A standardized mean difference statistic (d) measures effect size by showing how much two group means differ using standard deviation units.
Understanding the Standardized Mean Difference Statistic (d)
What Is the Standardized Mean Difference?
In social science research, we often want to know whether two groups differ meaningfully. For example, does a new teaching method improve test scores compared to the old method? Or does one political group have stronger attitudes about a social issue than another? To answer these questions, researchers often compare the averages, or means, of the two groups.
But raw differences in means can be hard to interpret, especially when data come from different scales. That’s where the standardized mean difference statistic, often called Cohen’s d, comes in. It helps researchers understand how large the difference between two groups really is by putting it into a common scale — the number of standard deviations that separate the two means.
A Cohen’s d of 0.5 means the two group means differ by half a standard deviation. A d of 1.0 means the difference is one full standard deviation. This allows researchers to judge the strength of an effect, no matter the original measurement scale.
Why Standardized Mean Difference Matters
The standardized mean difference is especially important when you want to compare results across studies, combine findings in a meta-analysis, or communicate your findings in a clear and interpretable way. Raw scores can vary in meaning depending on the measurement tool, but the standardized mean difference puts all comparisons on the same footing.
Social science researchers use it to:
- Quantify the size of a treatment or intervention effect
- Compare group differences in survey or test responses
- Translate findings across different types of studies
- Inform policy or practice decisions based on the magnitude of effects
By expressing mean differences in standardized units, Cohen’s d makes complex statistical results more accessible and easier to interpret.
The Formula for Cohen’s d
The most common version of the formula is:
Cohen’s d = (Mean of group 1 – Mean of group 2) / Pooled standard deviation
Let’s break that down:
- Mean of group 1 and group 2: These are the averages of the two groups being compared.
- Pooled standard deviation: This is a combined estimate of the standard deviations of both groups. It assumes that both groups have roughly the same variability.
The pooled standard deviation is calculated by taking the square root of the average of the two groups’ variances, weighted by their sample sizes. However, for most purposes, researchers rely on statistical software or published formulas to calculate it.
Interpreting Cohen’s d Values
The value of d shows how big the difference is between the two groups. Here are common benchmarks suggested by Jacob Cohen, though these are guidelines and not hard rules:
- 0.2 = small effect
- 0.5 = medium effect
- 0.8 = large effect
So, a study that finds a d of 0.2 means the two groups differ by only a small amount, while a d of 0.8 shows a much more noticeable difference.
These values are helpful when interpreting findings, especially when the raw numbers are not intuitive. For example, a test score difference of 5 points might sound big or small depending on the context. But if it equals a d of 0.5, we know it reflects a medium effect.
Examples Across Social Science Fields
Psychology
A psychologist compares depression scores between two groups: one that received therapy and one that did not. The group that received therapy had an average score of 18, while the untreated group had an average score of 22. With a pooled standard deviation of 4, the standardized mean difference is:
d = (22 – 18) / 4 = 1.0
This means the therapy group scored one full standard deviation better than the untreated group, which is considered a large effect.
Education
An education researcher tests two teaching methods on student performance. The traditional group scores 75 on average, while the new method group scores 78. If the pooled standard deviation is 10, the d value is:
d = (78 – 75) / 10 = 0.3
This suggests a small to moderate effect in favor of the new method.
Criminal Justice
In a study of sentencing outcomes, men receive an average sentence of 60 months, and women receive 52 months, with a pooled standard deviation of 20 months. The standardized mean difference is:
d = (60 – 52) / 20 = 0.4
This reveals a moderate effect size in sentencing differences between genders.
Political Science
A political scientist measures political trust in two ideological groups. One group scores 3.5, and the other scores 3.0 on a 5-point scale. If the pooled standard deviation is 0.75, then:
d = (3.5 – 3.0) / 0.75 = 0.67
This reflects a medium-to-large effect size, showing meaningful ideological differences in political trust.
Strengths of Using Standardized Mean Difference
- Unit-free comparison: Because it’s standardized, d allows you to compare results across different measures and studies.
- Widely understood: Cohen’s d is one of the most familiar and frequently reported effect size measures.
- Essential for meta-analysis: When combining data from multiple studies, standardized effect sizes like d are necessary for comparing results.
- Interpretability: With clear benchmarks, even non-specialists can understand what a small, medium, or large effect looks like.
Limitations and Caveats
- Assumes similar variability: The pooled standard deviation assumes the two groups have similar variance. If they don’t, the d value may be misleading.
- Sensitive to sample size: In small samples, d can be an unstable estimate. Some versions of d adjust for small sample bias (like Hedges’ g).
- Does not test for significance: A large d doesn’t mean the difference is statistically significant. Researchers still need to run t-tests or ANOVA to assess statistical significance.
- Context-dependent interpretation: A “small” effect might be important in some fields (like public health), while a “large” effect may be less meaningful in others.
Variations of d
Several alternatives or extensions of Cohen’s d exist:
- Hedges’ g: Corrects d for small sample bias.
- Glass’s Δ: Uses only the control group’s standard deviation in the denominator, useful when groups differ greatly in variability.
- Standardized mean difference in meta-analysis: Often uses weighted versions of d that account for sample size across studies.
Researchers should choose the version that best fits their design and purpose.
How to Report Cohen’s d
A good research report will:
- Present the raw group means and standard deviations
- Report the calculated d value
- Describe the effect size using words (e.g., small, medium, large)
- Include confidence intervals for d when possible
Example:
“The treatment group (M = 78, SD = 10) scored higher than the control group (M = 75, SD = 9). The effect size was medium (Cohen’s d = 0.30).”
This type of reporting helps readers understand both the size and meaning of the effect.
When to Use Standardized Mean Difference
Use Cohen’s d when:
- Comparing two independent group means
- Assessing the practical importance of a result
- Reporting outcomes for interventions or treatments
- Conducting or interpreting a meta-analysis
Avoid using d when:
- The groups have very different variances (consider Glass’s Δ)
- Your data are highly skewed (other measures may work better)
- You want to test for statistical significance (use a t-test instead)
Final Thoughts
The standardized mean difference statistic, commonly known as Cohen’s d, is one of the most important tools in the social scientist’s toolkit. It allows researchers to go beyond simply asking, “Is there a difference?” and begin answering, “How big is the difference, and is it meaningful?” Whether comparing therapies, teaching methods, or public policies, the use of d helps communicate the true impact of research findings.
By expressing group differences in terms of standard deviations, Cohen’s d turns raw scores into actionable knowledge that can guide decisions, inform practice, and improve the quality of research across disciplines.
Glossary Return to Doc's Research Glossary
Last Modified: 03/29/2025