Interpreting US labour market conditions has become unusually difficult in recent years. Payroll growth slowed markedly after 2022 and vacancies declined substantially, with job seekers increasingly reporting greater difficulty finding jobs. At the same time, the unemployment rate rose only modestly, wage growth remained above pre-pandemic norms, and many firms, especially small businesses, continued to report persistent hiring difficulties.

This contradictory combination of labour market signals has complicated the policy debate. Did the economy achieve a genuine ‘soft landing’ after the Federal Reserve’s tightening cycle, or are conventional indicators understating labour market weakness? Are observed short-term fluctuations obfuscate deeper structural shift in the labour market? More broadly, how should policymakers interpret a labour market in which different indicators appear to tell conflicting stories?

Recent VoxEU columns have highlighted the difficulty of interpreting labour market indicators in real time. Gottfries (2023) emphasises that movements in vacancies and unemployment do not always admit a simple interpretation in terms of labour market tightness alone, because the Beveridge curve itself shifts. Similarly, Stock et al. (2014) show that declines in labour force participation can reflect a combination of demographic, cyclical, and other structural forces. These examples illustrate a broader challenge: individual labour market indicators rarely map cleanly into a single underlying economic narrative. 

In the academic literature, Barlevy et al. (2024) document these shifts systematically, tracing the distinct roles of demand, supply, and matching efficiency in Beveridge curve dynamics over six decades of US data, and Brave et al. (2026) use them to motivate a sign-restricted VAR representation to identify their relative importance

Our recent research builds on this discussion by proposing a framework that organizes many labour market indicators into a small set of economically interpretable ‘narrative factors’ that can be used to expand and refine the interpretation of labour market developments (Brave et al. 2026).

Importantly, these narratives closely mirror the language policymakers already use. Federal Reserve officials routinely discuss “labour demand”, “labour supply”, and “matching efficiency” when explaining labour market developments. For example, Jerome Powell argued in June 2024 that “supply and demand conditions have come into better balance”, while the 2024 Economic Report of the President pointed to improving “matching efficiency” following the pandemic. Our framework helps to formalise the narrative concepts policymakers already use implicitly when interpreting labour market data and organizes them around generally agreed upon textbook labour market principles rather than having to rely on heuristics with only a few headline indicators. 

From many indicators to a few narratives

No single labour market indicator provides a clean and accurate signal about underlying economic conditions, especially in real time. Payroll employment, unemployment, vacancies, quits, labour force participation, wages, labour market flows, and business surveys each capture important information about labour market activity; but every series is noisy, revised over time, and influenced by multiple underlying forces simultaneously.

We combine information from 94 monthly and quarterly labour market time series spanning 1960–2026 and, using use principal components analysis (Stock and Watson 2002), summarise the common variation across these series. The resulting statistical factors are then rotated using a small number of economically motivated sign restrictions derived from standard textbook labour market models. The outcome of this rotation is a set of narrative factors that explain the same degree of variation as the original principal components, but where each narrative factor has a clear economic interpretation. Not all common variation in labour market indicators admits such a clean interpretation, however. Our framework therefore also retains a residual ‘kitchen sink’ factor that captures systematic movements in the data not naturally associated with the four standard textbook labour-market narratives we identify. Importantly, the economically interpretable factors still account for the overwhelming majority of cyclical labour-market dynamics.

Figure 1 illustrates the logic: each narrative factor corresponds to a distinct movement in the space of wages, employment, unemployment, and vacancies, with the implied sign restrictions summarised in the table in panel (e). Stronger labour demand, for example, should raise employment, vacancies, and wages while lowering unemployment; improved matching efficiency should reduce both unemployment and vacancies simultaneously. In our analysis, we further distinguish between long-run and short-run labour supply forces, motivated by recent studies (Elsby et al. 2015, Hobijn and Şahin 2022, Cairó et al. 2022) as well as policy discussions surrounding the post-2007 decline in labour force participation (Stock et al. 2014), all of which highlight differences in their labour market implications. While demographic trends such as growth in the working-age population primarily shape long-run labour supply, macroeconomic conditions influence labour supply in the short run by affecting workers’ search intensity, labour force attachment, and wage-setting behaviour. For instance, during recessions, more workers may lose their jobs and become willing to accept lower-paying positions, increasing short-run labour supply. As a result, higher employment growth in the initial phase of the recovery need not generate wage pressures.

Figure 1 Textbook illustrations of labour market narratives and the implied sign restrictions 

Our sign restrictions are intentionally sparse and do not uniquely determine the factors. The framework, therefore, characterises the full range of factor paths consistent with both the data and the identifying assumptions — i.e. the ‘narrative feasible set’. Put differently, instead of asking, What does the unemployment rate say?, we ask: What combination of labour demand, labour supply, and matching frictions can jointly explain the full set of labour market data? A striking feature of the results is that the estimated paths remain remarkably tightly pinned down despite this minimalist structure, delivering four factors with clear and economically intuitive time-series properties.

Figure 2 plots the narrative feasible sets for the four labour market factors over time. The shaded regions represent the range of factor paths consistent with both the observed data and the identifying restrictions. Despite the intentionally sparse restrictions, the feasible sets remain remarkably tight during most historical episodes, suggesting that the cross-sectional richness of the labour-market data strongly disciplines the admissible economic narratives.

Labour demand matters – but so does short-run labour supply

Labour demand and short-run labour supply explain most cyclical labour market fluctuations. The labour demand factor is strongly procyclical, declining sharply during recessions and recovering gradually afterward. It captures well-known business cycle episodes, including the sluggish recoveries following the 1991, 2001, and 2007–09 recessions. Long-run labour supply, by contrast, reflects slower-moving structural developments – demographic change, population growth, and immigration – and explains relatively little short-run variation.

More novel is the behaviour of short-run labour supply. This factor is strongly countercyclical: it tends to improve during labour market recoveries but deteriorate late in expansions and around recessions. Importantly, it should not be interpreted simply as labour force participation. Rather, it captures the effective elasticity of labour supply – the ease with which firms can attract additional workers without substantially raising wages. Early in recoveries, unemployed workers are readily available and reservation wages are comparatively low, so firms can expand employment without strong upward pressure on wages. Later in expansions, the pool of readily employable workers shrinks and reservation wages rise, making short-run supply increasingly inelastic.

This distinction matters for policy interpretation. Labour force participation and unemployment do not mechanically reveal whether developments are driven by supply or demand but instead reflect a combination of both. 

Understanding the post-pandemic ‘soft landing’

Our framework helps to interpret the puzzling dynamics of the post-pandemic labour market.  Following the COVID-19 recession, labour demand rebounded briskly, exceeding pre-pandemic levels by mid-2021. But labour demand alone cannot explain the historically elevated vacancy rates, elevated quits, and widespread hiring difficulties observed during the ‘Great Resignation’. Declining matching efficiency and unusual movements in short-run labour supply also played important roles.

Beginning in 2022, labour demand cooled steadily. The decline in magnitude resembled the weakening normally associated with recessions, but at a relatively gradual pace. Under ordinary historical circumstances, such a contraction would have produced a substantially larger rise in unemployment. Instead, unemployment rose only modestly, because short-run labour supply remained tight during the period. This joint contraction – cooling demand combined with declining short-run labour supply – is the defining feature of the soft landing, and is visible in Figure 2.

Figure 2 Narrative feasible sets

The quantitative implications of this unique set of circumstances are striking. The weakness in short-run labour supply reduced nonfarm payroll growth by roughly 90,000 jobs per month over the past year while simultaneously suppressing the unemployment rate by approximately 0.75 percentage points as of February 2026. Figure 3 shows these decompositions directly for the unemployment rate and payroll growth, making clear how much of the observed soft landing reflects supply-side factors rather than demand alone.

Figure 3 Post-COVID decompositions of the unemployment rate and nonfarm payroll growth

(a) Nonfarm payrolls 12-month growth

(b) Unemployment rate

This interpretation also resolves what otherwise appears paradoxical: households increasingly report difficulty finding jobs even as firms continue to struggle to fill vacancies. In the narrative framework, these observations are not contradictory. They reflect the joint effect of subdued demand and constrained short-run supply: a state in which neither side of the labour market finds conditions particularly favourable.

Interpreting imperfect real-time data

The broader contribution of the framework is methodological. Real-time labour market analysis is inherently difficult because labour market data are noisy, incomplete, revised frequently, and released at different times. Our framework exploits the cross-sectional richness of these data to address those problems. Because the factors are estimated jointly from many series, the approach naturally accommodates missing observations, publication timing differences, overlapping measures of similar concepts, and data revisions, all without requiring the analyst to commit to a single preferred series.

The results underscore a broader lesson: no individual labour market indicator should be interpreted in isolation. Payroll growth, vacancies, unemployment, and participation each reflect multiple underlying forces simultaneously. A decline in vacancies need not imply collapsing labour demand but can reflect declining turnover. A lower payroll growth rate need not signal labour market deterioration if short-run labour supply is weakening at the same time.

As labour markets continue to evolve through changes like demographic ageing, immigration shifts, remote work, and AI adoption, policymakers are likely to face increasingly complex and sometimes contradictory signals. In this environment, the central challenge is to organise the data into coherent economic narratives without relying too much on just a handful of indicators, and our framework provides a systematic way of doing so. 

References

Barlevy, G, R J Faberman, B Hobijn and A Şahin (2024), “The Shifting Reasons for Beveridge Curve Shifts,” Journal of Economic Perspectives 38(2): 83–106.

Brave, S A, R A Butters, R Cole, and M Gillet (2026), “The U.S. Labor Market: A Long-Run Perspective”, Economic Perspectives No. 1, Federal Reserve Bank of Chicago. 

Brave, S, E Crust, S Eusepi, B Hobijn and A Şahin (2026), “Making Sense of Labor-Market Indicators Amid Data Imperfections”, paper prepared for the Brookings Papers on Economic Activity Spring 2026 Conference and FRB Chicago Working Paper.

Cairó, I, S Fujita and C Morales-Jiménez (2022), “The cyclicality of labor force participation flows: The role of labor supply elasticities and wage rigidity”, Review of Economic Dynamics.

Elsby, M W, B Hobijn and A Şahin (2015), “On the importance of the participation margin for labor market fluctuations”, Journal of Monetary Economics 72: 64–82.

Hobijn, B and A Şahin (2022), “Maximum Employment and the Participation Cycle”, Proceedings of the 2021 Jackson Hole Economic Symposium, pp. 273–372.

Gottfries, N (2023), “Reinterpreting the Beveridge Curve”, VoxEU.org, 12 December.

Stock, J, B Stevenson, J Coglianese, J Furman and S Braun (2014), “Understanding the decline in the labour force participation rate in the United States,” VoxEU.org, 18 August.

Stock, J H and M W Watson (2002), “Forecasting Using Principal Components from a Large Number of Predictors”, Journal of the American Statistical Association 97(460): 1167–1179.