908 Macroeconometric Modeling for Policy
2003
–8–40412
82004 2005 2006 20072003
–40412
82004 2005 2006 20072003
123457
62004 2005 2006 2007 2003
–15–10–50515
102004 2005 2006 20072003–15–10–50515
102004 2005 2006 200720037580859095105
1002004 2005 2006 20072003
–40412
82004 2005 2006 20072003
–10–5020
152004 2005 2006 20075102003
–5020
152004 2005 2006 2007510(a) Consumer price inflation (b) Unemployment rate (c) Wage growth(d) GDP growth(g) Money market interest rate (h) Real credit growth (i) Currency depreciation rate(e) Import price inflation (f) Real exchange rateFigure 17.15 Dynamic dEqCM forecasts 2005(1)–2007(3), with end-of-sample for estimation
of parameters in 2004(4)
Actual values are shown by solid lines, and forecasts by dashed lines. The distance between the two dotted
lines represents 70% prediction intervals
certain forecast for a target variable like inflation. Paradoxically, perhaps, the main
line of argument starts from the recognition that accurate forecasting is a near
impossibility in macroeconomics because of the inherent non-stationarity of the
economic time series included in the model. Non-stationarity takes different forms,
with different implications for macroeconometric modeling and forecasting, and
we have distinguished between unit roots and non-stationarity due to structural
breaks. We have demonstrated that macroeconometric models can be developed
theoretically and empirically in a way that is consistent with a unit root assump-
tion. At the modeling and estimation stage, non-stationarity due to unit roots can
in principle be handled by cointegration methods and, given that approach, unit
roots are unlikely to be a source of forecast failures. On the contrary, a correct unit
root assumption can help in concentrating on predictable functions of variables,
like growth rates and linear combinations of target variables.