Design and conduct of surveys (cont.)

Sample size

The larger the sample of units on which data are collected, the more likely it becomes that the resulting estimates will be close to those for the population as a whole. This is termed an increase in the precision of the estimates.

On the other hand, an increase in the scale of the operations can lead to the deployment of less motivated and less well-trained field workers and to less effective supervision of their work. If this happens, the number and severity of errors in the data are likely to rise, reducing the reliability and, probably, the validity of the results. These terms are explained here .

When the precision of a single estimate or the power to detect a key difference is crucial to the objectives of the survey, standard statistical formulae can be used to calculate the required sample size. With multi-purpose surveys, however, sample size decisions usually represent a pragmatic compromise reflecting the needs of diverse users of the data and the available budget.

Note though that:

• fieldwork costs are approximately proportional to sample size
• but precision increases in proportion to the square root of sample size

Thus, as one spends more on fieldwork, one gains incrementally less in precision from doing so.

It is important to allow for likely non-response (failures to make contact and refusals) when determining the number of units to attempt to include in a survey. In many surveys, a considerable proportion of the total costs of data collection are incurred in locating and making contact with the units included in the sample, rather than in collecting data from them once they have agreed to participate. Thus, if a high level of unit non-response is likely, this may add considerably to the cost of a survey.