In this paper the results of a theoretical and empirical investigation of different estimators for average hourly earnings and average weekly hours for the Bureau of Labor Statistics' Current Employment Statistics (CES) survey are presented. The investigations began in connection with the redesign of the survey. This is a large longitudinal survey of establishments that provides monthly estimates for, among other parameters, total employment, average weekly hours and average hourly earnings for production and non supervisory workers. Ten estimators which, for the most part, are ratios of two random variables are considered. The empirical study was done using national CES data that have responses for relevant characteristics for a period of thirteen months. Estimators were evaluated under both low and high nonresponse rates. In this study, nonrespondents are not missing at random.