Policy Analysis in Real Time Using IMF's Monetary Model

We investigate to what extent estimated relationships of the IMF's monetary model and their policy implications are sample dependent. This model constitutes the core of the IMF's financial programming models for developing and emerging economies. We observe that estimates of the model's key parameters and model-based measures of macroeconomic disequilibria are highly dependent on data vintage employed. Changes in parameter estimates solely due to data revisions are found to be much smaller than those owing to parameter instability, which may be due to model misspecification. Moreover, instability in parameter estimates contributes to more uncertainty in evaluations of macroeconomic excesses than data revisions. It is shown that analyses based on a version of the model in difference form are more robust across data vintages than those based on the model with variables in levels. Well specified models that take into account known data revisions may also have relatively stable parameter estimates and hence more robust policy implications.