12 Unit Roots, Cointegration and Error Correction
Most macroeconomic and financial series in levels, such as output, prices, money, exchange rates and share prices, do not return to a fixed mean or a fixed trend line. They behave more like the random walk of Figure 7 than like a stationary process. That has two large consequences. Regressions between such series can be entirely spurious, with high \(R^2\) and large \(t\) ratios between variables that are unrelated. And the usual large-sample theory of Section 4 fails, so the familiar \(t\) and \(F\) tables no longer apply. This section develops the theory of unit roots, the tests used to detect them, and the concept that rescues regression in levels: cointegration, with its companion, the error correction model.
12.2 Spurious regression and the asymptotics of a unit root
12.3 Unit root tests
12.4 Cointegration
12.5 Error correction models
12.6 Unit roots and cointegration in software
12.7 Practice Problems
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.