I assume you are meaning your Likert items are proposed to have 3 points. Do you have an idea about the expected distributions of each of the items? How many items do you have for your scale?

How can using a 3-point scale make your analysis any easier than a 5-point scale? In general, survey respondents prefer to have more options, rather than to be pushed into a narrower set of response categories.

There is no easy answer, but there is an interesting post here: https://www.rdatagen.net/post/the-advantage-of-increasing-the-number-of-categories-in-an-ordinal-outcome/

Note that the writer's simulations suggested that power increased sharply from two to five categories, increased a little more with six but thereafter not much happened. And this simulation study suggests that the optimum number of categories is between four and seven Article Effect of the Number of Response Categories on the Reliabili...

However, using actual parallel forms of the same questionnaire, these authors found that there wasn't a lot to be said for more categories : Article Optimal Number of Questionnaire Response Categories More May...

What the simulation studies cannot take into account is the effect of different numbers of options on test-taking because they start out by taking a distribution and slicing it up into categories. In real life, test-taking behaviours are likely to be influenced by the number of options and the ease with which respondents can distinguish between them.

More response anchors/options could enhance a Likert scale’s internal consistency and test-retest reliability. Therefore, you better go for a 5-point Likert scale. You might find the article by Simms et al. (2019) insightful.

Simms, L. J., Zelazny, K., Williams, T. F., & Bernstein, L. (2019). Does the number of response options matter? Psychometric perspectives using personality questionnaire data. Psychol Assess, 31(4), 557-566. https://doi.org/10.1037/pas0000648

Good luck,

The implication of limiting the Likert scaling to 3 is limiting the options of opinions to the research respondents to 3. This is a near close ended question. A close ended question is that which has a YES or NO answers. Another implication for limiting the answer options to 3 is the likely hood of running into a type1 error in your research. This is an error, which misinforms the researcher. That is, if the researcher did not get the truth from the research respondents. By limiting the answer options to 3, you are forcing the research respondents to be within the confinement of the 3 options. Rensis Likert Scaling of at least 4, in my view is just okay

You can always choose to collapse some of the answers (e.g., combine strongly agree with agree, and strongly disagree with disagree), so you can have a simple analysis regardless of whether you have 3 or 5 responses. There are two big differences. First, with 5 categories (or more), you have more information available to you, which may be helpful later on (especially if the collapsed categories show little variation). Second, with more categories available, people are less likely to choose the middle option, which could be either good or bad, but generally you benefit from getting people to express an opinion rather than being neutral.

Thanks, Mohialdeen Alotumi , for that interesting reference, which is posted on RG for those who don't have institutional access…

Article Does the Number of Response Options Matter? Psychometric Per...

Hi Jemima,

I am also curious about what type of analysis will you be using? Also, are you going to treat these variable(s) as continuous or categorical?

If you want to treat them as linear, with a three-point Likert scale (and considering the number of items), your Cronbach's alpha might be lower than expected.

Thank you.

5-point scale gives the respondents options to choose; and improves the reliability of your analysis. However, depending on the objectives of the research, 3-point scale should be matched with a large sample size. Regardless, 5-point scale is preferable to respondents and for data analysis.

Note: You should also take note of the non-parametric analytical tool to ensure a reliable result

Have you performed a reliability test for each item in your constructs?