Testing Mean Reversion with Unit-Root Tests and State-Change Regression

Summary

The document asks how to determine whether a time series is mean reverting, including whether the Ornstein–Uhlenbeck process provides a useful hypothesis. One response places relevant econometric methods under unit-root testing and notes that this is a broad literature with methodological pitfalls.

A second response gives an intuitive regression approach for a process expected to revert to zero: observe whether changes tend to be negative when the initial state is positive and positive when it is negative. This can be assessed by regressing each interval’s change on the process value at the start of that interval. The excerpt provides no equations, test statistics, sample requirements, or empirical results. The regression is a basic diagnostic, not a complete account of unit-root testing or proof of mean reversion; inference depends on the data and model assumptions.

Key ideas

• Unit-root tests are a major family of econometric methods for assessing mean reversion.
• A zero-mean-reverting process tends to fall when above zero and rise when below zero.
• Regressing interval changes on starting values can test for this directional tendency.
• The document gives intuition but no equations, inference procedure, or empirical validation.

Tags

# What methods are there for showing a time series is mean reverting?


# What methods are there for showing a time series is mean reverting?












What methods are there for showing a time series is mean reverting?

Is there a hypothesis relating to the Ornstein-Uhlenbeck process for example?

## Answer by Kiwiakos (score 3)

https://quant.stackexchange.com/a/21423

In econometrics all these tests are under the banner of 'Unit Root Tests'. There is a vast body of literature that deals with their formulation, treatment and pifalls.

## Answer by Alex C (score 1)

https://quant.stackexchange.com/a/21426

In essence, and without using any equations: A process that is mean reverting to zero will have a tendency to decrease when it is above zero and a tendency to increase when it is below zero. This can be tested by regressing the observed process changes during each time interval on the process state at the beginnning of each time interval.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.