Some of this conversation is motivated by an indicator proposed by economist Claudia Sahm and named after her as the Sahm rule. This indicator was originally created to identify recessions in real time so that it could be used by policy makers to trigger supporting policies. The indicator looks at how unemployment compares with recent levels. Historically, an increase of 0.5% relative to baseline is a strong indicator of the beginning of a recession. In the post-WWII period this indicator has produced two false positives although in both cases a recession followed soon after (see Sahm (2024) for a detailed description of the indicator).
While the indicator was not designed to predict recessions, its strong correlation with recessionary dynamics has been enough to trigger speculation about whether a recession is imminent, very much in the spirit of what happens when the yield curve inverts. Statistical regularities that preceded previous recessions might not have much power to predict the next recession because the number of business cycles is small and, in addition, each of them can be different in nature. Even if a certain indicator has always preceded all previous recessions, “this time could be different”. So far the labor market recovery from the 2020 recession has indeed been very different from any of the previous recoveries but as we look forward we face the question of whether it can continue doing so.
The US labor market during expansions
The Sahm rule is related to a set of broader patterns of the US labor market that are shared by all cycles in the post-WWII period. Outside of recessions, the US labor market is almost always in a state of healing, with the unemployment rate decreasing after a recession. Apart from these long and uninterrupted episodes of healing, we only observe very short periods of stable (and low) unemployment. During these periods unemployment moves sideways with some tendency to slowly creep up, until the Sahm rule is triggered, and the episode is interrupted by the next recession (see Figure 1).
To further understand this pattern, we take a close look at the most recent four US business cycles. We focus on the recovery period, defined to start the month after the peak of unemployment, which always happens a few months after the trough defined by the NBER. We then count the number of uninterrupted months where the unemployment rate was on a downward trend, we will refer to this period as the “healing period”. We define a downward trend as a negative change in the 3-month moving average of the difference between monthly unemployment and the average unemployment rate during the previous 12 months (this metric is close to the one used in the Sahm rule). This healing phase is always followed by an episode of stable or increasing unemployment, although at a very small rate in some cases, until the next recession starts. We refer to this phase as the “pre-recession” period. These latter episodes are very short relative to the healing period. Table 1 summarizes how the data splits between these two phases during each of the last four recessions.
Table 1. The two phases of the US labor market recovery
| Number of months | |
Recession | Healing period | Pre-recession |
1990 | 98 | 2 |
2001 | 44 | 5 |
2008 | 117 | 0 |
2020 | 32 | 13 (so far) |
The table above highlights several important and related facts:
1. In all four cycles most of the recovery is spent in the uninterrupted healing phase. The typical behavior of unemployment outside of a recession is decreasing.
2. The speed of adjustment during the healing phase is low so that it takes many years to return to a low unemployment rate. Compared to most labor market models, the speed of adjustment tends to be too low and too stable over time. Most theories predict that the speed of adjustment should be much faster in the early parts of the recovery. This supports theories of congestion in the labor market. (Hall and Kudlyak (2021), Mercan, Schoefer, and Sedláček (2024)).
3. Full employment does not seem to be a natural state of the labor market. Instead, the labor market seems to almost always be healing from a recession (Fatas (2021a)). The episodes of stable and low unemployment are very short. In fact, during the majority of what we call the “pre-recession” phase, unemployment is trending upwards. Once unemployment reaches the end of its healing period, a recession follows soon.
4. The US labor market does not display any slowdown periods or “mini-recessions”, where small shocks push the unemployment rate up by a small amount before returning to equilibrium. The dynamics of unemployment are dominated by the very long process of uninterrupted healing from large and infrequent shocks (recessions).
Do we observe the same patterns in other countries? Not as strong as the US. The US labor market is unique because the patterns described above are prevalent across all cycles. In other countries we also see dynamics that are not present in the US.
For example, in France, the expansion of the early 1990s started with the usual healing period but once unemployment reached low levels in the early 2000s, it remained broadly stable until the 2008 recessions alternating periods of increasing and decreasing unemployment (this figure uses AFSE (2024) recession dates for France).
In the UK we see a similar pattern during the same decades with unemployment remaining stable for a period of 8 years in the early 2000s with some fluctuations in both directions (for UK recession dates we use Broadberry et al (2023) and ONS (2022)).
These episodes that we observe in France and the UK where the labor market spends several years without a clear trend and small fluctuations in both directions are absent in the US labor market. This is the reason why the statistical observation that drives the Sahm rule works well in the US. When unemployment is not decreasing, it is likely that a recession is around the corner, and a short period of even the smallest increase sends an even stronger signal that a recession is imminent.
This time is different. For how long?
The last cycle in the US labor market is unique on many counts. From Table 1, we can see that the healing period was very short compared to the other cycles, more so if we consider the high level of unemployment from which it started. The speed of adjustment towards a low unemployment rate was much faster than in any other expansion.
But what is more relevant to understand today’s labor market is that the period of low and stable unemployment has already been much longer than anything we have seen in the previous cycles. The US labor market has already spent 13 months after the healing period ended and this represents about one third of the expansion time. This pattern is very different from previous cycles.
One could argue that the US economy has achieved, from a labor market point of view, a very successful and long-lasting soft landing. But now the question is about how long the US labor market can stay in that state. Previous cycles provide a pessimistic answer to that question as these periods always tended to be short in nature. But like with any other empirical regularity, such as the yield curve inversion, using patterns of unemployment to predict the next recession ignores the possibility that this cycle might be different.
From a theoretical point of view, we do not have much guidance. Why did previous periods of low unemployment end soon? One possibility is simply that a very long healing period combined with the inevitability of some large shocks along the way make it very unlikely to see long episodes of full employment (Fatas(2021b)). But, once again, we need to consider that this time “was different”. The healing period was much shorter and, as a result, while the post-healing period is longer than in other cycles, the length of the expansion is not. That allows for an optimistic reading even if one sticks to empirical regularities as there might be room for more years of a tight labor market while we wait for the next shock that sends the US economy into a recession. And as we have shown, we have observed those patterns in other countries of stable or even weakening at times labor market for a few years without a recession.
References
AFSE (2024), French business cycle dating committee.
Broadberry, Stephen, Jagjit S. Chadha, Jason Lennard, and Ryland Thomas (2023), "Dating business cycles in the United Kingdom, 1700–2010." The Economic History Review 76.4 (2023): 1141-1162.
Fatas, Antonio (2021a), “The short-lived high-pressure economy”. VoxEU blog post.
Fatas, Antonio (2021b), “The Elusive State of Full Employment”, CEPR Discussion Paper 16535.
Hall, Robert and Marianna Kudlyak (2021), "Why Has the US Economy Recovered So Consistently from Every Recession in the Past 70 Years?", NBER Macroeconomics Annual.
Mercan, Yusuf, Benjamin Schoefer, and Petr Sedláček (2024). "A congestion theory of unemployment fluctuations." American Economic Journal: Macroeconomics 16.1: 238-285.
ONS (2022), Communicating the UK Economic Cycle.
Sahm, Claudia (2024), “Sahm-thing more on the Sahm rule”. Blog post.