Tuesday, February 3, 2015

Phillips Curves and Deflationary Panic

Paul Krugman has an interesting post on Phillips curves and monetary policy. There are two pieces of "theory" at the heart of his argument: (i) the Phillips curve; (ii) the "deflationary trap." He shows us annual observations on the unemployment rate and core inflation for 2007-2014, estimates a linear Phillips curve, then argues, assuming anticipated inflation is 2% for the whole period, for an estimate of 4.9% for the natural rate of unemployment. Then, he draws a policy conclusion, which is:
Raise rates “too late”, and inflation briefly overshoots the target. How bad is that? (And why does the Fed sound increasingly as if 2 percent is not a target but a ceiling? Hasn’t everything we’ve seen since 2007 suggested that this is a very bad place to go?) Raise rates too soon, on the other hand, and you risk falling into a deflationary trap that could take years, even decades, to exit.
I'm going to use quarterly data, and link the points in the scatter plot so that we can see what's going on. Here, I'll use the pce deflator (excluding food and energy) and stick to 2007-2014, as Krugman does. This produces:
The Phillips curve is an elusive thing. Sometimes you see it in the data, and sometimes you don't. In this particular data set, the observation in the top left-hand corner is 2007Q1, and the line traces observations to 2014Q4. From 2007Q1 until unemployment reaches a peak in the recession, there is a decline in inflation that coincides with the increase in unemployment, which of course is consistent with a Phillips curve story. But, as many people have noted, the implied Phillips curve is quite flat. Over that period, there's a drop of about 0.8 percentage points in the inflation rate, coupled with an increase of more than 5 percentage points in the unemployment rate. Then, from the peak in unemployment (the observation furthest to the right) to 2014Q4, there's a drop of more than 4 percentage points in the unemployment rate, and the inflation rate does not change at all. You can see that, over this later period, the economy is hardly tracing out a Phillips curve. This makes me wonder why anyone would be confident in using the Phillips curve as a basis for forecasting inflation, or thinking about policy. I have some ideas, none of which has anything to do with sound scientific practice.

Just for good measure, we could look at a similar plot, but using the raw pce deflator without stripping out food and energy. That's the measure the Fed focuses on, of course. Here's what we get:
So, that shows a larger decline in the inflation rate during the period when the unemployment rate was increasing rapidly (in part due to the steep decline in commodity prices of course), but essentially the same story from the peak in unemployment - unemployment declines by more than 4 percentage points with essentially no change in the inflation rate.

So, look again at the first sentence in the quote from Krugman's post, above. Clearly, Krugman thinks that, if the zero interest rate policy (ZIRP) continues, and barring unforeseen negative aggregate shocks, the unemployment rate will continue to decline, and the inflation rate will rise - at worst we go above 2% for a short time. But what makes the experience of the last five years make Krugman think that inflation will start to rise? We've had five years of declining unemployment and ZIRP, but no net change in the inflation rate. Further, we might look at Japan, where the Bank of Japan is an old hand at ZIRP. Here's the overnight interest rate in Japan from mid-1995:
As you can see, the overnight rate has been below 0.5% for most of this period of almost 20 years, and below 0.1% for the last five years. Here's the Japanese CPI for the same period:
As you can see, on average the inflation rate has been about zero over the whole period (there's also a story about the last large increase in the CPI in April 2014 - that goes away if you take out the effects of the consumption tax). So, if I looked at the last two time series, that would not give me much confidence that a long period of ZIRP will produce more inflation.

Krugman's chief concern is the risk of deflation - a "deflationary trap." He's been worried about that before, and we know how that turned out. I've discussed deflation terror in an earlier post. The deflation trap is another myth, like the Phillips curve, that we would be better off without. This is in part what formal economics is for. We're forced to write down our theories in the form of internally consistent mathematical language, and subject these models to the scrutiny of our peers, and the public at large. That weeds out a lot of nonsense. The "deflationary trap," while it may sound like convincing economics, particularly when it comes out of the mouths of Larry Summers, or Paul Krugman, has no formal economics behind it. I have seen plenty of models (and have written down some myself) in which there are liquidity traps, and deflation, but I've never seen a model of what Krugman is describing - a vortex that sucks you in and won't spit you out for a very long time. On the empirical side, there's nothing to suggest that we have ever seen such a phenomenon. The Japanese experience of the last 20 years doesn't indicate that deflation is some self-sustaining trap - even given ZIRP, average inflation was zero over that period, as I already noted. Also, this paper by Charlie Plosser has some examples of how deflation need not be associated with horrible stuff.

If there is a deflation trap, or a low-inflation trap, I think it's a policy trap, as I argued in my last post. As such, Paul Krugman and Larry Summers are part of the trap phenomenon. We could start calling this the Krugman/Summers advice trap, I think.

20 comments:

  1. How does this even come close to meaning we should raise rates. The important thing is aggregate demand, in what crazy world is aggregate demand close to potential right now, in what crazy world is there not a considerable output gap? Raising rates when demand is still lackluster and when there is still overwhelming evidence for an output gap, plus the clearly overvalued dollar, is absolutely nuts.

    ReplyDelete
    Replies
    1. "The important thing is aggregate demand, in what crazy world is aggregate demand close to potential right now..."

      You're thinking about price and wage stickiness? Maybe you think that the prices and wages haven't adjusted from the shock we had more than 6 years ago? You seem to have a lot of insight into these output gaps. Tell me how you measure it.

      Delete
    2. I'm not talking about price and wage stickiness, not necessarily. I'm talking about excess capacity, I'm also talking about weak nominal growth which is below where it should be, and in terms of levels, far below its trend path.

      In terms of the output gap, look here: http://research.stlouisfed.org/fred2/graph/?g=Zt3

      Why should the Fed even think about raising rates now? The economy is not even close to overheating.

      Delete
    3. "I'm not talking about price and wage stickiness, not necessarily."

      Well, you must be talking about some inefficiency that you think sticking at the zero lower bound will correct. What is it? What's causing that excess capacity? Do you know where that "potential output" measure that's posted on FRED comes from? Do you think that we would be any farther from that potential output measure if the interest rate on reserves had been, say, 1.25%, for the last 3 years, instead of 0.25%? What's overheating? How would you define it?

      Delete
    4. Another thought: In February 2004, the unemployment rate was 5.6%, as it was in the last employment report, the inflation rate was 1.8%, and the fed funds rate was 1.0%. After the financial crisis happened, the Fed was criticized for keeping rates too low at that time (they were about to go into a tightening phase). What do you think? In February 2004, was policy just right, too accommodating, or was there excess capacity and the fed funds rate should have been zero?

      Delete
  2. "Well, you must be talking about some inefficiency"

    The 'inefficiency' is the long deleveraging process the economy is and must go through, given the high private and public debt burden in the US. This process sucks out demand and is generally a deflationary pressure, causing AD to be lower than it could be. Raising interest rates would be enormously unhelpful in this situation as it would increase debt burdens, making the deleveraging process even tougher; if these pressures causes falling prices and corresponding falling incomes, the debt burden in real terms increases even more so, causing AD to fall even lower and deleveraging to increase. Meanwhile, these deflationary pressures causing lower tax income, plus the higher interest rates, cause the public debt burden to increase, which will force austerity and fiscal contraction, causing even more deflationary pressure. This is a cycle that can cause one kind of 'deflationary spiral', if there is a risk of the above it would be tremendously terrible idea to raise rates now.

    Potential output is probably measured at the Fed involving the deviation of employment from full employment. Do you think we're at full employment now? Remember that the current quoted unemployment rate might be misleading, as there is still a lot of people out of the labour force. Do you have any reason to think we're at potential GDP?

    As for interest on reserves, I don't know is that the Fed's new policy instrument? Is it used to target the federal funds rate now? If so, then yes I think higher interest rates on reserves, if you consider that contractionary monetary policy, could have caused the US to be even further from potential output.

    Overheating is too much money chasing too few goods, it's inflationary pressure, where real output can no longer be increased to match nominal demand and therefore causes upwards pressure on prices.

    ReplyDelete
    Replies
    1. I can't say about 2004, I think employment is too imperfect an indicator, what if asset bubbles (like the housing bubble) are an indicator of overheating, of inflationary pressure being hyper-focused on one kind of asset class. Then maybe it should have raised rates. Otherwise, on the basis of inflation and unemployment indicators alone, I see no remarkable need to raise rates then. Therefore it essentially depends on whether you consider the housing bubble a form of economic heating, and also whether monetary policy is directly to blame for it.

      Delete
    2. "The 'inefficiency' is the long deleveraging process..."

      The ratio of household debt to GDP has fallen about 20% since the end of the recession. And house prices have increased about 20% over the same period. I guess you still think leverage is still inefficiently high. How do you know that? Seems unemployment has dropped like a rock. Why aren't all those overleveraged people out looking for work?

      I'm pulling your chain here a bit to see what you know. Potential output on FRED comes from the Congressional Budget Office. That's not a measure of what we might want to think of as efficient output. It's just based on projections of productivity, labor force, and capital stock - you shouldn't take it seriously. "Overheating" is a journalist's word, not an economist's. It's Phillips curve thinking, basically - output gaps predict inflation, which does not hold up empirically.

      "As for interest on reserves, I don't know is that the Fed's new policy instrument?"

      Effectively, yes:

      http://www.federalreserve.gov/newsevents/press/monetary/20140917c.htm

      Delete
  3. While I'm (obviously) British, so I don't spend that much time looking at US stats, I remain thoroughly unconvinced that the US is appropriately delevered, given how enormously levered it before the recession, even if it has declined 20%. There's still the public debt as well, which continues to grow.

    " based on projections of productivity, labor force, and capital stock", that seems like a good way to project potential GDP.

    "output gaps predict inflation, which does not hold up empirically.", it does hold up logically and in models, how can you have a positive output gap, that is essentially demand exceeding supply, and not expect increased prices? I expect the economitric specifications are wrong, or potential output is being miscalculated, if it really doesn't hold up empirically.

    ReplyDelete
    Replies
    1. "...that seems like a good way to project potential GDP."

      Trust me, this is close to useless.

      "...have a positive output gap, that is essentially demand exceeding supply, and not expect increased prices..."

      That's a Phillips curve. Look at the first chart (or the second). Presumably your output gap was falling while the unemployment rate was going down more than four points. Why didn't the inflation rate go up?

      Delete
  4. "That's a Phillips curve. "

    Depends how you specify it, again I think the headline unemployment figure can a misleading indicator of excess capacity.

    But I'm not thinking in terms of Phillips Curve, I'm thinking in terms of AS-AD; even if the Phillips curve relationship breaks down while the economy is not at full employment, even you must agree that pushing AD far enough will just result in more inflation and no more output, Friedman thought the same. The point is that I don't think we're anywhere near close to that point yet where the AS curve is sloping up asymptotically.

    Also, I think in terms of Nominal GDP, even though I'm not a market monetarist, I still think NGDP is a good way to think about policy, and we're below a reasonable NGDP path. Even if increasing AD more doesn't result in any more meaningful output (which I very much doubt) and increases inflation above target for a while (again, I doubt very much), all I'd say is 'so what?' My preferable target is some reflation of nominal income, I'd rather we be on trend for nominal income and above target with inflation, than below trend in nominal income while within the inflation ceiling.

    ReplyDelete
    Replies
    1. AS-AD is pretty limited as a policy tool, don't you think? We now have better ways to think about dynamic economies.

      Delete
    2. Last I checked, NK DSGE models still rely on an output gap. But they don't tell us a good way of measuring the output gap.

      Delete
    3. This illustrates a useful point, which is that you need a model to measure an output gap. Whether the NK model is a good one or not, it tells you how to do it. The output gap in the NK model is the difference between output in the underlying real business cycle model with flexible prices, and output as ground out in the sticky price equilibrium. So, I can estimate the parameters of the model, and then determine the time series of shocks that replicates the data exactly (this requires some assumptions, and some work). Then, I simulate the model under flexible prices with that sequence of shocks to get potential output. Then I'm done.

      Delete
    4. You can do that, but surely the Federal Reserve has already done this, and it probably tells them that the output gap is still too high to justify raising rates. But even then, what if there are other factors than sticky prices keeping the economy from its RBC equilibrium level, I'm sure there are all kinds of other frictions that could potentially be keeping the economy from its RBC level.

      Delete
    5. "...but surely the Federal Reserve has already done this..."

      Well, people in the Fed system think about all kinds of models, but the process by which the FOMC makes decisions is much more complicated than this - as is the case in any committee in which people have very diverse outlooks on the world. Sure there are all kinds of frictions, but there is a limited set that a central bank can do anything about. My impression is that, if you were making these decisions, that you would be in a panic most of the time, and your economy would look like Japan's - zero nominal interest rates forever.

      Delete
  5. "The deflation trap is another myth, like the Phillips curve, that we would be better off without."

    Was the Great Depression of the 1930s not a deflationary trap? This had nothing to do with sticky prices. Most industrial economies before WWII were price variable. (Roger Farmer had these price movements up on his blog a while ago and the Japanese case of the 1920s-30s is the same.) The causes of the Depression have been put down to reparations on Germany which were transmitted internationally (especially when they stopped paying), overly tight monetary policies associated with the Gold Standard and the fact it also did not allow a fiscal expansion, the collapse of the WWI boom which left a lot of distortions in many economies and a dysfunctional financial sector (which was not properly dealt with in Japan until 1927), a self-perpetuatng climate of "deflationary psychology" - i guess in more modern jargon downward expectations - low corporate profit expectations, high inventory levels and risk averse positions by banks towards lending, increased monopolisation of industry and the marginalisation of small firms, the weakening in the Sterling area as Britain was increasingly unable to manage crucial trade and supply routes etc I have seen all these explanations put forward. But sticky prices has definitely not been one of them. Keynes said that sticky prices can theoretically be an issue, but it was not an underlying problem. I think historians would say that it was a combination of events that causes phenomenon like the Japanese case from the 1990s or the Great Depression. A bit like a plane accident or a persons personality, or say the causes of WWI, - such events cannot be put down to one cause and you cannot really isolate one causal factor from another. It is the fact they come together that is important.

    How do you think then do you work with that, and have models that estimate an output gap which is defined as "the difference between output in the underlying real business cycle model with flexible prices, and output as ground out in the sticky price equilibrium" really going to help us understand the problem at all?

    ReplyDelete
    Replies
    1. If you think the Great Depression didn't have much to do with sticky prices, I'm with you. I'm by no means endorsing sticky prices as the primary friction a central banker should worry about, nor am I all that enthusiastic about NK models. However, we have to use the theory we have available to address the problems we are dealing with. What are the economic inefficiencies we're faced with, and what can monetary policy do about them? I've got some ideas about what is going on, which is what the last 2 posts in particular are about.

      Delete
  6. Feb 4 (Reuters) - Turkish President Tayyip Erdogan on Wednesday reiterated his view that a reduction in interest rates will cause the inflation to fall, a day after higher than expected inflation prompted the central bank to rule out an early rate cut.

    ReplyDelete
    Replies
    1. See: http://www.ft.com/intl/cms/s/0/f917986a-aad3-11e4-81bc-00144feab7de.html#axzz3QtzWN3GW

      He's a true neo-Fisherite.

      Delete