The measurement of migration, from census data and other sources
In the first part of the paper a general classification of migration measures is introduced. The rates corresponding to these measures are defined together with the associated populations at risk. A particular problem in migration analysis is then tackled. The questions asked in the British censuses of 1966 and 1971 about migration generate tabulations if migrants over the one year, and over the five years, preceding the census date. It is often observed that the numbers of migrants in the two periods are not linearly related; that is, the number of migrants over five years is less than five times the number over one year. A simple stochastic model embodying population-accounting principles is developed to show why this is the case, and that the relationship between one-year and five-year figures is a complex one involving some multiple migrations, some return migrations, and some deaths.