The objectives of this study are to find the optimum boundary of strata and to compare the two sampling designs for estimating the number of employees in establishments by stratified two-stage sampling, sampling with probability proportional to size at 1 st stage and 2 nd stage by (a) simple random sampling and (b) systematic sampling. A primary sampling unit is industrial code and a secondary sampling unit is establishment. Data used in this study are the number of establishments and employees in Bangkok Metropolitan in 1989, collected by Labour Studies and Planning Division, Department of Labour. The study shows that, for the optimum boundary of strata, population should be divided into 7 strata. The comparison of the efficiency between the two sampling designs by considering the coefficient of variation and different between the estimate and the true value, finds that two designs have nearly the same efficiency when a given error (d) is 10%. When d are 12.5 - 15% simple random sampling has higher efficiency and when d are 17.5 - 20% systematic sampling has higher efficiency. It is also found that in 7 th stratum (establishment with 300 employees and over), all industrial codes should be surveyed.