Respondent related correlates of response behaviour in audience research

This paper deals with unit nonresponse in audience research in theatres. Audience research in theatres fre-quently lacks the methodological rigour – due to high nonresponse rates and selective sampling – to be able to do scientifically valid statements on the composition, the aesthetical expectations and the evaluations of the members of the theatre audience. This may lead to seriously biased estimates of distributions and parameters. In this paper, we will explore nonresponse bias in audience research. More particularly, the focus will be on the explanation of nonresponse by socio-demographic and attitudinal characteristics of a theatre audience. The typical features of audience research make an attempt at exploration of the ‘silent minority’ a chal-lenging endeavour. First of all, people are not contacted in their homes, but in the playhouse itself. This gener-ates a series of specific problems faced by audience researchers and very specific reasons for nonresponse (non-contact/refusal/indirect noncooperation). Secondly, no real sampling frame is available : the population attending each playhouse as such, is unknown. Moreover, this population and the population of visitors to theatres are in no respect representative of the general population. Hence, this opens the opportunity to test the applicability of the seminal theoretical framework offered by Groves and Couper (1998) in a specific surveying context with an equally specific population. Two research questions will be addressed : 1) in what respect(s) do respondents differ from nonrespondents in audience research in a survey on arts participation ? Not only potential socio-demographic differences will be examined, but also possibly rele-vant correlates with regard to topic involvement or issue salience and experience with theatre will be put to the test; 2) can the hypotheses to explain nonresponse by means of respondent characteristics derived from a theory of participation in household surveys within a general population justifiably be used in the study of nonresponse within audience research? Before addressing these research questions, the specificity of the population and surveying situation are discussed. Cultural participants generally tend to be younger, higher educated and have to a higher SES than non-participants. There are no significant differences with regard to gender. These findings, based on a survey in the Flemish population (APS–2000 data), are concurrent with earlier population research on the composition of theatre audiences in Flanders. Pertinent features of the specific surveying situation that may affect response behaviour, are the severe time restrictions faced by the agencies establishing contact with members of the audience and spacial impediments. During the months February and March 2001 the University of Ghent carried out an extensive audience research in 3 theatre institutions in Ghent (Belgium) : the audience of 24 performances of 10 different plays – selected by means of time and place sampling – was contacted on site to participate in a survey on arts participation. The research design used for this attempt to model nonresponse consists of a two step procedure. In the first step, every xth member of the audience was handed over a 6-page questionnaire which had to be filled out on site. We opted for as short as possible a questionnaire with only questions on people’s socio-demographic characteristics so as to maximise response in this first step. At the same time every contacted and cooperative member of the audience was given a second 20-page questionnaire ¬– with a pre-stamped envelope – which was supposed to be filled out at home and returned by mail. Within the context of audience research, it is impossible to collect data on nonrespondents in step 1. Therefore, the composition of the audience in the first step is compared to a proxy of a population benchmark based on the weighted APS–2000 data. Aggregate socio-demographic differences between the population proxy and sample data are discussed. The validity of this best available method is questioned and discussed. However, our research design does make it possible to compare respondents with nonrespondents on a micro-level inspired by the three-step procedure used by Hox, de Leeuw and Vorst (1995): ignoring unit nonre-sponse in step 1, we use logistic regression to map selection in step 2 (response behaviour on the mail question-naire). The chance of completing and returning a mail questionnaire within this specific population of cultural participants in step 2 has been found to increase with age and educational attainment and vary according to occupational category. Moreover, involvement with survey topic is confirmed as a strong predictor of survey participation. Gender and experience with theatre remain insignificant in predicting response behaviour. These findings are compared with the socio-demographic correlates of response behaviour in general populations (Groves and Couper, 1998). Implications for statistically controlling for nonresponse bias in audience research and suggestions for further research are presented