We have already considered the problems of cost and time associated with simple random sampling, and cluster sampling is another probability sampling method which may be used to overcome these problems. It is also a useful means of sampling when there is an inadequate sampling frame or when it is too expensive to construct the frame. The method consists of dividing the sampling area into a number of small concentrations or clusters of sampling units.
Some of these clusters are chosen at random, and every unit in the cluster is sampled. For example, suppose you decided to carry out the bank survey using the list of all the customers as the sampling frame. If you wished to avoid the cost of simple random sampling, you could take each branch of the bank as a cluster of customers. Then you select a number of these clusters randomly, and interview every customer on the books of the branches chosen.
As you interview all the customers at the randomly selected branches, the sum of all interviews forms a sample which is representative of the sampling frame, thus fulfilling your major objective of a random sample of the entire population. A variation of this method is often used in the United States, because of the vast distances involved in that country (often referred to as area sampling). With the use of map references, the entire area to be sampled is broken down into smaller areas, and a number of these areas are selected at random. The sample consists of all the sampling units to be found in these selected areas.
- Advantages – the major advantages of this method are the reduction in cost and increase of speed in carrying out the survey. The method is especially useful where the size or constitution of the sampling frame is unknown. Nothing needs to be known in advance about the area selected for sampling, as all the units within it are sampled; this is very convenient in countries where electoral registers or similar lists do not exist.
- Disadvantages – one disadvantage is that often the units within the sample are homogeneous, i.e. clusters tend to consist of people with the same characteristics. For example, a branch of a bank chosen in a wealthy suburb of a town is likely to consist of customers with high incomes.
If all bank branches chosen were in similar suburbs, then the sample would consist of people from one social group and thus the survey results would be biased. This can be overcome to some extent by taking a large number of small clusters rather than a small number of large clusters. Another disadvantage of taking units such as a bank branch for a cluster is that the variation in size of the cluster may be very large, i.e. a very busy branch may distort the results of the survey.
0 Comments