Testando teorias para o consumo agregado no Brasil [Testing theories of aggregate consumption in Brazil]

In this paper, theories to describe aggregate consumption in Brazil are tested. Two datasets were utilized, one comprising yearly observations from 1947 to 2012, and another comprising quarterly observations, from the first quarter of 1991 to the second quarter of 2014. Our methodological approach was based on the use of BDS statistics to determine which theoretical and/or empirical model properly fits the features of time series. According to our results, the random-walk Permanent Income Hypothesis was not supported by the data, due to excess of sensitivity which can, to some extent, be explained by credit constraints. In this sense, a standard version of the consumption function with excess of sensitivity was estimated by assuming the error correction model. Once we accounted for (possible) multiple structural breaks in the latter consumption function, we cannot reject the null of IID residuals. Hence, we conclude that consumption in Brazil is better modeled as being sensitive to current income variations, which opens room for effective stabilization policies in the country.