ABSTRACT

This paper examines whether qualitative survey data can improve the power of ex ante short-term forecasts of inflation. The models are based on expert expectations from qualitative economic survey data. The survey series are derived by checking carefully the information in disaggregated and aggregated answers. Empirical evidence is presented that passing classical rationality tests is no necessary condition for expert expectations to improve forecasting facilities. Additionally, as proposed by Maddala (1991) the results indicate that testing for rationality in the cointegration framework is more appropriate in the presence of price expectations data. Hence our conclusion is that it is important to examine whether there is a possibility of incorporating a learning process in the form of errorcorrection-terms in the forecasting models. Thus without a priori restriction

of the survey expectations, cointegration analysis showed that the rationality restrictions are met by the equilibrium conditions between real and expected inflation. Despite other findings in the literature the results indicate that the predictive performance of vector-error-correction or autoregressive distributed lag models is superior to that of autoprojective models.