All Change! The implications of non-stationarity for empirical modelling, forecasting and policy

Journal article

In an age of congested transport systems, everyone knows what it is like to be stationary: stuck motionless in a traffic jam; a train standing still at a station long after the due departure time; an aircraft sitting at the departure gate several hours delayed. The same word is used in a more technical sense in statistics: a stationary process is one where its mean and variance are constant over time.1 As a corollary, a non-stationary process is one where the distribution of a variable does not stay the same at different points in time– the mean and/or variance may change for many reasons. Non-stationarity is like a statistical version of the changeover point in a relay race — as they all change, one team successfully transfers, while another drops the baton, and a third is reaching towards a future transfer with an unknown outcome.

Authors: Hendry, D; Pretis, F

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Oxford Martin School
• Journal: Oxford Martin Policy Papers
• Publication date: 2016-11-01
• Keywords: cointegration; forecast; nonstationarity; location shifts