Third-Party Research & Methodology Only

This section shares summaries of third-party academic research and descriptions of quantitative models. The content represents the findings of the original researchers, not the opinions or recommendations of Foxholm Financial. Foxholm Financial does not publish hypothetical or backtested performance metrics on its quantitative research pages. All content is restricted to methodology, signal construction, factor logic, and risk architecture. SEC rules require that investment advisers not present misleading performance data, and our methodology-only approach reflects that standard and the firm's fiduciary obligations.

P-Hacking

Research Risk Statistical Validity Hypothesis Testing

P-hacking is the practice of adjusting an analysis, consciously or not, until a result crosses the threshold of statistical significance. The name refers to the p-value, a number that summarizes how surprising a result would be if there were no real effect. By trying many variations and keeping the one that produces a small p-value, a researcher can manufacture the appearance of a discovery where none exists.

P-hacking sits at the heart of the broader reproducibility problems seen across many empirical fields, including quantitative finance. It is rarely deliberate fraud. More often it is the cumulative effect of small, defensible-looking choices, each nudging the analysis toward a significant result. The danger is that standard significance tests assume those choices were fixed in advance, and they were not.

Definition

A p-value measures the probability of seeing a result at least as extreme as the one observed, assuming the null hypothesis (the assumption of no real effect) is true. A common convention treats a p-value below 0.05 as "statistically significant." P-hacking exploits the flexibility in how an analysis is conducted to push a p-value below that line, even when the underlying effect is absent. It is the deliberate or inadvertent search for a significant-looking p-value rather than a test of a pre-specified hypothesis.

Key Principle

A p-value only means what it claims when the test was decided before looking at the data. Once a researcher tries many specifications and reports the one that happened to clear the 0.05 threshold, the reported p-value understates how likely the result was to arise by chance. The threshold has been gamed, not met.

Common Forms

P-hacking takes many shapes, and most look innocuous in isolation. What unites them is that each adds a hidden degree of freedom that standard tests do not account for.

Practice How It Inflates Significance
Selective stopping Collecting or extending data until the result becomes significant, then stopping
Trying many variables Testing numerous predictors and reporting only the ones that reached significance
Flexible outcome definitions Choosing among several ways to measure the result after seeing which one works
Subgroup mining Slicing the data into subgroups until one shows a significant effect
Optional covariates Adding or removing control variables until the p-value drops below the threshold

These practices are closely linked to data snooping and the multiple testing problem. The common thread is that running many analyses against the same data inflates the chance of a false positive, while reporting only the winner hides how much searching took place. In a trading context, the same dynamic drives overfitting and curve fitting.

Guarding Against It

The most effective safeguards remove the flexibility that makes p-hacking possible. Pre-registration, where the hypothesis, sample size, and analysis plan are specified before data collection, restores the meaning of the p-value by fixing the test in advance. Reporting every analysis attempted, not just the significant one, lets readers judge how much searching occurred.

When many tests are genuinely necessary, corrections for multiple testing raise the significance bar accordingly. In finance specifically, demanding that a result hold up in out-of-sample testing provides a check that no amount of in-sample p-hacking can fake, because the held-back data was never part of the search.

Known Limitations

Limitations to Keep in Mind

  • Intent is hard to judge. The same analytical choice can be legitimate exploration or p-hacking depending on whether it was planned. Because intent is invisible in the final report, honest and dishonest analyses can look identical on paper.
  • Some flexibility is unavoidable. Real analysis involves judgment calls about data cleaning, variable definitions, and model choice. Eliminating all flexibility is neither possible nor desirable, so the line between rigor and p-hacking is often a matter of degree.
  • Pre-registration has limits. Pre-registration constrains confirmatory tests but can discourage genuine exploratory work, where the most interesting findings sometimes emerge. It reduces p-hacking at some cost to discovery.
  • Correcting too aggressively hides real effects. Strict multiple-testing corrections lower false positives but raise false negatives, so genuine effects can be dismissed as noise. There is a real trade-off between catching false discoveries and missing true ones.

Further Reading

Glossary Research Risk Statistical Validity Hypothesis Testing False Positives
On This Page

Meet with a Fiduciary Advisor

Foxholm Financial is a fee-only registered investment adviser serving Georgia. We bring quantitative rigor to every client engagement. Explore our services or get in touch to discuss how we can help. To see how this kind of analysis informs real client work, explore a Strategic Portfolio Review.

Institutional Clients

Are you an institution or FinTech firm? Learn about our Quantitative Consulting Services.

Quantitative Fellowships

Foxholm Financial trains the next generation of quantitative analysts. Students and early-career researchers can explore our quantitative investment fellowships.

Disclaimer

This content is for educational and informational purposes only and does not constitute an offer to sell or a solicitation of an offer to buy any securities. Nothing herein constitutes investment advice or recommendations tailored to your individual situation. All investments involve risk, including the potential loss of principal. Past performance is no guarantee of future results. Information presented is believed to be factual and up-to-date, but Foxholm Financial does not guarantee its accuracy and it should not be regarded as a complete analysis of the subjects discussed. Before making investment decisions, consult with a qualified financial advisor who can evaluate your specific circumstances.