What’s wrong with Economath ?

Well here goes. I am going to claim that Paul krugman’s work on imperfect competition, increasing returns and trade was a step backwards from the earlier literature. I will attack Krugman from what is considered the left. In any case, I am definitely criticizing him for excessive orthodoxy. I base my case on the work of notorious red Larry Summers (OK really on the work of the less notorious red Larry Katz). This radical way beyond Krugman stance is a description of a paper written by my two PhD supervisors.

Krugman (my bold)

Yes, there was an old tradition arguing that increasing returns made the case for infant-industry protection; underlying this tradition was the intuitive notion that increasing-returns industries are the “good stuff” you want to get, and keep away from other countries.

But that was not at all the point of New Trade Theory, which ended up suggesting that concentration of production due to increasing returns is generally beneficial to importers as well as exporters of increasing-returns goods, that it generally reinforced the case for open trade, rather than undermining it. (See the summary here (pdf)). And while there were a few people who got that point before the models — Bela Balassa was right on point — there weren’t many. It turned out — and still turns out — that people’s economic intuition, if untutored by models, missed a major possibility that is in fact probably the main story.

Ah “probably” may be a word rather like “might”. The discussion might be improved if people were required to write a number when they write probably as in “probably (by which I mean the probability is at least 70%)”. Most of the time the word is used, it has nothing to do with mathematical probability. Basically it has two meanings “plausibly” and “approximately certainly”. Roughly it is used to argue that if I can explain how something might be true, I can assume that it is true and consider the claim it is false to be an error.

Note how Krugman’s claim that a little math improves thought is all argued with weasel words (bolded). There is no reference to evidence. There is certainly no claim that no model can be written in which the older view makes sense. Such models have been written (for example by Krugman).

My main objection is that Krugman’s models included the assumption that the labor market clears. This means there is full employment (because the question being addressed is not unemployment) but also that the same worker would get the same wage if he or shee worked in a different sector. One mostly harmless assumption is that there are no compensating differentials (all work is equally unpleasant). The key assumption is that there aren’t labor market rents — that there aren’t good jobs which pay good wages and bad jobs which pay bad wages (so either workers are indifferent between working in an auto plant and a McDonalds or McDonald’s workers lack some skill needed to assemble autos).

This assumption is absolutely standard in the economics literature, with one exception. No ordinary person doubts that there are good and bad jobs and workers in bad jobs qualified to do the good jobs. The glaring exception in economics is the literature on wages and, in particular, sectoral wage differentials. That literature shows evidence for the non-economist’s view which overwhelms everyone except for the genius (Kevin Murphy) who can reconcile any possible data with a market clearing model.

Now imagine a model in which some jobs are better than others because they pay higher wages. Imagine that jobs in heavy industry are better than jobs in light industry (steel mills vs textile mills) In this case, infant industry protection can benefit the protecting country, because it gets more of the good jobs and less of the bad jobs. This is the conventional idea of the economically illiterate man in the street. The key issue is whether jobs in high wage industries are better for the same worker (they would not be if they just compensate for say greater risk or if workers in low wage sectors are not capable of doing the jobs in high wage sectors). The evidence is overwhelmingly on the side of the economically illiterate man in the street.

This is not a new argument, nor is it one that hasn’t been made by famous economists. In this post I can’t pretend to be a daring radical breaking with the orthodoxy which I was taught, because the paragraph above is a summary of a small part of the work of my PhD supervisors Larry Katz (who did most of the work) and Larry Summers (yes that Larry Summers)

Can Inter-Industry Wage Differentials Justify Strategic Trade Policy?
Lawrence F. Katz, Lawrence H. Summers
NBER Working Paper No. 2739 (Also Reprint No. r1308)
Issued in November 1989
NBER Program(s): ITI LS IFM
This paper examines the relationship between labor market imperfections and trade policies. The available evidence suggests that pervasive industry wage differentials of up to 20 percent remain even after controlling for differences in observed measures of workers’ skill and the effects of unions. Theoretical analysis indicates that given non-competitive wage differentials of this magnitude policies directed at encouraging employment in high-wage sectors could significantly enhance allocative efficiency. For the United States and other developed countries, such policies are more likely to involve export promotion than import substitution. Increased international trade flows (at least through 1984) have been associated with increased employment in high-wage U.S. manufacturing industries relative to low-wage U.S. manufacturing industries.

I note in passing the weasel “more likely” hmm Larrys how much more likely ? at least 20% like the wage differentials ? I also note that the question answered “yes” (with data) by Larry & Larry is on the Booth economists questionaire “Question A: The federal government would make the average U.S. citizen better off by using policies that directly focus more on increasing manufacturing employment than employment in other sectors.” The results of the poll 5% agree 51% disagree and 15% strongly disagree. There is view widesread among non economists. The data say it is correct (as far as the data can ever say anything). There is a very clear consensus among economist that it is not correct. In this case the conventional assumption totally stomps the results of analysis of the relevant data.


Economists (other than the one’s who study wages) insist on an assumption which most people find implausible and which has repeatedly led to testable predictions which have been rejected by the data (OK “lead” is a weasel word since always they hypothesis of interest must be combined with auxiliary hypotheses to be tested — this isn’t a problem with the inter-industry wage differential litarature it is a problem with science). Why ? I see three reasons
1) economic debates are settled by authority. A very famous economist made an assumption for convenience. Over time, this becomes a dictate which must be obeyed (the Dixit dixit principle).
2) admitting that economists might be wrong and the man in the street might be right is intollerably humiliating
3) shopping for Occam’s razor at the convenience store. It is understood that any data can be fit (an nothing explained) if one relaxes standard assumptions. Therefore the economists version of Occam’s razor is to fit the facts currently being discussed with the fewest possible deviations from standard assumptions. This may be a necessary rule to make a theoretical discussion work at all. But it gives great power to clearly false assumptions made for the convenience of the economists who first tried to think about the issue

OK so that’s one argument that, I think, justifies the claim that non-free trade policy can increase money metric “welfare” in a country. I am very confident that this is true (I am willing to offer 10 to 1 odds). Also the claim is useless. The fact that there is a non free trade policy does not mean it would be implemented if the pro free trade dogma were relaxed. I am also confident that actually possible deviations from free trade would hurt the USA (OK about 2/3 1/3 on that one). Also, and conventionally for non econmists too. I think protection would lead counter protection and end up bad for everyone.

Here’s another argument. Marshallian spill overs. Here the idea is that there are some industries where being near your competitors is good, because people learn from each other. This can explain the extreme geographic concentration of industries even in cases where access to natural resources isn’t key (that means not just why are steel mills near coal). This argument was made by Marshall, but it was mathed up and rediscovered by uhm Krugman. The key point is that the spillovers are an externality, so the market outcome is not Pareto efficient. Marshallian spillovers provide a justification for infant industry protection.

Krugman’s work on economic geography strongly strongly supports the case for infant industry protection. The argument that knowledge spillovers justify infant industry protection was made vaguely and with words by Schumpeter.

Again and again there have been plausible sounding arguments which don’t work when you do the math — with standard assumptions. Then decades later they do work when you do the math after relaxing an assumption which was always known to be false.

Finally a challenge. Krugman quotes straw man ““Well, government borrowing drives up interest rates, and higher interest rates depress private investment, so increasing government spending in a slump is actually contractionary.” People who try to do economics without any kind of mathematical modeling do indeed say things like that — and it’s very hard to explain why it’s self-contradictory nonsense without a bit of math.

In fact standard models don’t have this effect. The final impact on say GDP of something like a government spending increase has the same sign and smaller magnitude than the impact holding other endogenous variables constant. This is pretty much a requirement if one wants determinate equilibria (so the model yields forecasts). Macro economists don’t like models with indeterminate equilibria (assumptions are often explicitly added to prevent indeterminacy). Note the conclusion follows from what macroeconomists like and dislike. It is simply just not true that the claim that government spending causes lower output by causing lower investment is logically “self-contradictory”.
It is possible to write down such a model and make the math work (in the end real interest rates are lower too). Alessandra Pelloni and I have published such a model

Consider case 2 and assume as is standard in these cases that the spending will be financed with lump sum taxes.

Here the conclusion is the direct result of assumptions made, because they are convnetional and needed to prevent embarassing results. Based on two examples, I claim that this is always the contribution of economic theory to economic thought.