What the Index of Qualitative Variation measures
The Index of Qualitative Variation (IQV), sometimes called the index of variation or the variation ratio's cousin, measures how spread out a set of observations is across a set of nominal categories — categories that have no natural numeric order, such as religion, ethnicity, blood type, marital status, industry, or eye color. Because these categories are not numbers, you cannot compute a mean, variance, or standard deviation for them. The IQV fills that gap: it tells you whether cases pile up in one or two categories, or whether they are spread evenly across all of them.
The IQV is scaled to a fixed range. A value of 0 means there is no variation at all — every observation falls in a single category. A value of 1 means variation is at its maximum — the observations are divided perfectly evenly across every category. This 0-to-1 scaling is what makes IQV comparable across different datasets, even when they have different numbers of categories or different total counts.
The formula
The most common definition of the IQV is:
IQV = k(N² − Σf²) / [N²(k − 1)]
where:
- k = the number of categories,
- N = the total number of observations (the sum of all category frequencies), and
- Σf² = the sum of the squared frequency (count) of each category.
The quantity N² − Σf² in the numerator measures the observed variation. The denominator N²(k − 1)/k is the maximum possible value that N² − Σf² can reach — which happens when the N cases are split as evenly as possible across the k categories. Dividing observed variation by maximum possible variation forces the result into the 0-to-1 range. An equivalent and often-quoted form is IQV = (1 − Σp²) × k/(k − 1), where p is each category's proportion; both give identical results.
A worked example
Suppose a survey of 100 people records religious affiliation in four categories with counts 40, 30, 20, and 10. Here k = 4 and N = 100. Then Σf² = 40² + 30² + 20² + 10² = 1600 + 900 + 400 + 100 = 3000, and N² = 10000. So IQV = 4 × (10000 − 3000) / (10000 × 3) = 4 × 7000 / 30000 = 28000 / 30000 = 0.9333. That is high variation: the four groups are fairly evenly represented. If instead 97 people were in one category and 1 each in the other three, Σf² = 9409 + 3 = 9412, giving IQV = 4 × (10000 − 9412) / 30000 = 2352 / 30000 ≈ 0.0784 — very low variation, because almost everyone falls in one group.
Why the index is useful
IQV is widely used in sociology, demography, ecology, and market research to compare diversity or heterogeneity between groups. A city with an IQV of 0.90 for ethnicity is more ethnically mixed than one with an IQV of 0.35. A product line whose customer industries score 0.85 is drawing from a broad base, while one at 0.20 is concentrated in a single sector. Because the index is normalized to 0-to-1 regardless of sample size or category count, these comparisons are fair even when the underlying tables differ.
Common reference points
- IQV = 0: all N cases in one category (no diversity).
- IQV near 0.1–0.2: one dominant category with a few outliers.
- IQV around 0.5: moderate concentration — some categories clearly larger than others.
- IQV near 0.9–1.0: categories are close to evenly sized (high diversity).
- IQV = 1: exactly equal counts in every category — maximum possible diversity for that number of categories.