Avaliação empírica da projeção da taxa de poupança agregada das famílias brasileiras com dados da Pesquisa de Orçamentos Familiares: Decomposição de Neuman-Oaxaca no modelo de Heckman

The Household Budget Survey from IBGE is the only source of microeconomic data on savings of Brazilian families. As this research has been carried out at intervals of at least five years, it is important to discuss and empirically assess whether the estimated models with latest lagged research data are able to produce reliable projections of aggregate current savings rate and future of Brazilian families from current observations of predictive variables that are available from annual frequency for researches.Alternative projection models are estimated with predictive variables suggested by the literature on consumption and savings. One is based on the procedure used by the IBGE to project the aggregate savings rate. Other uses of non-linear structure of Heckman model to take into account the fact that a proportionally high amount of families have zero savings. As the work is focused on the aggregate savings rate, an important methodological aspect is that the models need to be estimated with the weight of each family given by its sample participation on aggregate income.The decompositions of Blinder-Oaxaca and of Neuman-Oaxaca show that the difference between the saving rates estimated with the versions of the POF 2002-2003 and 2008-2009 has been explained in greater extent by changes in the estimated coefficients of the projection models, leaving a secondary role for the effect of changes in sample means of predictive characteristics. This result undermines the use of models estimated with POF data in the projection of aggregate savings rate of Brazilian households, since only the second effect is observed in years without availability of POF.