Comparison
Cramér's V vs Phi Coefficient — Which Effect Size
The phi coefficient and Cramér's Vare both chi-square-based effect sizes for categorical association. For a 2×2 table they are identical; for larger tables phi can exceed 1 and is no longer meaningful, so Cramér's V — which always stays between 0 and 1 — is the right choice.
Reviewed by the crosstabs.com methods team · Last updated
Run this on your own data — free, no signup
Upload a CSV or XLSX. Everything runs in your browser; your file never leaves your device.
Open the workspace →They're the same for 2×2 tables
Both measures start from the same chi-square statistic. The phi magnitude is √(χ²/n), and Cramér's V is √(χ² / (n·(min(rows, cols) − 1))). For a 2×2 table, min(rows, cols) − 1 = min(2, 2) − 1 = 1, so the extra divisor is just 1 and the two formulas collapse into the same expression.
In other words, on a 2×2 table phi and Cramér's V are mathematically the same number. The only thing phi adds is a sign: computed as (ad − bc) / √(…)it can be negative to show the direction of association, whereas Cramér's V is always reported as a non-negative magnitude.
Formula
Definition
|φ| = √(χ²/n) and V = √(χ² / (n·(min(r, c) − 1)))
- χ²
- = the Pearson chi-square statistic for the table
- n
- = the grand total (sum of all cell counts)
- r, c
- = the number of rows and columns in the table
- min(r, c) − 1
- = equals 1 for a 2×2 table, so the two formulas coincide there
Worked example
Worked example
Take the 2×2 table [[30, 10], [10, 60]] with χ² = 40.55 and n = 110.
min(2, 2) − 1 = 1
Phi magnitude: √(40.55 / 110) = √0.3686 = 0.607.
Cramér's V: √(40.55 / (110 × 1)) = 0.607.
Because the extra factor (min(2, 2) − 1) equals 1, both land on exactly the same value: 0.607 (≈ 0.61) — a strong association.
When to use it
Use it when
- Report phi when the table is 2×2 (both variables binary) and you want a signed measure that shows the direction of association, not just its strength.
- Report phi (signed) when you specifically want to know whether the two "yes" categories occur together more or less than chance.
Not the right tool when
- Report Cramér's V instead when the table is larger than 2×2, because phi loses its 0–1 bound and can exceed 1.
- Report Cramér's V instead when comparing association across tables of different sizes, since it rescales to a common [0, 1] range.
How to interpret it
Rule of thumb
For a 2×2 table phi and Cramér's V are interchangeable, so pick whichever your audience expects. Above 2×2, always use Cramér's V: phi loses its 0–1 bound and can exceed 1, which makes it impossible to interpret as a 0-to-1 effect size.
Why they diverge for bigger tables
The phi magnitude √(χ²/n) has no upper limit once a table is bigger than 2×2. On a larger table the chi-square statistic can grow large enough that √(χ²/n) climbs above 1, at which point phi is no longer a meaningful 0-to-1 effect size.
Cramér's V fixes this by dividing by the extra (min(r, c) − 1) factor, which rescales the statistic back into the [0, 1]range no matter how many rows and columns the table has. That is exactly why Cramér's V, and not phi, is the standard effect size for tables larger than 2×2.
Which to report
For a 2×2 table where you want to convey direction, report phi (signed) — it tells you both how strong the association is and which way it points. If direction does not matter, either measure works because they are numerically identical.
For anything larger than 2×2, or whenever you want to compare association strength across tables of different dimensions, report Cramér's V. Its guaranteed [0, 1] range keeps the numbers comparable and interpretable.
Frequently asked questions
- Are Cramér's V and phi the same?
- For a 2×2 table, yes — they are mathematically identical, because Cramér's V divides by (min(rows, cols) − 1), which equals 1 for a 2×2 table. For larger tables they differ, and Cramér's V is the one that stays bounded between 0 and 1.
- Why can phi be greater than 1?
- The phi magnitude is √(χ²/n), which has no upper bound once a table is larger than 2×2. On bigger tables the chi-square statistic can be large enough that √(χ²/n) exceeds 1, so phi stops being a usable 0-to-1 effect size. Cramér's V avoids this by dividing by an extra (min(rows, cols) − 1) factor.
- Which should I report for a 3×3 table?
- Report Cramér's V. For any table larger than 2×2 — including a 3×3 — phi can exceed 1 and loses its interpretation, while Cramér's V rescales to the [0, 1] range and remains comparable across table sizes.
References & further reading
- Cramér, H. (1946). Mathematical Methods of Statistics. Princeton University Press.
- Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.).
- Cramér's V — Wikipedia
Try it on your own data — free, no signup
Upload a CSV or XLSX. Everything runs in your browser; your file never leaves your device.
Open the workspace →