The four-fifths rule explained for small employers

What does the rule compare?

The four-fifths rule compares selection rates between groups. A selection rate is the number of people selected divided by the number considered in a defined group and stage. The screening ratio divides one group’s selection rate by the highest group selection rate. A ratio below 0.80 can flag a difference that deserves investigation. It does not, by itself, explain the cause of that difference or determine whether a selection process is lawful.

The distinction between a flag and a conclusion is essential. A dashboard can make a ratio look authoritative because the arithmetic is precise. The interpretation is more complicated. You need to understand the population, the stage, the sample size, the quality of the data and the relevance of the selection method. A ratio above the threshold is not a certificate of fairness, and one below it is not a complete legal finding.

The U.S. EEOC guidance on employment tests and selection procedures is a primary source for the broader legal context of employment testing. This article is an educational explanation of a screening calculation. It does not replace advice about a particular employer, jurisdiction or hiring decision.

Work through a hypothetical example

Suppose a clearly defined assessment stage includes two hypothetical groups. In Group A, 20 of 40 applicants advance, giving a selection rate of 50%. In Group B, 8 of 20 advance, giving a selection rate of 40%. Dividing 40% by 50% gives 0.80, or 80%. The example sits exactly at the conventional screening threshold. These numbers are invented for explanation; they are not HireValid customer outcomes.

Now suppose only 6 of the 20 applicants in Group B advance. Its selection rate becomes 30%. Dividing 30% by 50% gives 0.60. That ratio is below four-fifths and would prompt closer review. The calculation still does not identify whether the difference arose from the test, a prior screening stage, an inconsistent review practice, missing data or another factor.

Keep the denominator and stage consistent. If one group’s denominator includes everyone who applied while another includes only people who completed a test, the comparison is not measuring the same thing. Label the population and time period explicitly. A useful report should allow a reviewer to understand where each count came from rather than displaying a ratio with no underlying numbers.

Why small samples need care

In a small hiring round, one person can move a rate substantially. If a group contains five people, changing one outcome changes the selection rate by twenty percentage points. That does not mean the person’s experience is unimportant. It means the number is unstable and should not support broad conclusions without context. A responsible review treats uncertainty as part of the result.

Do not combine unrelated roles or periods simply to make the sample bigger. Different jobs may have different requirements and selection stages. Combining them can hide an important difference or create a misleading one. If you aggregate data, explain why the populations and processes are comparable. Keep the underlying views available to appropriately authorized reviewers.

A group with no applicants or a comparison group with no selections can also create undefined or uninformative ratios. Software should not quietly turn these into zero, a pass indicator or a confident warning. Show the actual counts and explain that the ratio cannot be interpreted in the usual way. A clear limitation is more useful than a tidy but misleading status badge.

Protect the information used in analysis

Demographic information can be sensitive. Decide whether collecting it is appropriate and lawful, how participation is explained and who can access the results. Voluntary responses may be incomplete, and missingness can affect interpretation. Do not infer protected characteristics from names, photographs or other proxies merely to populate a chart. That creates additional risks and can make the analysis less trustworthy.

The HireValid brief proposes optional, consent-based demographic collection stored separately from individual candidate scores. Fairness reporting is a later-release plan, not a live feature established by this website. The intention is to support aggregate review without presenting demographic attributes beside an individual’s assessment result to the hiring manager.

Before implementation, define access permissions, retention and minimum reporting conditions. Small groups may be identifiable even in a summary table. The person reviewing aggregate outcomes does not necessarily need access to every candidate’s sensitive information. Legal and privacy review should address the actual collection and reporting workflow rather than relying on the reassuring name of a feature.

Investigate the process behind a flag

Start by checking the data. Confirm that the group counts, stage definitions and dates are correct. Look for incomplete attempts, duplicate records and manual overrides. Check whether the same decision rule was applied throughout the period. A change in the cutoff or an undocumented exception can alter outcomes without appearing in the final summary.

Then review job relevance and possible barriers. Does the assessment measure a requirement of the role? Is unnecessary language complexity affecting performance? Are timing or equipment requirements appropriate? Were accommodations communicated and handled consistently? A ratio tells you where to look; these process questions help you understand what might need improvement.

Review the whole selection sequence as well as the specific stage. A test may be used after an earlier screening process that already changed the applicant pool. An interview after the test may introduce a different inconsistency. Avoid attributing every observed difference to the most visible tool. Keep a record of the investigation and obtain qualified advice before drawing legal conclusions or changing a consequential policy.

Avoid mechanical responses

An employer should not respond to a flag by silently changing individual scores or hiding the metric. Nor should it assume that replacing one test with another automatically resolves the issue. A meaningful response examines the requirement, the evidence supporting the assessment and whether an alternative could meet the business need with fewer unrelated barriers.

If the problem is unclear instructions, revise and test those instructions. If a requirement is unnecessary, remove it from the process. If reviewers are inconsistent, improve the rubric and calibration. These examples are process improvements, not a promise that any particular action achieves compliance. Document the reason for the change and what evidence would help you evaluate it afterward.

Make changes prospectively and thoughtfully. Moving a threshold after seeing the names or group outcomes can introduce new problems. A fair review needs consistent principles, transparent documentation and advice suited to the actual circumstances. The objective is a job-relevant process, not simply making a dashboard turn green.

Build a modest monitoring routine

A small employer can begin by documenting each stage, its purpose and the criteria for advancing. Keep the underlying counts and explain any limitations. Set a time to review the process after a hiring round rather than waiting for a complaint. Include candidate feedback and accommodation experiences, not only numerical outcomes.

Assign responsibility for follow-up. A report without an owner can become a passive record of the same unresolved concern. Decide who checks data quality, who reviews assessment design and who obtains legal advice when needed. Keep sensitive data access limited even if the business has a small team. Familiarity among colleagues does not remove the need for appropriate boundaries.

Key takeaways

• Calculate comparable selection rates before calculating the ratio.
• Treat the four-fifths threshold as a screening convention, not a compliance verdict.
• Show the underlying counts and acknowledge small-sample uncertainty.
• Protect demographic data and avoid inferring sensitive attributes.
• Investigate job relevance, barriers and the whole selection process.

Read the definitions of selection rate and adverse impact, then review the broader fairness approach.