Likert Scale

Why do we need a Likert scale? Let's assume you have formulated a hypothesis in which a variable that cannot be measured directly, a so-called latent variable, occurs. So how do you measure this variable in your questionnaire?

Latent variables are, for example, physical well-being or environmental awareness. Why are these variables not directly measurable? Of course, you could simply ask in your questionnaire, "What is your level of environmental awareness?"

However, you will find it very difficult or very inaccurate to measure environmental awareness with just this one question, because environmental awareness consists of several dimensions.

The first person thinks, "I am in an animal welfare organization and take care of animals in need," and ticks "very high," but also makes a long-distance trip by plane three times a year.

The next person thinks, well, my awareness of the environment is very high, I know that you should actually separate the garbage and not drive the car all the time, so my awareness of the environment is high, but I don't comply. Does that person then tick "high" or "low"?

You can, of course, just ask this single question to measure environmental awareness, but you won't be able to do much with the answers.

Scales

In order to be able to adequately measure latent variables such as environmental awareness, scales are used. A scale is a group of questions, such as questions 1, 2, 3, and 4, with response categories ranging, for example, from "strongly agree" to "strongly disagree." Together, these questions serve to measure a latent variable.

By using multiple questions, your measurement becomes more accurate. Such a scale is also called a multi-item scale.

The best-known scale is the Likert scale. This is a method used to measure personal attitudes. A Likert scale consists of several statements, called items. For each item, respondents indicate how strongly they agree or disagree using an ordered set of response categories. A single statement with these response categories is called a Likert-type item; the combined set of several items forms the Likert scale.

Often, four or five response categories are used.

4-level rating:

• applies (1)
• rather applies (2)
• rather does not apply (3)
• does not apply (4)

5-level rating:

• applies (1)
• rather applies (2)
• partly applies (3)
• rather does not apply (4)
• does not apply (5)

Whether a middle category is useful depends on the research question and context.

In some cases, it makes sense for respondents to have a "neutral" middle category available to avoid being pushed in one direction. On the other hand, there is a risk that many respondents will choose the middle category, which could make your results less meaningful.

Question with a Likert scale

Now the next question is, how do I come up with the right questions? Do I just make up the questions myself?

This is not a good idea, because in that case you would have to go to great lengths to check whether your questions really measure the latent variables correctly.

A better approach is to search the literature and find out whether there are already validated questions that you can use to measure your variables.

What is the easiest way to start a literature and scale search? We just start "quick and dirty" and search online for "environmental awareness questionnaire."

Likert scale and indexing

So, now you send out your questionnaire and have it filled out. Let's assume that you have received enough responses and are finished with the survey. What do you do with the answers? Let's take a look at that now.

So that we don't have too much text, let's just use this example again:

The participants of your survey have ticked something from "agree" to "disagree" for questions 1 to 4. Now you can export the finished survey to Excel or numiqo, then each row shows a respondent and the columns show the 4 questions. The first person has ticked "2" for question 1, "1" for question 2, "3" for question 3 and "2" for question 4.

Now you want to measure your latent variable with these four questions. The response to each individual Likert-type item is ordinal: the categories have an order, but the distance between adjacent categories is not guaranteed to be equal. A score created by combining several items is often treated as approximately metric in practice, provided that the items measure the same construct and are coded in the same direction.

A common approach is to build a mean index by calculating the mean of each row. This gives each respondent a score for the latent variable. Before doing so, any reverse-worded items must be recoded so that higher values have the same meaning for all items.

You can also easily create the mean value index with numiqo. To do this, copy your data into this table and click on "Prepare data". Now click on "Index formation" and select questions 1 to 4. Then click on "Mean value index" and enter the name for the new variable - done! You can then use Cronbach's alpha to assess the internal consistency of the items. A high alpha alone, however, does not prove that the items measure only one latent variable.