Sentimental Survey : Likert Data
Sentimental Survey : Likert Data
In 1932, Rensis Likert, an American Psychologist, developed a method for measuring attitudinal scale by classifying response to a statement on a scale given as Strongly disapprove, Disapprove, Undecided, Approve, Strongly approve. This data is now known as Likert data or Likert response.8
Likert Data is on a scale, and therefore is a scale type data, and is used mostly in sentimental surveys. It is also used in other surveys where opinion matters. Likert items are used to measure respondents’ attitudes to a particular conjecture or statement. The data is usually coded, for example as follows.
•1 = Strongly disagree • 2 = Disagree • 3 = Neutral • 4 = Agree • 5 = Strongly agree
General Properties:
A)- Likert data is an ordinal data, i.e. the choices or scores are having some sense of ascend (or descend if going in reverse). Because of this inherent order in options, they are ordinal data. Thus having one score is higher than the another.
B)- The options are associated with texts, which give sense to ordinality.
C)- There is a distance of some sense between two scores or options, which is equidistant in that sense, like “Strongly disagreed” is at at a same distance in a sense from “Disagreed”, as “Strongly agreed” is from “Agreed” in that same sense.
D)- The first option is totally opposite to the last option and the middle one is a neutral type.
Apart from the above, there are some general rules also, which are followed in order to have clarity in the data collected through Likert response format. Generally in the Likert data, total number of options or scores is in odd number like there will be either 5,7,9 … etc. options, may be, even three. This is done to keep the middle one at neutral place unless the Investigator wants a “voting type” opinion forcing respondent to choose either positive or negative ones and therefore, he insists on some kind of positive or negative opinion on the statement by omitting the middle neutral option. In that case, he will frame even number of options in Likert rating scale (options of Likert response) omitting the middle neutral part. In such situations, an even-point scale is used, where the middle option of “neither agree nor disagree” or “neutral” or “undecided” is not available to respondent. This type of choices or ratings in Likert items is also called a “forced choice” method, since the neutral option is removed.
Giving an option of neutral choice, makes many people to respond like that for difficult or controversial questions because they do not want to take risk of being classified as pro or anti to the statement in question.
Further the neutral option can be seen as an easy option to take when a respondent is unsure, and so whether it is a true neutral option is questionable in the eyes of Investigator. Therefore, he chooses even number of points for options of Likert items. However, there is a drawback in such type of "forced choice options" or "voting type response" driven Likert items that they invite lot of non-response and questions go unanswered many times, causing increased non-response.
Another, general convention is that distance between options are equidistance, but that might not be true. Like in the example given below under Q2, the scale does not seem to be equidistant but this type of data can still be analyzed the same way as that is used for Likert items data. So, though it is not Likert item in true sense, yet it, still, is called as Likert-type data.
Thirdly, generally zero is not opted for a value of options in Likert options. This is done to avoid ambiguity arising in analysis due to the value being zero for an important observed data option in the rating scale of a particular Likert item.
Distinguishing features:
The following is an example of Likert item, as it is ordinal, equi-spaced, 1st one opposite of last one and middle one is neutral type.
Q1 Regular exercise is essential for a healthy body.
(1) Strongly disagree (2) Disagree (3)Neutral (4) Agree (5) Strongly Agree
Now a example which is not Likert item but can be analyzed like Likert item. They are generally termed as Likert type.
Q2 Do you drink?
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often
Though above example is ordinal in nature, to some extent in some sense, it is equi-spaced too. Middle one, the option “(3) Average”, is neutral type as both ends revolve around it. However, the first and the last ones, the anchors of total spectrum of response, are not totally opposite to each other, as it was there in question Q1. Strictly, it might not be Likert item, but this example can still be analyzed by the same tools which are used for analyzing the Likert items. Therefore, we call such items Likert type item instead of Likert item.
Now another example,
Q3 Do you smoke?
(1) Never (2) Once in a while (3) One per day (4) One to six per day (5) More than six per day
Though ordinal in nature, it is not equi-spaced as per associated texts with options. The first and last ones are not opposite to each-other. Middle one is not neutral type. So it is better to analyze this data by methods other than those employed for Likert data.
Q4 How was training ?
Not Helpful 1□ 2□ 3□ 4□ 5□ Helpful
In the above example, categorical ordinal options are used, no text is attached to options. This is also not a Likert item, though there is a rating scale attached for the response spectrum. Better it is to be analyzed by other categorical methods.
There is a literature available for pros and cons of this type of data, Likert and Likert type items, and their analysis. One good article is “Designing and Analyzing a Likert Scale | Guide & Examples published on July 3, 2020 by Pritha Bhandari” at www.scribbr.com.
Examples:
Some examples of Likert items, sometimes called Likert ratings and also loosely called Likert scales, though "Likert Scale" is different from Likert items, are as follows:
A)- Feelings
Example: How happy were you with your stay at our hotel?
1)Very satisfied 2)Somewhat satisfied 3)Neither satisfied nor dissatisfied 4)Somewhat dissatisfied 5)Very dissatisfied
B)- Frequency of behaviors
If one is looking for how often a consumer purchases a product, repeats a particular action, say purchasing, reading, smoking etc, Likert scale can help uncover these behaviors.
Example: How often do you read articles on your phone vs in a newspaper?
1)Much more 2)Moderately more 3)About the same 4)Moderately less 5)Much less
C)- Agreement statement
Perhaps the most popular Likert scale in survey questions is a scale of agreement-disagreement, where a respondent is asked to select the answer that best reflects their belief about a statement provided.
Example: Please select how much you agree or disagree with the following statement: Cats make better pets than dogs.
1)Strongly agree 2)Somewhat agree 3)Neither agree nor disagree 4)Somewhat disagree 5)Strongly disagree
D)- Pain or Enjoyment or Meaningful
Ex.D.1: How much physical pain you feel ?
Ex.D.2: How much do you enjoy your life ?
Ex.D.3: How much you feel that your life is meaningful?
1)Not at all 2)A little 3)A Moderate amount 4)A very much 5)An extreme amount
Positive and Negative Question:
Note here, the questions (D.1) is a negative question, while (D.2) and (D.3) are positive question. In positive questions, when the scale increases from 1 to 5, feeling is that of better one, means ordinality is in the same direction, while in negative question, the ordinality is in opposite direction. In question (D.1), pain increases as scale increases, therefore the better feeling is towards lower side of scale. This point is important for the sake of combining the scores of the many Likert items into one for measuring a quality which is governed by many such Likert items as its factors. While combining many likert items into one, all Likert items scores should be transformed properly so as to have same direction as per the combined measuring aim for which they are being combined.
E)- Measurement of quality
Example: How would you rate quality of service?
1)Very poor 2)Poor 3)Neither poor nor good 4)Good 5)Very good
F)- Measurement of satisfaction
Example: How much you are satisfied with your preparations?
1)Very dissatisfied 2)Dissatisfied 3)Neither dissatisfied nor satisfied 4)Satisfied 5)Very satisfied
Biases of Likert items:
Responses on Likert items are generally found having following biases.
》Respondents avoid using extreme response categories (central tendency bias), because s/he does not want to be categorized as having extremist views
》Respondents avoid response to the statements (acquiescence bias) because of culture of institutionalization that encourages and incentivizes eagerness to please people of their community, institution or culture.
》Respondents disagree with statements out of having a defensive desire to avoid negative consequences that respondents may fear will result.
》Respondents provide answers in faking good mode so that the response should be evaluated as showing strength or lack of weakness.
》Respondents provide answers in faking bad mode so that the response should be evaluated as indicating weakness or presence of impairment or pathos in order to draw benefit out of it.
》Respondents portray themselves or their organization either in faking good or faking bad defying their own true beliefs for having social desirability or preventing norm defiance.
Despite having all the above biases, this method of having Likert items is very much prevalent in market, social and psychometric studies or suyveys, because they help in measuring quantitatively the qualitative characteristics, traits, attitudes, sentiments, themes or echo systems.
Qualitative research & Likert items:
Qualitative research is very popular these days as they unfold many thematic characteristics by pointing to physical parameters of causations defined inside such statements.
To know how the themes or echo systems characteristics are measured through Likert items, we should discuss Likert Scale too.
The word “Likert scale” is used in two sense. One, the options given in Likert items are called Likert scale in a very loose sense. Likert scales commonly have 5 or 7 items, and the items on each end are called response anchors. Some people call these options as ratings of the Likert item.
Likert scale and Rating scale:
Although these options of Likert items are loosely called Likert scales, but they are essentially special kind of rating scales, the opposite “that all rating scales are Likert scales”, is not necessarily true. A Likert scale is a specific type of rating scale that exclusively focuses on a range of answers on a spectrum, while a rating scale can consist of any number of rating choices, such as stars like rate by choosing stars ⭐⭐⭐⭐⭐ or numeric responses like shown as "Not Helpful 1□ 2□ 3□ 4□ 5□ Helpful" in discussion mentioned above under Q4, but these rating scales cannot be termed as Likert scale or options of Likert item, or more accurately a Likert item scale, unless equipped with the basics of Likert options viz ordinal nature, associated with texts, preferrably equi-spaced, spread over spectrum bounded by extreme anchors, preferably extreme anchors be opposite to each other, middle one preferably be neutral type.
Likert item scale or Likert scale:
Conventionally, the Likert item scale, or Likert scale or Likert options are kept in ascending order of the sense of the texts associated with them, so as to make analysis & comprehension simple. It simply means that more the value of option, more is the sense of text attached with them. Thus "1)Strongly agree 2)Somewhat agree 3)Neither agree nor disagree 4)Somewhat disagree 5)Strongly disagree " is not preferred scale as in this scale "agreement " is decreasing from 1 to 5. Though number attached is increasing, yet the sense of text, which is here "agreement", is decreasing. Therefore, conventionally, the scale "1)Strongly disagree 2)Somewhat disagree 3)Neither agree nor disagree 4)Somewhat agree 5)Strongly agree " is preferred. It will ease analysis. Other valid example is "1)Very dissatisfied 2)Dissatisfied 3)Neither dissatisfied nor satisfied 4)Satisfied 5)Very satisfied" as sense of attached text, "satisfaction" and numbers are in ascending order. Similarly "1)Not at all 2)A little 3)A Moderate amount 4)A very much 5)An extreme amount" is conventionally valid because underlying text sense "how much" is in the ascending order along with numbers attached with them. In the same spirit, "1)Very poor 2)Poor 3)Neither poor nor good 4)Good 5)Very good" is also valid because underlying text sense "quality" is in the ascending order along with numbers attached to it.
Likert Scale:
When we want to use word “Likert Scale” in its strict sense, it means combining many Likert items into one thematic concept, which overall may also be having options of the same kind.
“Likert Scale” is, thus, the sum of responses of several related or contributory Likert items, whereas a “Likert item” is a statement that the respondent is asked. Likert item is at a more granular level. The Likert item is simply a statement, which respondent evaluates. He gives a quantitative value to it by choosing one, out of many (generally 5 or 7) level of rating scale having some subjective or objective texts attached to such options. These texts, in some literature, is also named as dimensions, the most popular and most commonly used is “disagreement/ agreement” as is given below :
•1 = Strongly disagree • 2 = Disagree • 3 = Neutral • 4 = Agree • 5 = Strongly agree
Likert Scale is combining the scores of many such related Likert items, which, theoretically or axiomatically , are considered connected with the theme for which measurement is being devised. Note here that on Likert item, data or response is collected, while in Likert Scale, the value is derived through computation from values gathered in the associated Likert items.
Likert Scale is generally the sum or average of Contributing responses collected on several constituent Likert items on a proper scale.
Likert Scales can also be combination of Likert Scales too or proper combination of Likert Scales and Likert items.
Therefore, Likert data or Likert items and Likert Scales are two different things.
Likert Scale can be used in customer satisfaction surveys to determine how customers felt about their experience, for a product, or service.
Workout of Likert Scale from Likert items:
To further understand, let us take an example.
Say we have to find an answer to "how do you rate quality of life." One method is just ask it on a rating scale 1 star, 2 star, 3 star … 5 star. We can also ask it in the form of Likert item as ○very poor, ○poor, ○average, ○good, ○very good. Note here the difference between a "general rating measuring" (1 star⭐, 2 star⭐⭐, 3 star⭐⭐⭐ etc) and "Likert rating measuring" (○very poor, ○poor, ○average, ○good, ○very good). Likert rating measuring is a special kind of rating where the response is given on a Likert scale, wheras rating is a general term for capturing response.
If we want to measure Likert rating with more insights into it, then we have to generate Likert items for the factors of our question "how do you rate quality of life". For example, say there are five factors or dimensions on which we want to devise Likert measuring for the question:
1) To what extent pain is felt that prevents from doing essential work.
2) How much medical help is needed for keeping your life running properly.
3) How much life is being enjoyed.
4) To what extent life is being felt meaningful.
5) To what extent you feel financially independent.
These five questions, whether positive or negative, can suitably be combined to arrive at a combined score as a measure to the Likert Scale for the question "how do you rate quality of life."
Let these Likert items are framed like:
A) To what extent pain is felt that prevents from doing essential work.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often
B) How much medical help is needed for keeping your life running properly.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often
C) How much life is being enjoyed.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often
D) Do you agree that your life is meaningful.
(1) Strongly agree (2) agree (3)Neutral (4) disagree (5) Strongly disagree
E) Do you feel yourself financially independent.
(1) Strongly disagree (2) disagree (3)Neutral (4) Agree (5) Strongly Agree
Let the options chosen in (A) is "a", (B) is "b", (C) is "c", (D) is "d", and (E) is "e". Let us see how they contribute to our main thematic question (Q) given by "how do you rate quality of life", measured by a scale as "very poor (lowest) to very good(highest) as [very poor, poor, neither poor nor good, good, very good ]". If we have more pain, that is heigher value of "a", then quality of life should be towords poor. So for "a", we should take contribution as (5-a+1) for question Q. When we consider B, lesser medical help need will rate quality of life to better side. Thus for B=b, the contribution to Q will be (5-b+1). For C, the contribution to Q will be "c" itself, because heigher enjoyment will be higher contribution to Q. For the factor D, higher meaningful life will contribute better rating for life for Q. But here, in D, the D is being measured in opposite direction, as scale is written in opposite direction. So, contribution for each "d" of D will be (5-d+1). Lastly for fifth factor of financial independence E, the higher the financial independence, the better will be rate of quality of life. Thus for E, for each value of "e", contribution will be "e" itself. Thus measure "q" for Q, based on A, B, C, D, E will be
q = (5-a+1) + (5-b+1) + c + (5-d+1) + e.
This q can be computed from values of all the five Likert items A to E. This is aggregating method for finding Likert Scale and this "q" for question Q, will range from 5 to 25, where very poor rate of life will have score 5, and very good rate of life will have score as 25.
Another measure for thematic question (Q) given by "how do you rate quality of life", measured by scale as "very poor (lowest) to very good(highest) can be given by average formula as
h = sum of contribution / factors = q/5 = ( (5-a+1) + (5-b+1) + c + (5-d+1) + e ) / 5.
This "h" will be in the range of one to five. But , this "h" can take decimal values too. So it becomes a continuous variable and is able to measure question Q (rate of quality of life) in between the very poor and poor scale points too.
If we do this way as done in A to E above, there are two type of adjustments, we have to do. Firstly , if the factors A to E are positive or negative to original thematic question Q and then whether the factors are being measured in ascending order or descending order in the spectrum of response. Such arrangements might become cumbersome in analysis and may create confusion to figure out the real insights. So what we do, conventionally, is to frame questions in proper ordinality unless it is imperative on the part of Investigator to ask the question in a way, where normal ordinality rules can not be insisted upon. What are these conventions. We feel better if increase in the score from lower value to higher value fetch more underlying text sense. Therefore, answer to "how do you rate quality of life." in a Likert item as ○ 1.very good, ○ 2.good, ○3.average, ○4.poor, ○5.very poor. may be reframed to as ○1.Very poor, ○2.Poor, ○3. Average, ○4.Good, ○5.Very good This way of measuring the quality of life will mean, higher score we have, better will be the quality of life.
We can imagine that there is some kind of magnitude in the scale, we see that this weight is increasing as the associated numbers of the options are increasing. With this logic, we will modify scale of question (D) from "(1) Strongly agree (2) agree (3)Neutral (4) disagree (5) Strongly disagree" to "(1) Strongly disagree (2) disagree (3)Neutral (4) Agree (5) Strongly Agree". Thus magnitude of agreeing is increasing in the order of associated numbers of the options. Once reframed this way, it gives ease to analysis. However, after having framed like this too, we have to differentiate positive questions and negative questions in order to consolidate them into a combined score for the overall question or theme. Reframing this way, making options in the ascending order of underlying text sense (uts), we have questions rewritten as :
A) To what extent pain is felt that prevents from doing essential work.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often. [uts - feeling]
B) How much medical help is needed for keeping your life running properly.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often. [uts - frequency]
C) How much life is being enjoyed.
(1) Never (2) Sometimes (3) Average (4) Often (5) Very Often. [uts - quantity]
D) Do you agree that your life is meaningful.
(1) Strongly disagree (2) disagree (3)Neutral (4) Agree (5) Strongly agree [uts - agreement]
E) Do you feel yourself financially independent.
(1) Strongly disagree (2) disagree (3)Neutral (4) Agree (5) Strongly Agree. [uts - agreement]
After doing this, contribution to question Q for rating of quality of life, only the thing that we need to look into is positivity and negativity of the factors for thematic question Q. We need not to look into ordinality of the scale, because all scales are in ascending order of their underlying text sense. In this rewritten (A) to (E) factors, we have "pain" as in (A), "needed medical help frequency" as in (B) are the negative factors for question (Q), others are positive ones. Therefore, measure "q" of Q can be defined as
q = (5-a+1) + (5-b+1) + c + d + e.
And h = q/5
Using Likert Scale in a matrix question type
Matrix style question types allow Likert scales to be used to ask about many items in one go at a place. Suppose we want to provide an insight about how much good a hotel is, we may have factors like price, parking, condition of bedroom, commute facility from transit point, nearby markets, food nearby etc. We may have likert items on all this on rating scale of dis-atisfactory/ satisfactory. These all can be combined into matrix as
