Error Term

What Is an Error Term?

An error term is the component of a regression analysis equation that represents the variation in an outcome that the model does not explain. In a simple linear model, the relationship is written as:

Y = β0 + β1X + ε

Here Y is the dependent variable, X is an independent variable, β0 and β1 are estimated coefficients, and ε (epsilon) is the error term. Whatever the model cannot account for, such as omitted factors, measurement noise, or pure chance, is folded into ε.

How It Works

Analysts never observe the true error term directly. Once a model is fitted, the gap between each actual value and the model's predicted value is called a residual, and residuals serve as the practical stand-in for the error term. Coefficients are often estimated using ordinary least squares, which chooses the line that makes the squared residuals as small as possible overall.

For standard inference methods to be trustworthy, the error term is usually assumed to have an average of zero, a constant spread across observations, and no systematic pattern from one observation to the next. When these assumptions break down, the reported statistical significance of the coefficients can be misleading.

Why It Matters to Investors

Consider a regression of a stock's daily returns on the returns of a broad market index. The market term explains part of the stock's movement, while the error term absorbs company-specific news, earnings surprises, and other factors the model leaves out. A large error term relative to the total variation means the model explains only a small share of the stock's behavior, which is commonly summarized by R-squared.

Reading regression output carefully helps investors judge how much weight to give a model's results. Key points include:

• A small error term relative to the variation in the data suggests a better fit.
• A large error term does not mean the model is wrong, only that other factors matter.
• Patterns in the error term can signal that important variables are missing.