Cronbach's Alpha: How to Run and Interpret It in SPSS
Cronbach's Alpha is usually the first output table you need to process after entering survey data into SPSS. This index checks whether the items within the same scale are consistent enough for you to continue with EFA, regression, or a PLS-SEM model.
Open your .sav file, select one group of items, such as the items measuring service quality, and run Analyze > Scale > Reliability Analysis. The output will include the Reliability Statistics and Item-Total Statistics tables. The difficult part is reading each column correctly, knowing when an item can be removed, and avoiding deletion simply because you want Cronbach's Alpha to increase.
When to use Cronbach's Alpha
Cronbach's Alpha is used when a construct is measured by multiple items, usually statements in a Likert questionnaire. For example, the construct “satisfaction” might be measured by SAT1, SAT2, SAT3, and SAT4. You check this SAT group separately, then move to the items for another construct.
This analysis fits research questions or models that ask whether several items measure the same latent construct. It checks the internal consistency of a scale. It does not establish convergent validity, discriminant validity, or whether one variable actually affects another variable.
You do not run Cronbach's Alpha for a single item. If a construct has only one item, there is no internal reliability for you to assess with this index. You should also avoid putting items from several different constructs into one run simply to obtain a large Alpha.
For a broader overview of the process, see scale reliability testing in SPSS. This article focuses on the procedure and on making decisions with Cronbach's Alpha.
Prepare the data before running the analysis
Before opening the Reliability Analysis dialog box, check the raw data. Each row usually represents one respondent, and each column represents one item. Items within the same scale need the same direction of interpretation, the same coding scheme, and the same Likert range.
Check variable type and coding
In Variable View, check Type, Measure, variable labels, and the coded values. Likert data may be entered as numbers from 1 to 5 or from 1 to 7. Do not enter text such as “Strongly agree” in an item column if you want SPSS to calculate the analysis directly.
Missing values need to be declared consistently. A blank cell, the code 99, and the code -1 may be interpreted differently by SPSS depending on how you entered the data. If 99 actually means “no response” but is still treated as a score, Cronbach's Alpha will be affected.
Check reverse-coded items
If the questionnaire contains a statement with the opposite meaning, reverse-code it before running the analysis. For a Likert scale from 1 to 5, the usual formula is new score equals 6 minus old score. For a scale from 1 to 7, the corresponding constant is 8.
In SPSS, use Transform > Compute Variable and create a new variable instead of overwriting the original. For example, use SAT3_R = 6 - SAT3 if SAT3 is a reverse-coded item in a 1 to 5 scale. Then use SAT3_R in Cronbach's Alpha and clearly note the replacement in the data file.
Check the item group
Each run should contain only items that measure the same construct. If your model has the scales TRUST1 to TRUST4 and SAT1 to SAT4, run TRUST separately from SAT. The EFA or regression that follows should also be based on this checked scale structure.
Steps to run Cronbach's Alpha in SPSS
Follow the sequence below in SPSS Statistics. Menu names may look slightly different across versions, but the main path remains Analyze > Scale > Reliability Analysis.
Step 1: Open Reliability Analysis
Choose Analyze > Scale > Reliability Analysis. Select all items from one scale and move them into the Items box. Under Model, leave Alpha selected because this is the model used to calculate Cronbach's Alpha.
Naming variables by group helps you avoid selecting the wrong items. If you are checking the “Service quality” scale, put only the items from that scale into the Items box. Do not add demographic variables or a general outcome variable.
Step 2: Select the Statistics to display
Click the Statistics button. Under Descriptives for, select Item, Scale, and Scale if item deleted. Under Inter-Item, you can select Correlations if you want to inspect the correlations between items.
The Scale if item deleted option matters because it creates the Cronbach's Alpha if Item Deleted column in the Item-Total Statistics table. This lets you see how Alpha would change if each item were removed. Click Continue, then click OK to run the analysis.
Step 3: Save the syntax and record every round
If you are writing a thesis, click Paste instead of clicking OK only at the first run. SPSS will create the corresponding syntax, for example:
RELIABILITY
/VARIABLES=SAT1 SAT2 SAT3_R SAT4
/SCALE('ALL VARIABLES') ALL
/MODEL=ALPHA
/STATISTICS=DESCRIPTIVE SCALE CORR.
Save the syntax in the same folder as the data. When you remove an item, record the reason, Alpha before removal, Alpha after removal, and the result of the recheck. The first EFA run rarely comes out tidy. Removing an item, running the analysis again, and recording each round is acceptable when you can explain the decision.
Step 4: Run each scale separately
Repeat the steps above for every item group. Do not combine all questionnaire items into one Alpha when the questionnaire contains several constructs. One overall Alpha can hide the fact that some items do not measure the same content.
After each run, save the output with a clear name, such as Alpha_SAT_v1 and Alpha_SAT_v2. When you write Chapter 4, you will know which item was removed and will not confuse the initial output with the final output.
Read the output
SPSS usually returns three main tables: Reliability Statistics, Item-Total Statistics, and, depending on your selections, Inter-Item Correlation Matrix. Read them in order. Do not reach a conclusion from one Alpha value alone.
The Reliability Statistics table
This table usually contains two columns, Cronbach's Alpha and N of Items. Alpha indicates the internal consistency of the whole item group, while N of Items shows how many items were included in the calculation.
The table below is illustrative output, not the result of a real study:
| Cronbach's Alpha | N of Items | Interpretation |
|---|---|---|
| 0.812 | 4 | The scale has good consistency under the selected reference threshold |
| 0.684 | 5 | It may be acceptable in exploratory research if the other indicators are suitable |
| 0.421 | 4 | Check reverse-coded items, item-total correlations, and item content |
A high Alpha does not automatically prove that a scale is good in every respect. If the items almost repeat the same statement, Alpha may be high while the measurement content is too narrow. Combine Alpha with Corrected Item-Total Correlation, the theoretical structure, and the next validation steps.
The Item-Total Statistics table
In the Item-Total Statistics table, focus on the key columns. Scale Mean if Item Deleted and Scale Variance if Item Deleted describe the scale after an item is removed. Corrected Item-Total Correlation is the correlation between that item and the scale total after that same item has been removed. Cronbach's Alpha if Item Deleted shows how Alpha would change if the item were removed.
Illustrative output:
| Item | Corrected Item-Total Correlation | Cronbach's Alpha if Item Deleted | Illustrative decision |
|---|---|---|---|
| SAT1 | 0.612 | 0.776 | Retain |
| SAT2 | 0.548 | 0.793 | Retain |
| SAT3_R | 0.184 | 0.861 | Check content and coding before deciding |
| SAT4 | 0.587 | 0.782 | Retain |
If an item's Corrected Item-Total Correlation is low, check whether it is reverse-coded, whether the data were entered correctly, and whether the item genuinely belongs to the construct. Do not remove an item simply because Cronbach's Alpha if Item Deleted is slightly higher. The removal needs both methodological and substantive justification.
Criteria for evaluating Cronbach's Alpha
State the threshold you use in the research methods section. A common threshold for Alpha is 0.7 or above (Nunnally, 1978). For a new or exploratory scale, Alpha from 0.6 may be considered (Hair et al., 2010). A Corrected Item-Total Correlation of 0.3 or above is commonly used (Nunnally and Bernstein, 1994).
| Criterion | Reference threshold | Source |
|---|---|---|
| Cronbach's Alpha, established scale | 0.70 or above | (Nunnally, 1978) |
| Cronbach's Alpha, new or exploratory scale | 0.60 or above | (Hair et al., 2010) |
| Corrected Item-Total Correlation | 0.30 or above | (Nunnally and Bernstein, 1994) |
| Observations per item | 5 to 10 | (Hair et al., 2010) |
These thresholds are reference points, not automatic deletion rules. If Alpha is 0.68 but every item has an acceptable item-total correlation, you can report the scale as acceptable in an exploratory context if you provide an appropriate rationale. If Alpha is 0.95, check whether the items duplicate each other's content instead of automatically declaring the scale excellent.
Cronbach's Alpha also does not replace validity testing. After the scale reaches an acceptable reliability level, you can continue with EFA in SPSS or use the appropriate measurement procedure for a PLS-SEM model. In SmartPLS, Composite Reliability and AVE belong to a different group of assessment criteria. See What is Composite Reliability.
How to write this in your thesis
In Chapter 4, report the initial number of items, any removed items, the reason for removal, the final Alpha, and the range of Corrected Item-Total Correlation. Do not write that the scale passes only because Alpha exceeds 0.7 while ignoring the other columns.
How to write this in your thesis: “The reliability test for the [SCALE NAME] scale showed that Cronbach's Alpha was [ALPHA VALUE] with [NUMBER] items. The Corrected Item-Total Correlation values ranged from [SMALLEST VALUE] to [LARGEST VALUE], all [greater than or equal to 0.3/meeting the selected criterion]. Therefore, the [SCALE NAME] scale [was retained/was adjusted by removing item ITEM CODE] for the next analysis step.”
If you remove an item, add the reason and run Alpha again after removal. Keep the complete output in the appendix. In Chapter 4, you can summarize the results by scale so the committee can follow the decisions more easily.
Common errors when running Cronbach's Alpha
The most common error is selecting an item from the wrong scale. You may also forget to reverse-code an item, enter a missing-value code as a valid score, or run Alpha on data that still include incomplete responses.
Another error is repeatedly deleting items to make Alpha rise. This creates a scale with missing content and leaves you with a decision that is difficult to defend. Read the original statement, check the coding, inspect Corrected Item-Total Correlation, and then decide.
If Cronbach's Alpha is negative, the cause is usually a reverse-coded item or variables with negative correlations. You can find specific guidance at negative Cronbach's Alpha in SPSS. If Corrected Item-Total Correlation is below 0.3, see what to do when Item-Total Correlation is below 0.3.
You should also distinguish Cronbach's Alpha from the assessment of the entire scale. Alpha does not tell you whether a regression model is significant, test causal relationships, or prove that an item belongs to the correct factor. Those questions require analyses that match your model.
Frequently asked questions
What is Cronbach's Alpha?
Cronbach's Alpha is an index for assessing the internal consistency of a group of items that measure the same construct. You usually read it in the Reliability Statistics table after running Analyze > Scale > Reliability Analysis.
Read Alpha together with Corrected Item-Total Correlation, item content, and scale design. A single high number is not enough to conclude that the entire scale is valid.
What is a good Cronbach's Alpha?
A value of 0.7 or above is commonly used for scale reliability (Nunnally, 1978). For a new or exploratory scale, a value from 0.6 may be considered (Hair et al., 2010).
State the source and the context in which you apply the threshold. You should also check each item's Corrected Item-Total Correlation, commonly using 0.3 or above (Nunnally and Bernstein, 1994).
Is Cronbach's Alpha below 0.7 acceptable?
It can be acceptable in some exploratory studies when Alpha is at least 0.6, the items have suitable item-total correlations, and the scale content has a clear basis. State the criterion you selected instead of raising or lowering the threshold after seeing the output.
If Alpha is much lower, check reverse-coded items, missing data, incorrectly entered variables, and whether different constructs were combined. Remove an item only when you have a clear reason.
Should I remove an item when Cronbach's Alpha if Item Deleted is higher?
Do not remove it automatically. Check whether the increase is meaningful, the item's Corrected Item-Total Correlation, the theoretical content, and the number of items remaining after removal.
If the item is important to the content of the construct, retaining it may be reasonable even when Alpha does not increase. If you remove it, rerun the analysis and record the decision in your thesis.
Why is Cronbach's Alpha negative in SPSS?
A negative Alpha usually appears when the average correlations between items are negative. Common causes include forgetting to reverse-code an item, using the wrong reverse-coding formula, or placing items from two different constructs in the same scale.
Check the questionnaire and coding before deleting an item. Then create the correctly reverse-coded variable, rerun the correct item group, and save the output from each processing round.
Do I need to run Cronbach's Alpha before EFA?
In a common workflow, you check preliminary reliability before running EFA to examine the factor structure. The final decision still depends on your model, scale, and method stated in the proposal.
Before running EFA, you also need to check KMO, Bartlett, eigenvalue, and factor loading against sourced criteria. Alpha answers the question of internal consistency only. It does not replace EFA.
Open your SPSS file, list every scale, run each item group separately, and save the Reliability Statistics and Item-Total Statistics tables before removing an item. If you need to run the analysis on your own .sav or .csv file and track every processing round, you can use DoThesis M4 data analysis.