• No results found

The Research Design: What Data for which Questions?

2.1 The Offline Data Set

The offline data set consists of answers to a questionnaire sent to a random sample of English and Dutch university researchers in one of the following eight disciplines.

HistoryMathematics

SociologyMechanical Engineering

Management ScienceChemistry

EconomicsPhysics

In both countries the use of a variety of different IDGs by university researchers is promoted. In the United Kingdom the Mailbase system (Mailbase 1999c) founded in 1989 promotes and supports the use of academic mailing lists by the UK Higher Education system. In the Netherlands, the Surfnet system (Surfnet 2001) includes a large number of mailing lists used by the academic system. As a consequence, since the end of the 1990s the use of IDGs for the ordinary user has been quite simple. The use of other academic mailing lists and newsgroups is also possible for researchers in both countries.

18 Chapter 2

The disciplines were selected in such a way that on the one hand important disciplines within the natural sciences, and on the other hand disciplines within the social sciences and humanities could be analyzed. The informal communication structures of the two groups of disciplines were expected to differ (see e.g. Becher 1989; Walsh and Bayma 1996f; Garvey, Lin, and Nelson 1979; Borden et al. 1992; Crane 1972;

Weedman 1993).

The selection of universities and departments had to ensure that it would be possible to distinguish between the hypothesized (individual level) effects of IDG use and various potential group-level effects, for example effects due to properties of the department or properties of the university. As a consequence, the selection procedure had to ensure that for every department there was a sufficiently high number of different researchers and that for every university there was a sufficiently high number of departments. It was accepted that such a sampling procedure has disadvantages for making exact inferences about the distribution of some variables;

however, this is not the aim of the study.

I chose all 13 universities within the Netherlands. Within England 23 universities were randomly selected (see below for details). Within every selected university I chose all departments that belong to one of the 8 disciplines. Within every selected department a random sample of the researchers was taken - including Ph.D. students and postdoctoral researchers - with one exception. A small number of departments had a very high proportion of Ph.D. students. In such departments I sampled researchers with a non-temporary position disproportionately often to ensure that the sample contained a large enough number of them.

Among the larger disciplines that have more than one department at the same university, only those departments which cover some core research areas within that discipline were selected, in order to ensure a high enough number of respondents from the same department. Within economics the departments that include micro- or macro-economics were selected, and within management science those of Human Relation Studies/ Organizational Behavior were chosen. The departments chosen within chemistry are related to organic or inorganic chemistry. Physics departments include research in the areas of theoretical or solid state physics. The mechanical engineering departments cover research on topics such as material science, production or design management or mechanics. Within sociology, mathematics, and history no selection was made. I performed the selection of departments in such a way that the

Nevertheless they still exist as important research units. However, there are a variety of research units that are sometimes known as “(sub)departments”, “clusters”,

“research groups” etc. The names of the chosen research units are listed in Appendix 2.1. The following Table 2.1 shows the universities and the chosen disciplines within them that were selected for the Netherlands. The last row shows the number of selected disciplines within the chosen universities. There were 66 departments within 13 universities selected.

Table 2.1: The Selected Disciplines and Universities within the Netherlands

Twente TUE Delft WAU RUG Utrecht UvA VU KUN Leiden Maas-tricht

Table 2.2 shows the selected universities and disciplines within England. There were 23 universities with 96 departments (or equivalent research units) selected.

Table 2.2: The Selected Disciplines and Universities within England

AngliaEast

20 Chapter 2

*: two departments refused to participate (see text); details: see Appendix 2.2 and text

The complete list of selected university departments can be found in Appendix 2.2.

The English universities and departments (or equivalent research units) were randomly sampled in the following way. A random list of 73 university numbers was created. Within every sampled university all disciplines belonging to one of the eight to be analyzed were chosen until for a given discipline 12 different universities had been selected. Thus for every discipline the 12 universities that were first in the sample were selected. This procedure makes it easier to distinguish between potential group-level effects at the level of the department and at that of the university since the average number of departments per university is increased. It has the disadvantage that departments of universities with a larger variety of disciplines have a higher likelihood of being sampled. Since it is not the aim of the analyses to describe the distribution of variables in the whole population of ‘all’ Dutch and English university researchers but to study (causal) relations between variables, this disadvantage was accepted. For the same reason, no weighting procedures were applied.

Wherever possible, the questionnaire was sent by (snail) mail to every researcher individually. A small number of English university departments hesitated to reveal the names of their Ph.D. students. In such cases the questionnaires were distributed to a random sample of them with the help of the secretary. In two such selected university departments the chair decided at short notice that the whole department would not participate due to time reasons.

The data were collected from September 1998 until March 1999. A usable questionnaire was returned by 1063 of 2688 researchers3 (~ 40%). Filling in the

3 Slightly more than 10% of the names/addresses retrieved via the WWW or official books were no longer valid.

Ranking of date of received filled in questionnaire

As Table 2.3 shows, the immediacy of response is not related to whether the respondent has an email connection, whether he makes use of IDGs, how many IDGs he uses, or how intensively he makes use of email. I interpret this as a hint that the use of email and IDGs is something ‘ordinary’ for those who use email/IDGs and hence does not give extra motivation for participation in the survey. As a consequence, there is no indication of a bias in the response rate in favor of IDG users.

The response rate does not differ between Dutch and English respondents (p>0.9). It does not differ between respondents of different disciplines, with the single exception of mechanical engineers who answered significantly less often. Just 30.4% of them returned a completed questionnaire (Chi-Square=15.1, df=7, p<0.05 for all 8 disciplines and Chi-Square=5.2, df=6, p>0.45 for the remaining 7 disciplines).

Further descriptive statistics shown in Appendix 2.3 reveal that 21.3% of the respondents are female. The proportion of postgraduate students is 38%, which implies that the sample contains a large number of 'peripheral' researchers who are assumed to profit disproportionately from the use of IDGs. Mechanical engineers account for 7.4% of the respondents. The very large majority of respondents report having an email connection (97%) and thus have the opportunity to make use of IDGs. The arithmetic mean of the age of the respondents is approximately 48 years;

22 Chapter 2

the youngest respondent is 22 and the oldest (retired) is 70 years old. The average length of research experience is approximately 11 years, with a range of 0 to 45 years.

There is a large variation in the degree of research activity. On average, each respondent wrote during the previous 12 months slightly more than 3 papers and visited slightly less than 3 conferences. A small minority, however, were much more active with regard to the number of conferences visited and papers written.

The respondents were asked to mention the names of a maximum of 5 most important mailing lists or newsgroups that they use for professional reasons. For every IDG they had to estimate both the average monthly number of messages sent to the IDG and the duration of their own membership in months. Moreover, I asked the researchers whether the IDG used had any relevance at all for their research field. A total of 23.3% of the respondents mentioned that they used at least one IDG for professional reasons. The mean number of IDGs used among those who use at least one is 1.90.

More information about the data set can be found in Chapter 3 and Chapter 4.