The standard errors of persistence (VOX)

A large literature on persistence finds that many modern outcomes strongly reflect characteristics of the same places in the past. Although these regressions feature unusually high t statistics, they usually also display severe spatial autocorrelation in residuals. Exploiting data from 28 persistence studies published in leading journals, this column runs artificial regressions where both variables are spatial noise and shows that this produces severely inflated t statistics. Replicating published studies shows that in most cases the main persistence variable frequently has lower explanatory power than spatial noise but can, in turn, strongly predict spatial noise.

Morgan Kelly

Lascia un commento

Il tuo indirizzo email non sarà pubblicato. I campi obbligatori sono contrassegnati *