Introduction to wage statistics in Japan

The gender pay gap in Japan narrowed with the average wage of women reaching 70% of that of men, the Nihon Keizai Shimbun reported in February 2012. We often see these kinds of news reports on wage trends, and when you open a labor economics textbook you will probably see a chart of long-term wage trends. There are approximately 55 million wage workers in Japan. How is their average wage calculated? Some people may think that the government, which collects taxes and therefore should know how much individual workers earn, calculates the average wage based on such income amounts. It may be surprising, but a database that would enable such calculations is not available in Japan, where a taxpayer identification number system is not yet in place. As such, average wages and other pertinent figures are calculated based on statistical survey data. In this article, I would like to introduce some major statistical surveys conducted by the Japanese Government on wages. Since these surveys are all based on randomly-selected probability samples, average wages calculated from their collected data do not represent the average wage of the entire population of wage workers in Japan. Therefore, the average wage and wage distribution of a sample are subject to deviations from those of the entire population of wage workers, depending on which individuals are selected into the sample. In statistics terminology, such deviations are called “sampling error.” In the case of the government statistics on wages to be introduced in this article, the problem of sampling error is not so serious, because samples used in the surveys include a large number of workers. What is more important is to understand the process in which statistical data is collected, and because of that process, what types of wage workers may be excluded from the samples to cause deviations. We call this type of deviations “non-sampling error.” In statistical surveys, information is collected by means of questionnaires, which are typically delivered by hand or downloaded in the Excel file format by individual respondents, who would complete the questionnaires by filling in their answers or entering selected item numbers. Thus, in order to clearly understand exactly what is being surveyed, it is also important to look at how the questions in the questionnaire are worded. By imagining how and in what process raw data are collected and processed into final data, you can have a better idea of the implications of statistical figures you handle. Those who have just started studying labor economics often fail to pay attention to basic questions concerning statistical