The Current Employment Statistics (CES) Survey is conducted monthly by the Bureau of Labor Statistics (BLS) for the purpose of estimating over-the-month changes in total employment, and in payroll earnings and hours for production and non-supervisory workers. In 1997, researchers noted the presence of fluctuations in the Average Weekly Hours (AWH) and Average Hourly Earnings (AHE) series, that appeared to be non-economic in nature and related to variations in the calendar. In this paper, we discuss the process used to identify and treat the observed distortions. First, the microdata were screened by an equal means test to indicate significant differences in months with varying numbers of days per pay period. These tests identified problematic industries and reporters. Second, we generated estimates without the problematic reports; this resulted in near elimination of the fluctuations in AWH and mitigated the distortions in AHE. Next, we contacted reporters to inquire about their reporting practices and confirmed the source of variation in the payroll figures. Finally, we developed time series models with variables designed to identify, measure, and treat the effects of the varying length of pay periods. The modeling resulted in the successful treatment of affected industries, which now display seasonal adjusted series with virtually no calendar-related fluctuations and significantly increased smoothness.