Twitter utilization in application of small area estimation to estimate electability of candidate central java governor

Abstract

A survey was designed to estimate electability of a governor candidate in the province of Central Java in 2018. The estimate was planned to be done at the province level. However, it is also interesting to do estimation at the district level. The problem is that the sample size at some district is too small or even zero so that the estimate would have inadequate precision. Using a small area estimation method, this research tried to have good estimation by utilizing information from Twitter and other auxiliary data in the modeling. To evaluate the estimation, we calculated the mean square error (MSE) of the estimated electability using resampling method. We noted that both indirect and direct estimations at the province level have equal precision, but the indirect estimation outperformed the direct estimation at the region level. We also compare our result to the election result of Governor Election.