Innovations in Digital Research Methods. Группа авторов
Many surveys in the public and private sector collect data on the attitudes of the UK population. A key source is the annual British Social Attitudes survey,62 which began in 1983. It captures the views of a representative sample of around 3,000 people. Alongside key demographics, questions cover a huge range of topics including: public expenditure, welfare benefits, health care, childcare, poverty, the labour market and the workplace, education, charitable giving, the countryside, transport and the environment, the European Union, economic prospects, ethnicity, religion, civil liberties and immigration. Access is free.
Longitudinal surveys are powerful tools for examining people’s attitudes over time. The British Household Panel Survey (BHPS) can capture changing attitudes over time. The BHPS began in 1991, initially covering around 10,000 individuals in over 5,000 households across Great Britain. In 2009, it was absorbed into the Understanding Society panel survey.63 Alongside the attitudinal data there is information covering: key demographics, employment, socio-economic circumstances, residential mobility, marital and relationship history, and social support. Some of the measures are comparable across equivalent European surveys.
Internationally, the European Social Survey64 and the World Values Survey65 provide access to data on public attitudes over 200 countries (see also the European Union Survey on Income and Living Conditions66). Again, access is free though a registration fee can be required.
Commercial organizations also conduct on-demand surveys, which include online panels. YouGov67 is an online opinion-polling organization, which uses signed-up panel members to respond to surveys in return for a small fee. Surveys can be commissioned and completed within 24 hours. There is no limit to the type of issue the survey can cover. Summary findings from previous surveys are sometimes made available via the YouGov website. The data is not available publicly but researchers can commission surveys and have targeted access to particular sample populations in the panels. YouGov have carried out quality assurance of the accuracy of their estimates using their large-scale self-selected panels compared to random sample surveys.68
Other sources of attitude data include social media such as Twitter and blog postings. Twitter data, both almost real time and archival, is available free only within certain limits, above which it must be purchased from Twitter or from approved secondary suppliers. However, as well as there being limits on the number of tweets that can be harvested, individual tweeters’ demographic information is often incomplete. Techniques are being developed to extract and collate demographic information from social media data and profiles (McCormick et al., 2011). Lots of other metadata can be hugely valuable, such as location, language and numbers and names of people they are following and being followed by. Such data has multiple potential uses beyond just the content of the particular tweet. In terms of attitudes, the coding of tweets and re-tweets and Internet search engine terms for attitudes and meaning can be of considerable value.
Blogging, Twitter accounts and websites can be automatically scraped using software and so-called web bots deployed online for collecting self-published attitudinal data. A recent online discussion in relation to the 2014 vote on Scottish Independence reads:
Personally I cannot see how breaking up the union right now can be a good thing. For example, SNP have said they will join the Euro ‘when the time is right’ which currently is probably never. In which case they will continue to use the pound and therefore have to accept that interest rates would be set in London. In addition, there’s no such automatic right of Scotland joining the EU. Again the SNP have said they are ‘confident’ that they would start talks from within, but what does that mean? What if they are wrong? Seems like an awfully big risk to me. It seems to me like Salmond is light on the details and the devil as we all know is in the details. (Anonymous, 2011)
In social science research terms the text could be coded for key variables including: gender, values, political knowledge and also linked to other posts. Language use and framing could also be analysed. Again, this data could be combined with traditional social science research tools, by gathering follow-up contact data and inviting authors of posts to participate in a more formal social science research study.
A key data quality concern here is the issue of fake and multiple Twitter accounts and Facebook accounts69, 70 (including the reported commercial market in creating and managing accounts71). In addition, there is the more substantive issue of the differences between people’s real life identities (that is, their socio-physical identities) and their online identities. This may be particularly problematic in relation to attitudinal data as attitudinal presentation could well be a key part of constructing an online persona. However, no reliable data exists on the prevalence of this phenomenon or how it impacts on the ‘real’ attitudinal data. There are some parallels here with the challenges of what is termed performance in existing social research methods, such as where a respondent provides an answer that they think the interviewer wants to hear, and issues around satisficing where a respondent chooses not to express a particular view and/or gives what they see as a socially acceptable response. For further discussion of data quality in survey research see Blasius and Thiessen (2012).
2.3.6 Data on Social Behaviour
Increasingly, social science research is attempting to analyse people’s actual behaviour alongside self-reports of their behaviour. In part, this is driven by evidence that respondents often misreport their behaviour. For example, many more people will state when asked that they voted in an election than actually did (Pattie et al., 2004) and many more report that they recycle household waste than actually do in practice (Whitmarsh, 2009). This might be a result of recall failure, or of social desirability bias reflecting perceptions of the civic duty to vote or be environmentally friendly.
As we have outlined above, there are different types of data sources that capture actual behaviour and data relating to real time activity is of great potential value here. For example, in relation to well-being and health, consequential data such as consumption information and administrative records of gym attendance and mobile tracking data on individual movement and exercise can be used. Mobile application data can also track social network activity. See, for example, Bucicovschi et al. (2013) who used data provided by the communications company Orange to map connections between 1216 cell phone towers. The volume of calls passed from one tower to the next revealed ‘communities’ and who talks to whom. Not all such data are easily available for social science research, however, and access is often reliant on effective partnerships between researchers and data holders in the private sector.
We consider a hypothetical example of how such a joined-up data approach may be designed. A social science researcher interested in anti-social behaviour has a range of possible data sources to use. They could combine consequential administrative data on anti-social behaviour with intentional data from the British Crime Survey. They might also analyse social media data from police forces and officers. As mentioned above, a police force recently posted tweets of all incidents dealt with in a 24-hour period. Example tweets include:
Call 215 stolen vehicle heading towards Manchester #gmp24 Thursday October 14, 2010 5:03;
Call 216 harassment report in Bolton #gmp24 Thursday October 14, 2010 5:03;
Custody update 101 in police cells at 5 am #gmp24 Thursday October 14, 2010 5:04;
Call 218 neighbour dispute in Wigan #gmp24;
Call