The Recent Ins and Outs of Unemployment: Using Flows to Study Labor Market Dynamics

The U.S. unemployment rate in August was 4.2%, a level that remains below the historical average. Yet the jobless rate has been gradually trending up since April 2023, when it last stood at a record low of 3.4%. Despite a modest decline in August, this overall upward trend has raised questions about the possibility of a soft landing and whether the U.S. may experience a recession.

In this blog post, we examine all potential flows into and out of unemployment to analyze the recent uptick in unemployment. We then compare the flows with those that occurred in past recessionary periods. By doing this, we hope to gain greater insight into the strength of today’s labor market and the state of the current business cycle.

Method and Data

We decomposed the monthly change in the unemployment rate into its inflow and outflow components using the following equation,

The variable u_t is the unemployment rate in month t, and n_t is the fraction of the population not in the labor force (NILF) in month t. EU_t and UE_t denote job separation and job finding rates, and those two variables determine the flows between unemployment and employment. Similarly, NU_t and UN_t determine flows between being unemployed and being outside of the labor force: NU_t is the share of people who were out of the labor force the previous month but are looking for work in the current month, and UN_t is the share of unemployed workers who decide to leave the labor force. Finally, the last term, o_t, is typically negligible for month-over-month changes and captures the changes in unemployment due to relatively small shifts in the labor force participation rate and population growth rate.

In our analysis, we used seasonally adjusted values of these variables constructed by the U.S. Bureau of Labor Statistics (BLS), utilizing monthly Current Population Survey (CPS) data between February 1990 and August 2024. Labor market flow rates were derived from individual workers’ month-over-month transitions between employed, unemployed and NILF status in the matched sample.See Harley J. Frazis, Edwin L. Robison, Thomas D. Evans and Martha A. Duff, “Estimating Gross Flows Consistent with Stocks in the CPS” (PDF), Monthly Labor Review, September 2005, pp. 3-9.

Unemployment Rate Increases Have Accelerated in Recent Months

The table below shows the average monthly change in the seasonally adjusted unemployment rate and the average monthly contributions of each flow component in the last three months, since January 2024, and since April 2023 when the unemployment rate started increasing, as well as between 2016 and 2019. The unemployment rate reached a low of 3.4% in early 2023 and has since been increasing, albeit not consistently, as tight monetary policy cools the economy. Since April 2023, the unemployment rate has increased by an average of 5 basis points per month. This increase accelerated to 9 basis points in the last three month, reflecting increases of 9 basis points in June, 20 basis points in July and a slight decline of 3 basis points in August.

Job Separations to Unemployment: The Force Behind Rising Unemployment

Our analysis of flow components using the previous formula reveals that the primary driver of the recent uptick in the unemployment rate is the increase in the job separation rate to unemployment.It is important to note that overall separations (not just to unemployment but also to NILF) have been gradually declining since 2021 and appear roughly constant in recent months. This trend is evident in both the CPS data and the Job Openings and Labor Turnover Survey (JOLTS) data (which additionally captures separations due to job-to-job transitions). The CPS data show a gradual downward trend for employment to NILF transitions, with an important decline in July 2024 and a modest recovery in August 2024. As a result, the observed decline in the total separations rate is not inconsistent with the uptick in the separations to unemployment that we highlight here, which primarily accounts for the increase in the unemployment rate. The average job separation component since April 2023 was 93 basis points, which then rose to 96 basis points in the year to date ending August 2024, and to 102 basis points for the last three months ending in August 2024. The importance of the job finding from unemployment component (u_t−1 × UE_t)—which reduces the unemployment rate—increased modestly (in absolute value) to -109 basis points in the last three months from a relatively stable average of -103 basis points since April 2023. However, this change was entirely driven by a larger stock of unemployed (u_t−1) rather than an improvement in the job finding rate (UE_t), which has hovered around 27% per month since April 2023. This pattern aligns with recent findings suggesting that a rise in the job separation rate leads the cyclical increase in the unemployment rate while the decline in the job finding rate tends to lag behind the rise in unemployment.For example, see Michael W.L. Elsby, Bart Hobijn and Ayşegül Şahin’s 2013 article, “Unemployment Dynamics in the OECD,” and Michael W.L. Elsby, Ryan Michaels and Gary Solon’s 2009 article, “The Ins and Outs of Cyclical Unemployment.”

We do not find a significant role for NILF flows in explaining the increase in the unemployment rate. Since April 2023, average flows from NILF to unemployment have been hovering around 105 basis points. Similarly, flows into NILF from unemployment remained relatively stable, ranging between -90 and -92 basis points. Consequently, there is no clear pattern for flows between NILF and unemployment that significantly contributes to the increase in the unemployment rate.

Historical Context: Comparing with Past Recessions

To provide perspective, we also compare the current period with the recessions in 1990-91, 2001 and 2007-09. The next four figures illustrate the smoothed evolution of transition rates between labor market statuses (EU_t, UE_t, NU_t and UN_t) around the onset of recessions. For example, the first figure shows the percentage of employed workers who lost or quit their job and found themselves unemployed in a reported month.

We designate as time 0 the peak month in economic activity in our sample of recessions, as identified by the National Bureau of Economic Research’s Business Cycle Dating Committee.We designate the peak as month 0, with the recession starting in the following month. Though the NBER defines the start of a recession as occurring sometime in the same month when economic activity peaks, we consider the start to be the first full month of economic contraction. As such, the first recession in our sample starts in August 1990, the second in April 2001, and the third in January 2008. We excluded the COVID-19 recession of 2020. We then contrast the evolution of flows before and during recessions with current trends, which covers the period from February 2023 to August 2024.

Flow Rate into Unemployment: Job Separations

Flow Rate out of Unemployment: Finding Jobs

Flow Rate into Unemployment: Joining the Labor Force

Flow Rate out of Unemployment: Leaving the Labor Force

SOURCES FOR ALL FIGURES: Bureau of Labor Statistics and authors’ calculations.

NOTES FOR ALL FIGURES: For past recessions, the dashed vertical line marks the peak month before the start of the downturn; for the current period, this line marks the last reported month, which is August 2024. Series are smoothed using a five-month centered moving average. Because of unavailable data, the 1990-91 time series doesn’t have a full 18 months of data before the downturn.

Although flow rates into unemployment are at historically low levels, the flow rate from employment has increased recently, as shown in the first figure above. Similar increases in the job loss rate have been observed before the start of past recessions. With regard to the job finding rate, its current level is comparable to those of previous periods. Moreover, its decline before the start of a recession is relatively muted—a pattern also seen in the current period. Typically, the job finding rate declines more sharply only after several months into a recession. As for the flows out of unemployment to NILF, we observe a sharp decline in the last six months—again, a pattern also seen in past recessions.

Keeping an Eye on Labor Market Dynamics

While current unemployment levels and flows into unemployment remain at historically low levels, the uptick in the last three months may signal more than a slight weakening in the labor market. It is crucial to keep a close eye on labor market dynamics—particularly job separation rates—in the coming months, as they may provide an early indication of a change in economic conditions.

Notes

1. See Harley J. Frazis, Edwin L. Robison, Thomas D. Evans and Martha A. Duff, “Estimating Gross Flows Consistent with Stocks in the CPS” (PDF), Monthly Labor Review, September 2005, pp. 3-9.
2. It is important to note that overall separations (not just to unemployment but also to NILF) have been gradually declining since 2021 and appear roughly constant in recent months. This trend is evident in both the CPS data and the Job Openings and Labor Turnover Survey (JOLTS) data (which additionally captures separations due to job-to-job transitions). The CPS data show a gradual downward trend for employment to NILF transitions, with an important decline in July 2024 and a modest recovery in August 2024. As a result, the observed decline in the total separations rate is not inconsistent with the uptick in the separations to unemployment that we highlight here, which primarily accounts for the increase in the unemployment rate.
3. For example, see Michael W.L. Elsby, Bart Hobijn and Ayşegül Şahin’s 2013 article, “Unemployment Dynamics in the OECD,” and Michael W.L. Elsby, Ryan Michaels and Gary Solon’s 2009 article, “The Ins and Outs of Cyclical Unemployment.”
4. We designate the peak as month 0, with the recession starting in the following month. Though the NBER defines the start of a recession as occurring sometime in the same month when economic activity peaks, we consider the start to be the first full month of economic contraction. As such, the first recession in our sample starts in August 1990, the second in April 2001, and the third in January 2008. We excluded the COVID-19 recession of 2020.