Coefficient of Determination (R²): Formula & Finance Use

What Is the Coefficient of Determination?

The coefficient of determination is used in regression analysis to evaluate how well an independent variable, or a group of independent variables, explains variation in a dependent variable. An R² of 0.80 means that 80 percent of the observed movement in the dependent variable is accounted for by the model, while the remaining 20 percent reflects unexplained variation.

In corporate finance, the measure appears most often in beta estimation for the Capital Asset Pricing Model, where it helps analysts assess how much of a company's share price movement is linked to the broader market. It also supports the review of regression models used in discounted cash flow forecasts when revenue, margins, or cash flows are modelled against external drivers.

How R² Works

R² compares the total variation in the observed data with the variation the regression model fails to explain. Total variation, known as SS_tot, measures how far each observed value sits from the mean of the dependent variable. Residual variation, known as SS_res, measures how far each observed value sits from the value predicted by the regression model.

When prediction errors are small relative to the overall spread of the data, the model produces an R² closer to 1. When the model performs little better than using the mean as a simple estimate, the result sits closer to 0. This makes R² a useful diagnostic because it converts the fit of a regression line into a single interpretable number.

Worked Example

Assume an analyst regresses a stock's monthly returns against the FTSE 100 index over 24 months to estimate beta for a weighted average cost of capital calculation. The total sum of squares is 0.0450 and the residual sum of squares is 0.0108.

The calculation is R² = 1 - (0.0108 / 0.0450), which gives 0.76. This tells the analyst that 76 percent of the stock's return variation over the period was explained by movements in the FTSE 100, while 24 percent was driven by firm-specific factors or other influences outside the model.

Finance Example

When analysts model NVIDIA's stock returns against the S&P 500 to estimate equity beta, the R² of that regression reveals how much of NVIDIA's price movement tracks the broader market rather than company-specific factors such as data-centre demand cycles, semiconductor margins, or AI chip adoption. A regression producing an R² of 0.45 would indicate that fewer than half of the return fluctuations are explained by general market movements.

That relatively modest fit matters because beta often feeds into the cost of equity, which then affects discount rates, net present value, and enterprise value. If the regression explains only a limited share of actual return movement, the valuation should acknowledge that the beta estimate carries a wider judgement range than the spreadsheet output suggests.

Key Considerations and Limits

R² is most useful as a diagnostic that helps determine whether a regression captures enough of a relationship to justify using its outputs. A high value can be reassuring, but the metric does not distinguish between genuine explanatory power and coincidental statistical fit. This is why analysts should read R² alongside the economic logic of the model rather than treating it as proof that the relationship is valid.

The measure also assumes that the relationship between variables is linear. A model may produce a respectable R² while still missing nonlinear behaviour, changing volatility, or structural breaks in the data. In practice, residual plots, sensitivity analysis, and out-of-sample testing are needed before a regression-based forecast can support a financing, valuation, or capital allocation decision.

• R² can rise when extra variables are added, even if those variables add little economic insight.
• A strong in-sample fit may still fail when tested on new data.
• A high R² can coexist with poor model design if the regression is based on unstable or poorly chosen inputs.

R² vs Adjusted R²

Adjusted R² responds to one of the main weaknesses of ordinary R² by penalising variables that do not improve the model's explanatory power enough to justify their inclusion. This distinction becomes important when a financial model uses several independent variables, because extra inputs can make a regression appear better fitted while adding little practical value.

For a model that links cash flow drivers to several macroeconomic inputs, adjusted R² is usually more informative than ordinary R² because it asks whether each added variable improves the model enough to justify the additional complexity. That discipline matters when a forecast will influence valuation, capital budgeting, or risk assessment.

Conclusion

The coefficient of determination is a practical way to assess how much explanatory strength a regression model has. In finance, that makes it useful when testing beta estimates, reviewing valuation drivers, and deciding whether a relationship in historical data is strong enough to support forward-looking analysis.

Executive use of R² should remain disciplined. A strong value can support confidence in a model, but it should never replace economic judgement, residual analysis, or testing under different assumptions. The metric is most valuable when it helps decision-makers ask better questions about model reliability before capital, valuation, or risk decisions are made.