Primary Scales of Measurement: Nominal, Ordinal, Interval, Ratio

The four primary scales of measurement—nominal, ordinal, interval, and ratio—were classified by psychologist S.S. Stevens. They form a hierarchy from weakest (nominal) to strongest (ratio). Each scale permits different mathematical operations and statistical analyses. The scale determines what you can meaningfully say about your data: whether you can count, rank, add, subtract, multiply, or form ratios. Choosing the correct scale is essential; using inappropriate statistics (e.g., calculating a mean on nominal data) produces meaningless results. Researchers should always measure at the highest possible level.

1. Nominal Scale

The nominal scale is the simplest and weakest level of measurement. It assigns numbers or labels to categories solely for identification or classification—numbers have no quantitative meaning. Nominal derives from the Latin nomen (name). Examples: gender (1 = male, 2 = female), marital status (single, married, divorced), customer type (new, returning), product category, region code, or employee ID numbers. Arithmetic operations are meaningless; you cannot average ID numbers. Permissible statistics include frequency counts, percentages, mode (most common category), and chi-square tests for association between nominal variables. You cannot calculate means, medians, or rank orders. Nominal scales are essential for grouping data and creating categories. In business research, they are used for demographic classification, market segmentation, and binary variables (yes/no, purchased/did not purchase). Despite their weakness, nominal scales are ubiquitous and fundamental.

2. Ordinal Scale

The ordinal scale ranks items in order (first, second, third) but does not specify the magnitude of differences between ranks. We know order but not how much more or less. For example, a customer satisfaction ranking (1st choice, 2nd choice, 3rd choice) tells preference order but not how strongly one is preferred over another. Other examples: education level (high school < bachelor’s < master’s < doctorate), letter grades (A > B > C > D), and Likert scales (“strongly disagree” to “strongly agree”) when treated as ordinal. Permissible statistics include median (middle rank), mode, percentile, rank-order correlation (Spearman’s rho, Kendall’s tau). Arithmetic means are controversial because intervals are not proven equal. In business research, ordinal scales are common for preferences, rankings, and attitude surveys. However, many researchers treat Likert scales as interval for convenience, risking inappropriate analysis. Ordinal scales preserve order but not distance.

3. Interval Scale

The interval scale has equal distances between successive values but lacks a true, meaningful zero point. Zero is arbitrary; negative values are possible. Because intervals are equal, addition and subtraction are meaningful. For example, the difference between 20°C and 30°C (10 degrees) equals the difference between 30°C and 40°C. However, 40°C is not “twice as hot” as 20°C because 0°C does not mean “no temperature.” Other examples: calendar years (year 0 is arbitrary), IQ scores, and many psychological test scores. Permissible statistics: mean, standard deviation, correlation, t-tests, ANOVA, regression, and factor analysis. In business research, interval scales are assumed for Likert scales (though debated), semantic differential scales, and composite scores from multi-item scales. Interval measurement enables powerful parametric statistics. The absence of a true zero is the key limitation. You cannot form ratios (e.g., “twice as satisfied”) without a true zero.

4. Ratio Scale

The ratio scale has all properties of interval measurement plus a true, meaningful zero point representing the complete absence of the attribute. Zero means zero. Because of the true zero, all arithmetic operations—including ratios—are meaningful. Forty sales is twice twenty sales; 0 complaints means no complaints. Examples: sales revenue in rupees, employee age, number of customer complaints, time in seconds, production units, distance, weight, and market share. Permissible statistics: any statistic—mean, standard deviation, coefficient of variation (standard deviation/mean), geometric mean, and all parametric and non-parametric tests. Ratio scales are the strongest level, retaining maximum information. In business research, most financial and operational metrics are ratio scales. Whenever possible, researchers should collect data at the ratio level, because ratio data can be transformed down to lower levels (e.g., categorizing into nominal groups) but not vice versa. Ratio scales are the gold standard for measurement.