Levels of measurement

We distinguish four different levels of measurement of a variable. We will present them in an order which reflects their precision, from the least precise to the most precise, as follows: 1) nominal; 2) ordinal; 3) interval; and 4) ratio.

Figure 1. A graphic illustration of the four levels of measurement

Nominal measures

Variables whose attributes are simply distinct from one another are called nominal measures. Nominal measures only assign names or labels for characteristics of a variable. Imagine you group people by their birthplace, hair colour, college major, political party affiliation, gender. These are examples of nominal variables.

Let’s suppose you want to explore learning platforms used in your faculty in the period of the pandemic. For that, you ask students in a lecture hall to gather in groups according to the main learning platform used in the academic year 2020-2021: all those using Google Classroom in one group, those using Microsoft Teams in another, those using Zoom, Cisco Webex, and so on also forming different groups. All of the students belonging to a particular group have at least one thing in common – the online learning platform used for workshops in the last academic year – and differ from the students in all other groups in that same regard. Keep in mind that how groups are located in the room has no relevance for the study. What matters is that all members of a specific group share the same learning platform and that each group has a different shared learning platform.

The only thing we can affirm about two individuals according to a nominal variable is that they are either the same or different (Babbie, 2013: 180-181).

In the course of a research, certain nominal measures play the role of independent variables, as gender, birthplace, religious denomination, ethnicity, because they (usually) do not vary or change over time. Other nominal measures can be dependent variables. For example, the status of partnership (unmarried, married, in a civil partnership, separated, divorced, etc.) might depend on the type of family of origin (nuclear, monoparental family or blended family) or on the religiosity of the person.

Ordinal measures

Variables whose attributes can be logically rank ordered are called ordinal measures. “The different attributes of ordinal variables represent relatively more or less of the variable. (...) An example is socioeconomic status as composed of the attributes high, medium, low.” (Babbie, 2013: 181). Comparing two people who are different in terms of an ordinal variable, you can affirm that one is ”more” than the other; that is, for example, more compassionate, more religious, or older.

If we are pursuing the earlier example of grouping students, for instance with an interest to find out the relative amount of formal education attained by their main caretaker. This time you will ask students whose caretaker graduated from college to stand in one group, those whose caretaker has only a high school diploma in another group, and those whose caretaker has not graduated from high school to stand in a third group. This mode of dividing students satisfies the nominal criteria of being different, but additionally we may arrange the three groups upward or downward in terms of the relative amount of formal education each of the main caretaker of students had. We may arrange the three groups of students in a row, ranging from most to least formal education their main caretaker had achieved. This arrangement can offer a physical representation of an ordinal measure. If we know which groups two individuals belong to, we can determine that one’s caretaker had more, less, or the same formal education as the caretaker of another student. In this example, it is irrelevant how close or far apart the groups constructed by the educational attainment of the main caretaker of students are from one another. The group of students whose caretaker graduated high school, however, should be placed between the less-than-high-school group and the college group, otherwise the rank order would be incorrect.

Let’s reflect together on this example. In a study carried out in Australia (Watson and Wooden, 2001), perceived prosperity was assessed using the following question: “Given your current needs and financial responsibilities, would you say that you and your family are …?” What is wrong with someone formulating the categories of variable as follows: Reasonably comfortable, Prosperous, Poor, Very Comfortable, or Just getting along? Most probably you noticed that these categories are mixed up. While the variable is ordinal, we know that we must arrange the classes of the variable in a logical sequence: Prosperous, Very comfortable, Reasonably comfortable, Just getting along, or Poor.

There is one more thing that is worth mentioning related to ordinal measures. Variables investigating the satisfaction with something or someone are usually ordinal measures, and the answers often use the Likert scales. Consider the following example: “On a scale from 1 to 7, how satisfied you are with your state of health, where 1 means not satisfied at all, and 7 means totally satisfied?”