A sensibilidade do ajuste sazonal no sistema de contas trimestrais brasileiro

Abstract

This article presents a series of exercises which analyze the sensibility of three statistical packages of seasonal adjustment and time series. The central point of this text consists of the analysis of the variation of the quarterly rates against the immediately previous quarter in the series of the Gross Domestic Product (GDP) at Brazilian market prices. Considering the importance of this rate for the economic planning, the conditions under which exchanges in the taxes may occur must be taken into consideration. The exercise carried out has evaluated the behavior of three software’s (X12-ARIMA, TRAMO/SEATS and X11), adopted by important institutions dealing with statistics and recommended by international manuals on Quarterly National Accounts. Two experiments were carried out with the purpose of verifying the sensibility of each method. The first one took into consideration the introduction of new quarters as from the generation of alternative scenarios and the second one compared the results initially projected by the models with the rates actually observed. The difference between adjusting a series of the GDP directly (direct method) and obtaining it through the linear combination of its components (indirect method) has also been analyzed.