Principles

For some reason little has been published on the subject of weighting. It is generally ignored in texts on statistics and on survey research. However, its prevalence, particularly in larger surveys, makes it essential that practising researchers be familiar with the purposes, principles and methods of weighting and with its implications.

Ideally a survey sample has all the characteristics of the population it represents, and the results it yields can be used as they stand after a straightforward count of the number of respondents falling into each designated category. In practice, however much effort may be put into the execution of a well planned sampling exercise, the sample will never be representative in all respects. A sample intended to be a ‘cross-section’ of the population may prove to be biased or deficient in some way. Or particular groups may be deliberately over-represented in the sample, as described in the preceding chapter.

The imbalances may be corrected at the analysis stage by a process known as ‘weighting’. Increased emphasis is given to the information from under-sampled groups and less to the information from over-sampled groups in calculating the final estimate. In practice it is normally done by attaching a ‘weight’ to each respondent in the survey.