In constant sum scaling, respondents allocate a constant sum of units, such as points, dollars, or chips, among a set of stimulus objects with respect to some criterion. As shown in Figure 8.5, respondents may be asked to allocate 100 points to attributes of a toilet soap in a way that reflects the importance they attach to each attribute. If an attribute is unimportant, the respondent assigns it zero points. If an attribute is twice as important as some other attribute, it receives twice as many points. The sum of all the points is 100. Hence. the name of the scale.
The attributes are scaled by counting the points assigned to each one by all the respondents and dividing by the number of respondents. These results are presented for three groups. or segments, of respondents in Figure 8.5. Segment I attaches overwhelming importance to price. Segment II considers basic cleaning power to be of prime importance. Segment III values lather. fragrance. moisturizing. and cleaning power. Such information cannot be obtained from rank
order data unless they are transformed into interval data. Note that the constant sum also has an absolute zero-10·points are twice as many as 5 points, and the difference between 5 and 2 points is the same as the difference between 57 and 54 points. For this reason, constant sum scale data are sometimes treated as metric. Although this may be appropriate in the limited context of the stimuli scaled, these results are not generalization to other stimuli not included in the study.
Hence, strictly speaking, the constant sum should be considered an ordinal scale because of its comparative nature and the resulting lack of generalization. It can be seen that the allocation of points in Figure 8.5 is influenced by the specific attributes included in the evaluation task
The main advantage of the constant sum scale is that it allows for fine discrimination among stimulus objects without requiring too much time. However, it has two primary disadvantages. Respondents may allocate more or fewer units than those specified. For example, a respondent may allocate 108 or 94 points. The researcher must modify such data in some way or eliminate this respondent from analysis. Another potential problem is rounding error if too few units (e.g., points) are used. On the other hand, the use of a large number of units may be too taxing on the respondent and. cause confusion and fatigue
Q-Sort and Other Procedures
Q-sort scaling was developed to discriminate quickly among a relatively large number of objects. This technique uses a rank order procedure in which objects are sorted into piles based on similarity with respect to some criterion. For example, respondents are given 100 attitude statements on individual cards and asked to place them into II piles, ranging from “most highly agreed with” to “least highly agreed with.” The number of objects to be sorted should not be less than 60 nor more than 140; 60 to 90 objects is a reasonable range. The number of objects to be placed in each pile is pre specified, often to result in a roughly normal distribution of objects over the whole set.
In the four primary scales, the level of measurement increases from nominal to ordinal to interval to ratio scale. This increase in measurement level is obtained at the cost of complexity. From the viewpoint of the respondents, nominal scales are the simplest to use, whereas the ratio scales are the most complex. Respondents in many developed countries. due to higher education and consumer sophistication levels. are quite used to providing responses on interval and ratio scales. However. it has been argued that opinion formation may not be well crystallized in some developing countries. Hence. these respondents experie’nce difficulty in expressing the gradation required by interval and ratio scales. Preferences can. therefore. be best measured by using ordinal scales. In particular. the use of binary scales (e.g preferred/not preferred). the simplest type of ordinal scale. has been recommended.I” For example. when measuring preferences for jeans in the United States. Levi Strauss
Qualtrics Question Library and Primary Scales
Access to Qualtrics is included with this book. Use the Qualtrics question library to electronically develop the following scales.
1. Gender measured on a nominal scale
2. Age measured on an ordinal scale
3. Age measured on a ratio scale
4. Income measured on an ordinal scale
Design your 0′” n question to measure income on an interval scale
Ethics in Marketing Research
The researcher has the responsibility to use the appropriate type of scales to get the data needed to answer the research questions and test the hypotheses. Take, for example. a newspaper such as the Wall Street Journal wanting information on the personality profiles of its readers and non readers.
Information on the personality characteristics might best be obtained by giving respondents (readers and non readers) several cards, each listing one personality characteristic. The respondents are asked to sort the cards and to rank-order the personality characteristics, listing, in order, those they believe describe their personality best first and those that do not describe themselves last. This process will provide rich insight into the personality characteristics by allowing respondents to compare and shuffle the personality cards. However. the resulting data are ordinal and cannot be easily used in multivariate analysis. To examine differences it the personality characteristics of readers and non readers and relate them to marketing strategy variables. interval scale data are needed. It is the obligation of the researcher to obtuin the data that are most appropriate, given the research questions. as the following example illustrates .
Scaling Ethical Dilemmas
In a study designed to measure ethical judgments of marketing researchers. scale items from a previously developed and tested scale were used. Afte. J pretest was conducted 0″ a convenience sample of 65 marketing professionals. however. it became apparent that some original scale item, were worded in a way that did not reflect current usage. Therefore. these items were updated. For example. an item that was gender specific. such as. “He pointed out that … “was altered to read “The project manager pointed out that. Subjects were requested to show their approval or of the stated action (item) of a marketing research director with regard to specific scenarios. Realizing that a binary or dichotomous scale would be too restrictive. approval or disapproval was indicated by having respondents supply interval-level data via 5-point scales with descriptive anchors of I = disapprove. 2 = disapprove somewhat. 3 = neither approve or disapprove. 4 = approve somewhat. and 5 = approve. In this way. scaling dilemmas were resolved
After the data have been collected. they should be analyzed correctly. If nominal scaled data gathered. then statistics permissible for nominal scaled data must be used. Likewise. when ordinal scaled data are collected. statistical procedures developed for use with interval or ratio data should not be used. Conclusions based on the misuse of statistics are misleading. Using the previous
personality example. if it decided to gather data By the rank order technique described. ordinal data would be collected. If. after collection. the client is hes to know how the readers and the non readers differed. the researcher should treat these data correctly and use non metric techniques for analysis (discussed in Chapter 15). When the researcher lacks the expertise to identify and use the appropriate statistical techniques. help should be sought from other sources. for example. from statisticians.