Background of the Likert scale
Likert Scale is one of the most used tools by market researchers when they want to evaluate the opinions and attitudes of their target audience. There are several types of measurement scales focused directly on measuring the attitudes of people, among them, one of the most used is the one we will talk about next.
What is the Likert scale?
Let us start with the name of the scale, which was developed by psychologist Rensis Likert. Likert established a distinction between a scale which materializes from a collection of responses to a group of items (maybe 8 or more), and responses are measured in a range of values.
Technically, a Likert Scale makes reference to the last-mentioned. The difference between these two concepts has to do with the distinction that Likert made between the phenomenon that is being investigated and the variables of the means of capture.
Likert Scale is a psychometric scale used mainly in market research to understand the opinions and attitudes of a consumer towards a brand, product or target market. It serves organizations to make measurements and know about the degree of conformity of a person or respondent towards a certain affirmative or negative sentence.
When responding to an item on the Likert Scale, the user responds specifically based on their level of agreement or disagreement. Frequency scales with Likert implement fixed response formats to measure attitudes and opinions.These scales allow determining the level of agreement or disagreement of the respondents.
Likert scale assumes that the strength and intensity of the experience are linear, therefore it goes from a total agreement to a total disagreement, assuming that attitudes can be measured. On this basis, there are two main types of Likert Scales: Even and Odd Likert Scale.
The answers can be offered in different levels of measurement, allowing scales of 2, 4, 5, 6, 7, 8 or 9 elements previously configured. In most cases, it is better to have a neutral element for those users who neither agree nor disagree.
- Level of agreement or disagreement of a sentence
- Frequency of completion of an activity
- Importance of a factor
- Likelihood of recommendation
- Satisfaction of using a particular product/service
Learn more about: Odd Likert Scale Questions
Examples of Likert Scales
- It is a scale that is easily application and design.
- likert items that have no clear relation to the expression.
- It offers a ranking of the opinion of the people surveyed.
- Very straightforward to answer.
- There are scientific studies that indicate that there is a bias in the scale since the positive answers always overcome the negative ones.
- There are also studies that prove that respondents tend to answer “in agreement” since it implies less mental endeavor at the time of answering an online survey.
- Difficulty in establishing precision with the amount of positive and negative responses.
Learn more about: Even Likert Scale Questions
Fundamental criteria for configuring items on a Likert scale
Items should be easily related to the answers in the sentence, regardless of whether the relationship between item and sentence is evident.
The items must always have two extreme positions as well as an intermediate item that serves as a graduation between the extremes. It is important to mention that although the most common Likert scale is that of 5 items, the use of more items helps to generate greater precision in the results.
The items on the scale should always be safe and reliable. To achieve reliability it is sometimes necessary to sacrifice the accuracy of the scale.
Likert scale and data analysis
Surveys are constantly used to measure quality. For example, surveys can be used to measure the client’s perception of the quality of the product or the performance of quality in the provision of services. Likert scales are a common classification format for surveys. The respondents provide their opinion about the quality of a product/service from high to low or better to worse using two, four, five or seven levels.
Researchers and auditors generally group collected data into a hierarchy of four measurement levels:
- Nominal data: The weakest measurement level that represents categories without numerical representation.
- Ordinal data: Data in which it is possible to sort or classify the answers, but it is not possible to measure the distance.
- Interval data: In general, whole data in which measurements of orders and distances can be made.
- Relationship data: Association between information about ordering, distance, decimals, and fractions between variables can be set up.
Data analysis using nominal data interval and ratio are generally simple and transparent. Ordinal data analyzes particularly in regards to Likert or other scales in the surveys, are not. This is not a new problem. The effectiveness of handling ordinal data as interval data continues to be debatable in survey analysis of a variety of applied fields.
The main reason why ordinal data is sometimes treated as interval data is the claim that parametric statistical tests are more powerful than nonparametric alternatives. Moreover, inferences from parametric tests are easy to interpret and provide more information than non-parametric alternatives.
However, the treatment of ordinal data as interval data without examining the values of the data set and the objectives of the analysis can mislead and misrepresent the results of a survey. To appropriately analyze scalar data, it is preferable to consider ordinal data as interval data and concentrate on Likert scales.
Basics of Likert scales
The Likert scale came into existence in 1932 in the form of the 5-point scale which these days, is extensively used. These scales range from a group of generic topics to the most specific ones that ask people to indicate how much do they agree or disagree, approve or disapprove, or believe it is true or false.
There really is no wrong way to build a Likert scale. The most important factor to evaluate is to include at least two or three response categories. The ends of the scale are often increased to create a seven-point scale by adding “very” to the top and bottom of the five-point scales. It has been shown that the seven-point scale reaches the upper limits of the reliability of the scale.
As a general rule, Likert and others recommend that it be better to use a scale as wide as possible. One can always collapse the answers into concise groups, if appropriate, for analysis.
By considering these aspects, scales are sometimes curtailed to an even number of categories (usually four) to eliminate the “neutral” option on a “forced choice” survey scale. The primary Likert record clearly states that there could be an intrinsic variable whose value marks the feedbacks or attitudes of the respondents and this underlying variable is the interval level, at best.
Analysis, generalization to continuous indices
A universal guideline suggests that the mean and the standard deviation are baseless parameters for detailed statistics when the data are on ordinal scales, just like any parametric analysis based on the normal distribution. The non-parametric test is done on the basis of appropriate median or range for inspecting data.
The Kruskall-Wallis models can provide similar results as an analysis of variance, but based on the ranges and not on the means of responses. Since these scales are representative of an underlying measurement continues, a recommendation is to analyze them as interval data as a pilot before collecting the continuous measurement.
Almost 60% of respondents recognize online learning as equal or better than face-to-face, but there is an insistent minority that believes, online learning as at least somewhat inferior. In case this information was collected and analyzed, with a scale of 1 to 5 from inferior to superior, this separation would be lost, giving means of 2.7, 2.6 and 2.7 for these three years, respectively. This would indicate a slightly lower than average agreement instead of the actual distribution of responses.
Best practices for analyzing the results of the Likert scale
Likert five-point scales are commonly associated with surveys and are used in a wide variety of settings. If a person has come across the Likert scale in case options such as – totally agree, agree, disagree or disagree, disagree or strongly disagree are asked about a particular topic.
Because the Likert element data is discrete, ordinal, and limited in scope, there has been a long dispute over the most valid way to analyze Likert data. The fundamental option is between parametric and non-parametric tests. The advantages and disadvantages for each type of test are generally described as the following:
- Parametric tests assume a normal and uninterrupted division.
- Non-parametric tests do not assume a normal or uninterrupted division. However, there are concerns about a lesser ability to detect a difference when one actually exists.
Which is the best option? This is a real decision that a researcher has to make when they decide to analyze information received from a survey where Likert Scale questions were asked.
Over the years, a series of studies that have tried to answer this question. However, they have been inclined to look at a limited number of potential distributions for Likert data, which causes the generalization of the results to suffer. Thanks to increases in the power of computing, simulation studies can now thoroughly evaluate a wide range of distributions.
The researchers identified a diverse set of 14 distributions that are representative of the actual Likert data. The computer program extracted self-sufficient pairs of samples to test all possible combinations of the 14 distributions.
In total, 10,000 random samples were generated for each of the 98 distribution combinations. The pairs of samples are analyzed using both the two-sample Parametric and the Non-parametric test to compare the efficacy of each test. The study also evaluated different sample sizes.
The results show that for all pairs of distributions the Type I error rates (false positive) is very close to the target quantities. In other words, if an organization uses any of the analysis and results are statistically significant, it does not need to be too worried about a false positive.
The results also show that for most pairs of distributions, the difference between the power of the two tests is trivial. In other words, if there really is a difference at the population level, any of the analyzes is equally likely to detect it.
There are some pairs of specific distributions where there is a power difference between the two tests. If an organization performs both tests on the same data and do not agree (one is significant and the other is not), this difference in power affects only a small minority of cases.
In general, the choice between the two analyzes is a loop. If an organization needs to compare two groups of five-point Likert data, it usually does not matter which analysis method is used.
Both, parametric and non-parametric tests, consistently provide the same security against fallacious negatives and also provide the same security against fallacious positives. These patterns are valid for sample sizes of 10, 30 and 200 per group.
Add the Likert scale in a survey with Questionpro!
The Likert-scale survey is a comprehensive technique for gauging feedback and information, which makes it significantly easy to understand and respond. This is a critical question to measure the opinion or attitude of a respondent towards a specific topic, so it will be of great help in the next step of an investigation.
With QuestionPro an organization can do an online survey using the Likert Scale. Start by creating a free account and start testing this type of question.
If an organization wants more information about this and other questions within our platform, consult our online chat, tell us about the project and we will gladly offer one of our research tools that best suitable for those needs.