What Is Relative Risk?
Relative risk (RR), also called the risk ratio, measures how much more (or less) likely an outcome is in one group compared to another. It's calculated as the ratio of the probability (risk) of the outcome in the exposed or treatment group to the probability in the unexposed or control group. If customers who see a product demo have a 40% purchase rate and those who don't see it have a 20% rate, the relative risk is 40%/20% = 2.0, meaning the demo group is twice as likely to purchase. Relative risk is more intuitive than odds ratios because it directly compares probabilities rather than odds. However, it can only be calculated from prospective (cohort) designs or experiments where you know the true incidence rate in each group, not from case-control studies.
Why Relative Risk Matters
Relative risk is the most natural way to communicate effect sizes for binary outcomes to non-technical audiences. "Customers who received the loyalty offer were 1.8 times as likely to repurchase" is immediately understandable. It's widely used in clinical trials, public health, and increasingly in market research experiments. Understanding relative risk also helps you spot when odds ratios are being misinterpreted, a common problem that can inflate perceived effect sizes.
How Relative Risk Works
Calculation from a 2×2 Table
| Outcome: Yes | Outcome: No | Total | |
|---|---|---|---|
| Exposed | a | b | a + b |
| Unexposed | c | d | c + d |
Risk in exposed group = a / (a + b)
Risk in unexposed group = c / (c + d)
Relative Risk = [a / (a + b)] / [c / (c + d)]
Worked Example
You run a survey experiment with 500 participants. Half receive a product testimonial before rating purchase intent (framed as "would buy" or "wouldn't buy"):
| Would Buy | Wouldn't Buy | Total | |
|---|---|---|---|
| Testimonial | 140 | 110 | 250 |
| No testimonial | 95 | 155 | 250 |
Risk (testimonial) = 140/250 = 0.56 (56%)
Risk (no testimonial) = 95/250 = 0.38 (38%)
RR = 0.56 / 0.38 = 1.47
Interpretation: Participants who saw the testimonial were 1.47 times as likely to say they'd buy the product, a 47% relative increase in purchase intent.
Confidence Interval
The 95% CI for relative risk is calculated on the log scale:
ln(RR) ± 1.96 × SE[ln(RR)]
Where SE[ln(RR)] = √[(1/a - 1/(a+b)) + (1/c - 1/(c+d))]
For our example:
ln(1.47) = 0.386
SE = √[(1/140 - 1/250) + (1/95 - 1/250)] = √[(0.00714 - 0.004) + (0.01053 - 0.004)]
SE = √[0.00314 + 0.00653] = √0.00967 = 0.098
95% CI for ln(RR): 0.386 ± 1.96(0.098) = (0.194, 0.578)
Exponentiate: 95% CI for RR = (1.21, 1.78)
The interval doesn't include 1.0, confirming the effect is statistically significant.
Absolute Risk Reduction and NNT
Two related measures provide additional context:
Absolute Risk Reduction (ARR) = Risk_exposed - Risk_unexposed
ARR = 0.56 - 0.38 = 0.18 (18 percentage points)
Number Needed to Treat (NNT) = 1 / ARR = 1 / 0.18 = 5.6
Roughly 6 people need to see the testimonial to generate one additional purchase, a useful metric for estimating campaign ROI.
Relative Risk vs. Odds Ratio
| Feature | Relative Risk | Odds Ratio |
|---|---|---|
| Measures | Probability ratio | Odds ratio |
| Intuitive? | Yes, "X times as likely" | Less so, "X times the odds" |
| Available from cohort/experimental designs? | Yes | Yes |
| Available from case-control designs? | No | Yes |
| Overestimates effect when outcome is common? | No | Yes |
| Used in logistic regression? | Not directly | Yes (exponentiated coefficients) |
The odds ratio approximates the relative risk when the outcome is rare (below ~10%). As the outcome becomes more common, the OR increasingly overestimates the RR:
| Baseline Risk | Odds Ratio | Actual Relative Risk |
|---|---|---|
| 5% | 2.0 | 1.91 |
| 10% | 2.0 | 1.82 |
| 20% | 2.0 | 1.67 |
| 40% | 2.0 | 1.43 |
| 50% | 2.0 | 1.33 |
This divergence is why you shouldn't say "twice as likely" when reporting an odds ratio of 2.0 for a common outcome.
Converting Odds Ratio to Relative Risk
If you know the baseline risk (p₀) and the odds ratio:
RR = OR / (1 - p₀ + p₀ × OR)
For OR = 2.0 and p₀ = 0.30: RR = 2.0 / (1 - 0.30 + 0.30 × 2.0) = 2.0 / 1.30 = 1.54
When to Use Relative Risk
- Experimental designs (A/B tests, survey experiments) where you can calculate the actual event rate in each group
- Cohort studies tracking outcomes prospectively
- Campaign effectiveness reporting where stakeholders need to understand the practical impact in probability terms
- Comparing interventions when absolute probabilities are meaningful (unlike case-control designs where they aren't)
Common Mistakes to Avoid
- Reporting odds ratios as relative risk: saying "2 times as likely" when the OR is 2.0 and the baseline rate is 30% overstates the true relative risk (which is 1.54)
- Calculating RR from case-control data: in case-control designs, the sampling scheme distorts the baseline risk, making RR uncalculable; use OR instead
- Ignoring absolute risk: a relative risk of 2.0 is more actionable when the baseline is 20% (absolute increase of 20 percentage points) than when it's 0.1% (absolute increase of 0.1 percentage points)
How Quali-Fi Supports Risk Analysis
Quali-Fi's Research plan ($1,061/month) calculates both relative risk and odds ratios for experimental and quasi-experimental survey designs, presenting confidence intervals and absolute risk differences alongside the ratios. The platform flags when odds ratios and relative risk diverge meaningfully, helping you choose the right measure for your report.
Measure intervention effects with Quali-Fi
Frequently Asked Questions
Should I report relative risk or odds ratio?
Report relative risk when your design supports it (cohort or experimental) and the audience is non-technical. Report odds ratios when you're using logistic regression or case-control data. When in doubt, report both and note the baseline rate so readers can contextualize the numbers.
Can relative risk be less than 1?
Yes. An RR below 1.0 indicates a protective or reducing effect. RR = 0.6 means the exposed group has 60% of the risk of the unexposed group, a 40% relative reduction. This is common in research evaluating interventions designed to reduce negative outcomes (churn, complaints, returns).
What's a "clinically significant" relative risk?
There's no universal threshold. In clinical research, RR > 2.0 or RR < 0.5 is generally considered a strong association. In market research, even RR = 1.2 can be meaningful if it applies to millions of customers or a high-revenue product. Always evaluate relative risk alongside absolute risk and business context.