a coauthor in which he argues that learning models (roughly adaptive expectations) will give you orthodox results. I played around with a simple version of that in my St. Louis Fed paper on this. What happens is that, if expectations and prices are sufficiently sticky, then low nominal interest rates will make inflation go up - forever. That is, low nominal interest rates produce instabilty, which we don't see in practice - e.g. Japan. So, to get the orthodox policy result, you produce something that violates what you see in the data. It's much easier to get neo-Fisherian results than orthodox ones, which tells me there's something wrong with orthodoxy.