Guest Contribution: “What Determines when a Recession is a Recession?”

Today, we present a guest post written by Jeffrey Frankel, Harpel Professor at Harvard’s Kennedy School of Government, and formerly a member of the White House Council of Economic Advisers. He was a member of the NBER’s Business Cycle Dating Committee for 25 years, with his term ending last fall. A shorter version appeared in Project Syndicate and in The Guardian.


The Business Cycle Dating Committee of the National Bureau of Economic Research declared on June 9 that US economic activity had peaked in February 2020, formally marking the start of the recession.

We all knew about the recession already and even the likely date when it started.  Looking at the numbers gave the same answer as “looking out the window.”  Measures of employment had fallen sharply from February to March.  Real personal consumption expenditures (PCE) and real personal income less transfers (which are  numbers that the NBER Committee looks at) both peaked sharply in February as well.  Official measures of GDP only exist on a quarterly basis; but the economic freefall in late March was enough to pull first-quarter GDP growth down to an annual rate of -4.8 %  (relative to the last quarter of 2019). Why did the NBER wait until now to declare something that had already been so clear?

Two critiques of the NBER Committee

Every time the NBER Committee declares a cyclical turning point, many react along the lines, “What took it so long?”  As it happens the 4-month lag between the event and the declaration last week was the shortest lag since the Committee was constituted in 1978.  The average lag across the 10 turning points since 1980 had been 11.7 months. The shortest lag had been 6 months.  The relative speediness this time is testament to the history-making suddenness of the pandemic-induced drop-off.

An equally common public reaction is skepticism — and often surprise — that the declaration of US recessions is left to the judgment of a panel of economists based on a variety of indicators.  “I thought a recession was defined as two negative quarters of GDP growth.”  The US is not quite the only country that does it this way.  The Japanese government also departs from the automatic two-quarter rule and considers other indicators in its official business cycle chronology.  But it is true that almost all other advanced countries use the two-quarter rule.

The US is unusual in that the NBER is a private (non-profit) research institution.  But the NBER chronology does have official status, as confirmed on the website of the Bureau of Economic Analysis of the US Department of Commerce.  [“The designation of a recession is the province of a committee of experts at the National Bureau of Economic Research (NBER), a private non-profit research organization.”]  Private committees in some other countries date business cycles by looking at a variety of economic indicators, including the eurozone, Canada, Spain and Brazil. But their decisions receive less attention from the media or official government bodies.

It is useful that everyone works off of a common officially designated chronology.  Whether it is an econometric scholar doing macroeconomic research or a politician giving a speech that tries to assign credit or blame for a recession, they use the NBER dates.

It is important to note that the two critiques — “why the lag?” and “why not the two-quarter rule?” — tend to be at odds with each other.  GDP statistics are only gathered with a lag and are always revised subsequently, particularly the following July.  Waiting for the definitive GDP numbers would sometimes require an even longer lag between the actual start of the recession and its official designation.

3 examples show the difference in procedures

The recent NBER declaration of the coronavirus recession is an example of how it helps to be free of the GDP rule.  Forecasts say that the second quarter of 2020 will show another plunge in US output (perhaps 30-40 % when annualized), much bigger than the fall in the first quarter.  Nevertheless, the second negative quarter will not be verified until the July 30 release by the Commerce Department’s Bureau of Economic Analysis. And even that number will only be the “advanced estimate.”

Another example of how much difference it makes: nobody questions the NBER ruling that there was a recession in 2001, even though the two quarters of negative GDP growth that year were not consecutive.  (That is going by the conventionally-reported output-side measure of GDP.  The NBER pays equal attention to the little-known income-side measure of GDP.)  As a corollary, everyone accepts that the longest US expansion on record until recently was the 10-year period (120 months) from March 1991 to March 2001.  Nobody tries to claim that because 2001 did not meet the criterion of two consecutive negative quarters, the record length of a US expansion is in fact the period of 201 months from March 1991 to December 2007.  That would be consequential:  it would deprive the just-completed 128-month expansion (from June 2009 to February 2020) of its title as the longest recorded expansion in US history.

A third example: at the time when the NBER committee announced that (what came to be called) the “Great Recession” had begun with a peak in December 2007, the government estimates still reported that the official GDP measure was actually higher in both the first and second quarters of 2008 than the last quarter of 2007.  Even though the announcement of the beginning of the recession was greeted as long overdue, the Committee would have had to wait another year and a half to get that crucial revision from the Commerce Department if it had applied the “two quarters of negative growth” rule.

Pros and cons

The two-quarter rule has both pros and cons compared to the NBER committee’s less mechanical approach. One advantage is that it appears more objective to have an automatic procedure that is simple and transparent, especially if the alternative is delegating the job to a committee of unelected unaccountable ivory-tower economists.

One major disadvantage of the two-quarter rule is that when the GDP statistics are revised subsequently, they may require a retroactive revision of the cyclical turning points.  For example, a 2011-12 recession in the United Kingdom  was subsequently erased from the record when the GDP numbers were revised in 2013.  Claims that in 2012 had appeared in the speeches of UK politicians and in the writings of researchers, made in good faith at the time, were subsequently rendered false.

Another disadvantage of using the rule of two consecutive quarters is that a single big long downturn can be recorded as multiple short recessions if it is interrupted by slight up-ticks.  Examples include Ireland, Finland and Italy in the aftermath of the global financial crisis.  The NBER waits until GDP has re-attained its preceding peak, or at least come close, before judging that a recession has ended.

This point could become an issue in the US and many other countries later in the 2020 recession.  As the coronavirus shut down the economy in March-April, the plunge left economic activity at levels far below where they were in 2019.  Some countries will have a positive 3rd quarter, which would put an end to their recessions according to the two-quarter rule.  Yet this will strike most people as the wrong answer, because the levels of GDP and employment are likely to remain far below their pre-pandemic levels.  This is especially true if the positive quarter is followed by a renewed downturn, as in the famous “W” scenario, which is all-too-likely.

The NBER waits before dating a trough or peak until it can be reasonably sure it will not have to revise the call in the future, after the dates have already entered the official chronology. The Committee would revise a date if it had to, and has occasionally considered it.  But a revision of dates would be confusing. And the Committee has never done it. Hundreds of other observers try to assess the odds of a recession like the current one in real time. The NBER committee sees its job, not as being the fastest, but as being definitive.

This post written by Jeffrey Frankel.

21 thoughts on “Guest Contribution: “What Determines when a Recession is a Recession?”

  1. Barkley Rosser

    Thoughtful piece, but the claim that we are likelier to see a larger decline in output during the 2020 second quarter than in the first quarter looks questionable, so questionable that in fact we might see positive growth in the second quarter and thus another example of an official NBER-called recession that violates the two quarter rule.

    I note in particular, as I did in a comment on another post, that retail sales rose 17.7% from April to May, more than offsetting the 14.4% decline the previous month, with retail sales now only 8% below the February peak. Even with growth cut in half we might find retail sales greater at the end of June than they were in Februarry.

    Certainly there are other components of GDP, with state and local governments apparently declining, and we just has another 1.5 million laid off last week, so some sectors seem to be still declining. But some services are almost certainly also rising, with increased hiring in dental services 10% of the job increase in May. And while I would think uncertainty would reduce new capital invrestment, construction was maybe the sector with the largest increase in hiring, possibly due to the restarting of ongoing investment projects.

    Anyway, this supposed massive decline in GDP for the second quarter looks far from being thoroughly and definitely cooked.

    1. Menzie Chinn Post author

      Barkley Rosser: Hard to see how Q2 GDP loss is smaller than the currently reported 5% loss in Q1 (SAAR). From IHS/Markit (formerly Macroeconomic Advisers): “Core retail sales rebounded significantly in May, well beyond our prior assumption, implying substantially more PCE in the second quarter. As a result, we raised our tracking forecast of second-quarter GDP growth by 4.9 percentage points to -37.1%. We also held our tracking estimate of first-quarter GDP growth at -5.2%.” Atlanta Fed GDPNow, incorporating the retail sales data, is at -45.5% as of 6/17.

      1. Barkley Rosser

        Well, Menzie, I confess to feeling kind of weird. I do not remember ever being in such disagreement with you, and not only you but a lot of other big official typrs, including Macroeconomic Advisers and the Atlanta Fed at least. But I think I am, and I do remind that lots of official forecasters were calling for a sharp decline of employment in May, which turned out to be massively wrong, whereas while I did not make a specific forecast, I was not surprised, and I was not surprised because i looked to me that US GDP was growing in May based on a few indicators, although I did not think that growth was all that grearr.

        Now things seems even more out of whack, and I am especially mystified by how little this new report on retail sales seems to be budging these forecasts of huge GDP decline for 2020 Q@, on the order of -37% to worse than -45%, if I am reading correctly.

        I mean look. Consumption id about 70% of GDP, and retails sales are about half of consumption. So retail sales are maybe a bit over a third of GDP. For April a third of GDP declined 14.4%, but then that third turned around and grew by 17.7% in May. We are talking about a third of GDP basically not declining over the first two thirds of the quarter. Am I not right on this? How on earth does one get overall GDP falling by the amounts for the second quarter by the amounts they forecast when a third of GDP has not dropped for the first two thirds of the quarter, and given that probably retail sales will grow more in June, even if at a smaller rate than in May, that third may actually end up net growing for the quarter. What is going on here?

        I have seen FRED out of St. Louis Fed. It shows several other variables growing in May, if not at the rate of retail sales, including manufacturing, with not a single one of those they show going down. They showed manufacturing declining 20% in April, but if there is growth in both May and June, that does not seem to be pushing towards these 37-45% deckubes, Again, what on earth is responsible for such declines when consumption might actually rise? Is there some massive collapse in services still going on? Not in dentistry.

        And then today the Philadelphia Fed released its manufacturing index. This was –43.0 for May 31, but this latest release for June 30 is 27.5, a massive turnaround. Just to show how huge that is, the index was higher than that for Feb. 28 at 36.7. But exceot fir tgat reoirtm bi itger monthly figure of this index was higher than the one issued today as one looks back until we get all the way back to mid-2018. Really, it looks like manufacturing is zooming right now.

        What on earth are Macroeconomic Advisers and the Atlanta Fed looking at that has them issuing these extremely negative forecasts that seem to be just wildly at variance with numbers on some very substantial components of GDP?

        Note that even if we end up having positive growth for the second quarter, which at this point I think is quite possible, that does not mean it will continue. We might see that infamous W. After all Covid-19 seems to be surging in many states, and many of those fiscal stimullus measures are goiing away. Third quarter may well not look as good as second, but for now I am completely mystified by these enormously negative forecasts for the second quarter, unless there is something going on here I do not get. Even if those numbers are measuring not the second quarter, but from the peak in February, I do not see how they are likely to come about, given what seems to be going on with manufacturing and rettail sales.

        1. 2slugbaits

          Barkley Rosser You make a number of fair points. I prefer to take the coward’s way out and note my distrust of some of the data. I’ll refer you to this Census Bureau report for retail sales, which provides the basis for the BEA release:

          A couple of things to note. First, all sales assume no price changes, only seasonal variations. My experience over the last few months is that we’ve seen lots of price changes that move all over the place. Some commodities saw steep discounts while other commodities were selling at a steep premium. The only thing this retail sales report captures is the change in the final cash register sales at normal prices and we don’t yet know how this shakes out in terms of quantities. And retail stores appear to have achieved those sales by drawing down inventory, which will tend to lower GDP. Table 1 of the Census report shows a detailed look at retail sales by subsector. Some sectors were hit very hard; e.g., furniture stores and clothing stores. But the electronic commerce subsector saw a big increase. And Table 3 shows the error terms around their estimates…and those are error terms for normal times. To top it off, we should expect that many of the stores used in the Census Bureau’s survey may not be in business anymore, further complicating retail sales estimates.

          Something else to keep in mind is that the retail sales numbers are not quite the same thing as retail sales in GDP. I’m not an economist, but my understanding is that the monthly retail sales figures represent gross sales at the cash register, but GDP measures value added by the retail sector. According to the BEA’s GDP-by-Industry tables the retail sector accounts for about 5.5% of GDP. OTOH, factors that might tend to support your argument are that other value added sectors do seem to be springing to life. For example, health care and social assistance accounts for 7.6% of GDP and manufacturing accounts for 11.0% of GDP. But then again, sectors that are seeing some increases in employment like the hospitality sector don’t contribute much to GDP even though they represent a lot of jobs. For example, Arts, Entertainment & Recreation is only 1.1% of GDP in normal times and hotels (Accommodation) is only 0.8% of GDP. Restaurants normally account for 2.3%. So even thought these sectors employ a lot of people, they don’t contribute as much to GDP. The big problem area will be to see what happens to state and local government spending, which accounts for 8.5% of GDP and is likely to take a huge hit over the next several quarters.

          Bottom line is that I don’t know what to expect about the GDP numbers. As I said earlier, I’m taking the coward’s way out on this one.

        2. baffling

          employment was worse than reported over the past few months, by several percentage points. retail sales have recovered some, but how much of this is amazon and other online related activities? these require fewer workers. this could explain why unemployment is still excessive while retail is increasing. people are not out in stores in mass, but they are purchasing online. since most of the country is still working, many of those folks still have jobs and income. but they are not eating out while shopping at the mall. our household is spending 50% compared to prepandemic. this will probably continue moving forward.
          “Is there some massive collapse in services still going on?”
          services and travel/hospitality are not recovering well. i have used telemedicine, but i won’t go to the dentist at this time. if you go to malls/shopping centers, you see more cars compared to march, certainly. but it is probably less than half of the activity i used to see pre pandemic. the most telling thing to me is, i can drive around the houston inner loop during evening rush hour at full speed, occasionally without another car within 100 yards of me. now many folks may be telecommuting, but this certainly says economic activity is still quite low, although traffic is busier than two months ago.
          the month of april and the first two weeks of may were completely dead. the reopening really did not catch on until memorial day. i think people question how that can occur and gdp grow? either the collection/data is not correct, or physical observations are not accurate.

        3. Barkley Rosser

          I fully accept that there are enormous amounts of noise in the data, a point that I think also applies to the models being used by those making these forecasts of humongous declines in GDP for second quarter that do not seem to be there according to the pieces of apparent data emerging. I accept that many are still being laid off, but I also note that this number has been declining. It was 1.75 million last week, down to 1.5 million this week, with this still way above a “normal” 200,000 or so.

          I am continuing to think about where else we might see a source for GDP declining. It certainly does not look like consumtpion, even if C ends up still lower on June 30 than it was on Feb. 29. it is almost certainly not declining right now, and is probably rising sharply, if very unevenlyu across sectors.

          There is capital investment, and there is certainly plenty of reason to think that it would decline and then not recover swiftly. High uncertainty is never good for new investment, although funding due to Fed policy seems reasonably available. But, as I have noted over on Econospeak, looking at the employment increase numbers for May that had so many surprised, construction was I think the sector with actually largest increase in jobs. That is almost entirely in what gets counted as investment, although my strong guess is that this is mostly the restarting of projects that were already going but got shut down temporarilyu at the height of the virus. Well, maybe that growth will slow, but offhand I do not see investment as contributing to a decline in GDP, much less some massive 37-45% decline, which, frankly, just looks absolutely absurd and impossible.

          Then we have exports. On this one I have no data (frankly have not looked), but while I doubt it is growing rapidly, I also doubt it is collapsing. Based on global carbon emissions data the world economy, moving somewhat ahead of the US one, bottomed out in early April. It has almost certainly been growing positively for the second quarter, even if in the end the US does actually end up with a negative second quarter number. And the strength of the rest-of-the-world economy is the most important element normally in determining exports. Yes, there have been a bunch of disruptions, and the dollar has been somewhat high, so all that could slow it down. But again, there is no obvious reason for exports to be leading the way to some quarter decline in GDP on the order of 37-45%, not remotely.

          Which leaves us with G, government, and there I think we do have some action, and it is my strong suspicion that the majority of the newer layoffs are coming from either state and local govenments themselves, constrained by budget balance rules and facing falling revenues in many locales, or else firms that do business with these governments. I am not aware of major layoffs occurring in the federal government. So here is an area that is probably pushing GDP downwards, but state and local governments together amount to something like 11-12% of GDP. They could completely shut down and not bring us to 37-45% decline of GDP. For these forecasts to come even remotely true, something else has to be driving this beyond the clear declines happening in the state and local governments.

          1. spencer

            In the Census trade report they have a table of real imports and exports and in May real petroleum imports dropped like a stone and the total real deficit widen sharply. Trade will not be a big positive in GDP as it has been recently.

        4. Menzie Chinn Post author

          Barkley Rosser: Over 2002M01-20M02 period, regression of real retail sales on real consumption (rsxfs on pce96 using FRED codes) in log first differences yields of coefficient of 2.2, t-stat of 8, Adj.-Rsquared of 0.44 – so in usual times gross CPI-deflated retail sales move more than real consumption. Now, in April, they matched about one-for-one, so maybe you’ll be right that consumption rebounds one-for-one in May.

          1. Barkley Rosser

            Thank you for this, Menziie. What is clear is that we are in an unprecedented period of massive uncertainty. Many normal relationships have broken down, and much data is highly questionable. There is a wide range of possible outcomes regarding the economy over the near term.

          2. spencer

            I believe the retail deflator calculated by the
            BEA when it does PCE and real GDP is a far
            superior price index for retail trade than the CPI
            used by the FRED. It is reported with PCE.

  2. Not Trampis

    I agree a thoughtful post.
    Never been a fan of the two quarter ‘rule’ which is soo popular down under. It needs to be more nuanced. I think the NBER has mostly got it right.

    does it have a definition of a depression. I know Barkley was surprised by a 10% fall in output.

    1. Barkley Rosser


      When was I “surprised by a 10% fall in output,” mate? I do not remember offhand.

      I am the one who was not surprised that there was an increase in employment in May, in contrast to the overwhelming majority of US economists making forecasts of this, who were calling for a sharp decline in employment. I based my lack of surprise on the considerable amount of data out there that the US economy was growing overall in May, which I forecast will prove to the be the case once we have the numbers.

      1. Not Trampis

        Sorry My friend. Clumsy writing.
        surprised that the definition of a depression is a 10% drop in output.

        1. Barkley Rosser

          Oh that. It also needs to last for at least three years. I have no problem with that.

        2. spencer

          There is no official definition of a depression.
          Recession, when your neighbor loses his job.
          Depression, when you lose your job.

      2. baffling

        barkley, how much of the may employment increase was related to ppe loans that went to pay for workers in may? those workers would have been unemployed otherwise, and many got paid for doing not much. those loans are expected to be forgiven if it paid for employees, so they acted like unemployment payments. i am not sure employment gains were directly correlated with increased economic activity for those firms. i think the ppe helped, but it may skew our interpretation of the strength of the economy.

        1. Willie

          Good point. Another aspect of it is the end of PPE, and any effects of a second wave of COVID. Or an extension of the first wave because we never dealt with it properly. Either of those will cause a second downturn.

  3. spencer

    In the big 1974 recession there was a school of thought that it actually was two recessions separated by a rebound in the summer months.

  4. 2slugbaits

    It will be interesting to see how much the drop in inventories will show up as a drag on GDP. A lot of businesses simply drew down inventories without replenishing stocks.

Comments are closed.