Dynamic Regressions with Variables Observed at Different Frequencies

We consider the problem of formulating and estimating dynamic regression models with variables observed at different frequencies. The strategy adopted is to define the dynamics of the model in terms of the highest available frequency, and to apply certain lag polynomials to transform the dynamics so that the model is expressed solely in terms of observed variables. A general solution is provided for models with monthly and quarterly observations. We also show how the methods can be extended to models with quarterly and annual observations, and models combining monthly and annual observations.