Reliability and sensitivity of research surveys
Elements of the design and operation of large scale surveys introduce sources of error not adequately dealt with in normal textbook formulae. Nevertheless, it is often possible by comparing the results of replicated surveys to arrive at objective and empirical estimates of accuracy and lack of bias in such surveys. Further, it is necessary to do this to convince our clients of the reliability of our work. Apart from the reliability of a survey, it should also be sensitive to changes in the situation. Examples of the sensitivity of opinion polls are given indicating how substantial variability and unpredictability in the way people's views respond to outside events implies that predictive models, especially in the course of an electoral campaign are fraught with danger. The unpredictable electorate reigns supreme and the opinion polls can only reflect, with as short a time lag as possible, the present rather than the future state of public opinion.
- This could also be of interest