For a rotation design or repeated sampling seen in many government surveys, composite estimation, where data from the current and earlier periods of time are combined, is often used to measure the level of a characteristic of interest. For the purpose of reducing the variability of the survey estimator, we construct a composite estimator for the Taiwan Labor Force Survey (TLFS) and compare its performance with the simple ratio estimator currently used in TLFS. The gains in relative efficiency are demonstrated to be about 17-35% for the following four characteristics: labor force, employed, non-labor force, and unemployed.