new data sources (and social media sources in particular), this issue is likely to be all the more salient. Social science researchers must be aware that the media and data are mutually embedded in a manner that affects the data we might use. The flip side of this challenge is that the gap between data and subject, between my-self and my-data or data collected about me, is closing both temporally (data about us is more closely contemporaneous to the activity/behaviour that has generated the data) and ontologically (more and more the data and the activity are one and the same thing).
So where does this take the social scientist? Is it just more choices about the data to collect, use and link? Or is it more fundamentally a step change in how people’s lives are being captured, documented and measured? What challenges are posed, for example, in terms of data access and data quality?
2.2.2 Evolving Traditional Data Types
To understand the changing data environment it is important to begin by reflecting on the expansion and enrichment of so-called orthodox intentional data and, in particular, social surveys. In the UK, longitudinal surveys, including the British Cohort Study,12 the English Longitudinal Study of Ageing13 and the Census Longitudinal Study14 now constitute very rich sources of information for understanding change over decades of people’s lives (see also Chapters 4 and 5). International surveys, such as the World Values Survey15 can provide insights into global opinion. Such surveys and the analyses conducted on them often include contextual data, such as area-level employment rates. The data can be analysed online through tools such as Nesstar.16 Nesstar allows users to access and analyse archived data, including government survey data such as the Labour Force Survey,17 the British Social Attitudes Survey18 and the British Crime Survey.19 The resource includes information on data origin, sampling and the coding of variables, as well as the original questionnaires. Similar data resources also now exist for qualitative and mixed methods data. Textual data from interviews, focus groups and observational studies can be accessed through, for example, the UK Data Service.20
Survey data can be highly detailed and, depending on the sample design, allow inferences to be made to a wider population. However, one of the limitations is the survey process itself in terms of the sample size, usually restricted by cost, which can limit comparisons between areas and groups due to low sample numbers. In addition, response rates can be low, which can introduce bias into the population estimates. The survey questions themselves can also have limitations in terms of constrained responses and self-reporting bias. The latter is where respondents give an answer they feel is expected, rather than what they really think. Moreover, there is also a well-established gap between reported attitudes and what people actually do in practice, and indeed between what people say they do and what they actually do (see De Vaus, 2002; Blasius and Thiessen, 2012).
Secure data that is not regarded as suitable for widespread, unregulated public use can be analysed in so-called ‘safe settings’ to ensure there are no risks to confidentiality (see, for example, the UK Secure Data Service,21 the HM Revenue and Custom’s (HMRC) Data Lab22 and the Ministry of Justice (MoJ) Data Lab23). The Secure Data Service allows access to individual-level data that is more detailed than that available under standard licensing (such as smaller geographic areas) and so provides potentially richer sources of evidence for social science research. The user analyses the data remotely rather than downloading it. The analytical outputs are then checked by the data provider. The conditions of use are based on licensing agreements with users as well as user accreditation, individual training and trust. HMRC’s data lab allows access to individual tax records but requires users to do so at the HMRC’s premises under controlled conditions. As part of the MoJ Data Lab, organizations that work with offenders can now apply for their offender data to be linked to the MoJ re-offending data. It is stated that MoJ analysts will prepare a report detailing the re-offending rates against a matched control group of offenders with similar characteristics, together with a conclusion on whether the service provided by the organization is associated with a change in re-offending behaviour. Compared to this, as we discuss below, the infrastructure and guidance for accessing and using new types of data such as Twitter and blogs are not well established but are developing.
The growing access to official administrative records of public service use also includes, for example, patient health records and school performance records through such initiatives as the UK Administrative Data Liaison Service (ADLS).24 These consequential data sources are, in theory, complete rather than being based on samples and can be of great value given their detail and coverage. However, like all datasets there are likely to be issues of missing data that the social science researcher needs to be aware of, including individuals who have not been traced or recorded. There are also likely to be duplicate records. We discuss an example of this in Chapter 3. Access to specific variables can also be restricted and the coverage, of course, is limited to the variables collected as part of the administration process and the time point of the data collection. As a result there are limitations on the number of research questions that can be addressed.
Research access to this kind of administrative data is part of the UK Government’s drive towards Open Data, whereby government departments and agencies are being required to provide greater access to service use and performance data for the purposes of transparency and accountability (Open Public Services White Paper, 2011).25 For an overview see Halford et al. (2013), Wind-Cowie and Lekhi (2012) and Shakespeare (2013). As well as providing an alternative to orthodox intentional data, administrative data also expands the range of available information. The use of such data is likely to be new to many social scientists and its properties, coding frames and terms of use can be very different and may require the acquisition of new skills, alongside new knowledge or greater interdisciplinary working. Nevertheless, it is notable that in a recent survey of over 300 (self-selected) social science researchers, nearly two-thirds had used administrative data in their research (Elliot et al., 2013), although a similar proportion (61 per cent) reported encountering barriers when trying to access such data.
2.2.3 Innovations in Linking Data
Methodologically, there are increasing opportunities to address research questions by data linking using statistical matching and drawing on multiple data sources. Well known examples of this include: the linking of hospital data to the Millennium Cohort Study26 (see Calderwood, 2007); the Work and Pensions Longitudinal Study (WPLS)27 which links benefit and programme information held by the Department of Work and Pensions (DWP) with employment, earnings, savings, tax credit and pension records from HMRC; and the Longitudinal Study of Young
