Adjusted Estimates of Worker Flows and Job Openings in JOLTS May 2008 Steven Davis, University of Chicago and NBER Jason Faberman, Federal Reserve Bank of Philadelphia John Haltiwanger, University of Maryland and NBER Ian Rucker, Bureau of Labor Statistics The views expressed are solely those of the authors and do not necessarily reflect the official positions or policies of the US Bureau of Labor Statistics, the US Bureau of the Census, the Federal Reserve Bank of Philadelphia, the Federal Reserve system or the views of their staff members.
Introduction The behavior of hires and separations is an important topic for labor and macroeconomics Relevant for several classes of models Labor search and matching, factor adjustment, etc. Useful in understanding important margins of cyclical employment fluctuations e.g., are recessions hires-driven or job loss-driven? JOLTS data are the best new source on hires and separations, but JOLTS data raise new questions Accurate measurement of magnitude, cyclicality of worker flows critical to understanding employment fluctuations
What is the JOLTS? JOLTS is a monthly survey of ~16,000 establishments Has employment, number of hires, number of separations (by quits, layoffs & discharges, and other separations) throughout month, and number of job openings at end of month Published statistics Begin in December 2000 Available by major industry and region for all nonfarm establishments Our study… Uses both published data and micro data for Jan-01 – Dec-06 Appeals to fact that JOLTS sample frame is the BED (QCEW) data Focuses only on private establishments
JOLTS Measurement Issues JOLTS has three notable measurement issues Two issues are observable in the published data JOLTS hires and separations estimates overstate growth relative to its benchmark estimates (CES)
Issue 1 – Overstatement of CES Growth
JOLTS Measurement Issues JOLTS has three notable measurement issues Two issues are observable in the published data JOLTS hires and separations estimates overstate growth relative to its benchmark estimates (CES) Magnitude of hires and separations data smaller than comparable estimates e.g., CPS gross flows data
Issue 2 – Worker Flow Magnitudes Employer-employee data from Davis-Faberman-Haltiwanger (2006) also suggest estimates are understated
JOLTS Measurement Issues JOLTS has three notable measurement issues Two issues are observable in the published data JOLTS hires and separations estimates overstate growth relative to its benchmark estimates (CES) Magnitude of hires and separations data smaller than comparable estimates e.g., CPS gross flows data Micro data show: JOLTS over-represents stable establishments and misses entry and exit Nature of the survey frame and sample nonresponse may play a role
Issue 3 – JOLTS Growth Rate Distribution BED distribution also shows greater shifts in moving between high-growth and low-growth quarters
Goals for this Paper Adjust JOLTS hires, separations, and job openings to reflect the universe growth rate distribution Adjustment uses universe quarterly growth rate distribution from BED as “true” underlying distribution in economy. After frequency adjustment, it interacts BED distribution with JOLTS data on mean worker flows and job openings rate by growth rate to construct adjusted measures. Compare adjusted and unadjusted JOLTS measures Quantify change in magnitudes Compare any differences in cyclicality
Summary of Results Adjusted worker flows are about one-third larger than published statistics Separations increase disproportionately through increase in layoff rate Adjustment alters relative volatility of hires and separations Volatility of hires declines, while volatility of layoffs doubles, making separations more volatile than hires overall Little change in Beveridge Curve (unemployment and vacancies) behavior
Adjustment Approach Adjustment based on identity that aggregate estimate is equal to weighted sum of values for each growth rate x t (b) = mean value over growth rate “bin” b Can obtain from monthly JOLTS micro data f t (b) = employment density for growth rate “bin” b Will replace JOLTS densities, after some adjustments, with BED values
Adjustment Step 1 – Estimating Mean Rates Estimate mean worker flow and job opening rates for fixed growth rate intervals using JOLTS microdata Rates show nonlinear relations to growth Relations vary little over time
Hires Rate
Job Openings Rate
Layoffs Rate
Quits Rate
Adjustment Step 2 – Density Creation Adjustment replaces JOLTS densities with BED densities by growth rate interval Estimation of monthly-quarterly relationship: Predicted JOLTS density based on BED data: Issues to Consider Only BED densities account for entry, exit Rescale JOLTS densities; add in estimates of entry and exit BED are quarterly, JOLTS are monthly Adjust BED based on monthly-quarterly relations observed in JOLTS
Results Magnitudes of hires, separations increase by one-third Volatility of separations increases (through layoffs) Net growth (H – S) decreases
More Results
Adjusted vs. Published Hires
Adjusted vs. Published Quits
Adjusted vs. Published Layoffs
Adjusted vs. Published Job Openings
Conclusions JOLTS data have been a major innovation in the study of labor dynamics Initial research revealed some measurement issues Adjusted estimates provide a different picture of labor market than published estimates Magnitudes of worker flows higher by one-third Cyclicality of separations relatively more important Our work stresses the importance of having a survey that is representative of both levels and growth when estimating flow statistics Work hopefully aids statistical agencies in achieving this goal
Quits and Layoffs Rates
Time-Series Shifts in Distributions
Adjusted Quits and Layoffs