Thursday, September 3, 2015

Historical Fiction

Steve Williamson has a very nice post "Historical Fiction", rebutting the claim, largely by Paul Krugman, that the late 1970s Keynesian macroeconomics with adaptive expectations was vindicated in describing the Reagan-Volker era disinflation.

The claims were startling, to say the least, as they sharply contradict received wisdom in just about every macro textbook: The Keynesian IS-LM model, whatever its other virtues or faults, failed to predict how quickly inflation would take off in the 1970, as the expectations-adjusted Phillips curve shifted up. It then failed to predict just how quickly inflation would be beaten in the 1980s. It predicted agonizing decades of unemployment. Instead, expectations adjusted down again, the inflation battle ended quickly. The intellectual battle ended with rational expectations and forward-looking models at the center of macroeconomics for 30 years.

Just who said what in memos or opeds 40 years ago is somewhat of a fodder for a big blog debate, which I won't cover here.

Steve posted a graph from an interesting 1980 James Tobin paper simulating what would happen. This is a nicer source than old memos or opeds from the early 1980s warning of impeding doom. Memos and opeds are opinions. Simulations capture models.

The graph:

Source: James Tobin, BPEA. 
I thought it would be more effective to contrast this graph with the actual data, rather than rely on your memories of what happened.

The black lines are the Tobin simulation. The blue lines are what actually happened. (I'm not good enough with photoshop to superimpose the graphs, so I read Tobin's data off his chart.)

The two curves parallel in 81 to 83, with reality moving much faster. But In 1984 it all falls apart. You can see the "Phillips curve shift" in the classic rational expectations story; the booming recovery that followed the 82 recession.

And you can see the crucial Keynesian prediction error: After the monetary tightening is over in 1986, no, we do not need years and years of grinding 10% unemployment.

So, conventional history is, it turns out, right after all. Adaptive-expectations ISLM models and their interpreters were predicting years and years of unemployment to quash inflation, and it didn't happen.


One can debate 1981 to 1983. Here reality followed the general pattern, moving down a Phillips curve. Perhaps that is the success.  But the move was much quicker than Tobin's simulation. One might crow that inflation was conquered much more quickly than Keynesians predicted. But perhaps the actual monetary contraction may have been larger than what Tobin assumed, and assuming a harsher contraction would have sent the economy down the same curve faster?

Tobin describes his simulation thus:
The story is as follows: beginning in 1980:1 the government takes monetary and fiscal measures that gradually reduce the quarterly rate of increase of nominal income, MV. It is reduced in ten years from 12 percent a year to the noninflationary rate of 2 percent a year, the assumed sustainable rate of growth of real GNP. The inertia of inflation is modeled by the average of inflation rates over the preceding eight quarters. The actual inflation rate each quarter is this average plus or minus a term that depends on the unemployment rate, U, relative to the NAIRU, assumed to be 6 percent. This term is (6/U(-1) - 1). It implies a Phillips curve slope of one-sixth a quarter, two-thirds a year at U = 6 and has the usual curvature.
So, I think the answer is no. A faster monetary contraction leaves the 8 quarter lag of inflation in place, so you'll get even bigger unemployment and not much contraction in inflation. If someone else wants to redo Tobin's simulation with the actual 81-83 inflation, that would be interesting. But it is a bit tangential to the central story, 1984. You can also see here in the highlighted passage (my emphasis) how adaptive expectations are crucial to the story.

Now, let's be fair to Tobin. Yes, as quoted by Steve, he came out in favor of "Incomes policies," which used to be a nice euphemism for wage and price controls, but have an even more Orwellian ring these days. But Tobin also wrote, just following this graph,
This is not a prediction! It is a cautionary tale. The simulation is a reference path, against which policymakers must weigh their hunches that the assumed policy, applied resolutely and irrevocably, would bring speedier and less costly results. There are several reasons that disinflation might occur more rapidly. When unemployment remains so high so long, bankruptcies and plant closings, prospective as well as actual, might lead to more precipitous collapse of wage and price patterns than have been experienced in the United States since 1932. Moreover, the very threat of a scenario like figure 6 may induce wage-price behavior that yields a happier outcome. A simulated scenario with rational rather than adaptive expectations of inflation would show speedier disinflation and smaller unemployment cost, to a degree that depends on the duration of contractual inertia, explicit or implicit.
My emphasis. Now, having seen only one big Phillips curve failure in the 1970s, it might be reasonable for policy-oriented people not to jettison their entire theoretical framework in one blow. And this Tobin piece, using adaptive expectations, does incorporate some of the lessons of the 1970s. In the 1960s, Keynesians used a fixed Phillips curve. Friedman famously pointed out that it would not stay fixed -- but even Friedman (1968) had adaptive expectations in mind. For policy purposes it might make sense to integrate over models and adapt slowly, an attitude I just recommended in present circumstances.

You can see Tobin clearly seeing the possibilities, and clearly seeing the conclusions that we would come to after seeing the "happier outcome." That he had not come to these conclusions before the fact is understandable.

That contemporary commentators should forget or obfuscate this history, in an effort to resuscitate a comfortable, politically convenient, but failed economics of their youth, is less forgivable.

I don't want to fully endorse the classic resolution of 1984. Lots of other things changed, in particular deregulation and a big tax reform in the air. There was a lot of new technology. Financial deregulation was kicking in. We may find someday that such "supply side" changes were behind the 1980s boom. And we may jettison or radically reunderstand the Phillips curve, even with the free expectations parameter to play around with. It certainly has fallen apart lately (here, here and many more). But ISLM / adaptive expectations as an eternal truth just doesn't hold up. It really did fail in the 70s, and again in the 80s.

PS: The chart using actual inflation FYI



24 comments:

  1. "The Keynesian IS-LM model, whatever its other virtues or faults, failed to predict how quickly inflation would take off in the 1970, as the expectations-adjusted Phillips curve shifted up. It then failed to predict just how quickly inflation would be beaten in the 1980s."

    -Failure to predict positive and negative supply shocks (which is mostly what these events were) is hardly an intellectual sin.


    "And you can see the crucial Keynesian prediction error: After the monetary tightening is over in 1986, no, we do not need years and years of grinding 10% unemployment. "

    -The chart shows unemployment declining from its peak at 10%+ to 5.3% in five-six years. Reality was closer to five years. Worse predictions have been made. Also, notice that inflation was higher in real life, which explains the difference between Tobin's predictions of elevated unemployment through the mid-1980s and reality. Volcker stopped disinflation quite a few percentage points short of Tobin's 1980 scenario.
    "Adaptive-expectations ISLM models and their interpreters were predicting years and years of unemployment to quash inflation, and it didn't happen."
    -Inflation wasn't as quashed as it was in Tobin's scenario, and unemployment took a much shorter time to rise (though not much less time than in Tobin's scenario to decline!). The 1980s really did have quite a few years of significantly elevated unemployment, leading to most of the Regan deficits.

    ReplyDelete
    Replies
    1. "Inflation wasn't as quashed as it was in Tobin's scenario"

      Inflation was quashed far faster than in Tobin's scenario, the fact that it stopped short due to Volcker is pretty much irrelevant for two reasons. One is that Tobin's graph predicts significant and long lived deflation- 6 years below zero with a trough at -3%. Two is that virtually all the employment gains occur during this period. Peak UE is in 1988 and trough UE is in 1995 with 100% of the deflation years in this band and only a single year of positive inflation.

      The relationships between UE and inflation in this graph simply didn't exist, and cannot be recused by invoking fed actions.

      Delete
  2. Mr. Cochrane:

    I think we all know that you and Mr. Krugman dislike each other ....a lot. However, to say that Mr. Krugman has been pushing a lot of "fiction" without acknowledging realities is just plain wrong. For starters, here is a link to a post by Mr. Krugman which has the same graph that Mr. Tobin came up with in his paper: http://krugman.blogs.nytimes.com/2015/09/01/the-triumph-of-backward-looking-economics/?module=BlogPost-Title&version=Blog%20Main&contentCollection=Opinion&action=Click&pgtype=Blogs&region=Body

    If your dislike for Mr. Krugman is so great that you don't want to even read what he has said, let me copy and paste something he did say in the above post:

    "Unemployment shot up faster than in Tobin’s simulation, then came down faster, because the Fed didn’t follow the simple rule he assumed. But the basic shape — a clockwise spiral, with inflation coming down thanks to a period of very high unemployment — was very much in line with what standard Keynesian macro said would happen. On the other hand, there was no sign whatsoever of the kind of painless disinflation rational-expectations models suggested would happen if the Fed credibly announced its disinflation plans."

    So, it turns out that Mr. Krugman does acknowledge that unemployment came down faster than Mr. Tobin's simulation.

    I'd say for starters: read other people's blog first - even if you dislike them a lot. I do... I read your blog and Mr. Krugman's.

    Ganesh

    ReplyDelete
    Replies
    1. "Historical Fiction" is the title of Stephen Williamson's original post that Cochrane is referencing. Click the link to Williamson's post- the whole point is that it's a response to the very same Krugman column you are linking to.

      As to whether the "basic shape" was in line with the Keynesian prediction as Krugman claims, I suppose it's a matter of interpretation. The two lines look pretty different to me. Tobin's simulation looks like it had the unemployment rate north of 9% for seven consecutive years. In reality, it ended up being above that level for just one year, even though it peaked slightly higher than he projected.

      Delete
    2. Also notice that Tobin's graph is plotting annual rates. The graph Krugman created for comparison is using quarterly data points. At first glance this makes it look like the high unemployment in the 80's lasted longer than it did.

      Delete
    3. So the operative question for you Zack (and all those Ratex proponents) is this. Did the Ratex guys have a simulation in early-1981 of what the US inflation-unemployment graph might look like? If so, share it. The point is which model did a better job of predicting the trajectory of the inflation-unemployment tradeoff.

      Ganesh

      Delete
    4. "The two lines look pretty different to me. Tobin's simulation looks like it had the unemployment rate north of 9% for seven consecutive years. In reality, it ended up being above that level for just one year, even though it peaked slightly higher than he projected."
      -That's because the Fed had a looser monetary policy than Tobin predicted, allowing inflation to hover around 4% instead of going into 2% deflation.

      Delete
    5. pithom, you are just as liberal with the facts as Krugman. Tobin has inflation to 4% between 1986 and 1987. In reality it took half the time (1983). Tobin has a 3% - 4% inflation with 10% unemployment (1987). In reality 3% - 4% inflation is accompanied by a 7% unemployment in 1992 and 1993.

      Delete
  3. I think we need to keep the ISLM model separate from the Phillips Curve. The ISLM model is a theory of the AD curve. It says nothing about how quickly or slowly wages, prices, and expectations adjust.

    ReplyDelete
  4. On thinking it over, I think your second diagram is not quite fair to James Tobin. Because Tobin was making a *conditional* forecast, about what would happen to inflation and unemployment *conditional* on what would happen to MV (or NGDP). And it looks like the Fed increased MV in 1884, while Tobin's conditional forecast assumed MV would continue to fall.

    I think the right way to do it would be to take the numbers for *actual* historical MV, plug them into Tobin's Phillips Curve, and compare Tobin's new "predictions" for inflation and unemployment with what actually happened.

    Eyeballing the graph, it seems that Tobin's predictions would look a lot better using actual MV, though still would show prices being less flexible than they actually were from 1884 onwards.

    ReplyDelete
    Replies
    1. What also happened, that the RE mob seem to want to ignore, is that the oil price fell significantly from the early 1980s onwards. Boxed in thinking yields hollow sounding results.

      Henry

      Delete
    2. Uninformative. Tobin's model has UE dropping only when inflation is approaching zero, in reality UE started dropping when inflation rates stabilized at 4. The price of oil drops until 86/87 while UE and inflation start handling themselves 2-3 years prior.

      Delete
    3. Except that oil prices, in real terns, rose between 2002 and 2011 just as much as they rose in the 1970s. Where is the inflation? Well, check if Mercury was aligned with Venus. Maybe that will improve the prediction of the model.

      Delete
    4. Everything is black and white with you guys. I'm not saying the oil price is the only factor. The oil price began dropping from a high of $40 in 1979. It reached a low of around $12 about five years later. A persistent and significant decline. It then averaged around $20 for the next 15 years. The roll over of oil prices would have contributed to changing inflationary expectations.

      Henry.

      Delete
    5. Constantine Alexandrakis:

      The relative importance of oil prices can be gauged from a) the extent to which oil was used for power generation, and b) fuel efficiency. Until the mid-1980s, the US used quite a bit of fuel oil (a crude oil derivative) to generate power. That changed as the US moved more and more to natural gas. Improving fuel efficiency meant that our consumption of oil did not grow nearly as fast in the 2000s. As you can see, both of these are supply-side factors i.e. a shifting out of the AS curve.

      Ganesh

      Delete
    6. Ganesh, what about other countries? Look, we can all agree that an oil shock is an adverse supply shock that should affect output and employment. But it is not at all clear why it should be strongly correlated with inflation, UNLESS the Fed tries to accommodate the shock. You show me a model that can explain how inflation depends on oil prices in conjunction with use in power generation that can match both US AND international data (did countries more reliant on oil experience high inflation in the 2000s?) and I will think about it. But to arbitrary proclaim that oil prices mattered pre-1985 and not after because that would be convenient to your story, well, you are assuming your conclusions. And that's bad science.

      Delete
    7. John, what are your thoughts about Nick's comment?

      Delete
  5. Whatever is the right theory, the facts on the ground were that even rising double-digit inflation subsided, evaporated, or retreated, or was beaten down by Volcker, rather quickly (even as Reagan was running primary deficits btw. Reagan moved to primary surplus only in year 7 or 8 of his Presidency).

    For me, this raises the question: What is even moderate inflation today regarded as an uncontrollable boogyman? There is hysterical prissiness afoot regarding even microscopic rates of inflation, usually under the banner that any rate of inflation threatens to "accelerate," "gallop" or turn into hyperinflation.

    The historical record shows good solid US growth with moderate inflation (3+% real growth, 3-% inflation 1982-2007), and that even escalating (there's a good scare word also) rates of inflation (late 1970s to early 1980s) are reversed quickly through tighter monetary policy.

    Gee, this is recent historical record, not theory.

    Gadzooks! The discussion today should be about "stimulus" and "expansion" and "job growth" and "real growth" and "we want boom times."

    Seems to me the Fed needs to gun the presses long and hard (QE at $50-$100 billion a month) until we see Japan-style rates of unemployment. We might, or might not see inflation---they are not in Japan.

    If we do see some inflation, so what? 3% is very livable, as recent history proves. If we see no higher inflation, all okay too.

    If we see higher inflation, say above 5%, it appears a moderation in monetary policy would retard inflation. Meanwhile, a lot business es would have made money, a lot of people would have gotten jobs, taxes could be cut thanks to higher revenues, trading partners would do better---are there any downsides to robust economic growth?

    Yes---central bankers do not like it.


    ReplyDelete
    Replies
    1. "For me, this raises the question: What is even moderate inflation today regarded as an uncontrollable boogyman? There is hysterical prissiness afoot regarding even microscopic rates of inflation, usually under the banner that any rate of inflation threatens to "accelerate," "gallop" or turn into hyperinflation."

      This is an interesting interpretation of those events. The inflation boogyman that was beaten down easily meant above average UE rates from 1975-1988. Even just starting in 1981 (Volcker took office in August 1979 so we are subtracting a year plus to get the best possible result) you still have 7-8 years of above average UE- and these were "boom" years according to many. Trying to avoid a decade averaging 7-8% UE is somehow "hysterical prissiness".

      Delete
  6. So if you make the series less noisy using core inflation, it looks like Tobin's stylized nautilus played out in ten years instead of twenty. That speed seems to be the main point of this post.

    A couple of tweaks to Tobin's assumptions (NAIRU, etc.) seem like they could easily explain the speed with which these economic effects played out.

    ReplyDelete
  7. Cochrane: "[Krugman asserts that] late 1970s Keynesian macroeconomics with adaptive expectations was vindicated in describing the Reagan-Volker era disinflation."

    I don't read Krugman that way. Instead, he often defends a middle ground between vindication and refutation. Such statements as "Keynesian economics is looking pretty good" have moderators and qualifiers built into them.

    Krugman is certainly open to the idea of rational agents who look both backward and forward. He is, however, opposed to ideological extremism, which seems to be in the nature of some academic purists.


    ReplyDelete
    Replies
    1. "He is, however, opposed to ideological extremism..?" Well, thanks for the laugh.

      Delete
  8. Krugman's rhetorical style is often extreme. When important people refuse to retreat from what he thinks are demonstrably flawed arguments, his criticism and scorn can be withering, to the anger of many.

    I see ideology but not extremism in his economics, however, and a sense of social justice that I admire greatly.

    ReplyDelete
  9. I looked at Britain's inflation/unemployment rate. Britain also managed to get inflation down pretty fast, but unemployment took much more time to get down, with the trough reached only around 1990. Perhaps the model fits better there?

    ReplyDelete

Comments are welcome. Keep it short, polite, and on topic.

Thanks to a few abusers I am now moderating comments. I welcome thoughtful disagreement. I will block comments with insulting or abusive language. I'm also blocking totally inane comments. Try to make some sense. I am much more likely to allow critical comments if you have the honesty and courage to use your real name.