straightforward extrapolation to provide a measure of prevalence, though the evidence suggests prevalence is significant. A more robust research design might ask respondents to report key demographics, change over time and to describe how their experiences compare with those of people they know in their social networks.
Another functional difference of many of the new data types is that the data is often generated directly by individuals and organizations themselves for their own purposes. A recent example of what we term self-generated data involves a UK police force using Twitter to announce all the emergency calls it received in order to highlight their work over a given period.36 This constitutes a potentially rich source of data for social science research, produced not via a traditional social science research design but self-reported by an organization. We consider this type of data in more detail below.
The United Nations is embracing the potential of digital evidence in relation to human rights and also examining policy impacts in almost real time.37 This can take the form of: monitoring food price discussions, money transfer patterns via mobile phone, or tracking health concerns expressed through Twitter or captured in Internet search records. Such techniques can include feedback loops where people’s attitudes and behaviour can be followed up. The data can be used to conduct almost real time research. However, data quality assessment practices need to be used and the verification of data still needs to take place. Quality assurance mechanisms are being developed which involve volunteers validating data.
There is a link here to what is termed citizen science and crowdsourcing, whereby people voluntarily allow the collation of their own data, or contribute data that they gather themselves, and also undertake data processing and coding. Data is being generated by citizens not only about themselves but also about issues they might have an interest in. For example, in the Satellite Sentinel Project, citizens are being asked to volunteer to observe and code images for evidence of human rights abuses (e.g. military activity or signs of explosions) in Sudan sourced from a network of private satellites.
In the changing data environment, there are increasingly detailed records of actual behaviour accumulating alongside survey data on people’s reported behaviour, and there is scope for reporting and monitoring behaviour in almost real time to complement more traditional social science research data gathered retrospectively through surveys, interviews and diaries. For example, purchase data collected by supermarkets could be used alongside food diaries; mobile phone movement data could be used alongside self-recorded time use data; and health monitoring data could be used alongside surveys of people’s self-reported health.
The step change for social science research lies in the potential, where appropriate, for identifying and bringing together the different data types described in our eight-point typology: orthodox intentional data, participative intentional data, consequential data, self-published data, social media data, trace data, found data and synthetic data. Synthetic data can be used as part of simulation and agent-based studies. For an example, see the recent UK project on the Social Complexity of Diversity, which uses computer-based simulations, and Chapter 6 in this volume.38
Moreover, almost real time data opens up opportunities for what may be termed ‘real time’ or ‘live’ social science, though this clearly challenges standard practices and timescales in research for data quality assurance and for peer review.
2.3 Combining Data and Mixed Methods – Key Research Areas
A useful way to examine and understand the new types of data in context is by comparison with existing forms of social science research data. Through a series of broad social science research policy areas, we will now consider some orthodox intentional data sources and research designs (such as social surveys) and identify other new data sources that might be used in combination with them as part of mixed methods studies. We compare different types of variables in each of the key areas. In doing so, we will use the UK as an example country, whilst acknowledging that data environments vary across countries and that some data sources transcend national boundaries.
2.3.1 Data on Economic Circumstances
Key sources for capturing data on people’s economic circumstances in the UK include the Census39 and the Labour Force Survey (LFS).40 The Census provides a profile of the UK population every 10 years. It collects information on people’s employment, health and family circumstances. It is a key tool in estimating population change. Data from the Census is available in summary tables as well as in samples of microdata. Specific data tables can also be requested for an administration fee. The questions on economic circumstances are, however, limited. It is notable that it is anticipated that the 2011 Census will be the last full census in the UK and there will be a shift in the future to smaller-scale data gathering and use of administrative records.41
The LFS is a quarterly survey of over 60,000 households. The LFS is now linked with the UK Annual Population Survey and includes increased coverage of urban areas down to local authority district level. The data includes a longitudinal component, with respondents being interviewed five times at three-monthly intervals. Questions cover such variables as people’s key demographics and occupation, training, health, earnings and benefit claims. Some of the measures are internationally comparable. Access to the data from such surveys as the LFS is often free (although usage is not completely unrestricted).
For many survey datasets, access to particular variables, geographic levels and detailed information is restricted because of concerns about confidentiality and statistical disclosure risks.42 For example, only samples of UK Census data and certain variable codings are released at particular geographic levels. This can inhibit analysis at lower geographies. For some government surveys and datasets, special licence use versions are available which contain more detailed variable codings and geographic information.
Other sources of data on people’s economic circumstances include income data available from commercial data providers. The data is updated from different sources, including surveys and other data gathering tools such as product warranty forms. Such data provides income estimates at the individual level, though these are often imputed. Many of the variables have bounded values, for example, age and income are in bands. Other variables cover people’s spending, savings and debts.
Consequential data, such as administrative data, including information on earnings, tax payments and benefits claims, are held by government departments and, if not released directly, can be available for social science research purposes under special agreements. Some commercial information is also available in the public domain. For example, organizations such as estate agents necessarily release data on properties on their books as part of their core business. If made available for research purposes, data from the Citizens Advice Bureau (CAB), and agency and bank consultations concerning debt advice, which includes anonymized client details, type of problem, advice given and outcomes, could also be examined alongside publicly available data from land records and on share ownership.
Open data resources such as OpenStreetMap43 can be used to map areas of deprivation and can be combined with official data such as the ONS Neighbourhood Statistics.44
