In every month on the first working Friday, the U.S. Bureau of Labor Statistics (BLS) releases its preliminary estimates of the total nonfarm business payroll employment and the change in total employment from the prior month. These estimates are produced using the Current Employment Statistics (CES) survey, which is an ongoing national probability sample survey of all nonfarm establishments in the United States. The preliminary estimates are revised a number of times to incorporate late reporting in the CES sample and also to incorporate the most recent benchmark population information. In this paper, we develop a statistical method that has the potential to minimize the amount of revision between the preliminary estimates and the second revision of the estimates. Based on the historical data, we first build an appropriate model that links these two estimates and then use this relationship to predict the second revision from the knowledge of the preliminary estimates for the current month. The preliminary results obtained from our study are encouraging.