Monthly data and short-term forecasting: An assessment of monthly data in a VAR model

Use of monthly data for economic forecasting purposes is typically constrai ned by the absence of monthly estimates of GDP. Such data can be interpolat ed but are then prone to measurement error. However, the variance matrix of the measurement errors is typically known. We present a technique for esti mating a VAR on monthly data, making use of interpolated estimates of GDP a nd correcting for the impact of measurement error. We then address the ques tion how to establish whether the model estimated from the interpolated mon thly data contains information absent from the analogous quarterly VAR. The techniques are illustrated using a bivariate VAR modelling GDP growth and inflation. Tt is found that, using inflation data adjusted to remove season al effects and the impacts of changes to indirect taxes, the monthly model has little to add to a quarterly model when projecting one quarter ahead. H owever, the monthly model has an important role to play in building up a pi cture of the current quarter once one or two months' hard data becomes avai lable. Copyright (C) 1999 John Wiley & Sons, Ltd.