Does X cause Y? Granger causality tests if X helps predict Y's future.

Does X cause Y? Granger causality tests if X helps predict Y's future.
Image: Intellect (generated)

In short: Granger causality is a statistical method that determines if one time series provides unique information to forecast another. Despite its name, it does not prove true causation. Instead, it measures temporal precedence, identifying whether past values of one variable significantly improve predictions of another variable's future beyond its own history.

Proposed by Clive Granger in 1969, this test shifts the focus from philosophical debates about 'true' causality to the practical utility of prediction. The core logic is simple: if a variable X contains unique information about the future of variable Y, then including X in a predictive model should yield better results than relying on Y's past alone. Econometricians often refer to this as 'predictive causality' or 'temporal relatedness' to avoid the logical trap of assuming that because one event precedes another, it must have caused it.

The mechanism relies on autoregressive models. Researchers first build a model for Y using only its own lagged values. They then augment this model with lagged values of X. If the inclusion of X significantly improves the model's explanatory power—verified through t-tests and F-tests—X is said to Granger-cause Y. While computationally efficient, the method has limitations. It struggles with non-linear relationships, latent confounding factors, and instantaneous effects, often leading to misleading results if the underlying system involves three or more variables that share a common driver.

The method has expanded far beyond economics. In neuroscience, it is used to map information flow between neurons by analyzing spike trains. Because traditional models required continuous data, a 2011 framework introduced point-process models to handle binary neural events directly. Despite these advancements, the test remains controversial. Critics argue that it fails to meet fundamental requirements for causality, noting that it can sometimes 'predict' the past from the future. Nevertheless, when used with caution and diagnostic tools like information criteria, it remains a cornerstone for analyzing dynamic, time-varying relationships in complex systems.

Opens the Intellect+ feed with this card first. Swipe up for the next one.

Sources

To cite this card, link to https://intellectplus.co.uk/app/science-granger-causality.

How we source, check and correct cards: Editorial standards.

General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.

Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?

Related

Finding something worth knowing…