Glossary

Spurious Regression

A phenomenon in which a regression between two independent non-stationary time series produces apparently significant results even though no genuine relationship exists. Cointegration tests and differencing are the standard safeguards.

Definition

Why It Matters

Spurious regression is one of the most insidious pitfalls in time-series econometrics. Granger and Newbold (1974) demonstrated that regressing independent random walks on each other yields high R-squared values and significant t-statistics with alarming frequency. Without testing for unit roots and cointegration, researchers risk publishing findings that are entirely artefacts of shared trends. This problem is especially relevant for emerging market macroeconomics, where many series trend persistently.

Example

Regressing Turkish GDP on the cumulative number of mobile phone subscribers over 2000-2023 produces an R-squared above 0.95 and a highly significant slope coefficient. However, both series are I(1) and not cointegrated; the result is entirely spurious. After differencing both series, the regression yields an insignificant coefficient and an R-squared near zero, correctly reflecting the absence of a genuine relationship.

Related Terms

Software Notes

SPSS: Not directly available; use R integration for unit-root and cointegration testing before regression.
R: Use adf.test() from the tseries package or ur.df() from urca for unit-root tests. For cointegration, use ca.jo() from urca. Always test for stationarity before running regressions on time-series data.
Stata: Use dfuller for the ADF test and vecrank for Johansen cointegration rank testing. Compare results from regress on levels versus first differences to detect potential spurious regression.