Conference Programme 2015

Conference floor plans and map
Tuesday 14th July      Wednesday 15th July      Thursday 16th July      Friday 17th July     

Thursday 16th July, 16:00 - 17:30 Room: HT-101

Measuring Social Networks in Large-Scale Surveys: Challenges and Practice of Ego-Centred and Complete Network Approaches 2

Convenor Mr Benjamin Schulz (Mannheim Centre for European Social Research and WZB Berlin Social Science Center )
Coordinator 1Mrs Kerstin Hoenig (Leibniz-Institute for Educational Trajectories (LIfBi), Bamberg)
Coordinator 2Professor Reinhard Pollak (WZB Berlin Social Science Center and Freie Universität Berlin)

Session Details

Survey researchers measure social networks in two fundamental ways: i) in an ego-centred manner that captures an actor’s ties and related characteristics, ii) in a broader way that captures complete networks within predefined boundaries. The latter approach gains increasing attention as recent projects in many countries and fields collect complete network data.
This increase largely follows from advances in survey instruments for complete networks and in statistical modelling, especially for network dynamics. Methodologically, complete network analysis makes it possible to separate selection and influence processes. These surveys, however, are mainly conducted in schools as meaningful network boundaries are easy to implement in this context.
Ego-centred measures offer the chance to collect network data in contexts where a complete network measurement is not feasible. To meet the challenge of reversed causality and endogeneity as a consequence of the non-random, often homophilous, formation of social ties, panel data including repeated, or prospective and retrospective, measures are promising. Several international large-scale surveys made significant progress in this domain.
By bringing together researchers from both camps, we seek to promote a discussion that allows for a better evaluation of the advantages and pitfalls of each approach. The session's focus shall be on longitudinal measurements and statistical modelling of social networks, especially on means to identify causal network effects. The second focus shall be on ways to assess the reliability and validity of ego-centred and complete network measures. For ego-centred network measures, such as name or resource generators, the reliability of these instruments will be of central interest. For complete network measures, the specification of network boundaries is a heavily debated issue. Contributions might also include studies on how to identify and deal with compositional changes or on inference errors that may follow from insufficiently specified boundaries.

Paper Details

1. A new tool to collect ego-centered network data in online surveys
Dr Tobias Stark (Utrecht University)
Professor Jon Krosnick (Stanford University)

This study presents a new data collection tool for ego-centered networks in online surveys. We make use of Web 2.0 graphical features to create a more engaging experience for survey respondents when answering the same question for all of their network contacts. We will present the finished tool and report results from a survey experiment with an U.S. online convenience sample (N = 430), comparing the data quality of the new tool with a classical ego-centered online survey.

2. How should we ask surveys questions about how many people do know you? Experimental evidence regarding the role of response categories and follow-up questions.
Dr Matías Bargsted (Institute of Sociology, Catholic University of Chile)
Dr Luis Maldonado (Institute of Sociology, Catholic University of Chile)

We report results from a survey experiment whose aim was to identify measurement conditions that provide more reliable data of the size of respondents’ acquaintanceship networks. The experimental design involved a test-retest component and two randomly assigned manipulations that altered key features of survey items asking respondents how many people they knew from a list of occupational groups. We argue that respondents’ asked with ordinal response scales and without follow-up items, compared to numeric scales and follow-up questions, will experience less response burden and, thus, will provide higher test-retest correlations for the network items. Findings provide

3. Interviewer Differences In Social Network Surveys
Dr Sören Petermann (Goethe-University Frankfurt am Main)
Dr Andreas Herz (University Hildesheim )

We investigate interviewer effects on ego-centric network measures. While studies consistently show that the size of reported networks varies by interviewers, we contribute to this line of research by a) analysing not only network size but other network features like density and composition, b) disentangling interviewer and locality variations, and c) researching homophily between respondent and interviewer.

4. Order effects in the position generator
Professor Bonnie Erickson (University of Toronto)

The position generator has been used widely and productively. However, no
previous surveys have varied the order of presentation of occupations in
the position generator, so we do not know whether order effects exist nor
what they are. This paper reports on a recent survey in which order of
presentation was randomized.