SUMMARY
In sample surveys there is often a need to estimate not only population characteristics, but subpopulation characteristics as well. We consider the problem of estimating the total value in domains (subpopulations). In this case, the Horvitz-Thompson estimator could be used. Nevertheless, it does not use any additional information about population units, which are usually known. To increase estimation accuracy we propose to use calibration estimators with auxiliary variables from the current and past periods. In the simulation studies based on real and generated data, we show the influence of using auxiliary information from past periods on the accuracy, and compare properties of two calibration estimators of domain totals in longitudinal surveys.