Causality and Cointegration between Consumption and GDP in 25 OECD countries: limitations of cointegration approach

A critical review of cointegration is presented in this paper, emphasizing some limitations of this approach to testing causal relations in Econometrics. Very often the usual way of analysing cointegration leads researchers to declare many important causal relations as spurious when they are obviously not. On the other hand that approach does not always avoid the peril of accepting as causal relations those that really are spurious. An application of the tests to the relation between Private Consumption and Gross Domestic Product in 25 OECD countries, during the period 1960-97, is performed and the results confirm the above mentioned limitations of cointegration tests. The most extreme case is that of the UK where, in one of the applications, the usual tests lead to rejection of cointegration between british Private Consumption and british GDP and acceptance of cointegration of british Private Consumption with GDP of all the other countries. Regarding cointegration we should take into account two important points: one is that many causal relations with autocorrelation coefficient of residuals near 0.9 show a result of uncertainty, or a result of near cointegration, and should not be counted as rejections. A second, equally important point, is that sometimes the problem of no cointegration among several variables, is due to small problems in the specification of the model, easily avoided with some changes in the form of the dynamic relation, and not to the existence of non-causal relations.