How to Calculate Adverse Impact: 9 Steps

If the phrase adverse impact makes you feel like you accidentally wandered into a meeting with three lawyers, two auditors, and one very stressed HR director, take a breath. The concept matters, but the math is not nearly as scary as the name. At its core, adverse impact analysis is a way to check whether a hiring, promotion, testing, or termination process is affecting one protected group at a meaningfully different rate than another.

In plain English: if one group keeps getting selected less often, promoted less often, or screened out more often, you need to know. Not guess. Not vibe-check it. Know it.

This guide walks through how to calculate adverse impact in nine practical steps, using the well-known four-fifths rule, also called the 80% rule. You will also learn when that rule is only the beginning, how to avoid common mistakes, and what to do when your numbers start waving little compliance red flags.

What Adverse Impact Means

Adverse impact happens when a seemingly neutral employment practice results in a substantially different selection rate for a protected group. This can show up in hiring, promotions, transfers, layoffs, training access, or any other employment decision that uses a selection process.

It is important to understand what adverse impact is not. It is not the same thing as proving unlawful discrimination all by itself. It is a warning signal. Think of it like the check-engine light in your car: maybe the issue is small, maybe it is serious, but ignoring it is rarely a winning strategy.

The four-fifths rule is the most common first screen. Under that rule, if the selection rate for one group is less than 80% of the selection rate for the highest-performing comparison group, adverse impact is generally indicated. But the rule is a guideline, not a magic force field. Small samples can distort the picture, and large samples can reveal meaningful issues even when the ratio stays above 80%.

Before You Start: Gather the Right Data

Before you calculate anything, decide exactly what employment decision you are analyzing. Are you reviewing:

• Applicants who passed an initial screening?
• Candidates who received interviews?
• People who were hired?
• Employees who were promoted?
• Workers selected for termination in a reduction in force?

This matters because adverse impact should be analyzed at the actual decision point. If you toss every stage into one giant spreadsheet smoothie, your results may look neat while telling you absolutely nothing useful.

You will usually need, for each group being compared:

• Total number of people in the group at that stage
• Number selected or advanced
• The specific job, requisition, or decision step
• Consistent demographic categories used for compliance purposes

And one more thing: use clean data. Misspelled categories, duplicate applicants, or missing dispositions can wreck the analysis faster than a hiring manager saying, “I just went with my gut.”

How to Calculate Adverse Impact: 9 Steps

Step 1: Define the Employment Decision You Are Testing

Start by identifying the exact action being measured. For example, “who got hired for sales associate roles in Q1” is clear. “Overall hiring stuff” is not. You should calculate adverse impact by job, decision type, and often by stage in the process.

If you are analyzing a positive action like hiring or promotion, you will compare selection rates based on who moved forward. If you are analyzing a negative action like termination, you still use rates, but you interpret the reference group differently because a lower termination rate is the better outcome.

Step 2: Identify the Groups You Will Compare

Next, break the data into the groups you need to review. In most cases, this means looking separately by sex and by race or ethnicity. Depending on the context, employers may also review other legally relevant groupings. The key is consistency. Compare like with like.

Do not mash all underrepresented groups together into one mystery bucket called “diverse candidates.” That might look tidy on a slide deck, but it can hide serious disparities inside the category.

Step 3: Count the Total Number of People in Each Group

For each group, count how many people were in the selection pool for that decision. This is your denominator.

Example:

• 80 male applicants
• 40 female applicants

If you are measuring the interview stage instead of the application stage, then only count people who reached that stage. Your denominator must match the exact point in the process you are studying.

Step 4: Count How Many People in Each Group Were Selected

Now count how many people in each group got the outcome you are measuring. This is your numerator.

Continuing the example:

• 48 male applicants advanced
• 12 female applicants advanced

Simple? Yes. Easy to mess up when the spreadsheet is messy? Also yes.

Step 5: Calculate the Selection Rate for Each Group

Use this formula:

Using the example above:

• Male selection rate = 48 / 80 = 0.60, or 60%
• Female selection rate = 12 / 40 = 0.30, or 30%

This is the heart of the calculation. If you get this part wrong, the rest of the analysis becomes premium-quality nonsense.

Step 6: Identify the Reference Group

For positive decisions like hiring or promotion, the reference group is the group with the highest selection rate. That becomes your comparison benchmark.

In the example above, the male group has the higher rate at 60%, so it is the reference group.

For negative decisions like terminations, practitioners often use the group with the lowest termination rate as the reference group, because that is the more favorable outcome. This is where a lot of people get twisted up, so slow down here and make sure the comparison matches the action being analyzed.

Step 7: Calculate the Impact Ratio

Now divide each comparison group’s selection rate by the reference group’s selection rate.

Using the same example:

• Impact ratio = 30% / 60% = 0.50, or 50%

This means the female applicants were selected at 50% of the rate of the reference group.

Step 8: Apply the Four-Fifths Rule

Here is the test everyone remembers:

• If the impact ratio is 80% or higher, adverse impact is generally not indicated by the four-fifths rule.
• If the impact ratio is less than 80%, adverse impact is generally indicated.

In our example, 50% is below 80%, so the result indicates adverse impact under the four-fifths rule.

That does not automatically mean the process was unlawful. It does mean you should investigate further, document what happened, and evaluate whether the selection procedure is job-related and consistent with business necessity.

Step 9: Check the Bigger Picture and Document Your Findings

This final step separates careful compliance work from lazy math cosplay.

Ask the following:

• Was the sample size large enough to make the result meaningful?
• Was the sample so small that one person would flip the result?
• Should you run a statistical significance test in addition to the four-fifths rule?
• Did the overall process show adverse impact, or is the issue concentrated in one stage such as a test, resume screen, or panel interview?
• Did recruiting practices discourage certain groups from applying in the first place?

When numbers are very small, the 80% rule can overreact. When numbers are very large, it can underreact. That is why good analysts often pair the impact ratio with statistical testing and a careful review of the actual process.

Document everything: your data source, group counts, selection rates, impact ratios, date of analysis, job group, stage reviewed, and any follow-up action. If a stage appears to create adverse impact, isolate that stage and assess whether there is a less exclusionary alternative that is substantially equally valid.

Worked Example

Result: Because 50% is below 80%, the analysis indicates adverse impact under the four-fifths rule.

Now imagine the company finds that the disparity begins at a pre-employment assessment stage, not the interview stage. That tells the employer where to look: the assessment itself, its cutoff score, how it is administered, whether it is truly job-related, and whether a different procedure could predict performance with less adverse impact.

Common Mistakes That Can Wreck the Analysis

Using the Wrong Denominator

If you divide hires by total applicants for one group but by screened applicants for another, your analysis is toast.

Combining Different Jobs

A warehouse role and a software engineering role may have very different applicant pools and selection methods. Combining them can hide real disparities or create fake ones.

Ignoring Individual Stages

Your overall hiring rate may look fine while one stage quietly causes the problem. Resume screens, tests, structured interviews, and background check steps can each produce different results.

Trusting Vendors Too Much

If a third-party assessment vendor says, “Our tool is fair,” that is nice. It is not the same thing as your organization verifying impact in your actual use case. The same warning applies to AI screening tools, algorithmic scoring systems, and automated decision aids. Fancy software can still produce old-fashioned legal risk.

Treating 80% Like a Final Verdict

The four-fifths rule is a first-pass screen. It is helpful, practical, and widely used, but it is not the only lens that matters.

What to Do If Adverse Impact Is Indicated

If your calculation shows adverse impact, do not panic and do not bury the spreadsheet in a folder named “misc final final v2.” Instead:

1. Confirm the data is accurate.
2. Determine whether the issue appears in the total selection process, an individual stage, or both.
3. Run additional statistical analysis where appropriate.
4. Review whether the selection procedure is job-related and supported by validation evidence.
5. Consider alternative procedures that are substantially equally valid and have less adverse impact.
6. Document the review and involve legal counsel or qualified compliance professionals where needed.

That response is far better than pretending the ratio did not happen. Numbers have a funny habit of reappearing when auditors, plaintiffs’ lawyers, or regulators show up.

Real-World Experiences and Lessons From Adverse Impact Reviews

In practice, the most eye-opening part of adverse impact analysis is not usually the math. It is what the math reveals about how decisions are actually being made. Teams often begin with the belief that their process is objective because it feels structured. Then the numbers show that one stage is filtering out a protected group at a much higher rate, and suddenly the room gets very quiet except for the faint sound of someone asking who built the spreadsheet.

One common experience is discovering that the problem is not the final hiring decision at all. The full process may look acceptable on the surface, but once the funnel is broken into stages, the trouble appears much earlier. Maybe a personality assessment is knocking out candidates before a human ever reviews them. Maybe a resume screen favors applicants with certain backgrounds that are not actually necessary for success. Maybe an interview panel is technically “structured,” but every manager uses the scoring guide differently. Adverse impact analysis turns vague suspicions into something measurable.

Another frequent lesson is that small process choices can have big consequences. A cutoff score that seemed perfectly reasonable can create a steep drop-off for one group. A physical test that sounded job-related during a meeting may turn out to be broader than the job truly requires. A background-check rule may exclude people in a way that is more dramatic than anyone expected. The experience many employers report is not, “We meant to discriminate.” It is, “We never stopped to measure the outcomes closely enough.”

There is also a very modern version of this story involving AI and automated hiring tools. Organizations sometimes assume that because the tool is data-driven, it must also be neutral. That assumption can age badly. If an algorithm scores candidates based on patterns tied to historical hiring data, the process can scale an old bias at digital speed. Adverse impact calculations remain useful here because they force the employer to test outcomes rather than fall in love with tech vocabulary.

Analysts also learn quickly that documentation matters almost as much as the ratio itself. When a company can show its data source, formulas, review dates, stage-by-stage results, validation support, and consideration of alternatives, it is in a much stronger position than a company that says, “We’re pretty sure the vendor handled that.” In real compliance work, confidence without records is just optimism wearing a blazer.

Finally, many teams come away from these reviews with a healthier mindset: adverse impact analysis is not just a legal exercise. It is a quality-control tool. When done well, it helps employers build selection systems that are more consistent, more defensible, and often better at identifying qualified people. In other words, the experience is not merely about avoiding trouble. It is about designing hiring and promotion processes that work better for the business and more fairly for the people moving through them. That is a much better outcome than discovering too late that your “objective” process had been making deeply subjective mistakes in a very mathematical way.

Conclusion

Calculating adverse impact is not about memorizing one formula and declaring victory. It is about understanding the decision being made, measuring who is selected at each stage, comparing rates carefully, and treating the four-fifths rule as the beginning of the conversation rather than the end of it.

If you remember the essentials, you are already ahead of the game: calculate the selection rate for each group, identify the reference group, compute the impact ratio, compare it to 80%, and then check whether sample size, stage-specific issues, or validation questions change the story. That is the real value of adverse impact analysis. It helps you move from assumptions to evidence.

And in HR, compliance, and talent analytics, evidence beats hunches every single time.