Guides / cronbach's alpha in spss
How to run Cronbach's alpha in SPSS
Reliability Analysis is a three-click job, but a negative alpha, an alpha that jumps when you delete an item, and an alpha above .95 all mean specific things — and only one of them is good news. This covers the path and the reading.
Go to Analyze > Scale > Reliability Analysis, move all items belonging to one scale into the Items box, leave the Model set to Alpha, then click Statistics and tick 'Scale if item deleted' plus Correlations under Inter-Item. Click Continue and OK. The Reliability Statistics table gives Cronbach's Alpha and the number of items.
Step by step
The click path
The dialog is small and most of it can be left alone.
One construct at a time. If your questionnaire has three subscales, that is three separate runs.
The other options — split-half, Guttman — answer different questions and are rarely what you want.
The first gives you the item-total table, which is where the useful information is. The second gives the inter-item matrix, which catches reverse-scoring errors.
SPSS will happily compute alpha on un-reversed items and return a number that looks like a finding. Recode negatively worded items first with Transform → Recode into Different Variables, and use the recoded versions here.
Output
Reading the output
The Reliability Statistics table gives you Cronbach's Alpha and N of Items. That is the headline figure.
| Alpha | Label | What to do |
|---|---|---|
| ≥ .90 | Excellent | Check for redundant items — very high alpha often means you asked the same question repeatedly |
| .80–.89 | Good | Proceed |
| .70–.79 | Acceptable | Proceed, note it as a limitation |
| .60–.69 | Questionable | Examine the item-total table before summing |
| < .60 | Poor | Do not sum the items — they are not measuring one thing |
| Negative | Error | Almost always an un-reversed item |
It depends on your sample as well as the items. The same instrument returns a lower alpha in a homogeneous group, because restricted variation reduces inter-item correlation. Report alpha for your sample rather than citing the figure from the original validation paper.
The commonest error
Reverse-scored items
A negative alpha, or one far lower than expected, is nearly always a reverse-worded item that has not been recoded. It is not a finding about your scale.
If one item correlates negatively with all the others, that is your culprit.
Select the offending item, give the output variable a new name such as Q7R.
On a 5-point scale: 1→5, 2→4, 3→3, 4→2, 5→1. On a 7-point scale the rule is new = 8 − old.
Alpha should now be positive and sensible.
RECODE Q7 Q11 Q14 (1=5)(2=4)(3=3)(4=2)(5=1) INTO Q7R Q11R Q14R. followed by EXECUTE. handles them in one step and leaves a record of exactly what you did — which is what you want when writing the methods section.
Diagnostics
The item-total table
Item-Total Statistics is the table worth reading. Two columns matter.
| Column | What it tells you | Threshold |
|---|---|---|
| Corrected Item-Total Correlation | How well the item fits the rest of the scale | Below .30 → the item is not pulling its weight |
| Cronbach's Alpha if Item Deleted | What alpha would become without this item | Higher than current alpha → the item is hurting the scale |
If deleting an item would raise alpha materially, you have a decision to make — and it is a substantive one, not a statistical one. Removing an item because the number improves, when the item is conceptually central to the construct, damages content validity to buy a better coefficient.
Deleting items until alpha crosses a threshold is a recognised way of producing a reliable measure of nothing in particular. If an item is theoretically necessary, keep it and report the lower alpha with an explanation.
Send your item set and output. A named researcher reviews the reliability analysis, checks the factor structure, and advises on which items earn their place.
Get a fixed quoteLimitations
What alpha does not tell you
If your software offers McDonald's omega, report it alongside or instead. SPSS provides it from version 27 in the same Reliability Analysis dialog under Model, and it relaxes the assumption alpha makes about equal item contributions.
Pitfalls
Five mistakes
1. Running all questionnaire items at once
Alpha is computed per scale. Mixing constructs produces a number that means nothing.
2. Not reverse-scoring first
It corrupts alpha and every total computed afterwards. A negative alpha is the giveaway.
3. Citing the original study's alpha
Alpha depends on your sample. Compute and report it for your data.
4. Reporting alpha above .95 as excellent
It usually indicates redundant items rather than a superior scale.
5. Treating alpha as evidence of validity
Consistency is not accuracy. They are separate claims needing separate evidence.
Reporting
How to report it
| Situation | How to write it |
|---|---|
| Standard | The eight items showed good internal consistency (α = .86). |
| With omega | α = .86, ω = .87 |
| Borderline | α = .68, which is below the conventional threshold and is noted as a limitation. |
| After removing an item | α = .84 following removal of item 7, which correlated poorly with the remaining items (r = .12). |
Send your dataset. A named statistician runs the reliability analysis, examines the factor structure, and advises what validity evidence your design supports.
See SPSS data analysisAnswers
Frequently asked questions
How do I run Cronbach's alpha in SPSS?
Analyze > Scale > Reliability Analysis. Move all items from one scale into the Items box, leave Model on Alpha, then click Statistics and tick 'Scale if item deleted' and Correlations. Click Continue then OK. Read the Reliability Statistics table.
Why is my Cronbach's alpha negative in SPSS?
Almost always because a reverse-worded item has not been recoded. Check the Inter-Item Correlation Matrix: an item correlating negatively with all the others is the cause. Recode it so that on a 5-point scale new = 6 minus old, then re-run.
What is an acceptable Cronbach's alpha?
Above .70 is conventionally acceptable and above .80 good. Below .60 the items should not be summed into a single score. Be cautious above .95, which usually signals redundant items rather than a better scale.
Can I run Cronbach's alpha on all my questionnaire items at once?
No. Alpha is computed per scale, so each construct or subscale needs its own run. Combining items measuring different things produces a coefficient with no interpretable meaning.
What does 'Cronbach's Alpha if Item Deleted' mean?
It shows what alpha would become if that item were removed. A value noticeably higher than your current alpha means the item is weakening the scale. Whether to drop it is a substantive judgement, not an automatic one — removing conceptually necessary items damages content validity.
Should I use Cronbach's alpha or McDonald's omega?
Omega where your software offers it. Alpha assumes every item contributes equally to the construct, which is rarely true; omega relaxes that assumption and is generally the more accurate estimate. SPSS provides omega from version 27 in the same dialog.
Keep reading