Ìndice de Atividade Econômica: Construção e Testes de Previsão dos Modelos de Filtro de Kalman e Box-Jenkins

This paper has as objective to present a practical method for building a composite economic activity index for local economies. As an example, the index was built using data for the city of Maringá-PR. In its development we employed factor analysis technique to find the principal components considering their common factor in order to build the index. This local index is then compared to national ones. As a result we found that national indexes move first comparative to our local index. Thus, the forecast of our local index should anticipate future economic changes. Two techniques are used in forecasting the local index; the first was the Kalman Filter and the second one the Box-Jenkins. The presence of outliers required that we use the technique proposed by Lin e Guttman (1993) in order to obtain stable coefficients for the Kalman Filter model. The two techniques are then compared using a statistic test developed by Diebold e Mariano (1995). As a final result the two models’ forecasting results showed to be equivalent.