Research work · 2004
Realization of the Congruence Postulate as a Method of Avoiding the Effects of a Spurious Relationship
Mariola Piłatowska
Dynamic Econometric Models
Abstract
Already at the turn of the 19 century the influence of the nonstationarity of economic processes on detecting the true relationships among them was noticed. Yule (1926) was the first researcher who raised the problem of the danger of a spurious relationship among processes monotonically rising or falling. In his simulation experiments studying the correlation between uncorrelated integrated processes (first and second order) he proved that the distribution of the correlation coefficient was not symmetric and not nearly normal, but it was closer to a semi-ellipse for a pair of I(1) series, and U-shaped – for a pair of I(2) series. The Yule’s experiments show explicitly that the correlation between processes with integral structure, i.e. between integrated processes, should be calculated on the data from which the nonstarionarity has been removed, e.g. on differenced data, i.e. first or second differences, but not on the original data, i.e. on levels. Such approach enables to avoid the danger of nonsense correlation as Yule called it. However as the seminal study referring to the importance of distinguishing the spurious correlation between economic processes from the true one is treated in the paper by Granger and Newbold (1974). Considering the relationship between unrelated white noises Granger and Newbold shown in simulation experiments that the true null hypothesis of no relationship was rejected 76 per cent of times using the standard t-statistic. They came to the conclusion that the standard significance tests are biased in the direction to reject the null hypothesis of no relationship even when it is true and at the same time to indicate an