Beware of econometricians bearing estimates: Policy analysis in a “unit root” world

Current statistical approaches to modeling many economic relationships are grounded in traditional ideas of deterministic trends. Some of the failures of these approaches are due to inappropriate models using time series with “unit roots.” After a shock, unit root processes do not revert to some time trend, but rather can drift up or down without bounds. A “random walk” is a well-known example of a unit root process. The purpose of this paper is to explain the importance of unit root processes to policy analysts who make or rely upon econometric models using time series dat. In particular, the presence of unit root processes in GNP, energy and electricity consumption exports, imports, and other variables suggests that modifications to the way economic relationships are estimated may be necessary. Once these modifications are made, many important parameters turn out to be much different, with substantive implications for both forecasting and policy.