The ADF Unit Root Test Explained for Economics and Finance Students

If you are using World Bank, CBN, or NBS data in your dissertation, your supervisor will ask about stationarity. Here is a plain-English explanation of the ADF test and how to write it up.

If you are writing an Economics or Finance dissertation using secondary data from institutions like the World Bank, CBN, IMF, or NBS, your supervisor will almost certainly ask about stationarity and the ADF test. This guide explains what it means in plain English.

What is stationarity?

A time series is stationary if its statistical properties — mean, variance, and autocorrelation — do not change over time. When you run a regression on two non-stationary variables, you can get a "spurious regression" — a high R-squared that looks significant but is entirely meaningless. Two unrelated trending variables will appear to be correlated just because they both move in the same direction over time.

This is why you must test for stationarity before running any regression on time series data.

What is the ADF test?

The Augmented Dickey-Fuller (ADF) test is the standard test for unit roots in a time series. A "unit root" is what makes a series non-stationary.

Null hypothesis (H0): The series has a unit root (it is non-stationary)
Alternative hypothesis (H1): The series has no unit root (it is stationary)

If p < .05, you reject the null hypothesis and conclude the series is stationary. If p > .05, the series is non-stationary and you need to difference it before using it in a regression.

How to read ADF results

The ADF output gives you:
- ADF test statistic (a negative number — more negative means more evidence against unit root)
- p-value (below .05 = stationary)
- Critical values at 1%, 5%, and 10% levels
- Number of lags used (selected automatically by AIC)

Example result:
ADF = -3.82, p = .003 (stationary)

The test statistic (-3.82) is more negative than the 5% critical value (-2.98), confirming stationarity.

What to do when your series is non-stationary

First difference the series (subtract each observation from the previous one) and test again. Most economic time series become stationary after first differencing. This is written as I(1) — integrated of order 1. If it takes two differences to achieve stationarity, the series is I(2).

Reporting ADF in your dissertation

Standard write-up:

"Prior to regression analysis, the stationarity of all variables was assessed using the Augmented Dickey-Fuller (ADF) unit root test. GDP growth was found to be stationary at levels (ADF = -3.82, p = .003). Inflation was non-stationary at levels (ADF = -1.24, p = .661) but stationary after first differencing (ADF = -4.17, p < .001), indicating integration of order one, I(1)."

ResearchScope and the ADF test

ResearchScope (bizscope.space/research) runs the ADF test automatically on Researcher plan datasets when a year column is detected. It uses statsmodels with AIC lag selection, presents critical values at all three levels, and gives a plain-English stationarity decision. The result string is APA-formatted and ready to copy into your Chapter 4.

Try it on your own data

Upload your CSV and get a full business intelligence report in 30 seconds. Dashboard is completely free, no account needed.

Analyze my data free