Can Earnings Calls Be Used to Gauge Labor Market Tightness?

Labor market tightness is a key labor issue that firms face and an important consideration in monetary policy decisions. Economists typically measure tightness as the ratio of job vacancies to unemployed workers.Data on the number of job vacancies are from the monthly Job Openings and Labor Turnover Survey conducted by the U.S. Bureau of Labor Statistics. Data on the number of unemployed workers are from the Current Population Survey, which is conducted by the U.S. Census Bureau for the Bureau of Labor Statistics. However, this measure of labor market tightness can only be disaggregated by industry, region and broad sizes of establishments. In this blog post, we employ a novel, textual analysis of corporate earnings calls to obtain a measure of labor issues that each publicly traded firm faces.The authors plan to publish a forthcoming St. Louis Fed working paper on this research.

Textual Analysis of Earnings Calls

Earnings calls are quarterly meetings where executives of publicly traded companies discuss the financials of their firms and factors that may impact current and future performance, as well as answer questions from stock analysts and institutional investors on information not contained in their earnings press release or other regulatory filings. As such, they are a useful source of data on firms discussing hiring challenges, union disputes, new innovations or outsourcing. For this analysis, we parsed the transcripts of 171,208 earnings calls from 6,456 U.S. firms between the first quarter of 2002 and the first quarter of 2024. The text, supplied by S&P Global, amounts to 60 gigabytes of data. Recent studies have used textual analysis of earnings conference calls to develop measures of firm-level concerns related to political risk, climate risk, and epidemic risk and sentiment.

In this blog post, we describe the construction of our measure of labor issues faced by firms. We built our measure by first identifying keywords, such as “labor shortage” and “labor costs.” These keywords, which we call “seed keywords,” could plausibly appear in text when firms are discussing labor issues. We then used a natural language processing model that identified other phrases with similar meanings—such as “wage pressures” or “salary inflation”—based on their context to the seed keywords.

To validate that these keywords indeed identify discussion of labor-related issues, we performed a human audit by reading examples of individual excerpts that make use of each phrase to confirm that they corresponded to actual talk of labor issues in the context of the transcript. We then constructed our labor issues measure by counting the frequency of the keywords and then dividing by the length of the transcript. While more aggregate data on hiring, quits or job vacancies can show labor issues at the industry or regional level, these transcript data are unique in that they allow us to explore firm-level labor issues.

The table below shows the keywords we identified, along with the number of earnings calls they appeared in and a sample excerpt demonstrating the context surrounding these keywords. Generally speaking, firms discussed difficulties finding workers and the need to pay existing workers more. Frequently, these mentions of labor issues appeared when firms discussed lower profitability or used the issues as a rationale for increasing prices. Some cited higher wages, while others discussed industry or even economywide issues.

Notably, the vast majority of excerpts in the data contained negative sentiment. As a result, we can interpret an increase in mentions of labor issues as a situation in which firms are facing more labor issues, rather than resolving existing ones.

Mentions of Labor Issues in Earnings Calls and Labor Market Tightness

While we have shown that our measure does actually capture firms discussing labor issues, another question is whether these discussions correlate with aggregate labor market tightness, as measured by the job openings-to-unemployed ratio. When this ratio exceeds 1, it means the number of workers that firms are trying to hire exceeds the number of workers without jobs.

The figure below compares aggregate labor market tightness with our measure of labor issues. For ease of interpretation, we normalize our index of labor issues to have an average of 1 in the pre-COVID-19 pandemic period between 2002 and 2019.We also chose to ignore firms from industries such as finance and real estate, where labor and hiring dynamics are substantially different from those in other industries.

U.S. Labor Market Tightness and Labor Issues Index

SOURCES: U.S. Bureau of Labor Statistics, S&P Global and authors’ calculations.

NOTES: Labor market tightness is measured using the ratio of U.S. job postings to unemployed workers. The labor issues measure, which counts the frequency of keywords in earnings calls and then divides them by the length of the transcript, has been normalized to average 1 between 2002 and 2019. Data begins with the first quarter of 2002 and goes to the fourth quarter of 2023 for labor market tightness and to the first quarter of 2024 for the labor issues measure. Shaded areas indicate a recession.

Over the entire period we examined, the correlation between the labor issues index from earnings calls and labor market tightness is quite high at 0.838. Both measures saw a decrease during the 2007-09 recession, and then an even sharper decline at the start of the pandemic due to shutdowns and decreased labor demand. Intuitively, this makes sense, since firms that are unable to be open or lack the necessary supplies are unlikely to need to hire new workers or pay existing workers more.

This decline was followed by a spike starting in late 2021 as employers struggled to keep up with high demand because of the economic reopening and stimulus payments to households. As many workers continued to stay home during the pandemic and firms faced strong demand for their products and services, employers found it difficult to hire enough workers and had to offer raises and bonuses to existing employees.

In the case of labor market tightness, the measure peaked in the second quarter of 2022 at 1.9, meaning that for each unemployed worker there were about two job vacancies. In the case of the labor issues index, its peak in the fourth quarter of 2021 represented a fivefold increase in mentions of labor issues relative to the pre-pandemic period. Consequently, our measure of labor issues seems to accurately track aggregate labor market tightness.

In conclusion, we constructed a novel, firm-level measure of labor issues from earnings calls and then showed that it closely tracked a measure of aggregate labor market tightness. Intuitively, firms more frequently discuss increasing wage costs, labor shortages and hiring difficulties when the labor market is tight, as was the case in the years following the start of the pandemic.

The key advantage of our methodology is that we can link this quarterly, firm-level measure of labor issues to the firms’ own employment and investment decisions—data available from Compustat—to investigate how they deal labor issues, such as those arising from a tight labor market. In our next blog post, we discuss whether a tight labor market prompts firms to automate some tasks, thereby reducing their future demand for labor.

Notes

1. Data on the number of job vacancies are from the monthly Job Openings and Labor Turnover Survey conducted by the U.S. Bureau of Labor Statistics. Data on the number of unemployed workers are from the Current Population Survey, which is conducted by the U.S. Census Bureau for the Bureau of Labor Statistics.
2. The authors plan to publish a forthcoming St. Louis Fed working paper on this research.
3. We also chose to ignore firms from industries such as finance and real estate, where labor and hiring dynamics are substantially different from those in other industries.