A statistical concept of causality where a variable X Granger-causes Y if past X values improve predictions of future Y.Granger causality is a statistical notion of cause and effect that asks a practical question: does knowing the past of one variable help predict the future of another, beyond what the other variable's own past already tells you? If yes, the first variable is said to 'Granger-cause' the second. It is not about physical causation in the deep sense—rather, it is a tool for detecting predictive influence in time-ordered data, often used in economics, neuroscience, and climate science.
Introduced by economist Clive Granger in 1969, who later won the Nobel Memorial Prize in Economic Sciences in 2003 for this work.
The idea is tested with regression models. You first build a model that predicts Y using only its own past values. Then you build a second model that adds past values of X. If the second model is significantly better at forecasting Y (typically judged by an F-test or similar), then X Granger-causes Y. The key is that X's information must come before Y's outcome in time, ensuring a temporal order.
Used in econometrics to test whether one economic indicator (like money supply) helps predict another (like inflation), and in neuroscience to infer directed influences between brain regions from fMRI or EEG time series.
Granger causality can be fooled by a third variable that drives both X and Y—if you omit it, you might falsely conclude X causes Y.












