Skip to content
Omnikalc

Cronbach's alpha: is your questionnaire reliable?

Compute Cronbach's alpha from item and total-score variances, or standardized alpha from the mean inter-item correlation, and the number of items needed to reach a target.

Updated on
Your inputs

Value out of range

I know

Value out of range

Value out of range

Value out of range

Value out of range

Results

Cronbach's alpha

0.778

Acceptable
Implied mean inter-item correlation
0.259
Spearman–Brown length factor to reach the target
1.14
Parallel items needed for the target alpha
12
Standard error of measurement factor √(1 − α)
0.471

What Cronbach’s alpha measures

Cronbach’s alpha estimates the internal consistency of a questionnaire or test: how strongly its items measure the same underlying construct, such as job satisfaction, anxiety or reading ability. It is the most widely reported reliability coefficient in psychology, education, marketing and health research.

The formulas

α = k ÷ (k − 1) × (1 − Σsᵢ² ÷ s²ₜ)

  • k: number of items;
  • Σsᵢ²: sum of the variances of the individual items;
  • s²ₜ: variance of the total score (the sum of the items).

If the items are highly correlated, the total score varies much more than the items separately, and alpha approaches 1.

Standardized alpha, from the mean inter-item correlation r̄: α = k r̄ ÷ (1 + (k − 1) r̄).

Example

A 10-item scale has item variances summing to 12 and a total-score variance of 40:

  • α = 10 ÷ 9 × (1 − 12 ÷ 40) = 1.111 × 0.70 ≈ 0.78: acceptable;
  • this corresponds to a mean inter-item correlation of about 0.26;
  • to reach α = 0.80, the Spearman–Brown formula gives a length factor of 1.14: about 12 items of the same quality.

Interpreting alpha

α Usual reading
0.9 and above excellent (above 0.95, items may be redundant)
0.8 to 0.9 good
0.7 to 0.8 acceptable
0.6 to 0.7 questionable
0.5 to 0.6 poor
below 0.5 unacceptable

A threshold of 0.70 is common for research instruments; decisions about individuals, such as clinical diagnosis, usually call for 0.90 or more.

Pitfalls

  • Alpha rises with the number of items, even if they are only weakly related. A long scale can show a high alpha without measuring a single construct.
  • A high alpha does not prove unidimensionality; use factor analysis for that.
  • Reverse-scored items must be recoded before computing alpha, or they will drag it down, sometimes below zero.
  • Alpha assumes that all items contribute equally. When they do not, McDonald’s omega is a better estimate of reliability.

The factor √(1 − α), multiplied by the SD of total scores, gives the standard error of measurement of an individual score.

Frequently asked questions

What is a good Cronbach's alpha?

0.70 is the usual minimum for research, 0.80 is good, and 0.90 or more is expected for decisions about individuals.

Why is my Cronbach's alpha negative?

Usually because reverse-scored items were not recoded, or the items measure different things.

Does a high alpha mean the scale is unidimensional?

No. Alpha measures internal consistency; use factor analysis to check dimensionality.

How many items do I need to raise alpha?

The Spearman–Brown formula gives the length factor: m = α_target(1 − α) ÷ (α(1 − α_target)).

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.