Written and maintained by CASRAI Editorial Board
Published
On this page
A nominal survey question asks respondents to choose a category that has no natural order, so the answers can be counted and compared by share but never ranked, averaged or subtracted. Marital status, sex at birth, employment status and county of residence are typical examples. This guide is for people designing or analysing questionnaires: it shows real nominal items with their exact wording, explains how to code and summarise them, and separates them from ordinal and yes/no items.
What is a nominal question in a survey?
A nominal question is a closed-ended item whose response options are labels for categories with no numerical ranking. The CDC’s epidemiology training defines a nominal-scale variable as one “whose values are categories without any numerical ranking, such as county of residence” (CDC, Principles of Epidemiology, Lesson 2). Penn State’s STAT 504 course gives the same test: for eye colour, the categories “brown, blue, green, and other have no intrinsic ordering.” If you can reorder the answer options without changing their meaning, the item is nominal. For the wider scheme, see levels of measurement: nominal, ordinal, interval and ratio.
Real nominal items from the CDC BRFSS questionnaire
The CDC’s Behavioral Risk Factor Surveillance System (BRFSS) 2022 questionnaire (published 29 August 2023) contains standard nominal items. The wording and codes below are quoted from that document, read 7 November 2026.
| Variable | Question wording | Response options and codes |
|---|---|---|
| MARITAL | “Are you…” | 1 Married; 2 Divorced; 3 Widowed; 4 Separated; 5 Never married; 6 A member of an unmarried couple; 9 Refused (not read aloud) |
| BIRTHSEX (module 25, MSAB.01) | “What was your sex at birth? Was it male or female?” | 1 Male; 2 Female; 7 Don’t know/Not sure; 9 Refused |
| EMPLOY1 | “Are you currently…?” | 1 Employed for wages; 2 Self-employed; 3 Out of work for 1 year or more; 4 Out of work for less than 1 year; 5 A homemaker; 6 A student; 7 Retired; 8 Unable to work; 9 Refused (not read aloud) |
Three design points are visible in these items. First, the numbers are labels, not quantities: “Retired” (7) is not “more” than “A student” (6). Second, the options are mutually exclusive and cover the common cases. Third, the BRFSS uses non-substantive codes (7 Don’t know/Not sure on the sex-at-birth item, 9 Refused throughout) alongside the substantive categories; recode these as missing before you tabulate categories. The employment list mixes work status (employed, self-employed) with out-of-work duration and life situation (student, retired); in analysis, treat it as an unordered set of categories. That reading is an editorial interpretation, not a statement from the CDC.
Nominal vs ordinal vs yes/no questions
Nominal vs ordinal survey questions
An ordinal question has options with a natural order, such as “poor, fair, good, excellent”; a nominal question does not. Penn State notes that the same variable can be treated as nominal or ordinal depending on the analysis, so the level is a property of how you use the answers, not only of the wording. Ordered items are collected in examples of ordinal survey questions.
Are yes/no questions nominal?
Yes. A yes/no item is a categorical variable with two unordered categories. The CDC lists alive/dead, ill/well and vaccinated/unvaccinated among its nominal-scale examples. Design issues specific to two-option items are covered in dichotomous survey questions.
Are Likert items nominal?
No. Agreement scales (strongly disagree to strongly agree) are ordered, so they are ordinal at minimum. See survey question types for the full set of formats.
Summarising nominal responses: a worked example
The data below are an illustrative hypothetical sample, not real survey results. Suppose 42 people answer the BRFSS-style MARITAL item. Two refuse (code 9) and are set to missing, leaving n = 40 valid answers.
| Category | Count (f) | Proportion (f / 40) | Percent |
|---|---|---|---|
| Married | 18 | 18 / 40 = 0.450 | 45.0% |
| Never married | 11 | 11 / 40 = 0.275 | 27.5% |
| Divorced | 5 | 5 / 40 = 0.125 | 12.5% |
| Member of an unmarried couple | 3 | 3 / 40 = 0.075 | 7.5% |
| Widowed | 2 | 2 / 40 = 0.050 | 5.0% |
| Separated | 1 | 1 / 40 = 0.025 | 2.5% |
| Total | 40 | 1.000 | 100.0% |
Check: 18 + 11 + 5 + 3 + 2 + 1 = 40, and 45.0 + 27.5 + 12.5 + 7.5 + 5.0 + 2.5 = 100.0. The mode, the most frequent category, is Married (18 of 40). Penn State describes the standard summaries for categorical variables as frequency counts and proportions, with a simple bar plot to show the relative sizes of categories, and notes that default alphabetical ordering is usually acceptable for nominal variables. A mean or median of the codes 1 to 6 would be meaningless here. For a step-by-step table build, see how to make a frequency table.
| You want to know | Use for nominal data | Do not use |
|---|---|---|
| Which answer is most common | Mode | Mean or median of the codes |
| How big each group is | Counts and proportions | Sums of the codes |
| Whether two categorical questions are associated | A cross-tabulation with a test for categorical data, such as the chi-square test | Pearson correlation on the codes |
The first two rows follow the counts-and-proportions approach in the Penn State material; the association-test row reflects standard statistical practice and is editorial interpretation rather than a quotation from the pages cited here.
Assumptions and common mistakes
- Treating codes as numbers. Store the label (or a factor), not a number that software will average.
- Leaving “Don’t know” and “Refused” in the category list. The BRFSS uses 7 and 9 as non-substantive codes; recode them as missing, or report them separately, so percentages describe substantive answers.
- Options that overlap. Categories must be mutually exclusive. If respondents can legitimately choose more than one, use a select-all-that-apply item and analyse each option as its own yes/no variable.
- Unclear denominators. State whether percentages use all respondents or only valid answers (above: 40 valid of 42).
- Calling an ordered scale nominal. If the options have a natural order, analyse them as ordinal and keep the information in the order.
How to report nominal results (APA-style example)
A reporting sentence for the worked example: “Of the 40 respondents who answered, 18 (45.0%) were married, 11 (27.5%) had never married, 5 (12.5%) were divorced, 3 (7.5%) were members of an unmarried couple, 2 (5.0%) were widowed, and 1 (2.5%) was separated. Two respondents declined to answer and were excluded.” Give the count, the percentage and the valid n, and state how missing responses were handled.
How to tabulate nominal questions in Excel, R or Python
Each snippet assumes the answers sit in one column, with refusals blank or recoded as missing.
Excel
=COUNTIF(A2:A43,"Married")=COUNTIF(A2:A43,"Married")/COUNTA(A2:A43)
The first formula gives the count for one category and the second its proportion of non-blank answers. Repeat for each category; the mode is the category with the largest count.
R
marital <- factor(df$marital) # refusals set to NAtab <- table(marital) # countsround(prop.table(tab), 3) # proportionsnames(tab)[which.max(tab)] # mode
Penn State’s STAT 504 uses prop.table() for this proportion step.
Python (pandas)
s = df["marital"].astype("category") # refusals as NaNs.value_counts() # countss.value_counts(normalize=True).round(3) # proportionss.mode()[0] # mode
In all three, missing values are excluded, so the proportions use the valid n.
FAQ
What are nominal questions in a survey?
They are closed-ended questions whose answer options are unordered categories, such as marital status or employment status. Answers are counted and compared by proportion.
Can I average a nominal survey question?
No. The numeric codes are labels, so an average of them has no meaning. Report counts, proportions and the mode.
Is gender or sex a nominal variable?
Penn State lists gender as a nominal example. The BRFSS sex-at-birth item above is nominal, with separate non-substantive codes for don’t know and refused.
How is a nominal question different from an ordinal one?
An ordinal question has options with a natural order; a nominal question does not. See types of variables and categorical variable examples for more cases.
Sources
- Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System 2022 Questionnaire (29 August 2023). cdc.gov/brfss. Read 7 November 2026.
- CDC. Principles of Epidemiology in Public Health Practice, Lesson 2, Section 2 (archived). archive.cdc.gov. Read 7 November 2026.
- Penn State Eberly College of Science. STAT 504: Analysis of Discrete Data, Lesson 1. online.stat.psu.edu/stat504. Read 7 November 2026.







