PhD Students' Research Group Networks. A Qualitative Approach Llms Coromina1, Aina Capo2, Germa Coenders3, and Jaume Guia4 Abstract This article examines the networks within the research groups where Spanish PhD students are pursuing their doctorate. Capo et al. (2007) used quantitative data to predict PhD students' publishing performance from their background, attitudes, supervisors' performance and research group networks. Variables related to the research group network had a negligible explanatory power on student performance once the remaining variables had been accounted for. In this article, a qualitative follow-up of the same students is carried out using extreme case sampling and in-depth interviews. The qualitative research shows networking as important for students. Out of the 115 aspects that students mention in the interviews as relevant to publishing in the qualitative research, 92 have to do with their supervisors, their research group or their network as a whole. Similarly, out of the 50 hindrances mentioned, 20 have to do with the networks or relations. The most commonly mentioned network-related topics are research group members pushing PhD students to publish, meeting researchers outside the research group, existence of other PhD students in the group, help with the PhD from group members, supervisor's interest in the thesis, the possibility of discussing with experts on the PhD's topic and frequent contact with the supervisor and research group members. Some of these characteristics were not, however, measured in the conventional quantitative social network survey. 1 Introduction This study belongs to a wider project designed to predict PhD students' academic performance carried out by the INSOC research group (International Network on Social Capital and Performance). The INSOC research group is composed of 1 University of Girona. Department of Economics. Faculty of Economics and Business. Campus Montilivi, 17071 Girona, Spain; lluis.coromina@udg.edu 2 University of Girona. Department of Economics; ainamaria.capo@gmail.com. 3 University of Girona. Department of Economics. Faculty of Economics and Business. Campus Montilivi, 17071 Girona, Spain; germa.coenders@udg.edu 4 University of Girona. Department of Organization, Management and Product Design. Faculty of Tourism. Pujada dels Alemanys, 4. 17071 Girona, Spain; jaume.guia@udg.edu researchers of the universities of Girona (Catalonia, Spain), Ljubljana (Slovenia), Giessen (Germany) and Gent (Belgium). A key point for the academic quality of higher education is that future professionals achieve the highest academic performance. These future professional candidates are today's PhD students. Therefore, finding the main reasons that make a PhD student successful, in terms of higher performance, is a fundamental aspect in generating quality in higher education. Furthermore, knowing which are these influential elements is relevant for research groups at universities in order to select and hire the most adequate future professionals, that is, PhD students, and to promote working conditions that foster and increase their performance. The aim of the INSOC research group is to study the determinants of PhD students' academic performance across the INSOC member universities. Performance in creative teams or working groups has been approached from both managerial/innovation and education perspectives and even from both disciplines simultaneously, as some key variables like mentoring operate in a similar fashion (Paglis et al., 2006). In the long run PhD students' performance is evaluated by the broader scientific community in terms of attended conferences and published papers. Our choice in this article is, thus, to consider performance from this view point in a way similar to Green and Bauer (1995). A first group of authors studying creative or academic performance stress the role of the personal background. For instance, Cohen and Levinthal (1990) related levels of education and experience to knowledge creation. Another group of authors focused on the role of attitudinal variables such as group atmosphere, job satisfaction or motivation. Ivankova and Stick (2007) found that self-motivation and an online learning environment were predictive variables of PhD performance. Wentzel and Wigfield (1998) pointed to the importance of the motivation of students. Similar findings are also found in the managerial field (e.g., Nonaka and Takeuchi, 1995). A third group of authors worked on the role of social network relationships within groups, including trust and communication among social network members (Hemlin et al., 2004; Wasserman and Faust, 1994). The basic idea behind this perspective is that an individual's success is strongly dependent on the relations with relevant others inside and outside the organisation (Burt, 2000). The importance of social relations in the network structure concerning individual performance can be captured by the concept of social capital. The key points are the relationship between students and supervisor (Cryer, 1996), with the research group as a whole (Hemlin et al., 2004) and socialization (Austin, 2002). On the other hand, being isolated in a research group can be one of the main problems for a PhD student (Rudd, 1984). Capo et al. (2007) used data from the INSOC project to predict PhD students' academic performance from their background, attitudes, supervisors' performance and research group networks using Slovenian and Spanish data. Variables related to the research group network had a negligible explanatory power on student performance once the remaining variables had been included. This was specially so for the Spanish data collected in the University of Girona. For that data set, frequency of supervisor contact was the only statistically significant network predictor on performance. Besides, against any previous expectation, frequent supervisor advice was found to be detrimental to student performance. Coromina (2006), following a different approach, found no significant network variables whatsoever. In this article we report a qualitative follow-up study of the same PhD students who participated in the initial quantitative research with the aim of understanding the unexpected results found regarding the effect of network variables (Capo, 2009). Sale et al., (2002) give strong arguments for combining quantitative and qualitative methods in a single study. Casebeer and Verhoef (1997) even argue we should view qualitative and quantitative methods as part of a continuum of research. As noted by Clarke and Yaros (1988), combining research methods is useful in some areas of research because the complexity of some phenomena requires data from a large number of perspectives. Closely tied to the arguments for integrating qualitative and quantitative approaches are the benefits that can be obtained from doing so, of which two are recurrent in the literature. The first is to achieve cross-validation by combining two or more sources of data to analyse the same phenomenon (Denzin, 1970). The second, which is more the case in this article, is to obtain complementary results by using the strengths of one method to enhance the other (Morgan, 1998). Research conducted by using different methods can be done simultaneously or sequentially within the umbrella of a main or common project (Tashakkori and Teddlie 2003). Usually one of the methods has more comprehensive relevance to the topic. The supplemental project using the second method may be planned to elicit information that the prime method cannot achieve or to inform in greater detail about some part of the dominant project. In this project, the quantitative approach was the core method, and the qualitative analysis reported in this article was carried out afterwards to supplement the former. The quantitative study in Capo et al. (2007) operationalized a set of relevant attitudinal, background and network variables and combined them into a single regression model predicting performance. This was done through a web survey of PhD students and their supervisors (see Coenders et al., 2007, for details). The goal of the supplementary qualitative study is to understand the PhD students' point of view and to know what or who fostered or hindered their research performance, especially regarding the network variables, which are reported to be relevant in the literature and failed to emerge as such in the quantitative study. In this qualitative study we conducted in-depth interviews with a subset of the students of the same quantitative sample who had been identified either as extreme cases or as typical cases in the original quantitative analysis. 2 Study design The reason to embark on a qualitative follow-up study was that network variables had failed to predict performance in the quantitative study, despite the empirical evidence in the academic literature and in the management field regarding creative jobs of a comparable complexity to that of a PhD. We collected data using in-depth interviews (Rubin and Rubin, 1995). Patton (2002) discusses three types of qualitative interviews: a) The informal conversational interview is completely unstructured and the questions spontaneously emerge from the natural flow of things during field work, b) in the interview guide approach, the topics are prespecified and listed on an interview protocol, but they can be reworded as needed and are covered by the interviewer in any sequence or order, c) the standardized open-ended interview is based on open-ended questions and neither the wording nor the sequence of the questions on the interview protocol is varied, so that the presentation is constant across participants. We used the interview guide approach because we wanted interviewees to talk in a natural way. Additionally, each student could report on issues especially relevant for him or her. The interview guide helps us stay on track; helps us ensure that important issues/ topics are addressed; provides a framework for the questions; and helps maintain some consistency across interviews with different respondents. Prior to designing the interview guide we had a conversation with the leaders of the two PhD student unions which are active at the University of Girona in order to identify hot topics. The interview guide contained only three questions, but respondents were encouraged to also provide additional details through extensive probing by the interviewer. The topics raised by union leaders were also taken into account when asking respondents for details. The questions were worded in such general terms that no clues were provided to the respondent that network variables were actually sought after. The three questions were: 1. Could you explain me your experience of doing your PhD at the University of Girona? 2. Everybody says that publishing is very important for PhD students. Could you explain me your publishing experience? 3. Could you tell me what advice would you give to a new PhD student? The interviews were conducted by one of the authors of this article between July 2007 and May 2008, four years after the quantitative study. The average duration of the interviews was twenty five minutes. We used the sampling techniques called extreme/deviant case sampling and typical case sampling. Using these purposive techniques we sought focus and minimized sample size, so as to select only those cases that best fit the research questions. The extreme/deviant case sampling involves seeking out the most outstanding cases, or the most extreme successes and/or failures, so as to learn as much as possible about the outliers. On the other hand, typical case sampling seeks those cases that are the most average or representative of the question under study. In our case, network variables failed to predict performance in the quantitative analysis because the nine cells in Figure 1 were in more or less equal proportions in the quantitative results. Research group networking potential Low Average High Performance Lower than expected Extreme extreme As expected typical Higher than expected Extreme extreme Figure 1: Typical and extreme cases regarding networking and performance. The qualitative analysis started with the identification of a few cases representative of each of the shaded cells in Figure 1. This was done in order to learn which unknown variables make a difference between higher and lower than expected performers given a particular network potential. In order to select respondents we thus need to construct a measure of research networking potential and a measure of meeting the expectation regarding performance. A measure of networking potential was computed with the assistance of judgement, correlation matrices and principal component analysis of the network measures obtained in the original quantitative analysis regarding the PhD student's research group. Finally the chosen standardized variables and the communalities found in a unidimensional principal component analysis are presented in Table 1. A scree plot showed a clear unidimensional solution: the first dimension explained 53.8% of the variance and the second a mere 15.1%. These results make the computation of networking potential in terms of the sum of the standardized variables in Table 1, reasonable. Table 1: Indicators of network potential of the research group. Communalities in a unidimensional principal component analysis. Indicator_Communality Research group size Number of different institutions the members of the research group belong to Sum of contact frequencies between PhD student and research group or external members with the aim of asking for scientific advice Sum of contact frequencies between PhD student and research group or external members with the aim of collaborating in research Sum of contact frequencies between PhD student and research group members with the aim of obtaining crucial information, data, software, etc. Sum of contact frequencies between PhD student and research group members with the aim of engaging in social activities outside working hours Sum of PhD student trust in research group members in a scale ranging from "complete distrust" to "complete trust" Sum of subjective probabilities of the PhD student to ask for emotional support to other research group members when confronted with serious problems Sum of getting on well feelings of PhD students towards research group members in a scale ranging from "very badly" to "very well"_ In order to compute a measure of how far performance is above or below prediction, we took the studentized residual in the regression model predicting performance in the former quantitative analysis. Furthermore, in order to include heterogeneity regarding styles of doing research, the selected students belonged to different fields of study. In the research tradition of the University of Girona two big families of fields of study are distinguished: natural science/technology (nt) versus arts/social sciences (as). Figure 2 shows the PhD students in the quantitative sample plotted according to the residual and the networking potential. The five areas of interest in Figure 1 are also approximately represented. Our initial aim was to select an equal number of students in each of them. However, some of the targeted students could not be contacted because they were no longer at the university and even their PhD supervisors did not know their whereabouts. The students whose label is within a box in Figure 2 were the ones finally interviewed for the qualitative study. The final qualitative sample size was 16. .880 .222 .703 .713 .592 .642 .925 .599 .925 3- s t u d H e n t i z e d i- i- r e s i d u a -2-I Et ,0 n- Et / ..-■■ „a.s -• äs nt HÜ- nt nt mi-" Et nt nt nt Ei Ei i -10 nt Et nt "nt Ei n\ Et... nt nt nt Et nt Et i0 "T" 15 research group networking potential Figure 2: PhD student plot according to the studentized residual and the networking potential obtained from the quantitative analysis. "as": arts/social sciences, "nt": natural sciences/technology. Labels within a box: students interviewed in the qualitative analysis. The interviews were tape recorded, transcribed verbatim and coded by one of the authors with the help of Atlas.ti software. Another of the authors reviewed the codes and the assignation of paragraphs to codes. We, then, classified the items reported by PhD students either as triggers or hindrances to publishing, and either as related to the student's network or not. Ei nt nt t nt Ei nt 0 i Ei nt nt nt nt nt nt 5 0 3 Results 3.1 Classification of major topics and student groups For the interpretation of the results, students were split into two groups, those who had a grant (6 students) and those who were academic teaching staff during their PhD (10 students). This distinction is of high relevance for PhDs in Spain at the time of conducting the study. 1. Some PhD students already belonged to the university staff prior to starting their PhD. At that time, the lowest categories of teaching staff did not require candidates holding a PhD. The members of these categories of course needed a PhD if they wanted to get promoted, which was the reason why many of them actually started a PhD; however no particular deadline is specified for finishing the PhD. Nothing required these PhD students to belong to a research group although in practice it was so in most cases. Teaching was usually their main job. Their average age was relatively high and some even carried out management tasks at university. 2. In the University of Girona, PhD students could obtain grants from the Spanish government, from the Catalan government, from the university itself or from a particular research project. These grants implied that the awarded PhD students had to be members of a research group. These PhD students had to teach no more than 60 hours a year and, therefore, research was their main job. Average age was lower, as most of these students started the PhD immediately after finishing a five-year degree called licenciatura in the Spanish university system of the time when the study was conducted. The grant did not imply that the students would later get a permanent position at the university and in fact most of them would end up doing a career in the private sector, notwithstanding their hope for the permanent position (Jacobsson and Gillström, 2006). 3. External PhD students did not fall into any of the two previous categories and were excluded from both the quantitative and the qualitative studies. This type of students usually does not belong to a research group and, thus, studying their research performance from their research group networks does not make sense. The codes and the count of students mentioning them are shown in Table 2. As we can see, it is easier for the students to speak about what helped them to publish (92+23=115) than about what hindered them from publishing (20+30=50). The fact that 92+20=112 out of the 165 mentioned items have to do with their networks suggest that networks are more important than as found in the quantitative analysis. Moreover, the classification of some of the items as nonnetwork is not completely clear. Visiting other universities during the PhD and lacking economic resources were classified as non-network items. However, the access to other universities or to economic resources can be facilitated by network members with external contacts and with fund-raising abilities. Table 2: Most often mentioned network and non-network factors that help or hinder performance. Times mentioned Overall Grant No sample n=6 grant n=16 n=10 High supervisor advice 12 5 7 Meets researchers outside research group 12 6 6 Easy meeting with group members 9 4 5 s Ö Group pushes to publish 7 2 5 te hsi Supervisor interested in PhD thesis 7 3 4 T3 3 Supervisor teaches to publish 7 4 3 o S eat p Group helps during PhD 6 4 2 le ot Other PhD students in the group 6 3 3 v M .5 Supervisor collaboration 5 2 3 ro ip l Talk with experts about student's topic 5 3 2 ^ (D (D ^ Group members are friends 4 2 2 eN Group with high scientific quality 4 2 2 Supervisor easy meeting 4 2 2 Supervisor trust 4 1 3 Subtotal 92 43 49 s Visit other universities during PhD 6 4 2 - m ot PhD thesis is the student's main task 5 4 1 - r it g si n o i n il Motivation for research 5 2 3 ž £