Measurement involves the identification of observations with attributes on variables. An example would be when you observe, or learn through a questionnaire response, that a subject of study is male. You would have identified that subject with an attribute (male) on a variable (gender).
Levels of measurement refers to the relationship among the attributes
making up a variable. Social researchers distinguish four such levels:
nominal, ordinal,
interval, and ratio.
| Level of Measurement | Relationship among attributes | Examples |
| Nominal | The same or different | Gender, political party, religion |
| Ordinal | Can be arranged in order, from lower to higher, from less to more, etc. | Religiosity from low to high, political orientation from liberal to conservative |
| Interval | Distance from one attribute to another can be expressed in standardized units | Fahrenheit and celsius temperatures, intelligence tests |
| Ratio | A true zero point exists | Age, income, years of education |
In the table above, each level of measurement has all the qualities of the levels higher in the table, plus those specified for it. For example, a ratio measure has a true zero point, plus the qualities specified for interval, ordinal, and nominal measures.
This means that the researcher sometimes can choose a level of measurement. Instead of taking full advantage of the qualities of a ratio measure, you could measure age by asking if subjects are: Below 21, 21-35, 36-55, 56-65, or Over 65. You would, effectively, be measuring age at the ordinal level by creating groups that can be arranged in order.