Jarque-Bera Test
What you'll learn: How to test whether data comes from a normal distribution by checking its skewness and kurtosis together in a single statistic.
What Is the Jarque-Bera Test?
You've learned several normality tests (Shapiro-Wilk, Kolmogorov-Smirnov, Anderson-Darling), but the Jarque-Bera test takes a unique approach: it specifically looks at two shape characteristics—skewness and kurtosis—and combines them into a single test statistic.
Remember that normal distributions have skewness = 0 (perfectly symmetric) and kurtosis = 3 (specific tail behavior). The Jarque-Bera test measures how far your data deviates from these two ideal values.
How It Works
The Jarque-Bera test statistic (JB) is calculated from your sample's skewness and kurtosis. If your data is truly normal, this statistic should be small (close to zero). Large values indicate departures from normality—either through asymmetry (skewness ≠ 0) or unusual tail behavior (kurtosis ≠ 3).
Think of it like a two-pronged quality check: you're simultaneously verifying both the symmetry and the tail thickness of your distribution.
The test returns a p-value: if it's below your significance level (typically 0.05), you reject the hypothesis that your data is normally distributed.
When to Use It
The Jarque-Bera test is particularly popular in financial modeling and econometrics because it's computationally simple and directly targets the two most common ways real-world data violates normality. However, like other normality tests, it becomes very sensitive with large samples, often detecting trivial deviations.
Key Takeaway: The Jarque-Bera test evaluates normality by measuring deviations in both skewness and kurtosis simultaneously, making it especially useful when you care specifically about asymmetry and tail behavior.