Economic data are regularly revised, leading to a number of alternative vintages being available fo a given data series over a particular time period. This paper confronts this problem, building on the model validity and data accuracy tests of Hendry (1994) to develop the concept of model and vintage coalescence (MVC). A taxonomy of MVC tests is established, and a test result unreported in Hendry (1994) is given. The apparent anomaly arising from the statistic is explained both analytically and by means of a Monte Carlo analysis of MVC. This framework is subsequently generalised to arrive at cross-vintage encompassing , where the comparison of models is extended to take account of the influence of data revisions. The analysis highlights the distinction between conventional encompasssing and cross-vintage encompassing, showing the power of the two to differ significantly. The paper concludes by noting the implications of the above analysis for econometric modelling. In particular, the definition of the data generation process and the nature of true progress in econometric modelling are discussed.
