Stratification in Representative Sampling

The results to be expected from stratification for various methods of sampling and for different sets of circumstances are described.

Abstract

STRATIFICATION is one of the methods that have been developed in recent years for increasing the accuracy of information obtained from samples and reducing the cost of obtaining it. The general procedure is quite simple; it consists of dividing what is to be sampled into its most important parts and then taking a sample of the right size from each part. For example, no marketing survey of the customers of a department store would be satisfactory if it were limited to purchasers interviewed at the jewelry counter or to people found in the bargain basement. To be fully representative of all the customers, the sample should include customers of each departmentof the store and, moreover,it should include them in their true proportions. If a fixed percentage of the customers of each department is taken, the sample will be spread in a representative way over all the departments and will tend to include each of the different kinds of customers in its true proportion. For this sample the departments are taken to be the most important strata and hence the sampling is stratified by departments. For other samples it might be more advantageous to stratify by income level, family size, type of residence, or some other pertinent characteristic by which the customers could be classified. The basis for the stratification is chosen with reference to the purposes of the sampling and the conditions under which it must be done. Like other methods employed in sampling, stratification can be used most effectively when the manner in which it affects the variability of samples is fully understood. The benefits it provides are not always large enough to compensate for the labor and expense of stratifying; indeed, there may be no benefits at all in some instances. This paper will describe the results to be expected from stratification for various methods of sampling and for different sets of circumstances.