Introduction
It is very common for surveys, especially the ones that aim to understand people’s opinion, to include questions in the format of Likert Scales. This includes studies regarding the impression and willingness to purchase a new product, to election polls or employee satisfaction. Likert Scales can be applied to a wide area of studies and hence it is the most typical question format you would expect to find in a questionnaire.
However, although researchers do not hesitate to include Likert Scales in their studies, they often find it difficult to present and interpret the results, at least in a comprehensive way. The aim of this article is to provide suggestions of how to analyze Likert Scales from a practical based view rather than theoretical.
What are Likert Scales
Firstly, what are Likert Scales and where did the name originate? Although Likert may refer to “like”, it is actually derived from the name of the developer; Rensis Likert. In 1932, the psychologist was interested in measuring people’s opinions and attitudes and developed a 7-point agreement scale that is very similar to the ones used today.
Likert scales technically are ordinal data for which strictly speaking the measures of mean or standard deviation cannot be applied because of the assumption of equal distance between categories is not valid. That is, the distance between “Strongly Agree” and “Agree” theoretically is not the same as the distance between “Agree” and “Neutral”. However, it is widely acceptable to apply parametric statistics for Likert Scales, especially for the cases where the scale is wide (i.e. more than 3 points) in order to take advantage of the depth of the information collected.
Common examples of 5-point Likert Scales are:
| Set | Example Question | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|---|
| Agreement | This article is useful | Strongly Disagree | Disagree | Neutral | Agree | Strongly Agree |
| Satisfaction | Are you satisfied with your job? | Not at all Satisfied | Not Satisfied | Moderately Satisfied | Very Satisfied | Completely Satisfied |
| Importance | How important is the weight of the smartphone? | Not at all Important | Slightly Important | Moderately Important | Very Important | Extremely Important |
| Frequency* | How often do you use this product? | Never | Rarely | Sometimes | Often | Always |
| Quality | How do you evalute this service? | Very poor | Poor | Fair | Good | Excellent |
| Awareness | Do you know this brand? | Not at all Aware | Slightly Aware | Moderately Aware | Very Aware | Extremely Aware |
| Approval | Do you approve this regulation? | Strongly Disapprove | Disapprove | Neutral | Approve | Strongly Approve |
*Frequency is relevant to the topic.
Further to the 5-point scale, the most widely used scales are 3 point, 7-point or 10-point. The benefits of using an odd number point is for the Neutral answer to be located in the middle. The wider the scale, the more information will be collected but wide scales could confuse or tire the respondent. It is important for the common questions to be grouped and asked within the same section and for the answers to be clearly demonstrated especially for wide point scales. The respondent should definitely understand what a low or high score is representing. Therefore, it is recommended to display a visual of the scale or repeat the scale characteristics often.
Common method – Top Box
The most common method to analyze Likert Scales, by not just simply displaying the frequency and percentage of each point, is by using the Top Box method. Usually this refers to the Top 2 Box Score in a 5-point Likert Scale or a 3 Box-Score in a 7-point Likert Scale. The top/bottom boxes refer to the sum of the highest or lowest ratings on a scale question. Taking as an example the Agreement rating, the answers of “Strongly Agree” and “Agree” would be grouped as “Agree” and the answers of “Neutral”, “Disagree” and “Strongly Disagree” would be grouped as “Not Agree”. There is also the option to keep “Neutral” as separate answer and divide into “Agree”, “Neutral” and “Disagree”.
In order to make this more visual, let’s see an example. For the question “It was correct for England to leave European Union” the answers are:
| Strongly disagree | disagree | neutral | agree | strongly agree |
|---|---|---|---|---|
| 16 | 41 | 20 | 22 | 9 |
By taking the Top 2-Box we will get that 22+9=31 “Agree” Vs 16+41+20=77 “Do not Agree”.
Else it can be shown as: 31 “Agree”, 20 “Neutral” and 57 “Disagree”
This method technically is correct and straight forward, but it limits the interpretation of the collected data since it is grouping the answers. By doing so, the summary is hiding the strength of the agreement and the reviewer is not able to understand if most of the answers are coming from the “Strongly Agree” or just the “Agree” box. At the same time, it is not able to see if the negative answers are originated from a soft “Disagree” or from a hard “Strongly Disagree”.
If the aim of the analysis was to group the answers, then the question could be set-up in that way so that the respondent shouldn’t spend time to decide between the “Agree” or “Strongly Agree” selection. There could have been just a 3-point scale; “Disagree”, “Neutral” and “Agree”.
Pro method – Score calculation
So, what can we do? How can we take advantage of the full information in the data, analyze Likert Scales and turn the effort of answering into valuable reports? The answer is by calculating Likert Scores!
Likert Scores are calculated by multiplying each frequency with the relative index of the answer. That is by multiplying for example the “Strongly Disagree” with 0, the “Disagree” with 1, the “Neutral” with 2, the “Agree” with 3 and the “Strongly Agree” with 4. The total score can then be weighted out of 100% so that the reviewer can directly perceive this score as a percentage. For the example above, if the total respondents were 25 then if all the answers were at the “Strongly Agree” box, then the perfect score will be 25×4=100 points. Otherwise, if all the answers were at the “Strongly Disagree” box, then the minimum would be a zero score (25 x 0 = 0). Let’s see some more examples:
| Set | Strongly disagree (0) | disagree (1) | neutral (2) | agree (3) | strongly agree (4) | total | top 2-box | total score |
|---|---|---|---|---|---|---|---|---|
| Question 1 | 10 | 5 | 5 | 5 | 0 | 25 | 5 | 30 |
| Question 2 | 5 | 5 | 10 | 0 | 5 | 25 | 5 | 45 |
The dynamic of this approach is clearly shown in the difference of the scores between Questions 1 and 2 where the Top 2 Boxes are the same, but the Total Score is different. This is because the answers are nearer to the agreement in Question 2 and hence the score is higher.
It is important when calculating Likert Scores to set the lowest box to zero index in order to directly set the minimum score to zero. It is also important to apply this method to at least 5-point Likert Scales.
The Likert Score calculation is very helpful when comparing between questions, that could include products, attributes or variables. Although non-parametric tests like Man Whitney test and Kruskal Wallis test are applicable, having a unique and solid score for each question enables the researcher to clearly rank the questions and use parametric analysis.
Even though this is a powerful method of analyzing the Likert Scale, the researcher should be very cautious when analyzing the data. The weight of each box should not be biased in order to force the score to be high, for example weighting the “Strongly Agree” with 10 points and the “Agree” with just 2 points. However, the researcher has the flexibility to use different weights for each box, taking an example from Eurovision points where the top box takes 12 points, the second top 10, the next 8, and then 7, 6, 5 etc.
How not to analyze Likert Scales
The most direct way of presenting the results of a Likert Scale is to display the frequencies or percentages of each box. It is often recommended to mention the box with the top score. Nevertheless, to analyze Likert Scales in this approach may lead to misinterpretation of the result if the top box is not subject to Normal distribution or there is an obvious ascending or descending order. For example, it would be correct to mention the top box of 47% in “Strongly Agree” for the below table since there is an obvious ascending pattern towards that box.
| Strongly disagree | disagree | neutral | agree | strongly agree |
|---|---|---|---|---|
| 10 | 9 | 12 | 22 | 47 |
On the other hand, if there is an equally high percentage on the other extreme, then not adding a remark would be a wrong interpretation of the results. This can be shown in the table below where the “Strongly Agree” also receives a 47% percentage, but there is a high selection (33%) for the “Strongly Disagree” box as well. For such case it is highly recommended to reference both percentages.
| Strongly disagree | disagree | neutral | agree | strongly agree |
|---|---|---|---|---|
| 33 | 4 | 6 | 10 | 47 |
Conclusion
Likert Scale is a very useful tool in capturing and understand the opinion or intention of the respondents. It is capturing not only the pure opinion but also there can be an understanding of the distance between the opinion and its strength. Although the measure is ordinal data, by carefully treating it as interval the researcher can apply weights to boxes and calculate Likert Score and be subject to parametric tests.
For further information and thoughts just contact me.
