2.4 Survey items with different levels of measurement scales

To select an appropriate response scale for each survey item, it’s important to understand the different levels of measurement scales. Stanley Smith Stevens, a professor of psychology at Harvard University, conceptualized four different levels of measurement scales: nominal, ordinal, interval, and ratio scales (Stevens, 1946^[1]).

2.4.1     Nominal scales

Nominal scales categorize items by assigning names or labels without implying any assumed rank-order among them. Since nominal scales do not involve rank-ordered ratings, they are not suitable for survey items intended to rate respondents’ perceived quality or value (e.g., questions like, “Was this terrible, bad, good, or excellent?”).

Instead, fact-capturing survey items often use nominal response scales. While numerical values (e.g., 1, 2, 3, 4, etc.) can be assigned to these options, these numerical values are used solely for identification purposes and do not imply any order or ranking. For example:

Here, the options can be coded numerically (Male = 1, Female = 2, Intersex = 3) for convenience in data processing. However, it doesn’t imply that Female and Intersex are higher or better than Male, or vice versa. Similarly, consider the following examples:

2.4.2     Ordinal scales

Ordinal scales consist of a group of rank-ordered level where the distance between a set of two consecutive levels is not necessarily the same as the distance between another set of two consecutive levels. For example, consider the following scale, which ranks frequency of appreciation. While the levels are ranked in order of how often someone feels appreciated, the distance between Never and Seldom may not be the same as the distance between Seldom and Often or between Often and Always. Nonetheless, this type of ordinal scales can measure people’s perceived quality or value.

2.4.3     Interval scales

Interval scales contain a group of rank-ordered levels where the distance between each pair of consecutive levels is consistent and equal. Interval scales can also be used to measure people’s perceptions. However, one important characteristic of internal scales is that they do not have an absolute zero point. Zero on an interval scale is an arbitrary zero value. For example, in the following response scale, the Very dissatisfied option is assigned the value 0. However, zero does not mean there is a complete absence of satisfaction—it simply represents the lowest point on the scale.

2.4.4     Ratio scales

Ratio scales have an equal distance between intervals, and they have an absolute zero value, which indicates the absence of the attribute measured (e.g., a value of zero in income = no income). Examples are measurements of age, length, weight, height, income, and time, as shown below:

2.4.5     Project the type of data you need before designing survey items

When designing survey items, it’s important to project the type of data you intend to collect before selecting appropriate response scales.

When capturing factual information, you may use nominal scales for categorical data (e.g., gender or ethnic groups, job titles), ordinal scales for rank-ordered data (e.g., military ranks), interval scales for rank-ordered and meaningful intervals (e.g., years).or ratio scales for precise measurements with a true zero value(e.g., income, age).

On the other hand, when measuring perceived quality or value, the design process is more complex as a variety of rating scales for measuring perceptions are available. Ordinal or interval scales are commonly used to measure perceived quality and value of something (e.g., job satisfaction, employee performance, or customer service).

In terms of flexibility in usage, ratio scales produce the most flexible data, followed by interval scales, ordinal scales, and nominal scales in that order. For example, after you collected actual income or age values (ratio), you can group the values into interval or ordinal categories and generate frequencies as shown below:

Similarly, after you have collected interval data such as satisfaction levels measured on a 11-point scale (interval data), you may group the data into several ordered categories such as low (0-3), medium (4-7), and high (8-10) satisfaction levels (ordinal data), if needed.

However, once you have collected nominal or ordinal data, you cannot report the data as interval or ratio data. For example, asking for age in categories like the survey item below (nominal scale) will yield categorical data, not continuous data. If you need exact ages for analysis, this survey item would not be suitable:

In conclusion, before designing your survey items, identify the type of data you need and select response scales that align with the appropriate level of measurement level.