Exploring Connections Between Political Parties and Personal Data Brokers in the UK
By Amber Macintyre
In Short: Insights into an investigation on the use of voters’ personal data in political campaigns through the lens of a Tactical Tech case study. It focuses on techniques used to understand political party spending on data about voters and on online marketing technology in the United Kingdom, during the 2015 and 2017 general elections and the 2016 ‘Brexit’ referendum.
Political campaigns are increasingly relying on our personal data, including our contact details, political preferences, our interests and behaviours: travel routines, shopping and online browsing habits. This data is used by political campaigners to optimise their public communications and mobilisation efforts to influence our attitudes toward issues and candidates and, ultimately, to influence how we vote. They use data to profile people’s opinions, personalities and preferences and to find the most engaging messaging tactics for individual voters through testing different language styles and channels of communication, from TV to social media to door-to-door visits.
The collection of this information is conducted both by so-called personal data brokers and digital campaign consultants – people and agencies equipped with technology and skills to amass people’s profiles, preferences, frustrations and to process this information for commercial gains. While not an illegal activity by default (unless the data is obtain fraudulently), this practice is controversial and has raised various concerns, including about violations of campaign laws and privacy regulations, manipulative messaging, exclusion of social groups through selective message targeting, and the unethical profiling of individuals. However, due to a lack of transparency on what these actors do, how they do it, what software they use and how it all works, it is difficult for voters to decide or control how much personal data they would share, even if they had a choice.
This case study is based on a subject from the United Kingdom and shows how our team at Tactical Tech investigated issues related to the use of citizens’ personal data in recent UK political campaigns. Specifically, it gives you an idea of how we set our goal and asked the initial questions, where and how we searched for sources of information and evidence, and what resources, tools and techniques we used in order to make sense of the information we collected.
With a focus on methods, the case covers a small fraction of the entire research we did as part of a long-term project on Data and Politics. It illustrates a few of the many steps we undertook to explore new territories of how political influencing works in data saturated environments where people’s most private information is now used by data brokers as a resource in the race for successful political campaigns.
The research we describe below is part of The Influence Industry, a project developed by Tactical Tech to investigate the use of people’s data in political marketing and campaigns around the world, in countries such as Brazil, Kenya, India, Mexico, US, Canada and many others.
Screenshot from Tactical Tech’s project “The Influence Industry.”
The aim of our investigation was to find out what kind of data was being collected and used by political parties, interest groups and candidates involved in the UK’s 2015 and 2017 general elections, and in the 2016 ‘Brexit’ referendum. We wanted to examine how politicians and political parties purchased voters’ data and profiles from data brokers, how they paid for services and ads on platforms such as Facebook, and hired digital communication experts who used – or instructed the them on how to use – data to develop targeted and personalised content for their political campaigns.
How we collected the data
As with any other research, we started by scanning what was already out there about the topic. This generally required extensive reading, talking to knowledgeable people and even seeking specialist advice, as well as collecting information about people and organisations we were investigating. Given the political nature of the research, we needed to be mindful of any potential risks involved in accessing sensitive material or approaching public individuals, companies and political parties.
While any investigation involves various sources of data and techniques, the following list outlines our process for gathering and working with relevant material over the course of this project:
Reading other investigations and existing articles on the topic - to develop a sense of what is already out there, what cases have been uncovered and what questions haven’t been addressed or answered yet.
Collecting and analysing data gathered by UK institutions, especially election regulators, and other publicly available sources.
Understanding party practices from a political supporter perspective, such as by attending political events or using online political party platforms and campaign apps.
Collecting and analysing materials published by the parties and political campaigners – to understand specifics about campaign practices and voter targeting strategies they adopt.
We worked with the premise that interviews or other forms of direct access to parties and campaigning firms would not be possible in most cases. We assumed this not only because campaigning practices and the use of voters’ data are largely opaque and generally qualified as campaigning secrets during active election campaigns, but also because the issue had already escalated with the Cambridge Analytica scandal, which emerged during our investigation.
For this research, we looked at the campaigning practices in the 2015 and 2017 general elections (elections for the Members of Parliament in the UK House of Commons) and the European Union (EU) membership referendum of 2016 that led to Brexit negotiations. This helped to reveal how digital campaign tactics and spending changed over time and to understand how they differed between general elections and a one-off, high-stakes referendum.
It is worth mentioning that the level of access to public information that we had in the UK is not the same everywhere in the world, especially concerning political party practices, campaign advertising and public spending on such activities. The UK Electoral Commission publishes financial reports and invoices of campaign spending, as well as results of its own investigations into election-related matters, among other resources. However, our investigative process should prove relevant to other contexts too, and, as an investigator, you will need to constantly adapt your tools and techniques to find evidence in a particular situation, while also making sure you are fully aware of any potential legal, ethical and safety risks.
How we analysed the data
Other investigations and existing articles
There is no need to repeat work that has already been done, and the media generally scrutinises activities of political parties during every election and referendum.
In our case, for the UK’s 2015 election, there were several articles about political spending on online platforms – such as this report in the Guardian. The media’s interest in this topic grew by the time the 2016 EU referendum and the 2017 election arrived, as major news outlets continued to publish articles on the subject, such as these ones on the BBC and The Guardian.
A heightened media scrutiny into political spending was accompanied by several official investigations conducted by the UK government. One such investigation took place in a series of parliamentary hearings between 2017 and 2018 conducted by the UK Department for Digital, Culture, Media and Sport (DCMS). The hearings looked into the impact of fake news on the political process. Among other findings, the hearings demonstrated how widespread and normalised data-driven practices across political campaigning in the UK have become.
When gathering information from media sources and other reports, it is crucial to check the accuracy of data and conclusions as an essential step before using any of this information in your investigation. Here are some measures you can take:
Check the reporters’ and researchers’ sources. Sometimes the news outlet conducts their own investigation; sometimes they cite content from elsewhere. Check if this is referenced in the article and, if possible, find the original source. This extra effort can help place the facts in their original context and show how important details are sometimes manipulated or excluded from one report to the next.
For example, in May 2017, the Guardian reported, ‘It was with AggregateIQ that Vote Leave (the official EU Leave campaign) decided to spend £3.9m, more than half its official £7m campaign budget,’ a figure that was then repeated in various other sources, including Business Insider and Bloomberg. This figure, however, is not consistent with the Electoral Commission spending records which show an overall spend of £3.5 million from Vote Leave (£400,000 less than the original Guardian article claimed), including all other Leave campaigning groups. More about this in our UK campaigning report.
Cross-reference your sources to determine how reliable they are and to verify data and facts. Make sure that the same numbers, statements, datasets or affirmations are supported by more than one source (media, person, institution etc.). If possible, use various sources to corroborate the same piece of evidence.
Finally, it is important to consider any political or ideological agendas or biases a news source may have. A newspaper may be explicit about their political preferences, for example, by declaring their endorsement for a particular candidate in an election. Otherwise, it might be useful to research the history and ownership of the newspaper through available company records, academic sources or a thorough online search. In some cases, if you are not familiar with the region you are researching, it is worthwhile talking to a reliable source who has knowledge of how a newspaper is perceived locally.
Data from UK regulators: Political spending in the UK
Many countries have formal regulations and monitoring of election campaign activities and spending. This monitoring requires collecting data that is then often made available to the public and can be very useful in an investigation. Unfortunately, not all countries make the data public, but it is worth checking if there are any government or related monitoring bodies that either publish election-related information or can be directly contacted to request such data.
In the United Kingdom, the Electoral Commission is tasked by law to set spending limits in elections and to monitor activities of political parties. To document their spending, each campaign group must submit invoices, which the Electoral Commission verifies and publishes for public reference.
Screenshot of the UK Electoral Commission website: http://search.electoralcommission.org.uk/
The following is an overview of how we used the Electoral Commission’s data in our investigation.
THE UK INVOICES
Obtaining and reading the data
The first step of analysing the invoices for this investigation was to see what information the Commission acquired and to download the dataset for further analysis. We did this by going to the Electoral Commission’s database of election spending, which can be found here: http://search.electoralcommission.org.uk/.
This database provides a wide variety of financial information on all official campaigning individuals and groups across all elections and referenda since 2001.
Since the Electoral Commission makes a vast amount of data accessible, it may be necessary to first narrow the scope of the search by selecting relevant criteria. The Commission’s database lets users narrow your queries by selecting the various types of political actors: parties, campaigners, referendum participants and regulated donees (individual members of political parties, elected office holders and members’ associations). In the same section, there are options to choose one type of financial information: donations, loans, spending, registration details and accounts related to all these individuals and groups.
Showing the criteria selected on the Electoral Commission database. Screenshot from http://search.electoralcommission.org.uk/
Here, we chose to search for the spending of all the groups included in the list. We can then refine our search further, according to the type of election, the specific event and whether to include spending incurred outside of Section 75 (a regulation that limits how much can be spent based on the number of voters). As we wanted to know all spending costs, not only those within certain limits, we left it ticked, as can be seen in the image below. While we sought the same spending information for both general elections and the EU referendum, we wanted to keep each event’s data separate for easier comparison. We ran the search three times for each event and ended up with three collections of data.
Showing the filters selected on the Electoral Commission database. Screenshot from http://search.electoralcommission.org.uk/
At this point, the database shows embedded in the website and from here it is relatively easy to search for individual instances and examine some of the spending. It is not possible, however, to sort this information or analyse it in any significant way. At the bottom of the page, the data can be exported to a .csv file, a file format that makes it easy to exchange datasets from different applications. The .csv file we obtained is a type of spreadsheet that can be read in Excel, LibreOffice Calc, or Google Sheets.
After opening the .csv file in the free and open-source program LibreOffice Calc, we could use the ‘sort’ and ‘filter’ functions of the spreadsheet editor. Firstly, we could filter by different company names, such as Facebook or Google. Then we could sort the ‘amount’ column by either the largest or smallest financial values. The table below shows the types of categories provided in the spreadsheet, by which the information can be sorted and filtered:
|Reference Number||Identifying number for each invoice given by the Electoral Commission|
|Reporting Period Name||The election or referendum in which the invoice was submitted|
|Regulated Entity Type||Political party or third-party organisation|
|Total Expenditure||The total amount on the invoice (for 2015, anywhere between £0.12 and £528,670.21|
|Date Incurred||Date of the invoice|
|Expense Category Name||The Electoral Commission’s expense categories include: overheads and general administration, transport, market research/canvassing, unsolicited material to electors, advertising, campaigning broadcasts, rallies and other events, manifesto or referendum material|
|Supplier Name||The name of the company or individual supplier|
|Supplier Address||The business address of the supplier|
|Amount per region||Split by England, Scotland, Wales and Northern Ireland|
|Invoice||Attachment with an image of the invoice|
By using pivot tables (tables that summarise and collate data from other tables) to access data from the larger dataset, we could then add up costs for certain categories and parties. This is a quick, useful way of determining how much a party or organisation spent on advertising or market research, for example, or how much money was paid to different companies like Facebook or Google (see images below).
Using a pivot table to chart political party spending by category, such as on advertising or market research
Using a pivot table to list political party spending by supplier, highlighting spending on Facebook and Google
Analysing the data
While the Electoral Commission’s datasets are very useful at providing a general account of campaign spending, we had to make several adjustments to refine the data in a way that was more useful to the focus of our investigation. In order to account for spending that was specifically for either digital campaigns or data, we had to make our own category: “companies that received political party money for working with personal data.”
While creating a custom category involved more research, the Electoral Commission database was still able to provide helpful information in order to refine the spreadsheet data to our needs. By examining the invoices that must be provided to the Electoral Commission, and serve as the basis for the data in the spreadsheet, we were able to determine more accurately whether the spending involved digital campaigns or data purchases. We achieved this by a two-step process:
1. Searching the invoice for context clues
The Electoral Commission database includes copies of invoices. These are available in the online database but are not directly linked in the downloaded csv. Examining these documents helped us understand what the companies and technologies hired by the parties do. However, while some invoices may provide more or less information about what services have been provided to the parties, others don’t. Sometimes, further research into the providers might be needed in order to shed light on their activities as well as their company history, client network and reputation.
For example, from invoice details and online searches of the money recipient’s (company or service) name, we could determine that Alchemy Social, a service employed by the UK Labour Party, specialises in social advertising campaigns. A simple search reveals that Alchemy Social is run by Experian plc, a large consumer credit reporting company that operates internationally by collecting and aggregating data on people and businesses, and that has allegedly generated detailed profiles on more than a billion people.
An invoice showing services charged to the Labour Party by Alchemy Social for "activity" on Facebook. Available at http://search.electoralcommission.org.uk/
2. Investigating the service provider
You may not always be able to build a complete picture from available invoices. Descriptions of the billed services can sometimes be obscure, while some invoices may be missing if the party has not submitted everything to the commission or has channelled expenses through other campaign partners. In the case of the Brexit campaign, for instance, the UK Electoral Commission fined Vote Leave (the pro-Brexit campaign) for breaking electoral law by funnelling over 600,000 GBP to pro-Brexit youth group BeLeave in order to avoid breaching its seven milion GBP campaign spending limit (this limit was imposed by the Commission on all campaigns).
The invoices we looked at didn’t provide explicit details, so we had to investigate more about the companies themselves. In these circumstances, we primarily relied on materials published by the companies themselves, usually on their websites.
An example showing how little detail can be contained in an invoice. This is a payment from the Conservative Party to the Messina Group, which advises on digital campaigning strategies. Available at http://search.electoralcommission.org.uk/
In cases where the company had no website, we found out more about their mission or activities on LinkedIn or other social media platforms. In some extreme cases, it was only through official business records that we could find information about the business and its activities.
In the UK, business records are available online and for free from the Companies House. You can search either for the name of the company or the name of a founder or director. This method does not always yield information about what the company does, although the main type of activity is often mentioned. It is, however, very useful to see the founders’ or directors’ names, which would allow you to then research whether they are connected to a political party or other businesses, for instance.
Following the campaign money
Researching the companies behind these invoices revealed a second layer of digital campaigning to investigate: the complex relationships among and between different data brokers, political marketing and tech companies.
In the case of Alchemy Social, we can draw some tentative conclusions from the fact that its Facebook services seem to be powered by the resources of its owner, Experian plc.
This example shows how additional research into the connections of money flows and organisations – which are not initially obvious in the Electoral Commission data – can provide us with a larger picture of the context of data spending as well as insights into the companies’ activities.
Political party gives money to Facebook or they give money to intermediary companies like Alchemy Social who are the go-between in money flows from campaigners, Facebook, and Experian
It is useful to keep in mind that another way money travels and is disguised from initial view is in the form of donated services from companies that are owned by or at least friendly to the political cause. There won’t be any trace of an invoice in the Electoral Commission’s database if a company simply provides a service for free. While it can be difficult to find such evidence, it is sometimes revealed by legal investigations. For example, the UK information commissioner found that another 2016 pro-Brexit campaign - Leave.EU - used customers data from an insurance company called Eldon Insurance to target them with political messages. This company is owned by Arron Banks who was, at the same time, a key financial supporter of Leave.EU. The case is detailed in this article from The Guardian.
Understanding party practices from a user and supporter perspective
In addition to following the trail of money through public datasets like those from the UK Electoral Commission, an investigation into digital and data-centered campaign services can benefit a lot from looking at the issue through the perspective of a political party supporter, or a user of various political campaign tools. These can include everything from signing up to party email lists, joining apps created by political parties, to browsing their websites and accessing campaign forums.
Another way to gain an insider perspective into data-driven services that are commissioned by political actors is by checking demos of the services themselves. Many companies provide trial or demo versions of their digital campaigning or marketing services. By trying out or watching demo videos of their products, it is possible to to see exactly what tactics and techniques they use to profile target groups.
For example, NationBuilder, one of the companies used by many smaller political groups and local politicians in the UK, has video demonstrations and images from their app as well as several free demos (also subject to terms and conditions that a researcher should note). These demos should provide an overview of their data-collection strategies and analysis capabilities.
A screenshot from NationBuilder’s website showing their offerings
Self-published materials from parties and political campaigners
Having already searched for news reports and investigations into our topic of interest, we turned to what the parties and service suppliers revealed about themselves.
Apart from taking part in trial versions of their apps and platforms, we can looked at other ways in which these companies communicate with potential clients and other industry players. Often, this information can be found in blogs, industry publications, interviews and press releases. These documents often include concrete information about campaign activities that can prove useful to the investigation.
For example, we researched the blog “https://dominiccummings.com/” by Dominic Cummings, campaign director of Vote Leave, which often focused on his campaign strategy and tactics. The methods he described there ultimately contributed to a successful campaign in the 2016 referendum. In one blog post, he provided a detailed insight into the use of data leading up to the vote:
“Instead of spending a fortune on an expensive agency (with 15% going to them out of ‘controlled expenditure’) and putting up posters to be ‘part of the national conversation’ weeks or months before the vote, we decided to: 1) hire extremely smart physicists to consider everything from first principles; 2) put almost all our money into digital (~98%); 3) hold the vast majority of our budget back and drop it all right at the end with money spent on those adverts that experiments had shown were most effective (internal code name ‘Waterloo’).”
While this is a subjective account of how one side won, it offers many historical and technical details that can substantiate the investigation and provide leads for further research.
Similarly, publications from the parties themselves can provide an overview of their latest outreach strategies. For instance, there is a blog about the activities of the Conservative Party and another that focuses on the practices of the Liberal Democrats, to name a few.
These are valuable resources because they are first-hand accounts of the parties’ activities. Nevertheless, it’s important to remember that any such information and statements could be biased towards the party’s agenda and needs to be thoroughly fact-checked in the same way information from a newspaper or any other sources should be verified.
Expectations and Findings
The techniques described in this investigation helped us get closer to identifying how much money politicians and political parties spent on leveraging personal data in UK political campaigning.
The research revealed large discrepancies in spending across different political parties, as well as an overall increase in spending on data. The research also highlighted the gaps in reported information.
An important finding was that sometimes an investigation doesn’t give you the results you expect. While at first this was frustrating, it later became an important finding in itself – the lack of transparency and the gaps in reporting are revealing in themselves. By identifying and reporting on it, we could highlight problems with the transparency of the system.
Published April 2019
Articles and Guides
Cambridge Analytica Files, from The Guardian. An investigation series on the controversial data mining and elections marketing practices of political consulting firm Cambridge Analytica.
Data and Democracy in the UK, from Tactical Tech. A research report on the practices of collecting, processing and using voters personal data in UK elections.
The Influence Industry, from Tactical Tech. A research project looking at the practices of collecting, processing and using voters personal data in elections around the world.
Pivot tables in LibreOffice from LibreOffice Help. A small guide on working with pivot tables and some useful tips to help you with data processing in LibreOffice Calc.
Tools and Databases
Data broker – a company or person using data as an asset by collecting data from various sources including from collating database records, polling and social networks. Brokers gather data from a variety of sources, including subscriptions, purchases and tracking cookies, for example.
Data-driven campaigning – any campaign activities that are reliant on people’s data to be carried out. For example, data gathered about users of social media is useful for creating voter profiles and developing targeted messages and social media ads.
Personalised content – messaging created to appeal to certain people and groups rather than broadER audiences. This content can be sent to certain segmented and targeted groups, such as everyone on an email list who shares a common interest, or everyone on a particular social media platform of a certain age, gender or location. The content of the message is usually personalised, by message, imagery and style, to be relevant to that group.
Political campaigner – a person that organizez and runs political campaigns for parties and individuals seeking to gain electoral support.
Political campaign – a process by which political parties and individuals carry out organized actions to share their policies and values with an aim to gain support, mainly for election day but also for support year-round.
Selective message targeting – messaging, including text and images, shared on channels meant to reach only certain groups selected based on details such as location, profession, various preferences etc.
Voter profiling – a technique used to understand the behaviour, personality and other characteristics of individual voters or groups of voters (or potential voters). Based on it, experts can create detailed profiles to help determine what political causes or parties people are likely to support or reject, and what messages they may to respond to.