What a Rating Response Represents
A survey may ask someone to rate a statement such as “I can easily find the information I need.” The response options might run from “Strongly disagree” to “Strongly agree.” These answers have a natural order, but it is not automatically clear that the steps between them are equal. That distinction matters when deciding what kind of variable has been recorded.
A single question of this kind is often called a Likert item or Likert-type item. A set of related items may be combined into a score, sometimes called a Likert scale. In ordinary usage, people sometimes use “Likert scale” for either the individual response format or the combined measure. For statistical classification, it is useful to distinguish the item from the score constructed from several items.
The word “ordinal” identifies what the response options support: a ranking. “Agree” indicates more agreement than “Disagree,” for example. The ordering alone does not tell us how much more, nor does it establish that the psychological distance from “Strongly disagree” to “Disagree” equals the distance from “Agree” to “Strongly agree.”
Codes Are Not Automatically Measurements
For data entry, a researcher might assign the codes 1, 2, 3, 4, and 5 to the five response options in order. This is convenient, but the numbers do not change the underlying information. They record the ordering of categories. The value 5 means a response in the fifth category, not necessarily five units of agreement.
The earlier material on nominal and ordinal categories distinguishes labels without an order from categories with an order. A Likert item is ordinal because its response categories can be ranked. It is not automatically quantitative just because the categories have numerical codes. In particular, calculating a numerical difference between codes assumes a meaning the response format itself does not supply.
Worked Example: Classifying a Five-Option Item
A hypothetical survey asks, “The clinic’s instructions were easy to understand,” with response options Strongly disagree, Disagree, Neither agree nor disagree, Agree, and Strongly agree. The data file records these as 1, 2, 3, 4, and 5 in that order. Classify the response variable and interpret the code 4.
Step 1: Identify the variable. For each respondent, the variable is the selected response category to this statement. Its possible values are the five named options.
Step 2: Identify the order. The categories are ordered from less agreement to more agreement. Therefore, the variable is ordinal categorical.
Step 3: Interpret the code. A recorded value of 4 denotes “Agree.” It does not mean that the respondent has four units of agreement, or that this response expresses twice as much agreement as a response coded 2.
The data-entry codes preserve the rank order, but the item remains an ordered categorical variable. The response categories, not the appearance of the codes, determine this classification.
A useful justification is to consider a different set of codes for the same responses. Suppose the five ordered options are coded 0, 1, 2, 4, and 9 rather than 1, 2, 3, 4, and 5. The ranking is unchanged: the fifth response is still above the fourth, and so on. But the numerical differences have changed. Under the first coding, the difference between the last two categories is \(5-4=1\); under the second, it is \(9-4=5\). Both codings represent the same ordered answers. Therefore, the response ordering alone cannot establish a meaningful numerical distance between categories.
Single Items and Combined Scores
A single item records one answer to one prompt. A multi-item rating scale instead uses responses to several related prompts to construct a respondent-level score. For example, a questionnaire about confidence in using a service might ask about finding information, completing a form, and asking for help. If the scoring rule adds the item codes, the resulting value is a summed score.
The score is not another response category to one question. It is a derived variable calculated from several recorded responses. Its meaning depends on the items and the scoring rule. Before calculating such a score, the rule should specify which items are included, how each item is coded, whether any item is reverse-coded, and how item responses are combined.
A sum can be useful as a compact summary, but it does not follow merely from the arithmetic that the score measures a construct perfectly or has equal intervals. Treating a summed score as approximately quantitative is a convention used in some analyses, not a fact guaranteed by the response options. That interpretation is more defensible when the items are intended to measure the same underlying characteristic, are coded in a consistent direction, and have a scoring rule established in advance. Even then, the score should be described as a score, not as a direct physical measurement.
Building and Interpreting a Summed Score
A common scoring issue arises when items point in opposite directions. If a high response means more of the intended characteristic on some items but less on another, the item with the opposite direction must be recoded before adding, if the scoring plan calls for higher total scores to represent more of that characteristic. For a five-category item coded 1 through 5, reverse coding can be done by replacing each value \(x\) with \(6-x\). This maps 1 to 5, 2 to 4, 3 to 3, 4 to 2, and 5 to 1.
This calculation is a scoring convention: it reverses the order of the codes so the item points in the intended direction. It does not prove that the distances between response categories are equal. The scoring plan must make clear which direction a high score represents.
Worked Example: Calculating a Four-Item Score
Suppose a hypothetical questionnaire has four five-category items, each coded from 1 to 5. Items A, B, and C are worded so that higher responses represent greater confidence. Item D is worded in the opposite direction, so its score is reversed using \(6-x\). A respondent records A = 4, B = 3, C = 5, and D = 2. Calculate the summed score and state its range.
Check the scoring rule. Each item has five ordered options coded 1 through 5. The scoring plan specifies that D must be reversed, so its scored value is:
Add the scored values. The other three items keep their original codes, so the total is:
Each of the four items contributes at least 1 and at most 5. Thus the minimum and maximum possible totals are:
The respondent’s score is 16 on the specified four-item scoring rule, whose possible range is 4 to 20 points. Given the direction of the coding, a higher score represents responses that are more aligned with greater confidence. The score does not mean “16 units of confidence,” and this one result alone does not show that confidence is measured without error.
This example depends on its stated rules. If item D were not supposed to be reversed, or if the scale used a different number of response options, the calculation would differ. Reversing an item is not something to infer from its numeric code alone; the wording and intended direction of the item must be checked against the scoring plan.
What a Total Can Hide
Combining answers into a sum compresses information. Different patterns of responses can produce the same total. A sum therefore cannot be read as a complete description of how a respondent answered each question.
Worked Example: Equal Totals, Different Response Patterns
Two hypothetical respondents answer the same four items, all coded from 1 to 5 in the same direction. Respondent A’s responses are 5, 5, 1, and 1. Respondent B’s responses are 3, 3, 3, and 3. Compare their totals and explain what the total does not reveal.
For respondent A:
For respondent B:
Both respondents have a summed score of 12, but their answer patterns differ. A selected the highest category on two items and the lowest on two; B selected the middle category on every item. The common total records the same sum under this scoring rule, but does not make their item-by-item responses identical. A careful description can report the total while retaining the individual responses when their pattern matters.
The total also depends on the set of items. A score of 12 from four items coded 1 through 5 is not directly interchangeable with a score of 12 from a different number of items or a different coding scheme. The score’s range and construction are part of its definition and should accompany its interpretation.
Why the Distinction Matters
Classifying a single item as ordinal prevents a common mistake: assuming that numerical labels automatically license arithmetic interpretations. It also encourages a clear account of what was actually observed. “The response was Agree” describes an item response. “The four-item score was 16 out of 20 possible points” describes a derived score under a particular rule. These statements refer to different variables and should not be substituted for one another.
When reading a rating questionnaire or its results, ask:
- What exact statement or question does each item present?
- What are the response categories, and in what order?
- Are numerical codes merely labels for those categories, or does a stated scoring procedure combine them?
- If several items are combined, which are included and how are opposite-direction items handled?
- What does a high resulting score mean under the stated rule, and what information is lost by combining answers?
A frequent pitfall is to say, “The answer is 4, so it is a quantitative measurement and is one unit above 3.” The correction is that, for a single Likert-type item coded in order, 4 identifies the fourth response category. It is one code higher, but the format does not establish that the category is one equal unit higher in the measured concept. A composite total may be useful, but its construction and interpretation must be stated rather than assumed.
Check Your Understanding
Try these questions without looking back.
- A response item has the categories “Never,” “Sometimes,” “Often,” and “Always,” coded 1 through 4. What is the variable type, and what information does the ordering provide?
- Why does coding the five responses to an agreement item as 1 through 5 not establish that the difference between 1 and 2 equals the difference between 4 and 5?
- A researcher adds responses to six related items. Is the resulting total a single item response or a derived score? Name two scoring details that should be specified.
- For a five-option item coded 1 through 5, a reverse-worded response is recorded as 4. Under the rule \(6-x\), what is its reversed score?
- Two respondents have the same summed score but different item responses. What does this show about information retained by a total?