Glossary

Granger Causality

A concept of predictive causality: variable X Granger-causes variable Y if past values of X contain information that helps predict Y beyond what is contained in Y's own past. Tested within a VAR framework using F-tests or Wald tests.

Definition

Why It Matters

Granger causality does not establish true structural causation, but it is an indispensable screening tool in time-series analysis. By determining whether one variable improves predictions of another, it helps researchers narrow down potential causal mechanisms, specify VAR models correctly, and justify policy interventions. It is particularly valuable in macroeconomics and finance, where controlled experiments are impossible and temporal ordering offers the first clue about causal direction.

Example

A researcher tests whether Turkish industrial production Granger-causes the unemployment rate. In a bivariate VAR with four lags, the F-test on the lagged production coefficients in the unemployment equation yields a p-value of 0.003, rejecting the null of no Granger causality. This suggests that past production values contain predictive information for unemployment beyond what unemployment's own lagged values provide.

Related Terms

Software Notes

SPSS: Not directly available; use R integration for Granger causality tests.
R: After estimating a VAR with VAR() from the vars package, use causality() to perform Granger causality tests.
Stata: After var estimation, use vargranger to perform Granger causality tests on all equations in the system.