Yes, race is a nominal variable. It places people into categories, but those categories have no inherent ranking or numerical order. You cannot say one racial group is “higher” or “lower” than another in any meaningful statistical sense, which is exactly what makes it nominal rather than ordinal, interval, or ratio.
What Makes a Variable Nominal
Psychologist Stanley Smith Stevens developed the four scales of measurement that statisticians still use today: nominal, ordinal, interval, and ratio. Nominal sits at the first level. A nominal variable assigns observations to categories, but those categories cannot be ranked, added, subtracted, multiplied, or divided. Each category simply has a unique identity, nothing more.
The University of Michigan’s statistics department puts it plainly: a nominal measurement can take on any element in a finite set of unordered values, and biological race is a commonly encountered example. Hair color, blood type, and yes/no survey responses also fall into this category. What they share is that no agreed-upon way exists to arrange them from highest to lowest.
Why Race Is Not Ordinal
An ordinal variable has categories that follow a clear sequence. Education level is a good example: a high school diploma comes before a bachelor’s degree, which comes before a master’s degree. You can rank them. Race has no equivalent logic. There is no basis for placing “Asian” before “White” or “Black” after “Native Hawaiian” on any meaningful scale. The categories are simply different from one another, not greater or lesser.
This distinction matters practically. Computing an average of a nominal variable makes no sense. If you assign “1” to one racial group and “2” to another, the average of 1.5 is meaningless. You can count how many people fall into each category and report percentages, but arithmetic operations on the categories themselves produce nonsense.
How Race Categories Are Defined
The U.S. Office of Management and Budget updated its standard race and ethnicity categories in March 2024. The current framework has seven minimum categories: American Indian or Alaska Native, Asian, Black or African American, Hispanic or Latino, Middle Eastern or North African, Native Hawaiian or Pacific Islander, and White. Respondents can select one or multiple categories from a single combined question, and all seven categories are treated as co-equal.
The American Psychological Association’s reporting guidelines emphasize that race refers to physical differences that groups and cultures consider socially significant, while ethnicity refers to shared cultural characteristics like language, ancestry, and beliefs. APA discourages language that treats race as a fixed biological essence and recommends using the terms that participants themselves use. When possible, naming a specific nation or region of origin (Bolivian, Salvadoran) is preferred over broader labels (Hispanic, Latino).
How to Summarize and Test Nominal Data
Because race is nominal, only certain statistical tools apply. The most basic summary is a frequency table showing how many people (or what percentage) fall into each category. The mode, which is simply the most common category, is the only meaningful measure of central tendency. A mean or median of race categories would be statistically invalid.
When you want to test whether race is associated with another categorical variable, the chi-square test is the standard choice. For example, researchers might use a chi-square test to see whether the distribution of a health outcome differs across racial groups. If the other variable is continuous, like a lab measurement or a test score, analysis of variance (ANOVA) lets you compare average values across racial groups to see whether the differences are statistically significant.
How Race Enters Regression Models
Regression analysis requires numeric inputs, so a nominal variable like race needs to be converted before it can be included. The most common method is dummy coding: each racial category gets its own binary (0 or 1) column, and one category serves as the reference group. The model then estimates how each group’s outcome compares to that reference.
A 2025 paper in the Annals of Epidemiology compared six different coding approaches for race in regression models, including dummy coding, simple effect coding, and deviation coding. Each method answers a slightly different question. Dummy coding, by far the most widely used, compares each group to the chosen reference. Deviation coding compares each group to the overall average across all groups. The choice of method and reference group can meaningfully change how results are interpreted, so it should be guided by the research question rather than selected arbitrarily.
In public health and clinical research, race is often included as a control variable with non-Hispanic White as the default reference group. Researchers are increasingly scrutinizing this practice, asking whether that default is always appropriate and whether simply “controlling for race” captures the complex social dynamics that race represents.
Best Ways to Visualize Race Data
Bar charts are the go-to visualization for nominal variables. Each category gets its own bar, and the bar’s height or length represents the count or percentage for that group. This works well because our eyes naturally compare bar lengths with high accuracy. Clustered bar charts can show race alongside a second categorical variable, like gender, in the same graphic.
Pie charts can also display nominal data, but they become hard to read once you have more than four or five categories. With seven federal race categories (and the possibility of multiracial responses), bar charts are almost always the clearer choice. Frequency tables work well in written reports where exact numbers matter more than visual impact.