"Dividing by 4": A feasible quarterly forecasting method?

One of the traditional motivations for building quarterly macroeconometric models is the demand for quarterly forecasts. Models based on annual data conceal higher frequency information and are not considered sufficiently informative to policy makers. Two difficulties may encumber quarterly macroeconometric modelling: the lack of observations, i.e. variables not being observed at the quarterly frequency, and the seasonal pattern in the data. Many methods have been suggested to deal with missing observations. Most of them employ smoothed series of approximations for the missing observations and necessitate seasonal adjustment of all other series in the models as well. Seasonal adjustment is no longer beyond criticism; the current view is that seasonal information should be employed rather than filtered out. An alternative method when confronted with missing quarterly observations is to generate forecasts with an annual model, to disaggregate these annual series into quarterly observations, and to add a seasonal pattern if required. This paper investigates under which circumstances this approach is preferable to forecasting with a quarterly model (partly) based on approximations of variables with missing observations.