Mathwords logoMathwords

Cronbach Alpha — Definition, Formula & Example

Cronbach's alpha is a number, usually between 0 and 1, that tells you how consistently the items on a test or survey measure the same thing. Values closer to 1 mean the items tend to rise and fall together across respondents.

For a scale of k≥2k \ge 2 items with scores X1,…,XkX_1, \dots, X_k and total score X=∑i=1kXiX = \sum_{i=1}^{k} X_i, Cronbach's alpha is α=kk−1(1−∑σi2σX2)\alpha = \frac{k}{k-1}\left(1 - \frac{\sum \sigma_i^2}{\sigma_X^2}\right). It equals the reliability of XX when the items are essentially tau-equivalent with uncorrelated errors. Under uncorrelated errors in general, it is a lower bound on that reliability. Its maximum is 1, and it can be negative when items correlate negatively.

Key Formula

α=kk−1(1−∑i=1kσi2σX2)\alpha = \frac{k}{k-1}\left(1 - \frac{\sum_{i=1}^{k} \sigma_i^2}{\sigma_X^2}\right)
Where:
  • kk = Number of items on the scale
  • σi2\sigma_i^2 = Variance of the scores on item i
  • σX2\sigma_X^2 = Variance of respondents' total scores (sum of all items)
  • α\alpha = Cronbach's alpha, the internal consistency coefficient

Worked Example

Problem: A 4-question survey asks students to rate their study habits on a 1–5 scale. Across all respondents, the item variances are 1.0, 1.2, 0.8, and 1.0. The variance of the total scores is 10.0. Find Cronbach's alpha.
Step 1: Add the item variances.
∑σi2=1.0+1.2+0.8+1.0=4.0\sum \sigma_i^2 = 1.0 + 1.2 + 0.8 + 1.0 = 4.0
Step 2: Divide by the total-score variance and subtract the result from 1.
1−4.010.0=1−0.4=0.61 - \frac{4.0}{10.0} = 1 - 0.4 = 0.6
Step 3: Multiply by kk−1\frac{k}{k-1} with k=4k = 4.
α=43(0.6)=0.8\alpha = \frac{4}{3}(0.6) = 0.8
Answer: Cronbach's alpha is 0.8. Many researchers use 0.70 as a rough cutoff for acceptable consistency, so this survey would usually count as reliable. The cutoff is a convention, not a mathematical rule.

Why It Matters

Psychology, education, and nursing research papers routinely report alpha for every survey scale they use. In a college research methods or psychometrics course, you will compute it to justify adding item scores into one total. Low alpha signals that some questions may be measuring something different and should be revised or dropped.

Common Mistakes

Mistake: Treating a high alpha as proof that all items measure one single trait.
Correction: Alpha rises as you add correlated items or as the items correlate more strongly, even if the scale covers several distinct traits. Use factor analysis to check whether the scale measures one dimension. Treat alpha as a consistency measure only.

Related Terms

  • Univariate Data — Each item's scores form one variable with its own variance
  • Bivariate Data — Pairs of item scores whose covariance pushes alpha upward
  • Scatterplot — Shows visually how two survey items move together
  • Consistency — Alpha measures a test's internal consistency, a different sense
  • Variance — Alpha is computed from item and total score variances
  • Covariance Matrix — Alpha can be calculated from covariances among test items