The Current Population Survey (CPS) is a monthly household sample survey consisting of eight rotation groups, so that each selected household will be interviewed for 4 consecutive months and another 4 consecutive months after resting 8 consecutive months. A composite type estimator is adopted in the CPS for the last several decades to estimate monthly employment and unemployment levels and rates, which combines sample information from the current month survey and previous months using the fact that 75% households have data for two consecutive months. In this talk, we introduce regression composite estimation and apply this method to estimate unemployment rates from CPS data for the period 2006-2012. We develop a Monte Carlo simulation experiment in order to compare the proposed, survey-weighted direct, and the AK estimates with the simulated true employment rates. Some numerical studies are conducted to illustrate the effectiveness of the proposed method.