The Bonferroni correction is a statistical technique that raises the bar for calling a result "statistically significant" when running multiple tests at once. It is one of the simplest and most widely used safeguards against false discoveries in quantitative research.

When researchers test dozens or hundreds of potential investment factors, some will appear significant purely by chance. The Bonferroni correction addresses this problem by dividing the standard significance threshold by the number of tests performed. The result is a stricter cutoff that makes it harder for random noise to masquerade as a genuine finding.

Definition

The Bonferroni correction adjusts the significance level (the probability threshold below which a result is considered meaningful) to account for the number of simultaneous comparisons. In statistics, a "significant" result typically means there is less than a 5% chance it occurred by luck alone. That 5% threshold is called the significance level, often written as α (alpha).

For example, if the standard threshold is 5% and a researcher runs 20 tests, the Bonferroni-adjusted threshold becomes 0.05 ÷ 20 = 0.0025, or 0.25%. A result must now have less than a 0.25% probability of occurring by chance before it can be called significant.

How It Works

The logic behind the Bonferroni correction is straightforward. Imagine flipping a coin 100 times and looking for unusual streaks. With enough flips, some streaks will appear simply because of randomness, not because the coin is rigged. The same principle applies to testing investment factors.

Consider a research team testing 100 potential stock-selection factors. At a 5% significance level, pure chance alone would produce roughly 5 factors that appear significant (100 × 0.05 = 5). Those 5 "discoveries" are false positives: they look real but reflect nothing more than statistical noise.

The Bonferroni correction handles this by lowering the threshold. With 100 tests, the adjusted significance level becomes 0.05 ÷ 100 = 0.0005, or 0.05%. Now a factor must clear a much higher bar before the researcher can claim it works. This dramatically reduces the odds that a false positive slips through.

Practical Example

Suppose a quantitative analyst screens 50 trading signals against historical data. Without any correction, using a 5% significance level, roughly 2 to 3 signals would appear to "work" by pure chance.

With the Bonferroni correction applied, the threshold drops from 5% to 0.1% per test. The expected number of false positives falls from about 2.5 to roughly 0.05. Any signal that passes this stricter bar is far more likely to reflect a genuine pattern rather than a statistical fluke.

In practice, this means the analyst's final shortlist will be smaller but more reliable. A signal with a p-value of 0.03 (3%) would pass the uncorrected threshold but fail the Bonferroni threshold. Only signals with very strong statistical evidence survive.

Known Limitations

Related Methods

For most quantitative finance applications involving tens or hundreds of factor tests, the Benjamini-Hochberg procedure is often preferred because it balances the risk of false discoveries against the risk of missing real effects. The Bonferroni correction remains the gold standard when the cost of a single false positive is very high, such as in regulatory reporting or clinical trials.

Academic Origin

The correction is named after Italian mathematician Carlo Emilio Bonferroni, who developed the underlying probability inequality in 1936. However, the method as applied in statistics was popularized by Olive Jean Dunn in her 1961 paper, which showed how to use Bonferroni's inequality for simultaneous confidence intervals and hypothesis tests. Dunn's formulation is what researchers use today.

In quantitative finance, the Bonferroni correction gained renewed attention after Harvey, Liu, and Zhu's influential 2016 paper, which argued that most published factor discoveries in finance fail to survive proper multiple-testing adjustments. Their work highlighted how many of the hundreds of "anomalies" in the academic literature may be false positives, a finding with direct implications for factor investing strategies.

Further Reading

• Bonferroni, C.E. (1936). "Teoria statistica delle classi e calcolo delle probabilità." Pubblicazioni del R Istituto Superiore di Scienze Economiche e Commerciali di Firenze, 8, 3–62.
• Dunn, O.J. (1961). "Multiple Comparisons Among Means." Journal of the American Statistical Association, 56(293), 52–64.
• Harvey, C.R., Liu, Y., and Zhu, H. (2016). "...and the Cross-Section of Expected Returns." The Review of Financial Studies, 29(1), 5–68.
• Benjamini, Y. and Hochberg, Y. (1995). "Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing." Journal of the Royal Statistical Society: Series B, 57(1), 289–300.

Related Terms