The University of Manchester · FACULTY OF ECONOMICS

ECON31031 Chap.9 Unit Roots and Nonstationary Processes

- one subject, every graph, every model, every mark
6 Chapters3-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 9 of 9 · ECON31031

Unit Roots and Nonstationary Processes

Unit Roots and Nonstationary Processes cover what happens when stationarity fails in the most consequential way. When the AR(1) coefficient equals one, back substitution never lets the past die out: the process is a random walk, its variance grows with time and it has no fixed mean to revert to. The supplementary reading stresses that unit roots have severe consequences for estimation and inference.

Higher-order autoregressions can also contain a unit root, revealed by a characteristic root equal to one. Differencing removes it, turning a random walk into white noise and an AR(2) with one unit root into a stationary AR(1) in differences, the logic behind ARIMA models. The syllabus then closes with an introduction to vector autoregressions. The practical response is transformation.

Differencing a series once applies one minus the lag operator; if that produces a stationary series, the original is integrated of order one and can be modelled as an ARIMA. Forecasting changes too: a random walk has no mean to revert to, so its forecasts stay at the last observed level.

In this chapter

What this chapter covers

  • 01

    The AR(1) with coefficient one: the random walk

  • 02

    Why variance grows and the mean does not settle

  • 03

    Unit roots in higher-order autoregressions

  • 04

    The differenced form of an autoregression

  • 05

    Integrated series and ARIMA models

  • 06

    The step to vector autoregressions

Worked example · free

Mean and variance of a random walk with drift

Q [6 marks]. A series follows y_t = 0.5 + y_{t-1} + e_t with y_0 = 10 and shock variance 2. Find the mean and variance of y_8 and the mean and variance of the first difference. This practice weighting is assigned by AskSia and is not an official assessment scheme.
Variance after t periods for two processes that both start at zero with unit shock variance.Variance after t periods for two processes that both start at zero with unit shock variance. The stationary AR(1) settles at 1/(1 − 0.64) = 2.78, while the random walk’s variance rises one unit every period without limit.01020300102030random walkAR(1), φ = 0.8limit 2.78periods since startvariance
Variance after t periods for two processes that both start at zero with unit shock variance. The stationary AR(1) settles at 1/(1 − 0.64) = 2.78, while the random walk’s variance rises one unit every period without limit.
  • 2Back substitution gives y_8 = 10 + 8(0.5) + the sum of eight shocks, so the mean is 14.
  • 2The eight uncorrelated shocks each have variance 2, so the variance of y_8 is 16 and keeps growing with t.
  • 2The first difference is 0.5 + e_t, with constant mean 0.5 and variance 2 at every date.
y_8 has mean 14 and variance 16, both depending on t, while the difference has mean 0.5 and variance 2, so differencing restores stationarity.
Sia tip — A moment that depends on t is proof of nonstationarity; compute one before asserting it.
Glossary

Key terms

Random walk
An autoregression of order one with coefficient one, in which each value equals the previous value plus a shock.
Integrated series
A series that becomes stationary only after being differenced a given number of times.
Differenced form
An autoregression rewritten with the change in the series on the left and the lagged level on the right.
FAQ

Unit Roots and Nonstationary Processes FAQ

What is a unit root?

It is a root of the characteristic equation of an autoregression equal to one. In an AR(1) it means the coefficient is one, the random walk case, where the variance grows over time and the process is not stationary.

How does differencing help?

Differencing applies one minus the lag operator, which cancels a unit root. A random walk becomes white noise, and an AR(2) whose characteristic roots are one and 0.3 becomes a stationary AR(1) in its first differences.

Where do VARs fit in the syllabus?

The syllabus lists an introduction to vector autoregressions as the last topic of Part II, and the learning outcomes describe VARs as the tool for modelling several time series together after the univariate material on stationarity and unit roots.

How can I spot a unit root in an AR(2)?

Solve the characteristic equation, the quadratic in lambda built from the two coefficients. If one root equals one, there is a unit root. The first stationarity inequality also fails at the boundary, because the two coefficients then sum to exactly one.

Why is a random walk not stationary?

Starting from a fixed value, its variance equals the number of periods times the shock variance, so it grows with time, and the covariance between two dates depends on the dates themselves. Both facts break the definition of weak stationarity.

What model do I fit after differencing?

Fit a stationary model to the differenced series, chosen with the identification tools for stationary processes. If one difference removes the unit root and the differences follow an AR(1), the original series follows an ARIMA with one autoregressive term and one difference.

Study strategy

Exam move

Derive the random walk moments by back substitution until you can show the variance growing with t in two lines. Practise factorising AR(2) lag polynomials when one root equals one, and write the resulting model for the differences. Rehearse the answer plan: definition, the condition that fails, the transformation that restores stationarity and the model you would fit afterwards.

Practise showing nonstationarity with a computed moment rather than a description, and practise factorising lag polynomials that contain one minus L. Each time, write the stationary model for the differences and state its mean, so the answer ends with a usable model rather than a diagnosis alone.

Add the random walk variance formula and the differenced form of an autoregression to your notes sheet, since both appear in diagnosis answers.

Working through Unit Roots and Nonstationary Processes in ECON31031? Sia is AskSia’s AI Economics tutor — ask any ECON31031 Unit Roots and Nonstationary Processes question and get a clear, step-by-step explanation grounded in how ECON31031 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 41 of your subjects at The University of Manchester, plus 1,000+ more Bibles across the AskSia library.
Sia - your ECON31031 tutor, unlimited, worked the way the exam marks it
The full 3-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works