The Royal Statistical Society (RSS) has just released a new Data Manifesto which looks at what the next UK Government could do to improve the use of data to improve policymaking, democracy and prosperity. One particular area of interest for policymakers is in sharing and analysis of data including ‘big data’. The RSS sees real potential benefits from data-sharing, such as improving the quality of statistics, improving evidence for decision-making, and for public benefits, such as finding more effective medical treatments. However there are also public concerns about the use of their data. These are in particular related to privacy issues, as most recently seen in the UK in relation to health data (‘care data’).
A recent survey by Ipsos MORI[1] for us suggests that there is a general 'data trust deficit', and that public support for sharing personal data depends very much on who it is being shared with - and for what reason.
The research found that the media, along with internet, telecommunications and insurance companies, all come at the bottom of a ‘trust in data’ league table. Only four to seven per cent say they have a high level of trust in these organisations to use data appropriately, compared with 36% trusting the NHS and 41% trusting their GP. However, nearly all institutions suffer a ‘trust in data deficit’, whereby trust in them to use data appropriately is lower than trust in that institution generally.
Broadly, the poll found more support for data-sharing across government than not sharing data at all, if it is ‘for the benefit of services and me’ and provided there are safeguards in place. The research also found low opposition (17%) to the government sharing anonymised data for public-funded research in universities and similar organisations; around half said they would either ‘tend to support’ (38%) or ‘strongly support’ (12%) this.
Other data recipients and uses, however, attracted a high level of suspicion. Four in ten of those polled said they would ‘tend to oppose’ (24%) or ‘strongly oppose’ (17%) the government sharing anonymised data with companies in order to ‘help improve their products and services, or develop new services’. Only a quarter said they would ‘tend to support’ (22%) or ‘strongly support’ (5%) this.
Health records being sold to private healthcare companies to make money for government prompted the greatest opposition (84%). However if records were instead ‘shared with private healthcare companies to develop more effective treatments’, opposition was far lower at 45%.
Our findings indicate there is a ‘data trust deficit’ whereby trust in institutions to use data appropriately is lower than trust in them in general. The research also indicates that when a case for public benefit is clearly stated and when there are safeguards in place, more of the public takes a positive view in favour of data use and sharing than disagree due to privacy risks. The addition of safeguards such as anonymisation of data, or punishment for data misuse, significantly improves the level of support from 33% to around 51%.[2]
The message for policymakers therefore is that they need to clearly communicate the value of any data sharing they wish to gain support for, and they need to put safeguards in place. It is also noteworthy that there is considerable opposition to sharing data for commercial purposes, and so this is an area where policy makers must tread very carefully.
This guest blog was written by Hetan Shah, Executive Director of the Royal Statistical Society. You can follow Hetan on his twitter handle: @HetanShah
[1] The results are based on two surveys, both from an online quota survey of GB adults aged 16-75. Fieldwork for most questions consisted of 2,019 interviews between 23 and 25 June. One question was placed on a later online omnibus of 1,000 GB adults between 15 to 18 July. To cut down on survey length, some questions were asked in smaller samples, of no fewer than 505. The data has been weighted by age, gender, region, social grade, working status, main shopper.
[2] 51% is the mean of a four-way split sample with c.505 looking at each safeguard – see p. 7 for chart. For each safeguard we can have 95% confidence that support among the general public would be within +/- 6.5% of what we found. It should be noted that whilst we tested public opinion in relation to ‘anonymisation of data’ this is not to indicate the technical feasibility of anonymisation which is a complex matter and indeed some would argue may not be possible at all depending upon the dataset.