The paper discusses a number of trends in the use of macro-economic models for acquiring information about the macro-economy. It is argued that a fundamental distinction should be drawn between models that are constructed to summarise the data and those which are used to interpret the data. By and large the former are statistical and so are judged on the basis of how well they fit the data. Nevertheless, while we have leaned a lot about the nature of macroeconomic data from the large number of statistical models that have emerged to carry out this task, we demonstrate that some of the perceived features may not stand up to a sustained investigation. In particular, it is argued that graphical analysis can often throw doubt upon certain conclusions stemming from formal statistical inference. The second part of the paper turns to models that have been set up with the aim of interpreting macroeconomic data. These are economic in nature, and represent stories that we tell about economic interactions with a view to understanding the origin of observable data characteristics. Given this orientation, goodness of fit is not regarded as the only criterion for assessing their success. Consequently, issues arise over how one can compare such models. A range of methods has evolved for ascertaining whether the proposed story is a good description of the data. We review these methods as well as making some suggestions about how one might improve on them.
