This section addresses the issues that arise from the fact that a number of factors will have interfered with obtaining the exact information, both in quality and quantity, from the survey data in spite of the analyst's best efforts to chose the appropriate survey method, to develop the best instrument possible, and to administer and execute the survey meticulously.
Why do we suspect that we will not quite get the information we want, and why should any corrections, adjustments and weightings be necessary? The answers to these questions fall into a number of categories. After the survey instrument was distributed, many of the analyst's conceptual, theoretical, and logical considerations were up against a test in the real world; namely the behavioural characteristics of the human beings from whom the survey information was to be obtained. And these human beings do not necessarily respond to our request in line with our wishes, expectations, and theories. Some of them were not able to respond to our request, others did not want to cooperate, others responded only partially, others again misunderstood some questions on the survey instrument.
Yet in spite of the less than perfect response that is likely to have occurred, the investigator still wants to, and has to, use the data to obtain information that is relevant for the survey population and not just for the subsample of people that responded "perfectly". It should be remembered here from the discussion of sampling theory in Chapter 4 that the original intent, based on this theory, was to develop population estimates on the basis of a carefully selected sample of that population. Unfortunately, in virtually all surveys, the population estimates have to be derived on the basis of a response of less than one hundred percent, in most instances from substantially less than this ideal target.
The purpose of this chapter then is to make the analyst aware of both the likely reasons for, and the consequences of, having to deal with only a subset of the desired sample. An awareness of these reasons and, particularly, of the effects of an imperfect response rate, can go a long way towards understanding the limitations of the survey results, the likely magnitude, direction, and implications of any biases resulting from them, and towards the developments of any adjustments and compensating measures that might be possible.
This chapter on Weighting and Expansion of Data covers the following topics:
9.1 POPULATION EXPANSION FACTORS
9.2 CORRECTIONS FOR NON-REPORTED DATA
9.3 CORRECTIONS FOR NON-RESPONSE