Adjusting for Calendar Effects in Payroll Hours and Earnings Series

Nicole Hudson and Jeannine Mercurio

Abstract

This paper examines the causes and treatments of calendar-related movements in the payroll hours and earnings time series from the Current Employment Statistics (CES) survey. Prior research has established that there is a correlation between the number of workdays in the month and fluctuations in CES hours and earnings. The strongest correlation was determined to exist for reporting establishments with a semi-monthly or monthly payroll. These predictable movements are related to respondent error in semi-monthly and monthly payroll reports and processing limitations for payrolls. Currently, the CES hours and earnings series are adjusted for variations in the number of workdays in the 1st through the 15th of the month to treat for these effects. This paper discusses the current methods used to adjust for this effect along with the methods used to monitor changes in these correlations over time. This paper also discusses methods to detect and evaluate any residual effects of these movements in the final seasonally adjusted series and future improvements. The results show that the length of pay period effect exists for the additional data types added to collect hours and payroll for all employees. The models currently used continue to perform well and provide an accurate means of adjusting for the length of pay period effect.