Why Your Questionnaire Needs a Reliability Check
You have spent weeks designing your questionnaire. The questions sound right, your supervisor nodded approvingly, and your pilot respondents finished it without complaining. But one question still hangs in the air: do your items actually measure the same thing, consistently?
That is exactly what Cronbach’s alpha answers. It is the single most reported reliability statistic in student theses, and examiners almost always ask for it. The good news is that checking it takes about ten minutes once your pilot data is in a spreadsheet. This guide walks you through what alpha means, how to calculate it step by step in SPSS, how to read the number you get, and the mistakes that trip up most beginners.
What Is Cronbach’s Alpha?
Cronbach’s alpha is a number between 0 and 1 that tells you how closely related a set of items is as a group. It was introduced by psychologist Lee J. Cronbach in 1951, in a paper titled “Coefficient alpha and the internal structure of tests” (Psychometrika, 16(3), 297–334), and it has been the standard measure of internal consistency ever since.
In plain English: if you have ten questions that are all supposed to measure job satisfaction, alpha tells you whether respondents who score high on one question also tend to score high on the others. If they do, your items are hanging together and your scale is internally consistent. If they do not, something is off — perhaps one question measures something different, or a reverse-worded item was never recoded.
The formula looks intimidating, but its logic is simple:
α = (k / (k − 1)) × (1 − (sum of item variances / variance of the total score))
Here, k is the number of items. The formula compares how much each item varies on its own with how much the total score varies. When items move together, the total score varies a lot relative to the individual items, and alpha climbs toward 1. When items barely relate to each other, alpha stays low.
Two things alpha is not: it is not a measure of validity (a scale can be reliably wrong), and it is not proof that your scale measures only one thing. A high alpha can hide two different dimensions mixed together. Alpha measures consistency, nothing more — but consistency is the foundation everything else is built on.
What Counts as a Good Cronbach’s Alpha?
There is no single magic cut-off, but the guidelines most examiners use come from Nunnally and Bernstein (1994), and they are the ones you should quote in your thesis:
| Alpha value | Interpretation |
|---|---|
| 0.90 and above | Excellent — but check for redundant items |
| 0.80 – 0.89 | Good |
| 0.70 – 0.79 | Acceptable |
| 0.60 – 0.69 | Questionable — acceptable only for exploratory research |
| Below 0.60 | Poor — revise or drop items |
A value of 0.70 is the usual minimum for a thesis questionnaire. Note the warning at the top of the table: an alpha above 0.95 is not a badge of honour. It usually means several items are near-identical paraphrases of each other. You are measuring the same question five times, not building a better scale.
Also remember that alpha grows with the number of items. A 20-item scale will almost always show a higher alpha than a 4-item scale measuring the same thing. So always report the number of items alongside your alpha — an alpha of 0.82 from 5 items is more impressive than 0.82 from 25 items.
Before You Start: 3 Checks
Run these three checks before you calculate anything. Skipping them is the most common reason beginners get a suspiciously low alpha.
1. Recode your reverse-scored items. If some statements are worded negatively (for example, “I feel drained at work” inside a job satisfaction scale), high scores on those items mean low satisfaction. Reverse them first — on a 5-point scale, a 1 becomes a 5, a 2 becomes a 4, and so on — or your alpha will collapse. In SPSS, use Transform > Recode into Different Variables.
2. Check that all items use the same response scale. Alpha assumes your items are on a comparable scale. Do not mix 5-point items with 7-point items in the same calculation without standardizing them first.
3. Use enough pilot data. Alpha is unstable with tiny samples. Aim for at least 30 respondents in your pilot before you trust the number, and report the sample size whenever you quote alpha.
How to Check Reliability Step by Step in SPSS
Follow these five steps. The whole process takes about ten minutes once your data is entered.
Step 1: Enter your pilot data. Each row is one respondent; each column is one questionnaire item. Make sure reverse-scored items are already recoded (see check 1 above).
Step 2: Open the reliability analysis. Click Analyze > Scale > Reliability Analysis. A dialogue box will appear.
Step 3: Move your items across. Select all the items that belong to one scale — only one scale at a time — and move them into the Items box. Leave the Model as Alpha. If your questionnaire has several subscales, run a separate analysis for each one. Never lump all items from different subscales into a single alpha.
Step 4: Ask for the extra statistics. Click Statistics, then tick Item, Scale, and Scale if item deleted under Descriptives for. These tell you which items are dragging your alpha down. Click Continue, then OK.
Step 5: Read the output. The first table, Reliability Statistics, shows your Cronbach’s alpha and the number of items. The Item-Total Statistics table shows the Corrected Item-Total Correlation for each item and what alpha would become if you deleted that item. Any item with a corrected item-total correlation below 0.30 is a candidate for revision or removal — but never delete an item on statistics alone. Read the item first: it may be badly worded rather than irrelevant, and a reworded item can save a dimension of your scale.
Worked Example: A 5-Item Job Satisfaction Scale
Let us make the maths concrete. Imagine a short job satisfaction scale with five items, answered by pilot respondents on a 5-point scale. After entering the data, you calculate the variance of each item:
- Item 1 variance: 0.80
- Item 2 variance: 0.95
- Item 3 variance: 0.70
- Item 4 variance: 0.85
- Item 5 variance: 0.90
The sum of the item variances is 4.20. The variance of the total score (each respondent’s sum across all five items) is 12.50. Now apply the formula:
α = (5 / 4) × (1 − 4.20 / 12.50) = 1.25 × (1 − 0.336) = 1.25 × 0.664 = 0.83
An alpha of 0.83 falls in the “good” range. You can report it with confidence — and because you have the item-level numbers, you can also explain in your thesis exactly how you arrived at it, which is precisely what examiners like to see.
Notice what happens to the numbers if the items do not move together: the total-score variance shrinks toward the sum of the item variances, the fraction inside the brackets approaches zero, and alpha collapses. That is the whole intuition in one picture — alpha rewards items that rise and fall together.
How to Report Cronbach’s Alpha in Your Thesis
Report alpha in the methodology chapter when you describe your instrument, and again in the results chapter if you re-check it on your final sample. Always give three things: the alpha value, the number of items, and the sample size. A good sentence looks like this:
“The 10-item job satisfaction scale showed good internal consistency in the pilot study (Cronbach’s α = .82, n = 35).”
If you dropped an item, say so plainly: “One item was removed because its corrected item-total correlation was .18; the alpha for the remaining 9 items was .84.” Examiners trust transparent reporting far more than a perfect-looking number with no explanation.
One formatting note: in APA 7th edition style, the alpha value is written with a leading zero omitted (α = .82, not 0.82), because alpha cannot exceed 1. The Greek letter α is used, not the word “alpha”, in formal reporting.
5 Common Mistakes to Avoid
1. Forgetting to reverse-code. The classic beginner error. A single unrecoded reverse item can drag a healthy alpha down by 0.20 or more. Always recode before you calculate.
2. Calculating one alpha for the whole questionnaire. If your instrument has three subscales, report three alphas — one per subscale. A single overall alpha hides weak subscales and means nothing if the subscales measure different things.
3. Treating alpha as proof of one dimension. Alpha can be high even when your scale secretly measures two different constructs. If you need to check dimensionality, run a factor analysis — alpha cannot do that job.
4. Chasing 0.95. An alpha above 0.95 usually signals redundancy. Three near-identical items inflate alpha without adding information, and examiners know it. Aim for the 0.70–0.90 range.
5. Reporting alpha from the original study instead of your own data. Quoting the published reliability of a scale is good practice in your methodology chapter, but your examiner also wants to see the alpha from your sample. Always compute and report both.
Frequently Asked Questions
How many items do I need for a good Cronbach’s alpha?
There is no fixed minimum, but very short scales struggle: with only 2 or 3 items, alpha is usually low even when the items are good. Most thesis scales use 5 to 10 items per dimension, which is usually enough to reach 0.70 if the items are well written.
Can Cronbach’s alpha be negative?
Yes. A negative alpha means your items are, on average, negatively correlated with each other. In practice this almost always means a reverse-scored item was not recoded, or that some items are pulling in opposite directions. Recode and re-check before anything else.
What should I do if my alpha is below 0.70?
First, check for unrecoded reverse items. Then look at the “alpha if item deleted” column and the corrected item-total correlations, and consider dropping or rewording the weakest item. If the alpha is still low, your items may be measuring more than one thing — in that case, split the scale or run a factor analysis rather than forcing a single number.
Is Cronbach’s alpha the only reliability statistic?
No. McDonald’s omega is increasingly recommended because it does not assume all items contribute equally, and the split-half method is a simpler alternative for short scales. That said, alpha remains the statistic examiners expect to see, so report alpha first and mention omega as a supplement if your supervisor asks for it.
Do I calculate alpha on pilot data or final data?
Both. Calculate it on your pilot data to catch weak items before the main study, then calculate it again on your final sample and report that value in your results. The pilot alpha guides revision; the final alpha is the number that counts.
The Bottom Line
Cronbach’s alpha is not a hurdle to clear; it is a quick health check on the instrument your whole thesis depends on. Recode your reverse items, run one scale at a time, read the item-total statistics before you celebrate or panic, and report the number honestly with its item count and sample size. Do that, and the reliability section of your thesis writes itself — and survives your examiner’s questions.
Next in this series, we will look at the practical companion to this post: how to translate a questionnaire using the forward-backward method, so your carefully checked scale keeps its reliability in another language too.