Granger Causality Test
Also known as: Granger causality test, Granger non-causality test, predictive causality test, Granger Nedensellik Testi
The Granger causality test, introduced by Clive W. J. Granger in 1969, assesses whether the past values of one time series help predict another beyond what the latter's own past already explains. It defines causality in a strictly predictive sense rather than as a structural or physical cause.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
+13 more
When to use it
Use the Granger causality test on continuous time-series (or panel) data when you want to know whether one series has predictive precedence over another, with at least about 40 observations. It requires that the series be stationary, typically verified with an ADF or PP unit-root test, and it is normally run as a follow-up to fitting a vector autoregression (VAR), which supplies the lagged structure. It is not appropriate for non-stationary series, where it produces spurious causality, nor for very short series, where the lag structure cannot be estimated reliably.
Strengths & limitations
- Provides a formal, testable definition of predictive precedence between two time series.
- Naturally extends to multivariate settings through the vector autoregression (VAR) framework.
- Does not require the data to be normally distributed.
- Captures predictive precedence only, not true structural or physical causation.
- On non-stationary series it produces spurious (false) causality, so stationarity must be established first.
- In short time series (n < 40) the lag structure cannot be estimated reliably and the test has low power.
Frequently asked
Does Granger causality prove that x causes y?
No. It only shows that the past of x helps predict y beyond y's own past — a statement about predictive precedence and forecastability, not about a true structural or physical cause.
Why must the series be stationary?
On non-stationary series the test tends to flag spurious causality driven by shared trends rather than genuine predictive links. Confirm stationarity with an ADF or PP unit-root test first; if the series are integrated, difference them or examine cointegration instead.
How do I choose the number of lags?
Select the lag length with an information criterion such as AIC or BIC, usually via the underlying VAR model. Arbitrary lag choices can change the conclusion and inflate or deflate the test's power.
What if my sample is shorter than 40 observations?
With fewer than about 40 points the lag structure cannot be estimated reliably and the test has low power. Consider fitting a VAR model or gathering more data before drawing conclusions.
Sources
- Granger, C. W. J. (1969). Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica, 37(3), 424-438. DOI: 10.2307/1912791 ↗
How to cite this page
ScholarGate. (2026, June 1). Granger Causality Test. ScholarGate. https://scholargate.app/en/econometrics/granger-causality
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- ARDL Bounds TestEconometrics↔ compare
- Cointegration TestEconometrics↔ compare
- OLS RegressionEconometrics↔ compare
- VAR ModelEconometrics↔ compare
- VECMEconometrics↔ compare