Likert Scale and other scales designed to measure attitudes, such as satisfaction, are ubiquitous in market research. They have their uses that is for sure, but there are a few caveats one should be aware of. In this post we will start this review by looking at number of scale points.
As researchers we have to maintain the balance between our client’s need for information and our respondents’ valuable time. In short, think about the scales you plan to use and ensure they are the right ones for the job. One of the first items to consider is should our scales have an even or odd number of points? Scales with an even number of points remove the “neutral” option – they force an opinion. This can make it more difficult on the respondent. Even scales are acceptable when we need an opinion and when we honestly believe the respondent can provide one. Testing imagery, such as a logo, is a valid use case for even scales.
Scales with an odd number of points (5, 7, 9, or 11) provide the respondent with a soft landing place if they truly do not have an opinion. Unlike the previous example, if you are asking a respondent to estimate the likelihood they will make a significant purchase during a specified time period – they may not truly know. In this case a soft landing neutral (or “unsure”) would be appropriate.
When it comes to measuring agreement we have to consider the possibility that respondents tend to agree more often than they would otherwise. This is known as acquiescence bias. Research on research shows that certain populations (lower education and lower-income) and some cultures are less likely to disagree. This bias alone has turned some researchers away from agreement scales altogether. At the minimum, you should test for the risk of bias in populations of interest.
If you are drawn to agreement scales consider an odd number of points, unless you are working with imagery and need a concrete decision. From a bias perspective test data results across key populations to assess the potential for acquiescence. If the results prove significant then they should be noted in your findings. From a question design standpoint, you can lead with “To what extent do you agree or disagree with the following items,” as a means of reminding the respondent that it is okay to disagree.
In future posts, we will examine other potential concerns with scales and alternatives to common agreement scales. Finally, we invite you to read our article: Sample vs Population.
