Unit Root: A Comprehensive Guide

A unit root is a property of a time series that makes it non-stationary — meaning its statistical characteristics drift over time rather than staying stable. Detecting whether a series has a unit root is one of the most important first steps in time-series analysis, because it determines whether standard forecasting and regression techniques can be trusted. This guide explains what a unit root is, why it matters, how it's tested, and what to do about it — in plain language. It connects closely to covariance stationarity and is a core topic in quantitative qualifications like the FRM.

What is a unit root?

To understand a unit root, picture a simple model where each value in a series depends on the previous value plus a random shock. The key question is how much of the previous value carries forward. If that coefficient is less than 1, shocks fade away over time and the series is pulled back towards a stable average — it's stationary. If the coefficient is exactly 1 — a "unit root" — shocks never fade; they accumulate permanently, and the series wanders without returning to any fixed level. A series with a unit root is therefore non-stationary: it has no stable long-run mean to revert to.

The classic example is a random walk, where each value equals the last one plus a random step. Stock prices are often modelled this way — today's price is yesterday's price plus an unpredictable move — which is why price series typically contain a unit root.

Why a unit root matters

A unit root matters because it breaks the assumptions behind many standard statistical methods. The most notorious consequence is spurious regression: if you regress one unit-root series on another, you can get a high R-squared and apparently significant relationship even when the two series are completely unrelated — simply because both are wandering over time. Acting on such a "relationship" would be a serious error. Unit roots also mean that the usual rules for statistical inference no longer hold, so conclusions drawn without checking for them can be badly misleading. This is why testing for a unit root is a routine and essential first step in time-series work.

How a unit root is tested

Several statistical tests are used to check whether a series contains a unit root:

• The Dickey–Fuller test and its more general version, the augmented Dickey–Fuller (ADF) test, are the most widely used. They test the null hypothesis that a unit root is present; rejecting it provides evidence the series is stationary.
• The Phillips–Perron test is an alternative that adjusts for certain complications in the data.
• The KPSS test reverses the logic — its null hypothesis is that the series is stationary — and is often used alongside the others for confirmation.

Because these tests can disagree or have limited power in some situations, analysts often apply more than one and combine the results with a visual inspection of the data.

What to do when a series has a unit root

If a series is found to contain a unit root, the standard remedy is differencing — working with the change from one period to the next rather than the level itself. Differencing a unit-root series usually produces a stationary one: stock prices have a unit root, but stock returns (the period-to-period changes) generally do not. A series that becomes stationary after differencing once is called "integrated of order one". Once a series has been made stationary, the standard toolkit of time-series modelling can be applied with confidence. (A more advanced approach, cointegration, handles special cases where two unit-root series genuinely move together in the long run.)

Why it matters for finance professionals

Financial time series — prices, indices, exchange rates — very often contain unit roots, so anyone modelling them must know how to detect and handle the problem. Understanding unit roots is what stops an analyst from being fooled by a spurious relationship between two trending series, and it's the foundation for building sound forecasting and risk models. It's a fundamental concept in financial econometrics and a regularly examined topic in professional qualifications.

Frequently asked questions

What is a unit root?

A property of a time series where shocks accumulate permanently instead of fading, making the series non-stationary with no stable long-run average. A random walk is the classic example.

Why is a unit root a problem?

It breaks standard statistical assumptions and can cause spurious regression — an apparently strong, significant relationship between two unrelated series that are both simply wandering over time.

How do you test for a unit root?

Most commonly with the augmented Dickey–Fuller (ADF) test, which tests whether a unit root is present. The Phillips–Perron and KPSS tests are alternatives often used alongside it.

How do you fix a series with a unit root?

Usually by differencing — using period-to-period changes rather than levels — which typically produces a stationary series. This is why analysts model returns rather than price levels.

Build your quant skills with Learnsignal

Unit roots and stationarity are cornerstones of financial econometrics. Learnsignal's tutor-led courses, including the FRM, develop the time-series and statistical understanding that topics like this build on — with clear teaching that makes the theory genuinely usable.