Cronbach's alpha calculator

Worked example: the Openness scale of the bfi data set

What Cronbach's alpha is acceptable? The usual threshold is .70 (Nunnally, 1978); values from .60 to just under .70 are accepted in exploratory research (Hair et al., 2010). The example below is a real result, not an illustration: the Openness scale (O1–O5) of the public bfi data set in the R package psych (Revelle, 2026), a 6-point scale, with the two reverse-worded items O2 and O5 recoded (7 − score). The calculator reproduces these numbers and matches R to four decimal places; see the verification page. To rerun it, download bfi.csv, reverse-score O2 and O5, and paste the five columns O1–O5 into the data box above.

Reading it in three steps:

1. Alpha = .603: below .70, so acceptable only by the exploratory-research standard; the lower bound of the confidence interval (.578) is below even .60.
2. O4 has a corrected item-total correlation of .220, below the .30 cut-off (Nunnally & Bernstein, 1994), and it is the only item whose “alpha if item deleted” (.614) is higher than the overall alpha.
3. Dropping O4 lifts alpha only to about .61, still below .70. The honest conclusion is that the scale has low reliability: report the figure and discuss it as a limitation rather than “fixing” the data. The results doctor walks through the legitimate next steps.

Checking by hand: the five item variances are 1.272, 2.440, 1.491, 1.481 and 1.756, so Σσᵢ² = 8.437; the variance of the total score is σₜ² = 16.289. Hence α = 5/4 × (1 − 8.437/16.289) = 1.25 × .482 = .603. If you forget to reverse-score O2 and O5, the same data give alpha = −.157: a negative alpha almost always means a reverse-worded item has not been recoded.

How to read the output

• Overall alpha. The most cited threshold is .70 (Nunnally, 1978); .60 to .70 is usually accepted in exploratory research (Hair et al., 2010). At .90 or above, check whether some items repeat the same idea (Tavakol & Dennick, 2011). These cut-offs are conventions rather than statistical tests, and authors attach different labels to the same value (Taber, 2018), so state the threshold and its source in your methods chapter.
• Confidence interval. Computed with the Feldt, Woodruff and Salih (1987) method, the same bounds SPSS prints in its Intraclass Correlation Coefficient table (two-way mixed, consistency, average measures). With a small sample, an alpha of .75 can come with a lower bound under .70, and the tool flags this.
• McDonald's omega does not assume equal loadings, as alpha does, and is usually higher when loadings differ. Hayes and Coutts (2020) recommend omega over alpha; because examiners still ask for alpha, the tool reports both. For two-item scales it reports the Spearman-Brown coefficient instead (Eisinga et al., 2013).
• Corrected item-total correlation below .30 is highlighted (Nunnally & Bernstein, 1994), as is any item whose deletion would raise alpha.
• A negative item-total correlation almost always means a reverse-worded item that has not been recoded: click Reverse before considering dropping it.

Drop one item at a time, starting with the weakest, and rerun; every number changes after each deletion. Read the item wording first, and keep at least three items per construct.

How to report Cronbach's alpha in APA 7

Give alpha to two or three decimals without a leading zero, with the number of items and the threshold you applied. For the example above: “Cronbach's alpha for Openness was .60, 95% CI [.58, .63], below the .70 threshold (Nunnally, 1978). Item O4 had a corrected item-total correlation of .22; deleting it would raise alpha only to .61, so the scale was retained and its low reliability is discussed as a limitation.” The calculator writes this paragraph for you, with the references, and the Word file adds one Item-Total Statistics table per scale.

Do not fabricate, filter or edit responses to reach the threshold

A low alpha is a result, not an error to hide. Generating responses that no real person gave, changing real answers so items “move together”, or deleting genuine respondents because they pull alpha down are fabrication and falsification, the core forms of research misconduct (Steneck, 2006). Legitimate steps are screening responses by rules written before you look at the results, recoding reverse-worded items, dropping an item for a content reason while reporting results before and after, and reporting alpha with its confidence interval and omega.

Frequently asked questions

What is a good Cronbach's alpha?

.70 is the most cited minimum (Nunnally, 1978), and .60 to .70 is acceptable in exploratory research (Hair et al., 2010). Lance et al. (2006) point out that Nunnally suggested .70 for early-stage research, .80 for basic research and .90 when scores are used to make decisions about individuals.

What does “Cronbach's alpha if item deleted” mean?

It is the alpha the scale would have without that item, the SPSS column of the same name. A value above the overall alpha means the item makes the scale less consistent, like O4 above (.614 versus .603). Consider dropping it only if its corrected item-total correlation is also below .30 and the construct is still adequately covered.

Why is my Cronbach's alpha negative?

Alpha is negative when the average covariance between items is negative, which almost always means a reverse-worded item has not been recoded. Find the item with a negative item-total correlation and click Reverse.

Will the results match SPSS and SmartPLS?

SPSS: yes, for the same data and the same handling of missing values (listwise deletion per scale, as the RELIABILITY command does). SmartPLS standardises indicators, so its alpha corresponds to the SPSS column “Based on Standardized Items” (bfi Openness: .609 versus .603). For PLS-SEM, also report rho_a and composite reliability (rho_c), with a range of .70 to .95 (Hair et al., 2019).

Should I compute alpha for the whole questionnaire?

No. Alpha is meaningful only for items that measure the same construct, so the tool computes it per scale and warns when a group has more than 12 items. A high alpha does not show that a scale is unidimensional either; that needs EFA, available in the survey data analysis tool.

References

• Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16(3), 297–334. https://doi.org/10.1007/BF02310555
• Eisinga, R., Grotenhuis, M. te, & Pelzer, B. (2013). The reliability of a two-item scale: Pearson, Cronbach, or Spearman-Brown? International Journal of Public Health, 58(4), 637–642. https://doi.org/10.1007/s00038-012-0416-3
• Feldt, L. S., Woodruff, D. J., & Salih, F. A. (1987). Statistical inference for coefficient alpha. Applied Psychological Measurement, 11(1), 93–103. https://doi.org/10.1177/014662168701100107
• Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2010). Multivariate data analysis (7th ed.). Pearson Prentice Hall.
• Hair, J. F., Risher, J. J., Sarstedt, M., & Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. European Business Review, 31(1), 2–24. https://doi.org/10.1108/EBR-11-2018-0203
• Hayes, A. F., & Coutts, J. J. (2020). Use omega rather than Cronbach’s alpha for estimating reliability. But…. Communication Methods and Measures, 14(1), 1–24. https://doi.org/10.1080/19312458.2020.1718629
• Lance, C. E., Butts, M. M., & Michels, L. C. (2006). The sources of four commonly reported cutoff criteria: What did they really say? Organizational Research Methods, 9(2), 202–220. https://doi.org/10.1177/1094428105284919
• Nunnally, J. C. (1978). Psychometric theory (2nd ed.). McGraw-Hill.
• Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.
• Revelle, W. (2026). psych: Procedures for psychological, psychometric, and personality research (R package version 2.6.9). https://CRAN.R-project.org/package=psych
• Steneck, N. H. (2006). Fostering integrity in research: Definitions, current knowledge, and future directions. Science and Engineering Ethics, 12(1), 53–74. https://doi.org/10.1007/PL00022268
• Taber, K. S. (2018). The use of Cronbach’s alpha when developing and reporting research instruments in science education. Research in Science Education, 48(6), 1273–1296. https://doi.org/10.1007/s11165-016-9602-2
• Tavakol, M., & Dennick, R. (2011). Making sense of Cronbach’s alpha. International Journal of Medical Education, 2, 53–55. https://doi.org/10.5116/ijme.4dfb.8dfd