Granger causality

El 27 de mayo falleció Clive Granger, Premio Nobel de Economía 2003 (otorgado junto con su cumpañero Engle). Granger nos explicó a los que usamos series de tiempo que nos ibamos a topar con algunos problemitas adicionales:

For want of better techniques, economists often applied statistics designed for stationary data to non-stationary data. But in 1974, Granger and his post-doctoral student Paul Newbold, building on the earlier work of the British statistician G Udny Yule, showed that pairs of non-stationary time series could frequently display highly significant correlations when there was no causal connection between them. For example, the US federal debt and the number of deaths due to Aids between 1981 and 2000 are highly correlated but are clearly not causally connected. Such "nonsense correlations" called into question the meaningfulness of many econometric studies.

Working with Engle, Granger realised that not all long-term associations between non-stationary time series are nonsense. Suppose, as the American academic Kevin D Hoover explained, that the randomly-walking drunk has a faithful (and sober) friend who follows him down the street from a safe distance to make sure he does not injure himself.

"Because he is following the drunk, the friend, viewed in isolation, also appears to follow a random walk, yet his path is not aimless; it is largely predictable, conditional on knowing where the drunk is," Hoover noted. Granger and Engle coined the term "co-integration" to describe the genuine relationship between two non-stationary time series. Time series are "co-integrated" when the difference between them is itself stationary – the friend never gets too far away from the drunk, but, on average, stays a constant distance behind.

Desde aquí, nuestro humilde homenaje a este grande de la econometría.

Fuente: telegraph.co.uk citado en el blog de Mankiw.