Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 16

DefINITIoN of MeasuReMeNT

Measurement may be defined as the assignment of numbers to characteristics of objects or


events according to rules. A point worth emphasising is that the definition of measurement does
not suggest the measurement of objects or events but the characteristics of interest in the
concerned
Measurement and 142
Scaling
objects or events. We do not measure a buyer or a product, but we measure a certain
characteristic of a buyer or a product. For instance, a buyer’s preference usage rate, income, or
attitude can be measured. Likewise, in respect of a product, we can measure its size, colour or its
suitability for a particular purpose.

Types of scales
There are four types of measurement scales. Table 9.1 shows characteristics of such scales.

Table 9.1 Characteristics of Measurement Scales


Scale Number System Marketing Phenomena Permissible Statistics
Nominal Unique definition of numbers Classification: Brands Percentages Mode
(0,1,2,…9) Male-female Store types Binomial test
Sale territories Rankings Chi-square test
Ordinal Order of numbers Attitudes Percentiles
(0<1<2….<9) Preference Median
Occupation Rank-order correlation
Social Class Two-way ANOVA
Interval Equality of differences Attitudes Range
(2-1=7-6) Opinions Mean
Index numbers Standard deviation
Product-moment correlation
Ratio Equality of ratios (5/10=3/6) Sales (Units/Rupees) Geometric mean
Units produced Harmonic mean
Cost Number of customers Coefficient of variation

Source: Kinnear, Thomas C. and James R. Taylor: Marketing Research: An Applied Approach (International Edition),

Singapore, McGraw-Hill Book Company, 1987, p. 292.

Nominal scales
Nominal scales are more widely used than any other scale for research in social sciences. In such
a scale, the numbers serve as labels to identify persons, objects or events. Thus, numbers may be
assigned to students in a class or patients in a hospital. We might further use the nominal scale
by counting students with a certain characteristic or attribute such as those who reside in the
university hostels and others. In a nominal scale, we split a set into subsets which are mutually
exclusive and collectively exhaustive. Consider the following exampe:
Number of Students

Undergraduates Postgraduates Total


Day scholars 800 200 1000
Hostellers 400 150 550
Total 1200 350 1550
Measurement and 143
Scaling
In this example, students have been identified and counted by two characteristics, namely,
whether they are pursuing an undergraduate or post-graduate course of study and their place of
residence. The numbers given in each of the four cells are mutually exclusive and the total of
these four cells gives the total number of students. In a nominal scale, the only operation
involved is the counting of numbers in each group. It is, therefore, the simplest of the four scales
and also the least powerful. The scale does not show any order or distance relationship nor does
it have any arithmetical origin. In view of these limitations of a nominal scale, it is unsuitable in
determining relationships but is very useful in preliminary or exploratory work, where it is
sufficient to know the broad dimensions of a certain phenomenon.

ordinal scales
Ordinal scales, as the name implies, are ranking scales. Besides having the unique characteristics
of the nominal scale, these scales also indicate the order. This is possible when one is able to
distinguish elements on the basis of a single direction. For example, one may rank two or more
households according to their annual income or expenditure. Suppose we have five households
with annual incomes as shown below:
Household A B C D E
Income (Rs) 6,000 10,000 4,800 12,000 11,000

If the household with the highest income is to be given No. 1 and the next to it as No. 2, and
so on, then the following order will emerge:

Household Order of households on the


basis of annual income
A D
B E
C B
D A
E C

This is the use of an ordinal scale, which involves the ordering of households on the basis of
their annual income. The point to be emphasised is that it is a mere ordering and does not
indicate the differences in annual income amongst the five households. On the basis of this scale
one cannot say whether the difference in annual income, between two households D and E, is
greater than, less than, or equal to the difference between any other two households, say, E and
B.
The use of the ordinal scale is possible when one is able to distinguish a certain product on
the basis of a particular attribute. The above example was numerical, dealing with the annual
income of households. It was simple as it did not involve any difficulty in ordering. Consider
another example where acceptability of a soft drink is involved. Here, one can ask the
respondents questions on the basis of one or more attributes such as flavour, colour, etc.
Respondents may be asked to indicate whether they like the soft drink or not. One can develop a
five-point scale such as given below:
I strongly like it +2
I like it somewhat +1
Measurement and 144
Scaling
I am indifferent 0
I dislike it somewhat 1
I strongly dislike it 2
In this manner, ranking can be obtained by asking respondents their level of acceptability. One
can then combine the individual rankings and get a collective ranking of the group.

Interval scales
The third type of scale is the interval scale. It possesses not only the power of the nominal and
ordinal scales but also additional strength which is the determination of the equality of
differences. The classic example of an interval scale is the measurement of the temperature.
Both the Fahrenheit and Centigrade scales belong to this type. One can say, on the basis of this
scale, that a temperature of 100 degrees is 20 degrees warmer than 80 degrees and 20 degrees
cooler than 120 degrees. It may be noted that differences between two values, say, on a
temperature scale are multiples of each other. Thus, the difference between 40°F and 20°F is
half the difference between 60°F and 20°F. Using the conversion formula from Fahrenheit to
Centigrade.
Degrees centigrade = 5/9 (Degrees Fahernheit – 32), we can find the corresponding temperatures
in Centigrade.
5
40° = (40 – 32)
95
40
= ¥8=
9
5 = 4.44°C
9
20° = (20 – 32)
95
60
= ¥ (– 12) = = – 6.66°C
– 9 9
40° – 20° = 4.44 – (– 6.66)
= 4.44 + 6.66 = 11.10°C
5 5 140
60° = (60 – 32) = ¥ 28 = = 15.55°C
9 9 9
60° – 20° = 15.55 – (– 6.66) = 15.55 + 6.66
= 22.21°C which is approximately twice of 11.10°C
The above example shows, that on a particular scale, equal differences indicate equal
differences in value with regard to that scale only.
Interval scales are more powerful than the nominal and ordinal scales. Also, they are quicker to
complete and researchers find them more convenient.

Ratio scales
Ratio scales possess the powers of the preceding three scales as also the concept of absolute zero
or origin. Thus, they have order, distance and unique origin and are the most superior amongst
all the scales. Examples of ratio scales are the commonly used physical dimensions such as
Measurement and 145
height, Scaling
Measurement and 146
Scaling
weight, distance, money value and population counts. Equal ratio on the ratio scale indicates the
equal ratio among the elements being measured. For example, 9 lbs and 45 lbs are in the ratio of
1:5. If we convert pounds into ounces, the same ratio will be obtained. Thus, 144 ounces and 720
ounces have the same ratio of 1:5, as earlier. In other words, one can change over from one unit
to another by using the relevant conversion factor. In the above example, a change from pounds
to ounces involved the multiplication of the two figures by 16. This facility of conversion from
one unit of measurement to another is available in the case of the ratio scale alone.
An example covering all the four types of scales is given

below: assessing a Respondent’s liking of soft

Drinks Nominal Scale

Which of these soft drinks do you like? Check all that apply
—Coke
—Pepsi
—Seven up
—Sprite

Ordinal Scale
Please rank the soft drinks on the following list according to your degree of liking each, assigning
your most preferred drink rank = 1 and your least preferred drink = 4.
—Coke
—Pepsi
—Seven up
—Sprite

Interval Scale
Please indicate your degree of liking of each of the soft drinks on the following list by checking
the appropriate position on the scale.
Dislike a lot Dislike Like Like a lot
—Coke
—Pepsi
—Seven up
—Sprite

Ratio Scale
Please divide 100 points among each of the following soft drinks according to your degree of lik-
ing for each.
—Coke
—Pepsi
—Seven up
—Sprite
100
From the foregoing account of the four types of scales, it should be clear that these scales are
Measurement and 147
in increasing order of sophistication from the viewpoint of data analysis.
Scaling
Measurement and 148
Scaling
DIffIculTy of MeasuReMeNT
Unlike physical sciences, measurement in social sciences (of which marketing is a part) is quite
difficult. The measurement of length, height, weight is a simple task involving the use of ratio
scale. But this type of situation is normally found in physical sciences. In contrast, measurement
in marketing is more difficult as it involves lower scales of measurement as compared to those
used in the physical sciences. This would be evident from Fig. 9.1.

Physical sciences

Social sciences

Length Preference Happiness


Weight Attitudes Creativity

Easy Very
difficult

Ratio Interval Ordinal Nominal


scale scale scale scale

Fig. 9.1 Difficulty of the Measurement Process

Source: Kinnear, Thomas C. and James R. Taylor: Marketing Research: An Applied Approach, New York, McGraw-Hill

Book Company, 1987 (3rd edition), p. 299.

The question that arises is: why measurement is more difficult in marketing compared to that
in physical sciences? This can be explained by taking an example. Suppose we have to conduct
a survey on the existence of medical facilities in certain localities in a city. Here, we have to first
design a questionnaire to collect the information. We will then interview sample respondents for
the information sought in the questionnaire. The point to note is that the attitude towards medical
facilities would vary from respondent to respondent. Even the questionnaire design may not be a
perfect instrument in eliciting the desired information. Thus, we see that measurement of attitude
poses a major problem. At this stage, it is worthwhile to know as to what could be the possible
sources of error.

souRces of eRRoR1
There are four major sources of error in measurement. These are:
(i) The respondent
(ii) The situation
1
Based on Cooper, Donald R. and Pamela S. Schindler. Marketing Research: Concepts and Cases, New Delhi,
Tata McGraw-Hill Publishing Company Limited, 2006, pp. 355–56.
Measurement and 149
Scaling
(iii) The measurer
(iv) The instrument for data collection
Each of these sources of error is briefly discussed here.

The Respondent

Measurement may get distorted on account of different opinions of the respondent on a given is-
sue. These differences arise on account of status of the respondent, level of education, social
class, nature of job, etc. Further, some respondents may be reluctant to express their negative
opinion on some issue while some others may express their attitude deliberately different from
others. Again, there may be respondents who may not know the topic on which their response is
being sought. In such cases, they may answer the question in a vague manner as they may not
openly confess their ignorance on that issue.

The situation

The situation in which an interview is being held also influences measurement. For example, an
investigator approaches a respondent who is discussing some matter with his friends. Naturally,
he may not provide answers by refusing to cooperate with the investigator. Alternatively, he may
give answers in a casual manner to cut short the interview time. Again, when the respondent
finds that his identity may not remain undisclosed, he may not express his actual feelings.

The Measurer

The measurer is no one else but the interviewer who is conducting interviews on the basis of a
questionnaire. Measurement errors may crop up on account of the method used by the
interviewer. It is possible that he may not be very careful while recording responses. He may
change the wordings or may not be able to record the complete response. This partial, careless
mechanical processing may lead to distortion of the responses.

The Instrument

Finally, a defective instrument for data collection may cause distortion in a number of ways.
These are explained below.
1. The questionnaire is too lengthy containing a number of questions.
2. The questionnaire has an element of ambiguity in the questions.
3. The language used in a question is suggestive of a particular response.
4. The layout of the instrument, the extent of space available for response, the type of
printing and the quality of paper used are inadequate and of poor quality.

cRITeRIa foR a GooD scale


There are two important criteria for ascertaining whether the scale developed is good or not. These
are reliability and validity. We first discuss reliability.
Measurement and 150
Scaling
Reliability
There are three major methods of estimating the reliability of measurement. These are:
(i) Test-Retest reliability
(ii) The Alternative forms reliability
(iii) Split-Half reliability

Test–Retest Reliability
This form of reliability involves repeated measurement of the same respondent or group using
the same scaling technique under similar conditions. This would involve administering a test at
two points of time to the same person or a group of persons. The scores of the two tests would
then be correlated. If the correlation is low, then the reliability too is less.
A point to note is that reliability is a necessary contribution to validity but it is not a sufficient
condition for validity. The relationship between reliability and validity can be better understood
by an example. Suppose a person takes his weight on a weighing machine which gives accurate
results, then the scale is both reliable and valid. If, on the contrary, the weighing machine always
records, say 2 kg. more than his actual weight, then the scale is reliable but not valid as it fails to
measure what it is supposed to. As the scale does not satisfy validity, reliability is hardly useful.

Alternative-Forms Reliability
Alternative-forms reliability involves the same respondent being given a set of two forms. The
forms are considered equivalent but are not identical. The results obtained on the basis of these
forms are compared to ascertain whether there is considerable difference between the two scores.
As each person in the group is given two forms and as the results of these two forms are to be
compared, this approach needs more time. Also, it is expensive.

Split-Half Reliability
Another approach to reliability involves administering the measuring instrument only once to
test the internal consistency. The split-half technique can be used in case of a multi-item
instrument. In fact it is another version of the alternative-forms technique. It involves splitting of
a multi-item measurement instrument into two equivalent groups. The item responses of the two
groups are then correlated to estimate reliability. A high degree of correlation indicates that there
is similarity or homogeneity among the items.

Validity
The second attribute that a scale should have is its validity.
While the concept of validity is simple to understand, it becomes difficult to apply the test of
By validity
validity is meant
in practice. the success
As Moser of thehave
and Kalton scalepointed
in measuring what
out, there areitfour
setsapproaches
out to measure, so be
that can
that differences
commonly between individuals’ scores can be taken as representing true differences
distinguished.
in the characteristic under study.
Measurement and 151
Scaling
1. Content validity
The first is the content validity, which implies that the contents of the scale correspond to the
con- tents of the attitude system and that they are comprehensive enough to cover the full range
of the attitude. The researcher should first define the problem clearly, identify the items to be
measured, and evolve a suitable scale for the purpose. Despite all this care, he may be criticised
on the ground that the scale developed by him lacks content validiy. This happens because,
whether or not a given scale contains the content validity will depend on the judgement of the
researcher and, this is likely to vary from individual to individual. To avoid this, it may be
preferable to approach a group of knowledgeable persons, rather than leaving it to one person
alone.

2. Construct Validity
One of the difficulties arising in attitude measurement is that it is perhaps impossible to measure
attitude directly. It can be measured only indirectly on the basis of answers given by the
respondents. In a situation of this type, the test of construct validity is used. The concept of
construct validity is more complex than that of content validity. In order to apply construct
validity, the researcher postulates the nature and extent of association between the attitude and
other specified variables. He then examines whether these relationships exist. If not, there could
be two possible explanations. First, his scale might be invalid as it does not satisfactorily
measure what it set out to measure. Second, his theory might be deficient in some way and it
may be difficult for him to identify it. The point to emphasise is that the construct validity is
based on theoretical considerations. For example, the status of an individual in a society may be
dependent upon such variables as the level of education, occupation or ownership of a car and a
house. Thus, on the basis of theory, an elite class should have a high degree of association
amongst these factors. The existence of a high degree of correlation in this case is a supporting
evidence and can be regarded as a test of validity.

3. Predictive Validity
Predictive validity signifies how best the researcher can guess the future performance, from his
knowledge of the attitude score. For example, an opinion questionnaire which forms the basis
for correctly forecasting the demand for a product has predictive validity. The procedure for
predic- tive validity first measures the attitude and then predicts the future behaviour on the basis
of this measurement.
This is followed by the measurement of the future behaviour at an appropriate time. Finally,
the obtained scores are compared with the earlier predicted scores. If the two series of scores are
closely associated, the scale is said to have predictive validity.

4. Concurrent Validity
In the case of concurrent validity, an attitude scale on one variable can be used to estimate scores
on another variable. For example, one may decide the social status of respondents on the basis of
their attitude toward savings. Here, the attitude scale as also the criterion measure are
administered almost at the same time. It may be noted that a high degree of concurrent validity
may sometimes be spurious as the collection of one set of data may influence the collection of
another set.
Measurement and 152
Scaling
practicality
The foregoing discussion shows that the scientific requirements of a research study should be
valid. In contrast, the operational requirements call for it to be practical. This implies that the
handling of a research study should be economical, convenient, and the results should be
interpretable.

economy
Often one has to strike a balance between the ideal research and the budget. Data not freely
avail- able and instrument length is one area where economic factors become dominant. The
choice of data collection method is also subject to economic factors.

convenience
Any device used for measurement should not be inconvenient to administer. A questionnaire or
a measurement scale should have a set of clear instructions to elicit better response. The more
complex the concepts and constructs, the greater is the need to provide very clear and complete
instructions.

Interpretability
When persons other than the test designers have to interpret the results, this aspect becomes im-
portant. It is usually an issue with standardized tests. In such cases, the researcher who is design-
ing the measuring instrument must provide important and relevant information so as to facilitate
interpretation in proper perspective.

DeVelopMeNT of MaRkeTING MeasuRes2


Suppose we have to develop a measurement for a particular concept such as customer
satisfaction in respect of a particular product, how should we proceed in this regard? We have to
proceed in a systematic manner, involving a number of steps. These steps are briefly explained
below.

1. specify Domain of concept


It is very important to develop a sound conceptual definition. For example, how do we define
customer satisfaction? It may be necessary not only to review the available literature but also to
discuss the concept of customer satisfaction with knowledgeable persons.

2. Generate sample of Items


This step requires that a list of specific items in the form of questions, phrases or statements be
prepared. It may be noted that items should cover each aspect of the concept as defined in Step
1. Here, too, a careful review of the literature and discussion with well-informed individuals,
consum- ers and executives of trade associations will be helpful in developing an appropriate list
of items.
Measurement and 153
2
Scaling
Tull, Donald S. and Del I. Hawkins: Marketing Research: Measurement and Method, New Delhi. Prentice-Hall

of India Private Limited, 1998, pp. 319–21


Measurement and 154
Scaling
3. collect Data
The items included in Step 2 should be placed in a proper format. This format should then be
used to collect information from the target group.

4. Improve the Measure


On the basis of the information gathered in Step 3, one can know that certain items do not
correlate with the total score for the overall measure. As such, such items need to be eliminated.
Sometimes one may find that the results are extremely unsatisfactory and unexpected. In such
cases, it is advis- able to go back to earlier steps relating to the conceptual definition and the
generation of sample items.

5. collect Data for Reliability and Validity assessments


Having eliminated superfluous items from the measurement instrument, one can now collect data
afresh on the basis of the revised instrument.

6. assess Reliability
Having collected new data on the basis of the revised instrument, this should be analysed for
reli- ability. It is advisable not to depend on one assessment, but two or more reliability
assessments should be used. On the basis of this assessment, unrealiable items should be
eliminated from the overall measurement instrument.

7. assess Validity
After having ensured the reliability of the measurement in Step 6, its validity needs to be
examined. While any method of assessing validity can be used, construct validity is preferable.

8. Develop Norms
This is the final step that is concerned with the administration of the measurement instrument to
various groups of people such as demographic groups, users, non-users, etc. Averages and
variances should be calculated as these facilitate better interpretation of data collected. When
measurement is to be used in subsequent studies (or multiple studies), these calculations will be
all the more useful. No doubt, adherence to all the above-mentioned steps would require not only
more time but also more money. If one cannot wait that long or is unable to bear the increased
cost, one can make necessary adjustments to reduce these steps in accordance with the
availability of time and finance.

1. Indicate which type of scale (nominal, ordinal, interval and ratio) is being used in each of
the following questions. Justify your answer.
(i) What is your monthly income?
(ii) How much time do you spend in the university library every day?
—Less than 1 hour
—1 to 2 hours
—2 to 3 hours
Measurement and 155
—3 to 4 hours Scaling
—More than 4 hours
Measurement and 156
Scaling

(iii) Your college/university brings out a quarterly magazine. How satisfied are you with it?
—Very satisfied
—Satisfied
—Neither satisfied nor dissatisfied
—Dissatisfied
—Very dissatisfied
(iv) What is the level of your education?
—Upto middle level
—Matriculation
—Higher secondary
—Graduation
—Post-graduation
2. Mention the type of scale represented by each of the following items or series of items:
(i) a metre
(ii) a series of value judgements ranging from zero to ten indicating zero as the poorest and ten
as the best
(iii) a series of value judgements ranging from “extremely dissatisfied” to ‘‘fully satisfied”.

You might also like