Nominal data is data that labels observations into named categories with no natural order, such as gender, nationality, blood type, or eye colour. Common examples of nominal data include marital status (single, married, divorced), mode of transport (car, bus, bicycle, walking), and religious affiliation. You can count how many observations fall into each category and identify the most frequent one (the mode), but you cannot rank the categories or perform arithmetic on them.
Nominal data is the simplest of the four levels of measurement, but that does not make it unimportant. A huge proportion of variables collected in social research, healthcare, education, and market research is nominal, and handling it incorrectly is one of the most common statistical mistakes in student work. This guide explains exactly what nominal data is, gives plenty of real examples, shows how it differs from ordinal data, and sets out which analyses (mode, chi-square and more) are appropriate.
“Nominal scales are used only for qualitative classification. The numbers serve only as labels for the classes, and the categories cannot be ordered or quantified.” — S. S. Stevens, On the Theory of Scales of Measurement, Science (1946)
What Makes Data Nominal
Nominal measurement classifies observations into distinct, mutually exclusive categories. The word “nominal” comes from the Latin nomen, meaning “name”, because the categories are simply names or labels. The categories have no natural order or hierarchy, which means you cannot meaningfully rank them, and you cannot calculate a mean or median from them. Nominal data has three defining characteristics.
- Categories are mutually exclusive — each observation belongs to exactly one category and never to two at once.
- Categories are exhaustive — every possible observation has somewhere to go, which is why an “Other” category is so common.
- No ordering is implied — one category is not more or less than another. “Spanish” is not greater than “French”.
A special case of nominal data is the binary (or dichotomous) variable, which has only two categories — for example, yes/no, pass/fail, or alive/dead. Binary variables are still nominal because the two categories have no inherent rank. Nominal data is a type of categorical variable, sitting alongside ordinal data within the broader family of qualitative data.
A simple test helps you spot nominal data quickly. If you can rearrange the categories into any order without losing meaning — for instance, listing nationalities as British, French, Nigerian, or as Nigerian, British, French — the variable is almost certainly nominal. If reordering the categories destroys the meaning (as it would for “low, medium, high”), the data is ordinal instead. Because the order of the labels is arbitrary, the only thing you can legitimately do with nominal data is count how many observations fall into each category and compare those counts.
Not sure how to analyse your nominal data?
ResearchProspect to the rescue!
Our statisticians can run the right tests — from frequency tables to chi-square — and interpret the output for you. Explore our statistical analysis service.
Examples Of Nominal Data
Nominal data appears everywhere in research and everyday life. Here are some quick, everyday examples before we look at research-specific ones:
- Gender / sex: male, female, non-binary.
- Nationality: British, French, Nigerian, Indian.
- Blood type: A, B, AB, O.
- Eye colour: blue, brown, green, hazel.
- Marital status: single, married, divorced, widowed.
The table below maps common nominal variables to the research areas where you are most likely to meet them.
| Research Area | Nominal Variable | Possible Categories |
|---|---|---|
| Education | Type of school attended | State comprehensive, grammar, independent, sixth-form college |
| Healthcare (NHS) | Method of referral | GP referral, A&E, self-referral, emergency transfer |
| Housing | Tenure type | Owned outright, mortgaged, rented (private/council/HA), other |
| Employment | Employment status | Full-time, part-time, self-employed, unemployed, student, retired |
| Transport | Main mode of commute | Car, train, bus, bicycle, walking, working from home |
| Social research | Religious affiliation | Christian, Muslim, Hindu, Jewish, No religion, Other |
| Marketing | Preferred brand | Brand A, Brand B, Brand C, no preference |
“In the 2021 Census of England and Wales, 46.2% of the population described themselves as ‘Christian’ and 37.2% reported ‘No religion’ — a classic example of a nominal variable used at national scale.” — Office for National Statistics (ONS), Census 2021
Where Nominal Data Sits: The Levels of Measurement
Statisticians group all variables into four levels of measurement, first proposed by psychologist S. S. Stevens in 1946. Each level builds on the one before, adding a new mathematical property. Nominal is the foundation — it has categories but nothing else. The table below shows how the four levels compare.
| Level | Named categories? | Ordered? | Equal intervals? | True zero? | Example |
|---|---|---|---|---|---|
| Nominal | Yes | No | No | No | Blood type, nationality |
| Ordinal | Yes | Yes | No | No | Survey rating (poor–excellent) |
| Interval | Yes | Yes | Yes | No | Temperature in °C |
| Ratio | Yes | Yes | Yes | Yes | Height, weight, income |
Knowing the level of measurement of each variable is the single most important step in choosing the right statistical test. Nominal data restricts you to a small but reliable toolkit, which we cover next.
Appropriate Statistics For Nominal Data
Because nominal categories have no mathematical relationship, most standard statistics do not apply. The two workhorses are the mode (the most frequent category) for description and the chi-square test for testing associations. Here is what you can and cannot do:
| Statistic / Test | Appropriate? | Notes |
|---|---|---|
| Frequency count | Yes | The primary descriptive tool for nominal data |
| Percentage / proportion | Yes | Convert frequencies to proportions of the total |
| Mode | Yes | The only appropriate measure of central tendency |
| Mean | No | Categories have no mathematical relationship |
| Median | No | Cannot order categories without imposing an arbitrary ranking |
| Chi-square test of independence | Yes | Tests association between two nominal variables |
| Cramér’s V | Yes | Effect size for chi-square; measures strength of association |
| Pearson correlation | No | Requires at least interval data |
| Logistic regression | Yes (as outcome) | Use for a binary or multinomial nominal outcome instead of a t-test |
Step 1 — Observed counts. Inner zone: 30 car, 70 public transport. Outer zone: 70 car, 30 public transport.
Step 2 — Expected counts (if there were no association) = (row total × column total) ÷ grand total. Each cell expects (100 × 100) ÷ 200 = 50.
Step 3 — Compute χ² = Σ (O − E)² ÷ E = 4 × [(30 − 50)² ÷ 50] = 4 × (400 ÷ 50) = 4 × 8 = 32.
Step 4 — Compare to the critical value for df = (2 − 1)(2 − 1) = 1, which is 3.84 at p = 0.05. Since 32 > 3.84, we reject the null hypothesis: transport choice is significantly associated with city zone.
Visualising Nominal Data
Because nominal data is about counts within categories, the best charts emphasise frequency rather than trend.
- Bar chart — the standard choice. Each bar represents a category; the height shows frequency or percentage. Bars should be in a meaningful display order (e.g. descending frequency, or a conventional order such as alphabetical).
- Pie chart — useful when you want to show proportions of a whole, but it becomes hard to read with more than four or five categories.
- Stacked bar chart — helpful for comparing nominal distributions across groups (e.g. mode of transport broken down by gender).
- Avoid histograms and line charts — these imply order or continuity, which nominal data does not have.
Working with nominal data
- Frequencies & percentages
- The mode
- Bar chart
- Pie chart
- Chi-square test
- Cramer’s V
Coding & Entering Nominal Data
In quantitative data files, nominal categories are usually assigned numeric codes: England = 1, Scotland = 2, Wales = 3, Northern Ireland = 4. This is necessary so that software can store the data, but remember — those numbers are codes, not quantities. During data collection and entry, keep a clear codebook so every label maps to one code.
- In SPSS: set the ‘Measure’ column to ‘Nominal’ to tell the software how to treat the variable.
- In R: wrap nominal variables in
factor()to define them as categorical. - In regression with 3+ nominal categories: use dummy coding, creating separate binary (0/1) variables for each category except one (the reference category).
Nominal Vs Ordinal
The most common confusion is between nominal and ordinal data. Both are categorical, but only ordinal data has a meaningful order. Before treating data as nominal, ask: can these categories be meaningfully ranked? If yes, the data is ordinal, not nominal.
- Employment status (employed / unemployed / student) — no inherent ranking. Nominal.
- Highest qualification (GCSE / A-level / undergraduate / postgraduate) — clearly ordered. Ordinal.
- Type of accommodation (halls / private rent / family home / other) — no natural ranking. Nominal.
- Satisfaction with accommodation (very dissatisfied … very satisfied) — clearly ordered. Ordinal.
The practical consequence is in analysis: ordinal data can use medians and rank-based tests, whereas nominal data is limited to the mode, frequencies and chi-square. To see the full hierarchy, read our guide to the levels of measurement in statistics.
