How biased are measures of cyclical movements in productivity and hours?

The movement of hours worked over the business cycle is an important input into the estimation of many key parameters in macroeconomics. Unfortunately, the available data on hours do not correspond precisely to the concept required for accurate inference. We study one source of mismeasurement--that the most commonly used source data measure hours paid instead of hours worked--focusing our attention on salaried workers, a group for whom the gap between hours paid and hours worked is likely particularly large. We show that the measurement gap varies significantly and positively with changes in labor demand. As a result, we estimate that the standard deviations of the workweek and of total hours worked are 25 and 6 percent larger, respectively, than standard measures of hours suggest. We also find that this measurement gap is an unlikely source of the acceleration in published measures of productivity since 2000.