摘要：The increasing availability of microdata (the unit of observation being either a person or a household) over the last two decades is a promising development for obtaining better insight into the migration process. Such data typically contain information on personal attributes and are relatively flexible for defining theoretically more meaningful migration variables. Two types of such data are now used widely: longitudinal (for example, the well-known Income Dynamics Panel of the University of Michigan) and cross-sectional (for example, a sampie taken from a population census). Longitudinal data are usually considered better than cross-sectional data because, with their temporal depth at the individual level, they are particularly helpful for studying the effects of previous migration and employment experiences on current migration propensity, as demonstrated by the research results in the United States since the early 1970s (for example, Morrison 1971; DaVanzo 1976b, 1978; Morrison and DaVanzo 1986) and in Canada since the mid-1970s (for example, Grant and Vanderkamp 1976, 1984). The usefulness of longitudinal data files, however, is often limited by their relatively small sample size and a sampling frame not representative of the individuals composing the whole system. A consequence of the smal1 sampie size is that the research results based on longitudinal data have so far revealed little about the spatial patterns of the migration process.
