Visualizing the Distortions of Statistical Maps Caused by the Increase in Unreported Cases in Japan’s Population Census

The purpose of this study was to visualize how regional differences observed from the population census deviate from their “true” forms due to an increase in unreported (missing) cases by considering multiple scenarios in which there was an increase in the number of unreported cases in the census. For municipalities in the Tokyo metropolitan area, we estimated the percentages of unmarried people and those with short-term residence from the 2015 Population Census for different scenarios of unreported rates using 1) proportional distribution by age-group and 2) weighted distribution based on the ratio of unreported rates among attributes. In both indicators, the results showed that there were clear regional differences with high and low percentages in urban centers and suburban areas, respectively, in a scenario where the unreported rate became zero. However, when we examined the maps estimated under scenarios where the unreported rates increased by 1.5- or 2-fold, such regional differences appeared to be reduced or eliminated. This means that further increases in the number of unreported cases will increase the probability of spurious regional differences in the future. Careful interpretation of the results and use of statistically corrected data are required when analyzing the census data for items and regions with high unreported rates.