In Section 4 the determination of the degree of mobility was seen as a problem of measurement rather than as the statistical estimation of an underlying model. The correlation coefficient, for example, was seen as a sufficient measure in its own right, rather than as a parameter in a fully-specified model of earnings generation. In this section we review the results obtained in econometric studies of earnings dynamics, inter preting them in the light of the theoretical discussion in Section 2. We begin with what we have called the ‘statistical’ models of earnings mobility, that is the models derived from the Galton-Markov equation. It will be shown that available results seem to invalidate the basic statistical assumptions behind that model, whether it is formulated as a first-or as a second-order auto-regressive process. We then move on to a class of econometric models which intend to explain earnings mobility through identifying systematic determinants of earnings and earnings profiles, mostly following the human capital approach. Although more satisfactory than the purely statistical models, those explanatory models still leave a substantial part of earnings mobility largely unexplained — even when they rely on fairly complex stochastic specification of individual earnings processes. We conclude the section by examining the possible impact on the results of the previous models of the main imperfections that affect available data sources on individual earnings profiles, most noticeably sampling errors, measurement errors and attrition.