Exogeneity and measurement of persistence

This paper argues the fact the empirical evidence on persistence is mixed is not very surprising, as economic theory is bound to be drawn upon in order to specify the statistical model. This is illustrated in two ways. Firstly, we highlight the fact that concept of persistence is model dependent, i.e. it is a function of the maintained model adopted on the basis of economic theory. Secondly, we analyse the related issue of the definition of a set of parameters of interest. In particular, consider a simple bivariate case. Given weak exogeneity of the regressors, the parameters of the unique cointegrating vector can equivalently be estimated within a full system or a single equation framework. On the contrary, if persistence is being measured , weak exogeneity of the regressor does not hold any longer, as the parameters of interest cannot be written as a function of those of the conditional model only, and the concept of model (rather than weak) exogeneity becomes more relevant. Once again, economic theory can be seen to play an essential role in model specification.