in traditional face-to-face, telephone and mail surveys, alongside the opportunities that technical innovations provide for enhancing the quality and efficiency of survey research methods. Drawing on a selection of major social science surveys, Murphy offers examples that point toward the continuing importance of survey-based methods in the social sciences.
Murphy observes that, despite the proliferation of born digital data, recent years have nevertheless witnessed an explosion in the quantity and diversity of data generated through survey research. This has been facilitated by developments in e-Infrastructure, an example being the unprecedented opportunities for the recruitment and retention of respondents afforded by the public’s mass adoption of email and, subsequently, social networking sites. Similarly, survey researchers have benefited from the increasing availability of paradata, that is, data about survey transactions and interactions with respondents, which can be used to gain insights into their motivations and the meaning behind their responses. Such e-Infrastructure affordances have made significant advances in the capture, analysis and dissemination of survey data possible. They lead Murphy to argue that, contrary to predictions that new sources of data will make surveys redundant, they offer ways both to make surveys more effective tools and to meet the challenges that have threatened their value. For example, the availability of administrative data and methods for matching it with survey data hold great promise for minimizing respondent burden and cost. In a different vein, Murphy observes that virtual worlds, such as Second Life, offer new ways of conducting interview-based surveys.
Nevertheless, Murphy reminds us that social media brings new challenges, in particular, the problems of bias through samples of unknown representativeness, and quality assurance. The prospect of using social media as a substitute for traditional surveys – for example, the use of Twitter as a way of measuring public opinion through sentiment analysis – is often heralded as a sign of their imminent demise. Murphy, however, warns of the dangers of relying on such data where there is ‘… no standardization or check on the validity of the information being shared’. He argues, instead, for more research into the value of Twitter as a means to recruit respondents, citing as an example a recent study where it was used in diary data collection. Finally, Murphy discusses the potential of mobile devices for SMS-based survey delivery, noting its efficacy for administering them at predetermined times or in the context of specific events or – when used in conjunction with GPS – specific places.
Murphy’s conclusion is that survey methods are continuing to play a major role in social research, and pessimism about their survival is misplaced. This role, however, is increasingly being shaped by people’s use of communication technologies. Given the rapid pace of innovation of these technologies, the future for survey methods remains hard to predict.
Chapter 5: Advances in Data Management for Social Survey Research
As argued in the previous chapter, despite the availability of new sources of social data, making optimal use of more conventional data sources such as surveys remains of critical importance to social research. However, using survey research data can present major challenges for data management. For example, pursuing a particular research question may require linking different datasets, extracting variables, combining them and recoding their values before statistical analysis can start. In this chapter, Lambert argues that data management practices have failed to keep pace with these challenges and explains how e-Research can advance the state of the art, drawing on examples of working with quantitative datasets generated through social surveys taken from the DAMES (Data Management through e-Social Science) project.2 He argues that enhanced facilities for file storage and linkage, for using metadata to describe data, and for the capture of data preparation routines (‘workflows’) can raise standards in data management and help researchers share their experience and expertise with one another. (Exercises illustrating each of these facilities can be found at the book’s website.)
Lambert concludes by examining the prospects for the adoption of more advanced data management tools and practices. Using an example where ‘bottom-up’ and ‘top down’ innovation processes might successfully complement one another, he notes how the push from journals and funding agencies for researchers to publish metadata about their data management is likely to have a decisive influence.
Chapter 6: Modelling and Simulation
Quantitative simulation and modelling are perhaps the most obvious examples of the potential for e-Research methods and tools to revolutionize the study of complex socio-economic problems, and their applications are becoming increasingly widespread. New sources of data and more powerful computational resources have made possible the development of more complex and sophisticated techniques and, of course, larger-scale models. As Birkin and Malleson point out in this chapter, while modelling and simulation in the social sciences have been around for fifty years, prompted by an earlier wave of innovations in computation, recent advances in both data and computation are now having a profound effect.
This chapter provides an introduction to the state of the art in four model classes that are of particular interest to social scientists – systems dynamic models, statistical and behavioural models, microsimulation models and agent-based models. Examples are presented of each of these classes – a retail or residential location model (spatial interaction model or mathematical/systems dynamic model); a traffic behaviour model (discrete choice or statistical model); a demographic model (microsimulation model); and a crime model (agent-based model). Birkin and Malleson observe that while building ever more sophisticated models of social systems has never been easier, the task of demonstrating that such models faithfully represent an underlying social reality remains the key challenge. They then relate some experiences and lessons from building a prototype social simulation infrastructure capable of providing support for the whole research lifecycle, and they stress, in particular, the importance of model reproducibility, reusability and generalizability. They conclude with a summary of some of the – as yet – unexploited opportunities for social simulation presented by new sources of data (e.g., using mobile phone data to update in real time models of population movements) and the challenges (e.g., data ownership and ethics) that will have to be met if these are to be realized.
Chapter 7: Contemporary Developments in Statistical Software for Social Scientists
In this chapter, Lambert, Browne and Michaelides examine the prospects of the quantitative social sciences being in a position to exploit the power of new social data, computational resources and tools to achieve advances in statistical analysis. They review the range of statistical software packages currently available to social researchers and the factors influencing their patterns of adoption. They illustrate their review with examples of the application of statistical methods in domains such as education, health inequalities and epidemiology. They argue that the profusion of statistical tools, while having the benefit of offering choice to researchers, nevertheless raises significant barriers, both social and technical (and, indeed, socio-technical), that need to be addressed if the power of the tools is to be fully exploited by the social science research community.
Regarding social barriers, the authors note that in the UK there is a lack of capacity in statistical skills within the social research community. Regarding technical barriers, they observe that the proliferation of statistical tools has been at the cost of inter-operability and has created a situation that they describe as ‘balkanization’. This can deter researchers from using the tool most appropriate for a particular analysis – rather than the one they are most familiar with – and may also inhibit experimenting with new tools. Echoing the concerns raised by Purdam and Elliot, they also point to problems with transparency, replicability and robustness of statistical analyses using computer packages whose algorithms are not accessible to the user. Drawing on the principles of e-Research for their inspiration, Lambert et al. conclude by presenting some ways of overcoming the social and technical barriers, which they exemplify
