A large scale surveillance study was carried out to verify how trends in income inequality depend on kinds of income survey data in Japan. Gini coefficients were computed for a long time period using thirteen kinds of existing income survey data. The trends of the Gini coefficients were statistically tested by using time-series regression. The effect of sampling errors in income survey data and of consolidating income classes on the trends of income inequalities was also examined. This study revealed that the trends of income inequalities heavily depend on the data employed, and therefore it is necessary to extensively consider the results of many kinds of income survey data to draw conclusions about trends in income inequality.