Why do BLS hours series tell different stories about trends in hours worked?

Hours worked is an important economic indicator. In addition to being a measure of labor utilization, average weekly hours are inputs into measures of productivity and hourly wages, which are two key economic indicators. However, the Bureau of Labor Statistics' two hours series tell very different stories. Between 1973 and 2007 average weekly hours estimated from the BLS's household survey (the Current Population Survey or CPS) indicate that average weekly hours of nonagricultural wage and salary workers decreased slightly from 39.5 to 39.3. In contrast, average hours estimated from the establishment survey (the Current Employment Statistics survey or CES) indicate that hours fell from 36.9 to 33.8 hours per week. Thus the discrepancy between the two surveys increased from about two-and-a-half hours per week to about five-and-a-half hours. Our goal in the current study is to reconcile the differences between the CPS and CES estimates of hours worked and to better understand what these surveys are measuring. We examine a number of possible explanations for the divergence of the two series: differences in workers covered, multiple jobholding, differences in the hours concept (hours worked vs. hours paid), possible overreporting of hours in CPS, and changes in the length of CES pay periods. We can explain most of the difference in levels, but cannot explain the divergent trends.