Is macro making progress?
At a recent blogger conference, I was asked to name the most important paper published in my field (which is macroeconomics) over the past ten years. I couldn’t think of any.
In one sense, that’s a reflection of the fact that the field has moved on from the 20th century macro research with which I am most familiar. My ignorance may say more about me than it does about the field of macro. In desperation, I suggested that Paul Krugman’s 1998 Brookings paper (It’s Baaack . . . ) was the most recent one that I recall having a decisive impact on how we think about macroeconomics. A few years ago, I wrote a paper discussing how the important “Princeton School” of monetary economics was heavily influenced by this paper.
Many brilliant economists continue to do quite sophisticated research in money/macro. And yet I rarely see new papers that seem interesting to me, at least in the way that many papers from the last half of the 20th century seemed interesting when they were first published. And it’s not just macro. Casual art fans like me are familiar with hundreds of famous painting from the period from 1880 to 1924, but very few famous paintings from the period 1980 to 2024. Why is that?
Tyler Cowen recently linked to an NBER working paper by Joel P. Flynn and Karthik Sastryx, which looks at how optimistic and pessimistic narratives could contribute to the business cycle. At a technical level, the 134-page paper is far above anything I ever did, featuring literally hundreds of mathematical equations, some fairly complex. Here’s an excerpt from the conclusion:
When we calibrate the model to match the data, we find that the business-cycle implications of narratives are quantitatively significant: measured declines in optimism account for approximately 32% of the peak-to-trough decline in output over the early 2000s recession and 18% over the Great Recession. Finally, we show that the interaction of many simultaneously evolving and highly contagious narratives, some of which are individually prone to hysteresis, can nevertheless underlie stable fluctuations in emergent optimism and output. Taken together, our analysis shows that narratives may be a significant cause of the business cycle.
Their work employs a “real business cycle” framework, of which I’m generally somewhat skeptical. It’s not that these RBC models don’t tell us important things about the economy, rather I believe that (at least in the US) real shocks are primarily important as a determinant of long run trends, not business cycles. (With Covid being the obvious exception.)
I only skimmed the paper, so I won’t offer an opinion on their empirical estimates, but this caught my eye:
Our analysis leaves open at least two important areas for future study. First, we have analyzed how firms’ narratives matter and abstracted away from studying households’ narratives. It seems reasonable that similar mechanisms could operate on the household side of the economy, where contagious narratives might influence spending and investing decisions. Moreover, co-evolving narratives on both the “supply side” and the “demand side” of the economy might have mutually reinforcing effects. From this perspective, narratives have the potential to explain even more of the business cycle than our analysis suggests.
I like this observation, as I’ve long believed that the most important impact of supply (real) shocks is the way they interact with demand (nominal) shocks. Thus a real shock in housing/banking during 2007 probably depressed the natural rate of interest. The Fed fell behind the curve and cut rates too slowly (especially in 2008.) This led to a fall in nominal GDP (less demand), making the recession much worse.
They conclude with the now almost mandatory call for further research:
Second, there remains much more to study about what “makes a narrative a narrative”—that is, in the language of our model, what microfounds the set of narratives and their contagiousness? A richer study of these issues would cast further light on policy issues, including both the interaction of standard macroeconomic policies with narratives and the potential effects of directly “managing narratives” via communication. Moreover, probing these deeper origins of narratives could further enrich the study of narrative constellations beyond our analysis, to account for the full economic, semantic, and psychological interactions between narratives in a complex world.
Will that follow up research answer those questions? I’m skeptical. I worry that the next brilliant pair of young macroeconomists will think to themselves, “Flynn and Sastryx have already done that, let’s develop a different model.” There’s probably enough truth in almost any plausible macro model that you can find some empirical support for the theory (at least if you sufficiently “torture” the data set.)
It’s possible that my skepticism about modern macro merely reflects an old guy who is out of touch with recent developments. I plead guilty. But in the second half of the 20th century, one didn’t have to read 100 page research papers to understand that macro was generating lots of revolutionary ideas. I am not seeing interesting new ideas being explained in non-technical papers for the layman.
Here’s one way to think about my pessimistic mindset. I’ve done macro research for almost my entire life. Fairly early on, I came to the conclusion that US business cycles were pretty simple. In most cases (not Covid) it was simply a question of monetary policy mistakes driving fluctuations in NGDP, and real GDP being highly correlated in the short run due to sticky wages.
If you were going to explain why the paintings of 1880-1924 seem more memorable than the paintings of 1980-2024, you might point to the fact that painters in the earlier period discovered lots of interesting new styles, and that there simply weren’t as many interesting new styles to discover in the latter period. Another view is that I’m wrong, and that future generations will discover even more masterpieces in the art of painting from the 1980-2024 period than from 100 years earlier. Time will tell.
Thomas Kuhn said that in science we often make progress by developing models, then discovering that there are certain “anomalies” not explained by these models, and then developing new and improved models to explain those anomalies. Perhaps our best late 20th century macro models do a pretty good job of explaining business cycles (and note that the “Fed mistake” theory I just gave could explain why explanation doesn’t imply prediction). Perhaps the remaining anomalies are simply very hard to explain.
But this doesn’t fully explain my skepticism about modern macro. You can argue that we invented too many good macro models in the second half of the 20th century. We have Keynesian models, monetarist models, real business cycle models, Austrian models, MMT models, and many variations within each category. Flynn and Sastryx are employing a RBC framework in their paper. Because my own view is that this framework is not very useful for understanding business cycles, from the outside this whole line of analysis seems a bit off target. And that skepticism doesn’t just apply to RBC models, from my perspective any non-market monetarist model is somehow missing the point. They all appear to be trying to explain something that has already been adequately explained. They aren’t addressing anomalies in the model where Fed mistakes drive NGDP and create cycles due to sticky wages; they are usually working with an entirely different framework.
This is why to a grouchy old guy like me, macro no longer seems progressive. We are not filling in the gaps; we are continually trying to reinvent the wheel.
Again, it’s very possible that I’m out of touch. All I can say is that I no longer read papers and think, “I always wondered why certain macro variables (M,Y, P, i, U, etc.) showed this pattern, now it makes more sense.” I don’t see the progress.
But hey, people in 1890 didn’t yet see the brilliance of Van Gogh, so it’s quite possible that I’m missing something important.
PS. Here’s a Kandinsky painting from 1925. What was there left to say?
PPS. Here’s one of the 247 mathematical equations in the paper:
