A major hotel chain desires to evaluate the potential offered by the incentive travel market (e.g. trips that companies award as prizes in sales contest). The chain has a list of 21,476 companies that might have used travel prizes as incentives. It wishes to estimate what percentage of this prospect list used travel incentives in the past year.
It is projected that about 40 percent of the prospective customers used incentive travel in a year. Based on this judgment, the chain desires to estimate the actual percent by means of a sampling from the list of prospects.
How can a suitable sample be chosen? How long big should the sample be in order to be virtually certain that the sample based estimate will be within 10 percentage points of the actual percentage, who used incentive travel in the past year?
Because much marketing information is obtained by sampling situations such as the above are common in marketing research. Therefore, the student should understand the advantages and limitations of collecting data via samples. This discussion of sampling will be limited to an examination of basics concepts. The reader is assumed to have only a limited knowledge of mathematics.
Lower total cost and greater timeliness are the major reasons data are collected by sampling rather than by complete enumeration. However, samples may possess other advantages. For example, greater accuracy of individual measurements may be possible. Fewer filed workers or observers will be needed, and these may be more carefully selected trained and supervised. Considering these several aspects, it will often be found that sampling provides faster, at less cost – than would be possible by attempting to collect data from all units of interest.
Confusion between Sampling Errors and Data Collection Errors:
Data collected from a sample will reflect both the ‘underlying reality’ and the method used to collect the data. Researchers are interested in measuring this underlying reality: however, they must be aware of possible errors in their data collection method. When data are collected from a sample, an additional source of error is introduced that of sampling. If the results of a sampling study is found to be incorrect, the tendency is to attribute this to bad sampling technique when, in fact, the method of data collection might have been inaccurate.
Example: The research objective is to find the percentage of Denver households owning a VCR. In a count of all households an incorrect answer would be obtained if the method of determining ownership were wrong. If only a sample had been studied, the inaccurate measurement method would still be present; and in addition, there would be sampling error because the sample would not be an exact replica of the universe.
Both possible sources of error – sampling and data collection must be considered in designing research. Although this discussion focuses on sampling error, the reader must keep in mind non-sampling issues as well.
Three problems must b addressed in any sampling operation. The first problem is to define the universe being studied. The universe is the entire group of items the researchers wish to study and about which they plan to generalize. For a given project, the universe might consist of women older than 40 residing in the United States, all families within the corporate limits of the city of Chicago, or all grocery stores in the New York metropolitan area. Thus, the definition of the universe, in any particular case, is determined solely by the research objectives.
Many decisions must be made of the universe is to be sharply defined. For example, to define the universe of grocery stores located in the New York metropolitan area, such questions as the following must be answered: What is a grocery store? Are both chains and independents to be considered? Are stores which sell primarily cooked foods (delicatessens) to be considered? What point or time period is involved? Exactly what geographical area is to be considered?
Failure to define the universe appropriately, in accord with study objectives, may yield misleading results.