What is Adjusted R-Squared?

Adjusted R-squared is a refined version of the R-squared statistic used to judge how well a regression model fits the data. Crucially, it corrects a misleading quirk of ordinary R-squared, making it a better tool for comparing models — especially those with different numbers of explanatory variables. This guide explains what adjusted R-squared is, why it's needed, how to interpret it, and why it matters — in clear, plain language. It's a core topic in quantitative qualifications like the FRM.

First, a quick recap of R-squared

R-squared (the coefficient of determination) measures the proportion of the variation in the outcome variable that a regression model explains, on a scale from 0 to 1. An R-squared of 0.70 means the model accounts for 70% of the variation in what you're trying to predict. Higher generally looks better — but there's a catch that adjusted R-squared exists to fix.

The problem adjusted R-squared solves

Ordinary R-squared has a significant flaw: it never decreases when you add more variables to a model, even if those variables are useless. Adding any explanatory variable — however irrelevant — will leave R-squared the same or, more usually, nudge it up slightly, simply by chance. This creates a temptation to throw in variables to inflate R-squared, producing a model that looks better but isn't genuinely more useful, and may be "overfitted" to the quirks of the sample data. Relying on R-squared alone, an analyst could be fooled into preferring a bloated, worse model that performs poorly the moment it meets new data.

What adjusted R-squared does

Adjusted R-squared fixes this by penalising the model for each additional variable. It only rises when a new variable improves the model's fit by more than would be expected by chance; if a variable adds little genuine explanatory power, adjusted R-squared falls. In effect, it asks not just "does this variable add fit?" but "does it add enough fit to justify the added complexity?" This makes it a more honest measure of a model's quality, and unlike ordinary R-squared, it can decrease — and can even be slightly negative for very poor models.

How to interpret and use it

Adjusted R-squared is most valuable when comparing models with different numbers of variables. The practical rules of thumb are:

• If adding a variable increases adjusted R-squared, the variable is likely earning its place in the model.
• If adding a variable decreases adjusted R-squared, it's probably not pulling its weight and may be better left out.
• When choosing between models, the one with the higher adjusted R-squared generally offers the better balance of fit and simplicity.

Adjusted R-squared is always less than or equal to ordinary R-squared, and the gap between them widens as you add more variables — a useful signal in itself.

Why it matters

Adjusted R-squared embodies an important principle in modelling: more complexity is not automatically better. By rewarding genuine explanatory power and penalising needless variables, it helps analysts build models that are both well-fitting and parsimonious — and that are more likely to perform well on new data rather than just fitting the past. It's a guard against one of the most common modelling mistakes, overfitting, which makes it a valuable everyday tool for anyone who builds or relies on regression models.

Why it matters for finance professionals

For anyone building or interpreting regression models — in forecasting, risk, or financial analysis — adjusted R-squared is an essential measure. It allows for fair comparison between competing models and discourages the temptation to inflate apparent fit with irrelevant variables. Understanding it, and how it improves on ordinary R-squared, is fundamental to sound quantitative analysis and a regularly examined topic in professional qualifications.

Frequently asked questions

What is adjusted R-squared?

A refined version of R-squared that measures how well a regression model fits the data while penalising the inclusion of unnecessary variables — making it a fairer measure for comparing models.

How does it differ from ordinary R-squared?

Ordinary R-squared never decreases when you add variables, even useless ones. Adjusted R-squared can fall if an added variable doesn't improve the model enough to justify the extra complexity.

Why is adjusted R-squared useful?

It's especially valuable for comparing models with different numbers of variables, helping you choose one that balances good fit with simplicity and guards against overfitting.

Can adjusted R-squared be negative?

Yes — for a very poorly fitting model it can be slightly negative, whereas ordinary R-squared cannot fall below zero. It's also always less than or equal to ordinary R-squared.

Build your quant skills with Learnsignal

Adjusted R-squared is a key tool for sound regression modelling. Learnsignal's tutor-led courses, including the FRM, develop the statistical understanding that topics like this build on — with clear teaching that makes the methods genuinely usable in practice.