An Opinion II

This is my second comment on Noah Smith’s substack. Today he wrote about something I should know about — macroeconomics. As always, I am amazed by Noah’s knowledge (it is his former former field of academic research). I don’t know for sure if his article is available only for subscribers, so I will try to summarize a little.

His thoughts:”the U.S. is confronting another unexpected macroeconomic shock. Although the economy is doing great in terms of jobs and growth, it’s plagued by worryingly high inflation,” is it due to a demand shock or a supply shock. Although economists use much fancier models we can understand this with a model with an upward sloping aggregate supply curve and a downward sloping aggregate demand curve. This suggests the issue is high demand (from extreme stimulus). Also note the US is importing a lot, so it isn’t something that started with supply chains.

My thoughts: The AS/AD model is too simple, but the fancy models do not address the key point it omits at all. In the 70s and 80s macroeconomics changed to become what I hate and Noah used to hate (he has gone a bit soft). It was considered key that the model be populated by agents with well defined objectives and rational expectations. On the other hand, except for that extreme simplifications were used. The old models had many equations because they considered different sectors of the economy to be different. I assert that this was correct and necessary and the agreement to abstract from this in a failed quest to obtain tractable models was a terrible mistake. I will return to this again and again.

In particular, the AS/AD model has GDP on one axis so the goods and services produced (and demanded) in a country are summarized with one number. The model abstracts from the detail that goods are different from services. This distinction is crucially important right now in (among all other countries) the USA. Covid caused two demand shocks — reduced demand for services and increased demand for goods. This is a well known fact. It means that for a given level of total demand, many more tons of goods had to flow through supply chains. This means that high inflation can be caused by supply chain issues even if total aggregate demand is consistent with low inflation. It is totally standard to ignore the distinction (among many) but it makes no sense. It is possible that ports and trucks are strained causing high costs and high prices even if they are managing more than they did in 2019. In fact, this is obviously the case. As is usual, I cite Krugman

Now I agree that many economists predicted high inflation would be caused by the American Rescue Plan and here we are. It is entirely possible that the current high inflation was caused by overheating (provided the Phillips curve has a high slope — see below). But I do not think it reasonable to ignore the difference between goods and services when discussing current US inflation.

I note that abstracting from such details is absolutely standard in academic macro. I think this is unacceptable (also see below). I think in the 70s and 80s we threw out the baby and kept the bathwater (my tone is partly due to this phrase which I do not like “The critics made many good points, but we were a little too harsh. A few macroeconomic”, I do not think we (Noah, I, and others) were harsh enough.

Actually Noah (who writes better than I do as well as knowing a huge amount) started in 2008 noting the total failure of then standard macro to predict the great recession or guide policy. He notes that Bernanke was an exception (and we were very lucky he was Fed chairman). I actually disagree about the great recession.

Noah writes ” Some, like Paul Krugman, criticized the academic discipline of macroeconomics for ignoring the importance of the financial sector and the zero lower bound on interest rates,  and for building models that were mathematically beautiful but practically useless in a crisis.” From then on, he considers the problem to have been leaving out the financial sector. Since then macroeconomists have concentrated on adding financial frictions. I do not agree that this was the only fatal defect of 2008 era academic macro models.

I think it safe to say that the great regression had a lot to do with housing, residential investment, and house prices. I note that 2008 academic macro models do not have a housing sector and that 2022 academic macro models do not have a housing sector. I ask is there something else surprising that one could not have predicted given the relative price of houses in 2006 (and now by the way). The data say no. I think there are three separate events which have to be discussed separately, and that only one is discussed.

First there is the housing bubble which burst in 2006. One might try to understand how it happened and what effect the bubble bursting would have had without a financial crisis. Then there was the financial crisis triggered by the Lehman bankruptcy. Then there was the extremely slow recovery. Essentially all of academic macroeconomists’ attention has focused on the crisis. Fortunately, economists have models of financial crises and basically understood what had happened and what had to be done. The shock was dramatic, but it didn’t last very long. RIsk premia returned to normal during 2009.

It is hard to argue that the extremely slow recovery was caused by financial frictions. One reason is that all interest rates were very low. The only argument is that mortgage lending standards switched from nonexistent to very tight, causing a slow recovery of residential investment. Somehow bankers extreme fear corresponded to extremely low mortgage interest rates. It is possible to write a model where high fear of default coexists with extremely low mortgate interest rates (it is always possible to write a model with any implication one pleases). But it is a stretch. Another possible explanation of low demand for housing was that the expected future change in the relative price of housing shifted from rapid growth (making a house a very good investment) to no growth. This would reduce demand for houses. It is also undeniably what one sees in all of the (tiny set) of data on such expectations. I see no need to look for another explanation. Here again a lot can be learned with a little disggregation. The slow recovery was a slow recovery of two sectos, residential investment and government consumption and investment. The second is not related to financial frictions. I am not convinced that the first was either. I think the issues are homeowners speculating and austerians. I don’t think either has been addressed by academic macroeconomic research.

I see two fatal problems with academic macroeconomists’ analysis of the great recession. FIrst there is the attachment t to mathematically beautiful models. Models of financial frictions are applied game theory models of asymmetric information. Economists love that. Models of manias, panics and crashes or of irrational policy makers are not beautiful. Second one decision made in the 70s and 80s has not been reassessed at all. It was decided that small models with few equations are better than sprawling models with many. The decision to consider a single good (really single goodandservice) and a single kind of investment (business fixed capital investment) was not reconsidered.

Before going on, I note that Noah (like say Krugman and Blanchard) explains his points with a 60s era AD-AS model. The fancy models that central banks use have this concept embedded somewhere deep inside them. But to understand the basics, we can just look at a very simple model — aggregate demand and aggregate supply, or AD-AS.” Here Noah is a bit diplomatic. He allows the possibility that the fancy models add something of value to AD-AS. In fact, they add testable implciations (which are false). Economists use AD-AS to understand what the fancy models will say and to explain what they said. There is supposed to be something of value added by the fancy model analysis in between. I see no basis for the statement that the fancy models add anything of value. In contrast, they are pretty hideous already even with extreme simplifications such as ignoring the housing sector and ignoring the distinction between goods and services. I think Noah is being diplomatic.

Finally Noah considers the disinflation of the 80s. He quotes “The Slope of the Phillips Curve: Evidence from U.S. States”, by Hazell, Herreño, Nakamura & Steinsson 

“We estimate the slope of the Phillips curve in the cross section of U.S. states using newly constructed state-level price indexes for non-tradeable goods back to 1978. Our estimates indicate that the slope of the Phillips curve is small and was small even during the early 1980s.”

and

“Our results imply that the sharp drop in core inflation in the early 1980s was mostly due to shifting expectations about long-run monetary policy as opposed to a steep Phillips curve,”.

This is all I have read of the paper (written by macroeconomists who are recognized as outstanding by Noah and pretty much everyone else). It is very reckless of me to criticize a paper which I haven’t read written by top economists. I am very reckless.

My objection is simple. The evidence discussed concerns the slope of the Phillips curve. Until the quoted sentence there was no mention of ” expectations about long-run monetary policy” and there is no hint (in the little bit I read) of analysis of evidence about those expectations. In standard macro models price changes are determined by the level of demand and expectations about future price changes which depend largely on long-run monetary policy. I suspect that “expectations about long-run monetary policy” here stands for “everything but the slope of the short run Phillips curve”.

Even later, Noah wrote “In fact, a number of macroeconomists are currently tackling this question, using tools such as surveys to gauge the inflation expectations of businesses and households, and relate these to actual price-setting and consumption behavior. For example, D’Acunto et al. (2019) look at very detailed data on consumer purchases and find that consumers’ inflation expectations are driven by the prices of the things they buy from day to day (not very surprising). And Candia et al. (2021)Coibion et al. (2018) and other papers find that if surveys are to be believed, neither households nor corporate managers pay very much attention to inflation or monetary policy!”

I want to discuss two things other than “expectations about long-run monetary policy” and ” a steep Phillips curve,”. first note “non-traded goods”. This is key, because there was a dramatic appreciation of the dollar during the period of disinflation. This dramatically reduced inflation of the prices of traded goods (also if produced in the USA as they had to try to compete). It is clearly wise to look at non-traded goods, but this does not eliminate the effects of the exchange rate for several reasons. First inflation expectations might depend on actual CPI inflation (and not on thoughts about monetary policy). Second wages depend on the CPI. In the 80s a significant fraction of wages were indexed to the CPI. Producers of non-traded goods use traded goods as inputs. So the extreme shift in exchange rates affected both the costs of producers of traded goods and the expectations of those producers and their consumers.

I’m going to take a step back (way back into the early 80s literature actually). It is not true that price setting depends only on current costs and future expected wages and prices (whether of suppliers or competitors). It also depends on the estimate of the current price level. This should not be a problem as the CPI is published every month. However, it is absolutely clear that people don’t believe the BLS (or don’t know what it said). Perceived inflatoin (as reported in surveys) is not the same as offically measured inflation. But perceived (past) inflation is what affects actual current inflation.

In particular, the price of gasoline has a huge effect on perceived inflation (even more so recently than back then when other prices were also interesting). Inflation in the 70s and 80s was strongly correlated with lagged increases in the price of petroleum. This fits using gasoline prices as a crude price index, delayed adjustment due to wage and price stickiness due to contracts and implicit contracts, and myopia.

In particular, I note that, in the 80s, inflation declined more rapidly that Livingston Survey expected inflation and the forecast errors were dramatically high. This does not fit a shift driven by changed expectations about long run monetary policy.

If anyone has read this far, my conclusion is that Noah (who is amazingly knowledgeable as always) has become diplomatic and I sure haven’t.