Non-responses: the curse of preelectoral and electoral surveys

Abstract

The problem of non-response or missing data is central for present-day social surveys. Especially important it is for pre-electoral and electoral surveys where the response rate usually is very low. The aims of this presentation are: i) to describe the sources of non-response; ii) to discover the reasons of non-response; iii) to summarize the methods for overcoming the problem, and iv) to give real examples from Bulgarian recent elections. Additionally, the methods for calculating the response rate are summarised.
The comparison is made between the response rates (RR) for different types of electoral and pre-electoral surveys, and the impact of missing data on the estimates is investigated.
Special attention is given to exit-polls where the response rate is extraordinary low. A comparative analysis is given how the non-response rates change over the past 10 years. Real data from Bulgarian municipal elections are used for this analysis.
The conclusion is that the problems with low RR in electoral and pre-electoral surveys remain unresolved and investigations are necessary to discover the determinants of non-response.