Interview: Emi Nakamura, macroeconomist

In which we talk all about macro.

If you ask any macroeconomist to tell you who the stars of their profession are right now, Emi Nakamura’s name will surely be at or near the top of the list. In 2019, Nakamura won the John Bates Clark medal, one of econ’s two most prestigious awards — and one that very rarely goes to a macroeconomist. Originally from Canada, and now working at the University of California, Berkeley, she continues to amass top-journal publications at a fairly stupendous rate.

I first met Nakamura in 2011, when I was still a graduate student. She came to Michigan to gave a talk about her paper “Fiscal Stimulus in a Monetary Union: Evidence from U.S. Regions,” with frequent co-author Jón Steinsson. That paper provided some of the best evidence that fiscal stimulus boosts growth, and had a big impact on the academic debate then raging over whether to use fiscal policy to boost the economy out of its post-2008 doldrums.

Since then, Nakamura has continued to turn out work that’s highly relevant to current policy debates. As just one example, her recent paper with Hazell, Herreno, and Steinsson, “The Slope of the Phillips Curve: Evidence from U.S. States” has been the main paper guiding my own thinking about the current inflation. I’m also a big fan of her working paper “A Plucking Model of Business Cycles”, which I think has the potential to help us understand the current recovery. Her 2016 paper about China, with Steinsson and Liu, lent strong empirical support to the popular suspicion that China’s government cooks its books in order to make growth seem more consistent than it really is. Another recent paper of hers shows that when people are forced to move (in this case by a volcano), their economic situation improves; this has obvious implications regarding America’s declining geographic mobility. And so on.

Meanwhile, Nakamura has also been working to improve macroeconomics itself. The macro field was in deep crisis after 2008; the financial crisis and recession showed that something was wrong, and many people (myself included) pointed out how divorced from empirical reality the theories had become. But instead of complaining about it, Nakamura simply started working to fix it. She’s done work on how to identify monetary policy shocks with high-frequency data, how to tease out the effects of policies by looking at their different effects on different regions, and other methodological innovations that allow macro to be more scientific. In 2018, she and Steinsson wrote an overview paper about new empirical macro that I think will end up having a very big influence on the direction of the profession.

In the following interview, I talked to Nakamura about inflation and what to do about it, about how theory and evidence interact in macroeconomics, and about the future of macro.

---

N.S.: I guess we should start out by talking about inflation, which is kind of the big macroeconomic topic on everyone's minds right now. What do you think is the main cause of the inflation we're now experiencing in the U.S. (and to a lesser degree, in other rich countries)? Can we expect it to just sort of go away on its own, or do we have to take some kind of policy steps to get rid of it?

E.N.: The recent increase in inflation is much more than historical experience would have predicted (which is about an increase in inflation of 1/3% for every 1% decrease in unemployment). I think several factors have been important. 

First, after a long hiatus from playing a major role in inflation, supply shocks are back! The most dramatic of these is the disruptions to the labor market. US labor force participation is down by roughly 1.5%, and so far the decline is pretty persistent. And the shocks to labor supply go far beyond that: many workers are out sick, or quarantined (or are at risk of this). The decline in labor force participation is much larger than in the Euro area, and this may be partly because Euro area countries implemented a lot of policies that kept workers in place at their jobs during the pandemic. https://venance-riblier.shinyapps.io/Employment/. Few people anticipated the persistence of the pandemic effects on labor force participation, and it's really hard to tell how this will play out over the next year or two. Some of these workers may come back to work, but others, particularly those who have retired, may not. There have also been other important supply shocks: it's more expensive to operate a daycare or a factory than it used to be due to safety restrictions due to COVID. It used to be hard to come up with good examples of negative supply shocks in teaching undergraduate economics classes, but COVID certainly counts as one!

Second, there has been a historic shift in demand from services to goods: https://fred.stlouisfed.org/graph/?g=LnYU. In the Great Recession, the fraction of expenditures spent on goods fell. The opposite happened during COVID: the fraction of spending on goods rose pretty dramatically. This is another tectonic shift in the economy that I think is putting enormous pressure on supply chains among other things. Many more people are working from home, and they all need computers, and the semi-conductors needed to build those computers. All those goods have to be shipped to the United States and to people's houses. This is a supply "pressure" but not really a "supply shock" because its ultimate cause is an increase in demand (at least for certain kinds of goods). But a recent Jackson Hole paper points out that secular shifts in demand can lead to the same inflationary pressures as supply shocks. Again, I have a lot of uncertainty about how long it will take for people's consumption patterns to return to normal. I think some of these shifts in consumption are related to the changes in labor supply I mentioned before: when you go to work, you buy services complementary to work-- the coffee on the way to your office, and the salad for lunch-- to the extent that people shift to a lot more work from home (and less people are working at all) this could make some of the changes in demand patterns quite persistent. 

Third, there has been a very rapid recovery and a lot of government support for spending. Households have a huge buildup in savings https://fred.stlouisfed.org/graph/?g=Lo1j, and spending this down is no doubt contributing to demand.  Conceptually, one might expect these demand pressures to be captured by the unemployment rate. The unemployment rate is still somewhat higher than before the COVID crisis. But, there is a lot of evidence that the unemployment rate is providing an incomplete picture of labor market tightness right now:  vacancies and quits are quite elevated relative to pre-COVID, despite higher unemployment, and lower employment rates.

One graph that strikes me as interesting in assessing the role of these different factors is this one: https://fred.stlouisfed.org/graph/?g=Lo1x. This graph shows the inflation rate for the shelter and non-shelter components of the CPI along with the unemployment rate. I graphed these series back to 1990 because that's roughly when long-run inflation expectations started to stabilize in the US. Here you see very clearly the fact that the shelter component of the CPI is quite cyclical, whereas the non-shelter component is much more volatile (for example, the big commodity-driven fluctuations in 2008). Even during the COVID period, the shelter component of the CPI has a reasonably stable relationship versus the unemployment rate whereas the non-shelter component has increased pretty dramatically. It's important to recognize that the shelter component of the CPI in the US is based on rental costs: so this isn't exposed to either supply chains or labor market shortages. So one interpretation of this graph is that it suggests a big role for the first two factors I emphasized, as opposed to only conventional aggregate demand factors. However, some are predicting a big catch-up in rent inflation soon, and perhaps some of these patterns also have to do with demand shifts related to housing. We have to be humble in extrapolating past relationships to the present given the fundamental shifts that COVID has imposed on the economy.

One thing that hasn't contributed much to inflation so far is an unhinging of longer run inflation expectations. Both survey and market-based measures of longer run inflation expectations look pretty stable (see below). So far, the Fed has been very successful in stabilizing long run expectations and this is a big achievement. The goal is, of course, to avoid these supply shocks and shocks to relative prices becoming the kind of self-fulfilling high inflation we saw in the late 1970s.  There has been a notable uptick in longer run inflation expectations in the very recent past, but so far it is small. It's one of the Fed's primary goals these days to keep it that way. 

N.S.: I'm a little confused about the labor supply shock story...If reduced labor supply is a big part of the inflation story, then shouldn't real wages be rising? Usually when there's a shortage, prices rise, but instead real hourly compensation has been falling. How can we explain that?

E.N.: That's a very good point! The hourly (not real hourly) compensation has gone up a lot (5.75% year on year) but not enough to adjust for inflation, and cause real wages to rise much. We haven't seen much daylight between the nominal and real hourly compensation series over the past couple decades (see this graph) but it has really opened up over the last year. Other measures of nominal wages, such as the Wage Growth Tracker, which looks at changes in wages for the same workers over time, look similar. 

One important point is that real wages have gone up the most at the bottom of the wage distribution-- see this plot (and select “wage level” at the top to see things disaggregated by quartiles of the wage distribution). This speaks  to the unequal nature of the COVID shock. One positive thing about the COVID recovery is that it seems that it has somewhat narrowed inequality. But even in this plot, the lowest quartile of wages have not increased much in real terms. One would expect this pattern if wages are particularly "sticky" in response to macroeconomic shocks, particularly relative to more flexible prices such as food and energy. The older academic literature on wage rigidity was critiqued a lot for predicting countercyclical real wages, but perhaps we are actually seeing this as part of the COVID recovery. 

N.S.: Also, let's talk about inflation expectations. I've been basing a lot of my own thinking on your recent paper with Hazell, Herreno and Steinsson, in which you explain the 70s inflation as partly due to the oil shocks but partly due to a regime shift in beliefs about Fed hawkishness. Which means I'm relieved that both market expectations of inflation and longer-term survey expectations look contained. Does that mean the Fed is doing enough right now? Or should they be actively trying to push inflation back toward its 2% target?

E.N.: I am also relieved that longer-run inflation expectations remain contained (see my graph above of the TIPS 10Y and SPF 5Y/10Y expected inflation rates). There is a notable uptick at the end of the series, but so far, it remains small. But market expectations are predicated on what the market expects the Fed to do. There is a self-fulfilling prophecy element in this, as in many things in macroeconomics. So long as the market expects the Fed will do what it takes to contain inflation, we won't see much movement in longer run inflation expectations.  The Fed is working very hard to preserve this. But we can't take this for granted. In other places, and other times, even in the US, longer run inflation expectations have been much less anchored. A big challenge for the Fed is that when you're successful-- like, say, if their current shift in policy manages to avoid the “bad equilibrium” where people lose confidence about longer run inflation expectations, then you never get to see the counterfactual---the  crisis you have averted never actually materializes. For example, the Fed liquidity policies early in the pandemic may have averted a financial crisis. But we'll never see the counterfactual. 

N.S.: Now, a lot of people are talking about this Jeremy Rudd paper, challenging the notion that we really understand how inflation expectations influence inflation itself. What do you think about that? Do macroeconomists believe too strongly in the power of expectations? After all, you have a paper, with McKay & Steinsson, showing that forward guidance is probably less effective than people think, because consumers and companies have limited power to respond to things that happen far in the future. That paper seemed to fit with a growing theoretical literature questioning the notion that everyone is making their economic decisions based on very long-term thinking. Should this shape how we think about inflation now?

E.N.: I think it's hard to make sense of really big inflationary episodes, without a role for inflation expectations. How does inflation rise to be tens or hundreds of percent per year, then fall back to zero, based on the Phillips curve slope alone? From a theoretical standpoint, incorporating a role for inflation expectations just means that firm managers think about the future in setting their own prices-- how much will wages go up? How much will competitors raise their prices? How much will suppliers raise their costs? These issues are quite salient in high inflation periods, for example, in wage negotiations-- there's a great anecdote about this in this interview with Paul Volcker, where Volcker recounts a meeting with a businessman who has just come back from a wage negotiation in which the businessman is excited to have locked in 13% wage growth for his workers (presumably predicated on high expectations of inflation). . 

When inflation gets really low, like in the US for quite a long time before COVID, there's a lot of evidence that people do not pay much attention to inflation.  After all, it doesn't matter much for either firms or workers. This is something I notice in teaching: American students often have very little idea coming into class of what inflation means, but Latin American students seem to understand it almost innately-- presumably due to the environment they have grown up in. 

I'm sympathetic to the idea that our models probably overweight the extent to which long-run equilibrium objects determine current behavior, along the lines you suggest. There's a lot of value to thinking about how the predictions of the models change with bounded rationality. 

N.S.: Pulling on that last thread, what do you think are the most interesting or important trends in macro theory these days?

E.N.: I'm an empirical macroeconomist, so perhaps it's no surprise that I am excited to see a stronger connection between theoretical work and evidence from microdata, as well as quasi-experimental empirical methods in many parts of macroeconomics. I think finding more ways to connect the theory to the data is a precondition for making the normative implications of our models more convincing, as Milton Friedman emphasized in his Nobel Prize lecture or in more detail here. The more progress we can make in convincingly establishing the facts, the better we will be able to evaluate what theories are most useful. This can also be a real inspiration for new models and theoretical ideas, when we find empirical results we don't expect. 

N.S.: It's interesting that you mention Friedman's "Essays in Positive Economics" as support for the idea of connecting theory to data. It was in those essays that Friedman made an analogy between macroeconomics and the shots made by an expert pool player. He argued that just as the pool player doesn't need to understand physics in order to make shots, macroeconomists don't need to understand the particulars of how economic decisions are made in order to model the resulting economic behavior. In other words, he seems to be arguing that we don't need macro models to fit micro data, only aggregate data. And yet in recent years, the profession seems to be strongly turning against Friedman's idea. In fact, your 2018 paper "Identification in Macroeconomics" is the best summary I know of this trend toward checking macro models against micro data. So do you agree that this has been a major shift?

E.N.: The pool player analogy has elements of truth, but I would not use it as an argument to throw away data. If we had large datasets with lots of random variation in the relevant variables, maybe we could focus on the macro data alone. But in practice, we tend to have small datasets with a lot of non-random variation. Macro data tends to be confounded by a lot of structural and institutional change, and on top of that, there are all the challenges that come with assessing causality because we can't use randomized trials. So, I think it makes sense, to try to combine micro and macro approaches, because we have more confidence that models with realistic assumptions will work in contexts where we have not been able to previously analyze their effects than other models with unrealistic assumptions, if the two models fit the observed data equally well.

To take an example from a different subject, if we want to convince ourselves of the negative effects of lead on people's behavior it is useful to observe not just the "macro" evidence on, say, crime in areas with more lead exposure, but also the biological evidence on what happens to cells when they are exposed to lead.

I am sure I am not the first person to express this view. I don't know enough about the history of economic thought to say how this perspective has ebbed and flowed in the past. I think a lot of the debate has focused on theory: for example, microfoundations for macroeconomic relationships. But there has been a huge increase in our ability to access and use microdata in recent decades, and that is a huge opportunity on the empirical side.

N.S.: Well, I'm definitely on your side here, and my impression is that most young macroeconomists are too. Using micro-data to validate the pieces of macro models seems like it could potentially lead to a golden age of rapid progress in macro.

So my next questions are: 1) What are some of the most exciting lines of research that you see academics pursuing out there in the macro world?, and 2) What are some important lines of inquiry that haven't gotten enough attention yet?

E.N.: Research on monetary economics has been particularly exciting in the years since the Great Recession, because the policy questions are so important, there has been a lot of new data, and the tools of monetary policy are changing. Regarding the path forward, I often find myself an advocate for the merits of "boring" work chipping away at basic questions. Just because a question has been studied before, doesn't mean that we are all convinced of the answer. Multiple studies coming to the same conclusion, using different, and hopefully increasingly convincing,  methods can have a lot of value in cementing our views on a subject (for example, think of the many papers that have contributed to changing economists' views on the marginal propensity to consume).  I'm also a big fan of work on macroeconomic measurement, which I think tends to be understudied.

N.S.: So if you could give advice to young macroeconomists at the start of their careers right now, what would it be?

E.N.: When I was an undergraduate at Princeton, I remember sitting in the office of one of my advisors, Bo Honore, and pondering the sign he had on the wall: "Question Assumptions." When I came for my job interview at Berkeley a few years ago, I was sitting in the office of the department chair at the time, Jim Powell. I looked up and saw exactly the same sign: "Question Assumptions." After recovering from the deja vu, I learned that the sign was acquired from a counterculture hippie when Bo and Jim were strolling around downtown Berkeley. I'm pretty sure the sign wasn't originally intended as research advice for aspiring economists, but I still think of this as some of the best advice I've gotten, and some of the best advice to pass on.