ABSTRACT
This chapter discusses spurious regression with integrated (unit root) time series and tests for unit roots. It cointegration of several integrated variables and test for cointegration and vector error correction model for cointegrated time series. If a dependent variable turns out to be significantly related with independent variables without any justifying reason, the relation is called a spurious or nonsense regression. If several variables are nonstationary but a linear combination of those variables becomes stationary, then they are defined as being cointegrated. The principle behind the VECMs is that there exists a long-run equilibrium relation among economic variables, such as the long-run average proportion of consumption in income. In the short run, however, they could be out of equilibrium. Using simulated data, the authors illustrate how regression analysis can be highly misleading when applied to variables containing stochastic trends. Based on the findings, they derive several rules of thumb for analysis of integrated variables.
