The Conference Board New Experimental Nowcast of Job Openings
March 5, 2024
March's JOLTS release expected to revise 2023 data upwards
The Job Openings and Labor Turnover Survey (JOLTS) has been published by the Bureau of Labor Statistics (BLS) since 2000. JOLTS provides a measure of job openings, hires, and separations. It employs a monthly rotational system wherein longitudinal panels of establishments (firms and businesses) are brought in and out of the sample. This survey design serves to uphold the crucial principle of randomness, fundamental for ensuring the robustness of statistical extrapolation. Response rates for JOLTS peaked at 68% in April 2014, but have recently dropped significantly, with a reported response rate of 32.4% in September 2023. Low survey response rates are an inherent weakness of all surveys, and its effect can be observed on JOLTS data.
The Conference Board is experimenting with a nowcast and innovative data initiative leveraging a proprietary database of job boards using Lightcast microdata to gauge the extent of job openings within diverse labor markets. This data initiative, with a present scope dating back to February 2015, harnesses advanced machine learning methodologies to generate estimations of historical, current, and projected jobs openings data.
Due to strong employment growth over the course of 2023, The Conference Board expects upward revisions to JOLTS openings in the same period. These revisions are expected to come forth on March 6, 2024, when the BLS conducts annual benchmark revisions to the JOLTS data. The experimental nowcast anticipates these revisions on the national aggregate level as the following graphic shows. The data series also has a two-months-ahead nowcast compared to JOLTS. For January 2024, we expect job openings at the national level, seasonally adjusted, to be 8.51 million, and in February 2024, 8.81 million.
Regarding industry-specific trends, the primary catalyst for these upward benchmark revisions appears to stem from the Manufacturing and Retail Trade sectors, both of which demonstrated notable deviations from JOLTS data throughout 2023, as illustrated in Figure 2 of the nowcast. While JOLTS indicates a steady decline within these sectors over the course of 2023, the Nowcast series diverges, suggesting a slight upturn in Manufacturing sector job openings and a stable trend in retail trade job openings during the same period. Despite both sectors exhibiting relatively stagnant payroll employment growth year-over-year (0.3% for Manufacturing and 0.9% for Retail Trade) alongside modest increases in corporate profit margins as per Q3 2023 GDP data, the productivity of Manufacturing workers remained unchanged from 2017 to 2023. Additionally, nonfarm business worker productivity experienced only marginal growth throughout 2023, and no noticeable economic factors account for the downward trajectory in job openings as estimated by JOLTS.
The key advantage of the experimental nowcast over conventional JOLTS reporting lies in its reliance on the analysis of job board data, which can provide more timely data releases. While JOLTS relies on a monthly survey-based data collection process, with insights lagging by 30-40 days, our experimental nowcast delivers estimations of job openings across labor markets in near real-time, approximately 5 weeks ahead of JOLTS releases. Our higher frequency data updates and independent models enable us to produce timely estimations of job openings at the state, industry (two-digit NAICS), MSA (top 18 by labor force population), and national levels.
The experimental nowcast differs by maintaining a consistent set of job boards over time. This is achieved through a meticulous and proprietary data cleansing procedure resulting in a selection of job boards with forward-looking reliability that provide consistent stratum representation throughout the sample period. The resulting subset of boards are inputs into our proprietary ensemble modeling process, which yields a model-based estimation of job openings.
The Conference Board will continue to analyze and monitor the performance of the experimental nowcast and JOLTS.
