Consider these two statements:First and in passing, I note that, after Einstein was introduced, "economic theory" was replaced by "scientific progress" . The implicit assumption is that there is some overlap between economic theory and scientific progress -- that economic theory has progressed. This view is not universally accepted. I am pretty sure that I have a problem with question A. For statement 1 to be true, "path breaking" has to be well defined.
1. The model in Lucas (1972), Expectations and the Neutrality of Money, made a path breaking contribution to economic theory. It is comparable in importance to the Solow model and the Dixit-Stiglitz formulation of monopolistic competition.
2. The model in Prescott and Kydland (1982), “Time to Build and Aggregate Fluctuations”, has no scientific validity.
Next, consider these two statements:
3. Einstein’s model of the universe based on his theory of General Relativity, made a path breaking contribution to theoretical physics, even though in his first application, Einstein built a model of a steady state universe.
4. Models of a steady state universe have no scientific validity.
In each case, the first statement in the pair is about the contribution of a mathematical model to scientific progress. The second is about the empirical validity of a specific model.
[skip] Here I want to point to a different indicator. Set aside the question of whether or not I am right that 1 and 2 are true. Think of some macroeconomist X that you know. Consider these questions:
A. Would X agree that there is an objective sense in which statements 1 and 2 can be said to be either true or false?
B. Would X agree that a reasonable person could conclude that statements 1 and 2 are both true?
C. Would X be able to examine dispassionately the evidence for and against these two statements and evaluate them independently?
A useful indicator of the degree to which macroeconomics has been infected by tribalism might the fraction of macroeconomists for whom the answers to at least one of the questions A, B, and C would be no.
regarding Romer's statement 1, I agree that, when originally presented, the Lucas 72 model, the Solow growth model and the Dixit Stiglitz example had similar scientific status. They were "path breaking" in that they left the existing path -- they were new and different. They were path breaking in that they were highly influential. My problem is that I don't see how that relates to "scientific progress". It was definitely scientific change, but only time could tell if it was an improvement or a worsening.
I also certainly agree with Romer's statement 2. I agree that a reasonable person could conclude that statements 1 and 2 are both true (I don't but I can see how a reasonable person might think that). I can evaluate the second independently from the first, but I need to understand what "path breaking" means to evaluate the first.
Before going on, I'd like to stress that Romer definitely did not compare Lucas '72 and Dixit-Stiglitz to General Relativity -- he compared it to the Solow growth model and the Dixit Stiglitz example of imperfect competition. However, I will contrast them. General Relativity explained an anomaly -- the precession of the perihelion of Mercury. It implied a prediction about how much gravity caused light to curve which was striking shocking and soon confirmed. Since then it has yielded a huge number of predictions which fit the data exactly (so far). It was easily modified to correspond to an expanding universe as the first formulation did, when this was pointed out Einstein added a fiddle factor to reconcile the theory with a steady state universe. Pysicists are quite sure general relativity is not the truth (because it is inconsistent with quantum mechanics and therefore a lot of data). But it is a very empirically successful theory.
In contrast, the Dixit-Stiglitz example did not attempt to explain anomalies not fit by earlier models of imperfect competition. The aim was to make models with imperfect competition tractable. The formulation is an example, and not one considered unusually plausible. They made a modelling choice not a hypothesis (neither would guess that people might actually have Dixit-Stiglitz preferences). Here I think the key cause of the enormous influence of the example was that it meant there was a standard way to handle imperfect competition.
Unfortunately, this is important not because other models are all intractable but because there are no general results. Models with imperfect competition can have a sunspot equilibrium with fluctuations which are not caused by shocks to taste and technology (this can occur if different goods are strategic complements). The set of equilibrium can be huge -- a multidimensional continuum. Together the assumptions of imperfect competition and Nash equilibrium imply almost nothing. The example made it possible to have the illusion that economic theorists understood imperfect competition, but this was discovery by assuming we have a can opener.
The example was fruitful because, once a lot of people decided to explore the same special case, they could discuss its interesting behavior. The fruits however, do not include any good reason to exclude the other problematic cases in which different goods are strategic complements. Theory can grow if people agree on core assumptions. This is progress if the assumptions are useful approximations. Once a field of economic theory has developed, its core assumptions are no longer vulnerable to data. I do not think the the development of a new branch of theory is necessarily scientific progress.
I think economic theory was massively improved by the Dixit-Stiglitz example, because it made economists outside of industrial organization willing to consider imperfect competition. But I think this can be seen as an accidental trick. It gave the impression that there were simple elegant results based on assuming imperfect competition similar to those based on assuming perfect competition. There aren't. Here I cite major Dixit-Stigitz user Paul Krugman
After a while, the new approaches came to seem too liberating; by the early 90s the joke was that a smart graduate student could devise a model to justify any policy. And while some important new theoretical work continued to be done, for example the Melitz work on heterogeneous firms or the Eaton-Kortum work on bilateral trade flows, I think you have to say that the field got tired of clever theorizing and wanted data instead.
I think the point is that this excessive liberation was already implicit given the acceptance of imperfect competition (and the emptiness of theory without data could conceivably have been recognized as soon as economists admitted that they couldn't prove that competition really is perfect, that is over a century ago). By the way, I heard that joke told by Robert Barro in 1988 or 1989 so before the early 90s.
There have been dead ends in natural science. Organic chemistry was once defined as the consideration of how the laws which governed chemical reactions inside living things were different from the laws which governed chemical reactions outside of living things. A Nobel prize in physiology and medicine was awarded for a theory of cancer propagation which is now believed to have no relationship to reality. Lamarckian evolutionary biology survived into the 20th century (and not just in the USSR).
A new branch of mathematics must be a contribution to mathematical progress (perhaps a small and boring one). But a radically new hypothesis which turns out to be totally false was not a contribution to scientific progress. Finding out that it was false was and such dead ends are inevitable in science. But in science development of new theory is not necessarily progress.
OK what about the Solow growth model and the Lucas supply function ? Like Dixit and Stiglitz, Solow mainly made a large number of extreme assumptions yielding a tractable model. Here I think Solow's assumptions were fruitful also in that they fit the data surprisingly well. There are excellent arguments for why one shouldn't be able to treat capital and labor as scalers (single numbers). But empirically, the Solow radical simplification fits the available data surprisingly well. There was no reason to think that the concept of disembodied technology would be useful. But it helps economists fit the data.
Lucas formalized an argument about price level missperception and fluctutations which had been made many times (for example by Keynes in "The General Theory" as Tobin explained to Lucas in 1971). Here again the theoretical change was to assume everything else away. In particular the Lucas model abstracts from the wage system and is population by self employed "suppliers" -- this was an extreme assumption at a time when trade unions were powerful even in the USA. Lucas definitely did not identify a previous conceptual error of treating the expectations unaugmented Phillips curve as a stable relationship -- this is a myth. He added a focus on expectational errors alone and the insistence that economists assume rational expectations. I think this too is unlike Dixit-Stiglitz. The reason is that, in 1972, the Lucas model was obviously grossly false -- it implies that output is a white noise and it was well known that economic fluctuations aren't. it requires that agents have very limited information on the price level when, in fact, they have a lot of information. I think it was clear that Lucas's new research program would be sterile. I think it was entirely sterile. I count new Keynesian models as part of Lucas's intellectual legacy (even though he never recognised the bastards). Here I think nothing was explained by the new models which hadn't been explained by the old models and economic theory did not progress at all. It grew and is a richer branch of applied mathematics, but I think the right direction to go now is back to before Lucas 1972.
I certainly don't think this of the Solow growth model or imperfect competition with a Dixit-Stiglitz preferences or a Dixit-Stiglitz aggregator.
Yes Lucas 1972 was path breaking, but the new path Lucas blazed lead to a dead end.