Économie et Statistique n° 483-484-485The overhauled Census: progress in methodology and contribution to knowledge
What is the advantage of using the local residence tax to extrapolate the number of main residences in the Population Census?
Since the census was overhauled at the beginning of the 2000s, INSEE has been required to publish the legal populations of all municipalities every year. For municipalities of fewer than 10,000 inhabitants, which are surveyed exhaustively every five years, the difficulty lay in updating population figures in the inter-census period. The use of local residence tax (TH) data meets this challenge in part and is a considerable innovation in the process of calculating populations. From the TH, the number of main residences can be estimated and thus population dynamics can be gauged. Every year, it helps to calculate the population of 40% of municipalities with fewer than 10,000 inhabitants, which in itself represents an estimate covering about 20% of the French population. Does using the TH generate any real value-added to the estimate of municipal populations and to the total population affected when it is applied? If the number of main residences is overestimated, the resulting error is low. Thus by extrapolating the number of main residences with the TH for five successive years, error is less than 5% for almost three quarters of all municipalities and the total number of main residences is overestimated by only 1%. In fact, since extrapolation is never over more than two successive years, error is in all likelihood less in those years when the TH is used to calculate populations. Nevertheless, there is a size effect: the largest of the small municipalities are more often overestimated. In addition, the legal populations of seven out of ten municipalities are estimated better with the TH than they would be by linear extension of the last censuses. Moreover, the TH reproduces in detail any trend-breaks observed in changes in municipal housing stock. As a statistical instrument, it is both accurate and flexible as it adapts to the different dynamics of change that may exist.