The Ins and Outs of Recessions

By Paul Gomme and Peter Rupert

For those who have been concerned that the U.S. has entered a recession, The Bureau of Economic Analysis’s October 27 release of its advance estimate for Quarter 3 real GDP brought good news: Output grew at an annual rate of 2.6%. This increase comes on the heels of negative output growth in the previous two quarters.

While two consecutive quarters of negative GDP growth are colloquially said to constitute a recession, there is no formal definition of a recession. The National Bureau of Economic Research (NBER) has a business cycle dating committee that (retroactively) determines the dates of entering and exiting a recession. The gray bars in the following figure are the recession dates as determined by the committee.

Recessions are recurring events that last for varying lengths of time. It appears that recessions were more frequent and lasted longer before the Great Depression in 1930’s than after. But what, exactly, determines entry and exit? The NBER states,

The NBER’s definition emphasizes that a recession involves a significant decline in economic activity that is spread across the economy and lasts more than a few months. In our interpretation of this definition, we treat the three criteria—depth, diffusion, and duration—as somewhat interchangeable.

NBER Business Cycle Dating Committee

More concretely, the NBER looks at some key variables,

Because a recession must influence the economy broadly and not be confined to one sector, the committee emphasizes economy-wide measures of economic activity. The determination of the months of peaks and troughs is based on a range of monthly measures of aggregate real economic activity published by the federal statistical agencies. These include real personal income less transfers, nonfarm payroll employment, employment as measured by the household survey, real personal consumption expenditures, wholesale-retail sales adjusted for price changes, and industrial production. There is no fixed rule about what measures contribute information to the process or how they are weighted in our decisions. In recent decades, the two measures we have put the most weight on are real personal income less transfers and nonfarm payroll employment.

NBER Business Cycle Dating Committee

Based on recent behavior of the six variables, cited by the NBER, it seems difficult to build a case for a recession. In the figure below, data for 2022 is colored red to focus attention on recent history. To be sure, real personal income (excluding transfer payments) has been flat through 2022, but it has not fallen as it did in the 2001 and 2007-09 recessions. Employment as measured by both the Employment and Household surveys is at or above its pre-pandemic levels. While real personal consumption expenditures fell sharply during the pandemic, they have since returned to a trend line starting in the mid-2010s. Although real retail sales have been flat through 2022, they are nonetheless running above their pre-pandemic trend. Finally, since 2012, industrial production has fluctuated around a flat line, and is currently above its 2012-2019 average. In summary, there are no obvious red flags, and it seems difficult to imagine that the NBER will decide that the U.S. was in recession at any time in 2022 based on the data at hand.

Sources: Federal Reserve Bank of St. Louis, Bureau of Economic Analysis, and Board of Governors, Bureau of Labor Statistics

Interestingly, the media and other analysts talk quite a bit about some things left out of the above two measures. Stocks, housing and the yield curve among others. Forbes, for example, on October 19 says: The Yield Curve, A Reliable Recession Indicator, Just Sounded An Alarm. The Conference Board’s Leading Economic Index (LEI) includes these and other measures.

The ten components of The Conference Board Leading Economic Index® for the U.S. include: Average weekly hours in manufacturing; Average weekly initial claims for unemployment insurance; Manufacturers’ new orders for consumer goods and materials; ISM® Index of New Orders; Manufacturers’ new orders for nondefense capital goods excluding aircraft orders; Building permits for new private housing units; S&P 500® Index of Stock Prices; Leading Credit Index; Interest rate spread (10-year Treasury bonds less federal funds rate); Average consumer expectations for business conditions.

Conference Board

Regarding the current outlook, the Conference Board states:

The US LEI fell again in September and its persistent downward trajectory in recent months suggests a recession is increasingly likely before yearend. The six-month growth rate of the LEI fell deeper into negative territory in September, and weaknesses among the leading indicators were widespread.

Ataman Ozyildirim, Senior Director, Economics, The Conference Board
US LEI declined further signaling an elevated likelihood of recession 

The figure below presents many of the series that comprise the Leading Economic Index. We could not locate a publicly-available source for the ISM Index of New Orders; hopefully new orders for consumer goods captures similar information. Likewise, absent data for the Leading Credit Index, we used the Chicago Fed’s Financial Conditions Credit Index. In 2022, the stock market is down as are housing starts and consumer sentiment. The increase in the financial conditions credit index indicate a tightening of credit conditions, a worrying development, but its level is still relatively low. The spread between the 10-year treasury yield and the Fed funds rate has narrowed, but does not reflect an inverted yield curve. Although average hours in manufacturing are down relative to the pre-pandemic period, they have been quite stable for the past year and a half. Further, initial unemployment insurance claims have come down since the onset of the pandemic, and have stayed low; in past recessions, claims tend to rise during a recession. Finally, new orders for both consumer and capital goods have risen through 2022, and are at historically high levels. In total, the individual series underlying the Leading Economic Index give a decidedly muddled answer to whether the U.S. is either on the verge of recession, or already in the midst of one.

Sources: Federal Reserve Bank of St. Louis, Federal Reserve Bank of Chicago, Bureau of Economic Analysis, Board of Governors, Bureau of Labor Statistics, University of Michigan

Conventional wisdom holds that the yield curve inverts — short rates exceed long rates — prior to recessions. Above, the LCI looks at one particular measure of inversion: the 10-year treasury yield less the Federal funds rate (an overnight rate). The LCI’s approach is useful when looking at the behavior of the yield curve over time. However, the yield curve includes a variety of maturities, and the LCI misses this dimension. Below, we plot yield curves for U.S. treasuries for select days in 2022. Early in 2022, the yield curve had the usual upward slope, although the short end of the curve was quite flat. Over the year, the yield curve has risen, reflecting a rise in yields across maturities. However, the short end has risen by more than the long end. While the yield curve was still positively sloped in early July, by early August the one year rate exceeded those out to 10 years, and the 20 year rate exceeded the 30 year rate. The August pattern prevailed through September and October. To reconcile these observations with the yield curve summary measure in the LCI, notice that the shortest maturity in the yield curves figure is one month while the LCI uses an overnight rate, the Federal funds rate.

Source: U.S. Treasury

Historical empirical regularities — such as, the yield curve inverts ahead of recessions — are likely to fall apart when relevant conditions change. In the case of the yield curve, what changed was the Fed’s Quantitative Easing program, and now unwinding. Exceptionally, Quantitative Easing saw large scale purchases of long term treasuries by the Fed. The figure below serves as a reminder that until the 2007-08 Financial Crisis, treasuries with maturities over five years constituted a small fraction of overall Fed treasury holdings; by 2013, they were the majority. While the share of longer term treasuries has since declined, they are still a much larger proportion of the Fed’s treasuries portfolio.

Source: Federal Reserve Bank of St. Louis

Looking at the changes in the Fed’s treasury holdings by maturity gives some idea of the magnitude of the Fed’s interventions in treasuries markets. The figure below reports monthly changes in the Fed’s portfolio of treasuries, calculated from the reported weekly changes. And we group the changes in the Fed’s portfolio into short (under 1 years maturity), medium (1-5 years maturity) and long maturities (over 5 years). The increases in overall Fed holdings of treasuries in 2009, 2011, 2013 and 2020 can be seen as increases in both short and long maturity treasuries. While it seems curious that between 2014 and 2018, large increases in the Fed’s holdings of short maturity treasuries were largely offset by decreases in its holdings of long maturity treasuries, this is not our current focus.

Source: Federal Reserve Bank of St. Louis

By reducing the supply of long maturity treasuries available to the public, the Fed pushed up the price, thereby lowering the yields on long maturity treasuries. In other words, Quantitative Easing likely flattened the yield curve when the Fed was actively accumulating long maturity treasuries.

The figure below zooms in on 2022. Starting in April, the Fed reduced its holdings of medium term treasuries. For reasons just discussed, these actions tend to increase medium term treasury yields. In the first half of 2022, the Fed accumulated both short and long maturity treasuries, thereby lowering their yields. The net effect of these changes in Fed treasury positions is to push up the `middle’ of the yield curve while pulling down the shot and long ends. In September and October of 2022, the Fed has reduced its holdings of short, medium and long maturity treasuries which serve to raise rates at all maturities, thereby raising the yield curve. The Fed may have good reasons for the observed changes in its portfolio of treasuries. The point is simply that these actions have likely introduced noise into whatever signal is coming from the shape of the yield curve.

At this point, readers may sympathize with U.S. President Harry Truman’s wish for a one-handed economist. There are, of course, many who attempt to boil the state of the economy down to a single number. For example, Jeremy Piger at the University of Oregon uses a dynamic-factor Markov-switching model first developed by Marcelle Chauvet at the University of California at Riverside, to determine the probability of being in a recession. Piger writes:

Monthly smoothed recession probabilities are calculated from a dynamic-factor Markov-switching (DFMS) model applied to four monthly coincident variables: non-farm payroll employment, the index of industrial production, real personal income excluding transfer payments, and real manufacturing and trade sales.

Jeremy Piger

To determine whether the model is good at predicting a recession, Piger then says:

Historically, three consecutive months of smoothed probabilities above 80% has been a reliable signal of the start of a new recession, while three consecutive months of smoothed probabilities below 20% has been a reliable signal of the start of a new expansion.

Jeremy Piger

Piger’s model produces the following graph, showing that as of August, 2022, the probability of a recession is just 1.22%.

Piger’s model lines up pretty well with the NBER recessions save for a couple notable exceptions. The NBER called a recession in March of 2001 but according to Piger’s model the recession probability never reached 50%. Indeed, during that recession, real GDP did not decline in 2 consecutive quarters but did have 2 quarters of decline during 2001. The 1990-1991 recession also never made it to 80% and again did not have 2 consecutive quarters of real GDP decline, but did have 2 quarters declining during the recession.

The message from the data seems pretty clear. The data that many people use to determine if we are in recession territory tell us that it seems unlikely. It is possible that the NBER, a few months from now, will deem that we were indeed in a recession at some point in the past, but the data today is not speaking the same language.

Solid September Employment Report

By Paul Gomme and Peter Rupert

The BLS announced that payroll employment increased 263,000 in September with an additional 11,000 revision to July’s numbers. Employment gains were widespread with only retail (-1,100), transportation and warehousing (-7,900) and government sectors declining (-25,000). The service sector added 244,000 jobs, of which health care (75,000) and leisure and hospitality (83,000) the prime movers.

Figure 1
Figure 2
Figure 3

Employment in retail has traditionally been higher than that in the leisure and hospitality sector. After the Great Recession leisure and hospitality employment grew much faster than retail and overtook retail employment in 2018. The pandemic had a devastating effect on employment in both sectors with leisure and hospitality taking a much larger hit. Recovery was pretty swift and the two sectors are now equal in terms of employment. It remains to be seen if leisure and hospitality employment will again outpace retail employment.

Figure 4

Average hours of work remained at 34.5 and with the employment increase lead to a 0.17% increase in total hours worked. Average hourly earnings increased from $32.36 to $32.46. Year over year CPI inflation is hovering around 8% and average hourly earnings over the year have risen 5%, leading to a 3% decline in real earnings as shown in Figure 6.

Figure 5
Figure 6

Figure 7 takes a longer view of the average real wage in the US. In this figure, we have normalized average hourly earnings, the CPI, and PCE deflator, setting each equal to 100 in March 2006 — the first available date for average hourly earnings. We have used a logarithmic scale for the vertical (y) axis so that a straight line represents a constant growth rate. The average real wage is given by the difference between average hourly earnings and a price index, represented by the gap between the average hourly earnings line, and that of a price index. In fact, since we have used a logarithmic scale, a constant gap implies a constant percentage gain in the average real wage, measured since March 2006. For the most part, there is a positive gap between average hourly earnings and either the CPI or PCE deflator. This means that, relative to March 2006, the average real wage has, for the most part, increased. The exception to this pattern of real wage gains is in the midst of the Great Recession (2007-09) when the CPI line lies above that of average hourly earnings which means that real wages declined relative to March 2003.

Figure 7

Figure 8 redoes the analysis underlying Figure 7, but focusing on the COVID-19 pandemic and its aftermath. To this end, we normalize average hourly earnings and the price indices to equal 100 in February 2020. As with Figure 7, the average real wage is the difference between average hourly earnings and either the CPI or PCE deflator. Once more, real wage gains are visually represented by the gap between the average hourly earnings line and that of a price index. Measured again by the CPI, the average wage was higher than its pre-pandemic level through 2020 and 2021. However, starting in early 2022, the CPI line rises above that of average hourly earnings which means that the average real wage has fallen, again relative to February 2020. An advantage of Figure 8 over a plot of the real wage is that it gives some perspective on the source of real wage movements. In particular, since mid-2020, nominal wages have grown at roughly a constant rate (the average hourly earnings line is almost a straight line). It’s the increase in the CPI, starting in 2021, that has eventually led to a decline in real wages. However, using the Fed’s preferred measure of the price level, the PCE deflator, shows that the average real wage has increased relative to February 2020, although these gains have eroded somewhat in 2022.

Figure 8

The household survey showed a 57,000 decline in the labor force. The number of unemployed persons fell 261,000, the number employed increased 204,000, leading to a decline in the unemployment rate from 3.66% to 3.49%. Note that the labor force participation rate remains lower than it was pre-pandemic.

Figure 9
Figure 10

Initial and continued unemployment claims were released last week and there was a hefty uptick in initial claims, climbing to 219,000 and a moderate increase in continued claims, rising to 1,361,000.

Figure 11
Figure 12

Although there has been a bit of a slowdown in some parts of the economy, the labor market continues to look pretty strong. Job vacancies have declined somewhat in recent months but there are still about 2 vacancies for each unemployed person.

Figure 13