128
(see Table A4; Appendix, for an overview of the key variables). Finally, the strategic social impact of the G1000 initiatives was assessed based on the extent to which G1000 participants or other individuals, groups and organisation in the local community used by social actors to support specific social actors or to oppose others. We based our assessment on information provided by interviewees and newspaper articles that followed the G1000 activities.
129 data collection process in Borne, this caused some lack of knowledge about the sampling method used and how well the data collection process was carried out. This had an influence on calculating an accurate response rate for Survey 1 and 2 in Borne since no survey software has been used which could have provided us with more information about the distribution process of the survey and thus the data collection process (e.g., number of ineligible/eligible participants).
We took some measures to enhance the quality of the Survey questionnaires 1 and 2. First, to avoid misunderstandings by respondents, we refined the wording of some survey questions in the original survey questionnaires in Enschede and Steenwijkerland.
Second, to gather more information from the respondents in Enschede and Steenwijkerland, we also added some open-ended questions and the ability to enter text for some questions to Survey 1 and Survey 2. Third, to make the three data sets comparable, we recoded the variables after the end of the data collection process and merged the different data files into one coherent final data set. Fourth, to complement a possible loss of information in the case of Borne, we used some interview questions to clarify a number of the research results. Finally, to compensate for the lack of knowledge about the data collection process in Borne, we contacted the “G1000 research team” to obtain more information.
6.4.2 Panel attrition and sample representativeness
As mentioned in Section 6.2.1, the survey response rates dropped by about half throughout the participation process.1 On the one hand, this implies that around half of the participants who participated in the first survey also completed the final survey wave. On the other hand, this conditional response rate also indicates a high panel attrition rate (or panel mortality). From previous panel studies, we know that attrition
1 Conditional response rates: Borne 54%, Enschede 46%, Steenwijkerland 46%.
130
(or (permanent) drop-out of the survey participants) is one of the major sources of non-sampling error in panel surveys since even modest attrition rates can affect the representativeness of the sample (Lugtig &
Research, 2014). From the literature on survey methodology, we also know that the process of attrition can vary across respondents and over time. For example, male respondents tend to drop out more often than females (Behr, Bellgardt, & Rendtel, 2005; Lepkowski & Couper, 2002).
Watson and Wooden (2009) found that people with a higher socioeconomic status (higher education and incomes) drop out less than those with lower socioeconomic status (although differences were small). Moreover, studies revealed that many people participate in multiple panel studies infrequently. One reason for this is, for example, a lack of commitment (Laurie, Smith, & Scott, 1999).
To examine whether attrition rates had a significant effect on the representativeness of our survey samples, we used one-sample chi-square tests. We determined how well the survey samples represented the G1000 population (participants attending the large-scale deliberation events) in all surveys in terms of gender, educational level, age, and participants’ role in the G1000 process. The results revealed that the survey samples did not differ significantly from the G1000 population in all three cases. Thus, we can conclude that attrition had no significant effect on the representativeness of the survey samples.
132
Chapter 7
133
7 R EPRESENTATIVENESS OF THE
G1000 INITIATIVES
In the selection and recruitment process of the three G1000 initiatives, a strong emphasis was placed on maximising the diversity of participants.
With the help of sortition, 10.000 citizens in Enschede and Steenwijkerland and 7.000 citizens in Borne were drawn at random from the population registers (excluding all citizens under 16 years of age) and invited to participate in the G1000 process. In all three G1000 initiatives, the purpose of using a random sampling method was twofold: firstly, the organisers wanted to ensure the inclusiveness of the participatory process by giving all citizens of the local community an equal chance to be selected and participate in the G1000 process. And secondly, the organisers used a random sampling method to select and recruit a diverse and large group of participants that represented the views and interests of the local population in the best possible way.
In addition to the randomly selected citizens, organisers also sought to achieve a “real representation of the [local] community” by bringing all stakeholders of a local community into one room (Van Dijk, 2020, p. 7, see also chapter 6). To this end, in all three initiatives, a smaller group of local stakeholders – officials, politicians, and experts/employers –
134
was recruited through a non-random targeted recruitment strategy (personal invitations by name).
Finally, all three G1000 initiatives also included some elements of self-selection: first, as participation in the G1000 is voluntary, only those who signed up for the G1000 could take part in the G1000. Secondly, to increase participation rates, randomly selected citizens who were unable or unwilling to attend the first large-scale deliberation events could pass on their invitations to someone else. Thirdly, organisers gave local citizens interested in the G1000 process the opportunity to participate as “free-thinkers”. And finally, in larger geographical areas (Steenwijkerland and Enschede), all invited residents were allowed to participate together with someone living in their neighbourhood.
Considering organizers’ efforts to maximize the diversity of participants using random and non-random selection methods, the question remains: To what extent were the organisers of the three Dutch G1000 initiatives successful in involving a diverse and inclusive group of citizens from the larger population in the participation processes? Were the three G1000 initiatives a good reflection of the local populations?
And how was the actual representativeness of the G1000 and its importance perceived by those who participated? Which factors have affected the representativeness of the G1000 initiatives?
In this chapter, we will answer these questions by first examining whether G1000 participants reflected the local municipalities descriptively and substantively. To this end, we will compare the socio-demographic characteristics and general attitudes and behaviour of G1000 participants with those of the wider local populations in Borne, Enschede and Steenwijkerland. Moreover, to get a better idea of whether the representativeness of the G1000 process has changed throughout the G1000 process, we will also analyse the socio-demographic characteristics and general attitudes and behaviour of G1000 participants over time. Subsequently, we will assess how the
135 representativeness of the G1000 was perceived by its participants. Did their perception differ greatly from the representativeness we assessed?
And do they think that the G1000 participants should be a good reflection of their local community? In the final sub-section of this chapter, we will first assess whether non-random sampling have contributed to or undermined the representativeness of the G1000 by comparing the socio-demographic characteristics and general attitudes and behaviour of the randomly selected G1000 participants with the non-randomly selected participants. Finally, we will assess whether self-selection undermined the representativeness of the G1000 initiatives by investigating how non-participation, no-shows and drop-outs influenced the composition of the G1000 initiatives? We also look at the question of who decided to end their participation in the G1000 initiatives and why?