Site Meter

Tuesday, August 18, 2015

Romer, DeLong and Krugman on Solow and Lucas

I can't resist typing this post. I am very sure it is not worth reading (that's why it's here not at

Paul Romer, Paul Krugman and Brad DeLong are having a polite inside baseball discussion of what went wrong with Macro in the 70s and early 80s. I agree with all three that it is probably time to move on to how to fix macroeconomics in the 2010s. Also civility is good. In particular, I think it would be useful to the reality based community to make sure Paul Romer feels welcome (I note that I get paid largely for trying to explain Paul Romer to students). However, I feel free to be as rude as I want to be on this blog. I don't share Romer's beliefs about the history of thought including his beliefs about the very recent thoughts of Paul Krugman and Brad DeLong.

I am to comment Romer's post brilliantly entitled "Solow's Choice".

update 2: Paul Romer himself was kind enough to answer my question for Paul Romer in a comment on that post ! I am thrilled.


Re your question, I hope that my posts about Solow's remarks in 1978 answer your question. But to be specific, I think that Solow did depart from the role of the scientist by using debating tactics to dismiss the critique by Lucas and Sargent. And I suppose he did not live up to Feynman's mandate in the sense that he does not acknowledge the problems that the big simulation models suffered from. But so did Lucas and Sargent, I suppose, by pushing the policy ineffectiveness result prematurely. Feynman integrity sets a pretty high bar.


end update 2

update: Romer has uploaded the conference volume here so you can read Solow's chapter as I just did.

end update:

Before wasting the time of the reader (if any) I state my conclusions.

1) I think that Solow made the right choice. It is true that academic macroeconomists (including those at MIT) almost all disagreed. I think subsequent events show that Solow was right.

2) I think that Brad DeLong absolutely agrees with Solow. I don't know much about much, but I know a lot about the thought of Brad DeLong. I think Romer's guesses about DeLong's beliefs are incorrect.

3) Romer makes guesses about Solow's motivation. I make different guesses (knowing much less).

Romer wrote

In the summer of 1978, Lucas and Sargent were making three claims:

(a) Existing multi-equation macro simulation models were not identified. That is, these models summarized correlations in the data but did not yield reliable statements of the form “if the government does X, this will cause Y to happen.”

(b) It was time to use SAGE models to address such fundamental questions about economic fluctuations as why changes in the supply of money influence economic activity; and

(c) SAGE models will imply that an active monetary policy cannot stabilize economic fluctuations.

Solow thought that Lucas and Sargent were wrong about the policy ineffectiveness claim (c). DeLong, Krugman, and I all agree. In the 2013 introduction to his collected papers, Lucas uses some asides about the Great Depression and the Great Recession to admit that now even he agrees. Claim (c) is what DeLong and Krugman have in mind when they say that Solow was right and Lucas was wrong.

Romer agrees with (b) but doesn't explain why he thinks that macroeconomic models must be SAGE models. I don't know if he considers the assumption that the economy is in general equilibrium a falsifiable hypothesis or not. I think if he thinks it is, he must consider the possibility that it shall be falsified (he wrote that theories must bow to facts). If it isn't, I don't see what is gained by assuming it. I think that GE implies the assumption of rationality. this is certainly true of the DSGE model presented by Arrow and Debreu 1954 and it is also assumed in all contemporary applied DSGE models of which I am aware. But this is an assumption which we can be sure is false and which is defended because it might be useful. I have never understood how anyone can argue that one must assume something because it might be useful. Lucas certainly insisted on the assumption of rational expectations. Solow certainly thought that it was a very bad idea to make that assumption (as did Friedman).

Personally, I think that the new Keynesian SAGE program has been sterile and that it was a mistake to start on it back in the 1970s.

I am quite sure that DeLong thinks Solow was right in rejecting (b). I am not expert on much, but I am quite expert on the thought of Brad DeLong. In 1983 he said that he was going to do fields exams in economic history and econometrics and not macroeconomics because he thought that macroeconomics had headed down a blind alley which it wouldn't leave for decades. I think subsequent events suggest he was right. Importantly, he knew about early new Keynesian work at MIT. He didn't like the rational expectations hypothesis (he is quite famous as a critic of it). Also more recently he wrote

But then Mike Woodford and company lost sight of the goal. Yes, New Keynesian models with more or less arbitrary micro foundations are useful for rebutting claims that all is for the best macro economically in this best of all possible macroeconomic worlds. But models with micro foundations are not of use in understanding the real economy unless you have the micro foundations right. And if you have the micro foundations wrong, all you have done is impose restrictions on yourself that prevent you from accurately fitting reality.

Thus your standard New Keynesian model will use Calvo pricing and model the current inflation rate as tightly coupled to the present value of expected future output gaps. Is this a requirement anyone really wants to put on the model intended to help us understand the world that actually exists out there? Thus your standard New Keynesian model will calculate The expected path of consumption as the solution to some Euler equation plus an intertemporal budget constraint, with current wealth and the projected real interest rate path as the only factors that matter. This is fine if you want to demonstrate that remodel can produce macroeconomic pathologies. But is it a not-stupid thing to do if you want your model to fit reality?

Krugman recently wrote

I’m actually mainly with Waldmann on this one, although Wren-Lewis’s analysis is nonetheless very useful. For the point he makes about the implications even of perfectly well-informed and rational consumers was and as far as I know still is totally misunderstood by freshwater economists [skip]

But aside from exposing the intellectual decline and fall of the Chicago School, is this the way we should go about modeling such things? Well, yes, sometimes, because rigorous intertemporal thinking, even if empirically ungrounded, can be useful to focus one’s thoughts. But as a way to think about the reality of spending decisions, no. Ordinary households — and that’s who makes consumption decisions — have no idea what the government is spending, whether it is temporary or permanent, whatever.

This is absolutely a critique (indeed a contemptuous dismissal) of claim (b) and not at all a comment on claim (c) .

I note that Wren Lewis agrees

I think it is clear that DeLong rejects the SAGE program and not just the policy ineffectiveness proposition. I don't claim to completely understand Krugman's view, but it seems to me to be very different from Lucas's and quite different from Romer's

Romer guesses that Solow dismissed Lucas and Sargent because he was worried that policy makers would take the policy ineffectiveness proposition seriously. I must stress that this is just a guess. Solow might have just had the impression that Lucas's approach was crazy. I my experience, most people do have that reaction.

Romer criticizes Solow for not writing down a model where fear of worker's anger prevents employers from cutting wages. In particular, he asks for a SAGE model with downward nominal wage rigidity. Solow did re-introduce the idea of efficiency wages and explicitly motivated the assumption that wages affect productivity through morale. I think Romer criticized Solow for failing to do what, in fact, Solow did (or perhaps for failing to work out a model during a conference).

Journal of Macroeconomics Volume 1, Issue 1, Winter 1979, Pages 79-82

Another possible source of wage stickiness


A number of hypotheses have been advanced to explain wage stickiness. This article explores another reason why wage stickiness might be in an employer's interest: the relationship between productivity and the wage rate. If the wage enters the short-run production function, a cost-minimizing firm will leave its wage offer unchanged, no matter how its output varies, if and only if the wage enters the production function in a labor-augmenting way.

A free pdf is available here

In the text of the paper, Solow explicitly refers to morale

Romer's problem might be that Solow didn't put his model of wage stickiness in a SAGE model but cited Malinvaud instead. Note the article is published in 1979 a very brief delay after 1978. Solow did send it to a low ranking journal (it wasn't rejected by a higher ranking journal -- decades later Solow had not ever had a manuscript rejected).

really pointless stuff after the jump

Romer hypothesized that Lucas ceased to engage with salt water economists because Solow dismissed his work with sarcasm and contempt. However, he admirably notes strong evidence against his hypothesis

Prior to 1978, Lucas had already used the kind of unhelpful debating language that I criticize Solow for using. Lucas wrote that a seminar audience would respond to Keynesian theorizing with “whispers and giggles.” His remarks were published in 1980, but in the introduction to his collected papers, Lucas indicates that they are from a talk he gave shortly after coming back to Chicago from Carnegie Mellon, which implies some time during the 1975-6 academic year. (Below, I’ll provide some context for this remark.)

Also Romer quotes Solow quoting Lucas and Sargent

Let me quote some phrases that I culled from their paper: “wildly incorrect,” “fundamentally flawed,” “wreckage,” “failure,” “fatal,” “of no value,” “dire implications,” “failure on a grand scale,” “spectacular recent failure,” “no hope.” Now if they were doing that just to attract attention, for effect, so that people don’t say “yes, dear, yes, dear,” then I would really be on their side.

He wasn't the first person at the 1978 conference to be uncivil. It is very odd to date the breakdown of normal scientific discussion at the talk which includes those quotations of things which were said earlier.

I think the hypothesis doesn't fit the facts.

I also note that when discussing how intellectually flexible Lucas was, Romer considers the writings of Sargent. They are different people.

Romer criticizes the large macroeconomic models criticized by Lucas and Sargent (the quoted claim (a)). Here I think that DeLong and Krugman do agree with Romer, Lucas and Sargent. Not that anyone cares, but I also agree. I think he should note two things. First Solow didn't work with such models. Second a grade of B or B- is not consistent with the claim that nothing much is wrong with the models. Solow always worked with small simple models. I don't think it is fair to lump him in with Fair. In any case, it just isn't true that the choices are those models or SAGE models. Solow used neither when discussing economic fluctuations.

I have only one objection to what Solow said in the passage quoted by Romer "identifying restrictions, whatever those words mean". Huh ? Why did he say that ? Surely Solow understood the importance of identifying restrictions. This is shown by ... damn I can't think of anything.

update: I think Solow's understnading of the importance of identifying restrictions is shown by something else he said at that conference

a modern disease, which a lot of papers suffer from, a tendency to build too much on a very thin econometric basis. In a complicated nonexperimental statistical situation, there are almost always several hypotheses which fit the data approximately equally well. We have hardly any way of distinguishing confidently among them. Even those horse races that everyone talks about don’t really do the discrimination job very well, again because of that ever-present possibility of patching up a mode! after the fact. The time-honored device in laboratory science for solving this problem is the controlled experiment - the critical experiment. We can’t perform experiments; so it is only prudent to be very leery of claims based on one or two t-ratios or on small reductions in standard errors of estimate. The significance tests we use have very little power against the next best competing alternative and I fear we tend often to forget that.

The shocking comment "identifying restrictions, whatever those words mean" appeared in the context of a discussion of how strange the conference would be to non-economists. It was also clearly a joke.

Solow also said

There is a very valuable and important point which is in very large part due to Lucas and Sargent, and one must give them credit for it, that what often looks casually like a change in structure is really the economic system reacting to its own past. It is possible that what happened between the 1960s and the 1970s is a kind of loss of virginity with respect to inflationary expectations.

And this in the discussion which Romer guesses caused the rupture in macroeconomics.

I also note the discussion of a change in inflation on change in unemployment relationship which might not yield any correlation between inflation and unemployment. This fits the data which were to be collected in Europe in the decades after the talk. He cites the speaker who pointed this out. This is what's called hysteresis. It does not imply a vertical long run Phillips curve -- there is no curve -- the set of points is two dimensional. This is 7 years ahead of the field

Geoffrey Moore’s findings. I would not regard Geof’s findings as very optimistic for received macro theory, either. What he reported is that the rate of inflation appears to accelerate when the unemployment rate falls in the upswings of growth cycles and to decelerate when the unemployment rate is rising. Franco pointed out yesterday, quite correctly, that this almost says that the rate of inflation is high when the unemployment rate is low, and low, when the unemployment rate is high. But it is not quite the same thing. If I can draw a diagram on a nonexistent blackboard for you, you can have a curve which moves through time up to the left, one measurement rising while the other is falling, and then comes back not right down the same curve, but at a slightly higher level, and then goes up again at a still slightly higher level and then comes back down again, once more at a higher level. It will always be true that x is rising while y is falling and y is rising while x is falling, but if you combine all those points you get a scatter that has essentially zero correlation as a whole. Clearly that sort of thing can and does happen. Geof Moore’s paper is certainly a problem for Lucas and Sargent but it is not an unmitigated blessing for the traditional macroeconomic view.

Solow also notes that downward nominal rigidity implies that inflation is required for relative wage and price changes. This is the basis for models of long run downward sloping Phillips curves developed in Akerlof, Dickens and Perry (1996) .

there is an asymmetry between upward and downward flexibility in wages and in many other prices as well. Then it almost automatically follows that there is a kind of inflationary bias in the system because the only way the system can generate the relative price changes that it has to bring about is by having the general price level float upward

In contrast, I noticed almost no defence of large macro models. Solow mentions them in passing but discusses mainly wage rigidity -- he is thinking of a very simple model. It is not a SAGE model because he asserts that people have bounded rationality and that sociologists have a point -- at least when they discuss the employment relationship. I really don't understand Romer's view of Solow's contribution to the conference.

end update:

Solow also made a radical critique not only of Lucas and Sargent's approach to understanding fluctuations but also of Samuelson's approach to microeconomics and almost every paper that Solow ever wrote (not including Samuelson and Solow 1960). He argued that the assumption of rationality is not useful

The assumption that everyone optimizes implies only weak and uninteresting consistency conditions on their behavior. Anything useful has to come from knowing what they optimize, and what constraints they perceive. Lucas and Sargent’s casual assumptions have no special claim to attention.

This is a critique of all economic theory then (or now) existing. I also note that it is a mathematical claim -- it is a claim about the almost complete lack of implications of an assumption not about how anything relates to the real world. It is not just a casual sarcastic dismissal. It is the statement of a well known theorem. I disagree only with the claim that the assumption implies any consistency conditions. This is not true.

The proof is trivial. Imagine a person whose sole aim in life is to act in a many inconsistent with utility maximization. If there were any actions inconsistent with utility maximization, one would be utility maximizing for this individual. This is a logical contradiction, so there must be no action inconsistent with utility maximizing. QED

Solow aknowledged that the assumption of rational utility maximization has no implications in a talk to incoming MIT graduate students in I guess it would be 1981 (Larry Katz told me this).


Anonymous said...

"....In 1983 he said that he was going to do fields exams in economic history and econometrics and not macroeconomics because he thought that macroeconomics had headed down a blind alley which it wouldn't leave for decades."

He was right , but I propose that people like DeLong and Krugman knew a lot more about what was going on than they shared with the rest of us.

That wasn't a dark alley they went down. Economists had the high beams on , along with multiple roof-mounted spots and floods. That alley had a nice , big , fat dead-end for them to back-and-fill within - for decades , as DeLong said. But it wasn't a wrong turn that got them into that alley. They knew exactly where they were going. A rare few , like Romer , are - belatedly - feeling some pangs of guilt for not speaking up about it. His recent "revelations" don't begin to make amends , IMO.

The only fairy tale of recent times that comes close to matching that of the economics profession is the one we're told about America "spreading democracy" in the Mideast and elsewhere.

Shameful deceptions , in both cases. Also criminal , if I had my way.


Robert said...

I think that Brad Was quite outspoken. He was very far from orthodoxy and his career temporarily suffered as a result. I don't know Krugman. I guess he didn't feel his field was going down a blind alley. I think we have the case of someone who didn't go down the nice big fat dead-end (brad) and someone who seems to have changed his mind about the usefulness of economic theory. I don't see any deception.

Again I don't know Krugman, but I am quite sure about Brad.

marcel proust said...

Irrelevant to the main point of the post...

"this is certainly true of the DSGE model presented by Arrow and Debreu 1954 ... "

How is AD(1954) a DSGE model? I don't recall anything D or S about it. My recollection is that it requires that each agent know every possible state of the world; does it also require knowledge of the probability of each state? I don't recall that, and w/o that, isn't stochastic not part of its structure? Also, in what sense is it dynamic? As I learned it, "every possible state of the world" means "every possible sequence of states", but I don't recall anything about probabilities and stochasticity: I thought that that was Lucas's contribution.

Robert said...

In Arrow and Debreu 1954 agents maximize expected utility by trading assets with state specific payoffs. The agents must know the probability of each state to do this. The model is an inter-temporal stochastic model with rational expectations.

So what was Lucas's original contribution ? Beats me.

here are two relevant wikipedia articles

see what is dynamic and stochastic here

So the complete Arrow–Debreu model can be said to apply when goods are identified by when they are to be delivered, where they are to be delivered and under what circumstances they are to be delivered, as well as their intrinsic nature. So there would be a complete set of prices for contracts such as "1 ton of Winter red wheat, delivered on 3rd of January in Minneapolis, if there is a hurricane in Florida during December"

"It was standard to assume rational expectations in game theory from Von Neuman and Morgenstern on and in general equilibrium theory from Arrow-Debreu 1954 on."

Now it is obviously insane to assume that markets are complete. In the standard DSGE models used in macroeconomicw, markets are assumed to be complete.

If markets are incomplete then, for an open and dense set of economies, a Pareto improvement can be achieved by restricting the choices agents rationally make when trading on perfectly competitive markets with price flexibility and market clearing. In other words, the first welfare theorem is an irrelevant extremely special case with, according to current general equilibrium theory, no relevance even assuming rational expectations, perfect competition and market clearing.

Fiura said...

Amazing! In this short replay to a comment there are tons of microfoundations of macro.

Anonymous said...

In the 1970s was a view that a microscopic approach in macroscopic theory was essentially wrong; it distracts scarce priority attention time and energy (TAPE). For that matter, the extent of the US-model already was a "their party."

Dutch uncles expressed it with their particular charm of course. No blesse, no blige.
An industrious staff-employee who elaborates the management-summary to 1776 pages gets a three word reaction.
In Dutch "foundation" is ladies underwear. Now picture a macroeconomist reporting in Amsterdam (close to the red light district) that in the USA research for microfoundation were hot and sexy. It happened.
One also can imagine the maritime tradition brought the remark that the noisy fleet of empty vessels makes the wider swings.
Read it with both eyes. The economice also accused the Vikings of a scandy navy; just envy.

In essence, macroeconomics ought to be macroscopic, and another discipline can cover some of the issues. Now we have expertise on how to run production and how to run projects; not all expertise, but some. In the course of an approach marginal utilityproductivity of that approach sinks relative to another approach, and so you can get the pattern of natural motion, a chain-reaction of interrupted falls. Equilibrium is horizontal, as in grave, and walking is vertical, and it swings. This is no excuse for drunks. There is overhead for a production run, so there is a kind of minimum duration for efficiency. (It is known from Adam Smith's pinmaker.) A career-term, some 40 years, sounds good. We then would get a Nebo-syndrome. Till 1939+40=1979 unemployment weighed heavily and in 1923+40 an inflation fear retired. Around 9/11 yield requirement returned to historically regular, but the 1970s trauma was still there. Man, there are nails allaround!!
I better can focus on economies than on economists; comparative costs.


Anonymous said...

Dear Robert Waldmann,
Have you commented on the new ISTAT model yet? Do you intend to? I would be curious to hear your thoughts.