Data-Driven Campaigning in the 2015 United Kingdom General Election

Assessing the Effectiveness of the Campaign

Our interview data suggested three ways in which effectiveness was defined by campaigners. First, in the heat of the campaign itself, effectiveness was measured by gathering metrics on things such as contact rates, the number of activists working for a party, and levels of support expressed through online donations (Conservative Party scholar, July 22, 2015; senior Labour Party campaign official, August 4, 2015; national Labour Party campaign official, August 6, 2015; regional Labour Party campaign official, August 17, 2015).

More broadly, after the election, effectiveness was assessed through interpretations of the outcome of the campaign. Here, the comments were predictable. The Conservatives (as well as some non-Conservatives) noted that the Tory campaign had been run very efficiently, and was certainly a highly targeted campaign (constituency Conservative party activist, August 2, 2015; senior Labour Party campaign official, August 27, 2015; senior campaign official from a political party, September 14, 2015; senior Conservative grassroots figure and commentator, September 15, 2015). Similar analysis emerged from the SNP (senior SNP campaign figure, August 18, 2015). In contrast, Labour supporters were keen to emphasize the danger of post hoc arguments, constructed with the hindsight of Conservative victory. As one senior Labour Party campaign official noted, “The victor, as they say, gets to write history. And obviously, you know, the Tories now claim this huge degree of sophistication” (interviewed September 16, 2015).

Finally, on occasions, interviewees attempted to fundamentally redefine the idea of effectiveness. One argument that recurred in the interview data (especially among parties that had less successful election outcomes) was how even a good data-driven campaign could not offset other communication, institutional, and strategic weaknesses (constituency Liberal Democrat activist, August 20, 2015; constituency Liberal Democrat campaign organizer, August 21, 2015; senior Labour Party campaign official, August 27, 2015; senior Labour Party campaign official, September 16, 2015). Therefore, it becomes possible to run an effective data-driven campaign within an ineffective election campaign. Other interviewees also spoke about the role of data in activist mobilization, and the legacy effects this might have on a political party in providing a pool of skilled activists that could be called on in the future (senior Labour Party campaign official, August 4, 2015).

Data Capabilities of UK Parties

As expected, the data capabilities of UK parties vary widely. Smaller parties have limited capabilities, often simply using Microsoft (MS) Excel spreadsheets (Green Party parliamentary candidate, August 19, 2015; local Green Party activist, September 10, 2015). They are only able to bring resources to bear on a limited number of seats, with choices largely being driven by where prominent leadership figures are running (senior UKIP campaign official, August 19, 2015; data specialist who has worked for multiple parties, August 26, 2015).

Even among the more established parties, there was inequality in terms of access to data. One interviewee summarized the ingredients that went into the databases that the parties were using:

There’s the electoral register, there’s the Mosaic data, there’s the census data which is fairly good. Not too bad, we’d use it for some stuff. More descriptive, not predictive. And we had polling data, and we had a sample of 100,000 and that’s pretty good. (Data specialist who has worked for multiple parties, August 26, 2015)

The electoral register provided the foundation for party databases (national Liberal Democrat official, August 17, 2015; constituency Liberal Democrat campaign organizer, August 18, 2015; senior Labour Party campaign official, August 27, 2015; senior campaign official from a political party, September 14, 2015). That said, it was also noted that the UK electoral register is a relatively imprecise document with little standardization (research seminar, September 15, 2015). Our interviews suggest that larger UK political parties are developing databases capable of matching multiple years of the electoral register, making for more accurate analysis (senior campaign official from a political party, September 14, 2015).

Both Census and Mosaic data offer geotagged information that can be used for modeling. The last UK Census was conducted in 2011, with data subdivided into 181,408 output areas in England and Wales (Office for National Statistics [ONS] 2012). As these data are freely available, it was unsurprisingly regarded as important by interviewees (national Liberal Democrat official, August 17, 2015; data specialist who has worked for multiple parties, August 26, 2015; senior Labour Party campaign official, August 27, 2015). Mosaic is a postcode-based database produced by the credit-rating company Experian (2015), dividing the population up into sixty-seven types, based on income, values, and levels of cultural capital. Access to Mosaic is costly, so was only available to larger parties, where it was used for modeling (Conservative Party scholar, July 22, 2015; data specialist who has worked for multiple parties, August 26, 2015; senior Labour Party campaign official, August 27, 2015). The Liberal Democrats tested Mosaic but decided not to use it (national Liberal Democrat official, August 17, 2015). Interestingly, the Green Party invested in Mosaic for the proportional European Elections, using it in the simplest way to identify types of voters who might support them and to locate the areas they lived in. They did not use Mosaic for the 2015 General Election (senior national and local Green Party campaign organizer, August 26, 2015).

The role of polls in data-gathering is complex, meaning that terminology needs to be clearly demarked, differentiating public polls, large sample polls, and surveys. For smaller parties, notably UKIP, but also the Liberal Democrats, public polls (especially those conducted at the constituency level) allowed them to decide which constituencies they might be viable in and to target resources accordingly (senior Liberal Democrat activist, August 17, 2015; senior national UKIP campaign official, August 19, 2015). ² Larger parties were able to conduct their own polling, either with large samples or focused on key constituencies (national Liberal Democrat official, August 17, 2015; data specialist who has worked for multiple parties, August 26, 2015; senior Labour Party campaign official, August 27, 2015). Research at this scale is obviously very expensive.

However, interviewees also revealed that parties were conducting surveys (i.e., nonrepresentative questionnaires conducted either face-to-face or online) for a very different reason—to learn about the specific responder (Conservative Party scholar, July 22, 2015; senior Liberal Democrat activist, August 16, 2015; senior campaign official from a political party, September 14, 2015). Parties were not interested in undertaking surveys to understand public opinion, but rather to augment their databases and to make future political communication more effective.

There seems little doubt that the Conservatives refined these techniques most, as it was the Conservative campaign that most effectively adopted the recent US innovation of targeting individual voters (Nickerson and Rogers 2014; Nielsen 2012). For instance, one Conservative activist described how letters were tailored to individual voters, containing messages designed to appeal to them:

We were given a couple of hundred letters each to go and deliver to people’s doors. They were all hand addressed on the envelope . . . I opened a few of the letters before delivering them to have a quick look and see what they’re about. And actually there were at least half a dozen different messages being delivered based on rich voter ID information. So because these people had had someone knocking on their door, had someone talk to them at least once, in some cases several times, about what interested them, what concerned them. (Senior Conservative grassroots figure and commentator, September 15, 2015; see also senior campaign official from a political party, September 14, 2015)

This contrasts with Labour, where targeting was still based on segments. One senior Labour Party campaign official (interviewed August 27, 2015), for example, spoke of targeting certain demographic groups such as students and pensioners with specific messages (also constituency Labour Party campaign official, August 20, 2015).

The purpose of all this is to increase turnout among would-be supporters. This is an area where newer data-driven tactics are bolted on to more traditional political practices. UK political parties have long run “get out the vote” operations on polling day (termed “knocking up” in British political jargon). This process can now be augmented by months and perhaps years of earlier data-gathering (senior Labour Party campaign official, August 27, 2015). This ensures that finite supplies of activists can be used efficiently to maximize vote share. In addition, data-driven campaigning also has the potential to change the very definition of the political base of a party. Although previously this might have been defined through geography, the move to individual targeting means that parties can target would-be voters whose political views are different to most of their neighbors, who previously would have been ignored (senior campaign official from a political party, September 14, 2015).

Implementing data-driven campaigning did produce challenges. Our interviewees highlighted two: gathering and entering data at the local level, and managing activists. Data-gathering was challenging for several reasons. In the first instance, members of the public may simply react angrily when approached by canvassers from political parties (Conservative Party scholar, July 22, 2015; constituency Conservative Party activist, August 1, 2015). Designing suitable data-gathering instruments also proved problematic for parties (research seminar, September 15, 2015; senior Labour Party campaign official, September 16, 2015; see also Ross 2015b). Traditionally, canvassers were asked to fill in returns on party preference and whether that preference was weak or strong. Labour retained this system in 2015. The Conservatives moved to an alternative system, which used a 1–10 ranking score to indicate strength of preference. While controversial when implemented, this model of data-gathering arguably allowed for much more granular targeting later. Some local parties also struggled to ensure that data were properly recorded (constituency Labour Party campaign official, August 20, 2015). In response to challenges of this kind, some parties experimented with smartphone apps and PalmPilots where canvassers could upload data in real time. This seems to have been particularly prominent among the Liberal Democrats (senior national Liberal Democrat official, August 5, 2015; senior Liberal Democrat activist, August 17, 2015; constituency Liberal Democrat activist, August 20, 2015).

A second challenge is managing activists. There are two issues here. The first relates to skills. As one senior Labour Party campaign official (interviewed September 16, 2015) noted, it is hard to run a modern data-driven campaign when your activists are “80-year-old, well-meaning . . . not massively switched on, a bit deaf.” This might be a caricature, but the evidence on political party membership suggests that party members are considerably older than the average citizen (van Biezen et al. 2012). ³ To combat this problem, parties had developed systems to train their activists in how to gather and handle data. For example, the Liberal Democrats had a series of online videos on the subject (constituency Liberal Democrat campaign organizer, August 21, 2015).

The second issue relates to geography, and especially where activists live and are willing to campaign. At its most extreme, this problem might involve vast numbers of party members wanting to campaign in safe seats. To overcome this, parties tried to persuade activists to be more mobile, setting up traveling campaign groups such as the Conservatives’ Team2015 (senior Conservative grassroots figure and commentator, September 15, 2015), or twinning safe and marginal constituencies (regional Labour Party campaign official, August 16, 2015; senior Labour Party campaign official, August 27, 2015). However, there were limits to what could be achieved, as political parties are voluntary organizations.

Data and the Institutional Context

This discussion provides some insights into the perceived effectiveness and data-driven capabilities of UK political parties. To better understand how these practices have evolved requires us to think about how they have interacted with the UK political system. Broadly, five explanations stand out as having particular significance: the data regulatory framework, the electoral system, the psephology of contemporary elections, the geo-institutional form that campaigns now take, and the underlying cultures of political parties. This list is not exhaustive, and it is certainly possible to identify additional factors. However, taken together, these factors have clearly played an important role in shaping data-driven campaigning practices in the UK

In comparison with the US, UK political parties have far less access to the types of data required to target voters. The most obvious omission is the lack of an electoral register that identifies voters by partisan preference, as exists in many US states (Hersh 2015). It is also important to note that the UK has much stronger data protection laws. Among those involved in data-driven campaigning in the UK, data protection requirements divided opinion. Some did not see them as onerous (senior Labour Party campaign official, August 4, 2015; national Labour Party campaign official, August 5, 2015), while others found them a major impingement (constituency Liberal Democrat activist, August 20, 2015; constituency Labour Party campaign official, August 20, 2015). There was also some awareness that parties in the US could access data more easily than their UK counterparts (senior Labour Party campaign official, August 27, 2015). To some degree at least, parties ensured that they had the necessary expertise and training in data protection regulation, and that data resources were kept secure (e.g., constituency Liberal Democrat campaign organizer, August 18, 2015; constituency Labour Party campaign official, August 20, 2015; constituency Liberal Democrat campaign organizer, August 21, 2015). An official from the Information Commissioner’s Office (interviewed September 11, 2015) also offered training to political parties before the election to ensure compliance with data protection legislation.

While the lack of available data may make data-driven campaigning harder in the UK, it is arguable that other elements of the British political system incentivize targeted campaigns. Perhaps the most significant among these reasons is the UK’s first-past-the-post electoral system, which can generate massively disproportionate results (Gallagher 2015). ⁴ The ramification of this is that some voters, depending on where they live, will wield relatively greater influence than voters who live elsewhere (Petts 2015). A political system where this is the case incentivizes parties to use their resources to target voters who are most vital to their success. In the interviews conducted for this project, this was evident first in the decisions made by the larger political parties (especially the Conservatives) about which seats to target, and then which voters to target within those seats. The Conservatives developed what they termed a 40–40 strategy (based on defending forty tight seats and winning an additional forty for a majority; Cowley and Kavanagh 2015). What was most striking about this list of seats is that it was not constructed based on their marginality. Rather, they were defined by what the party knew about the types of voters living in them, their propensity to swing to and away from the Conservatives, their reactions to certain messages, and other seat-specific factors, such as whether an incumbent was standing down (Conservative Party scholar, July 22, 2015). These seats were lavished with resources, and the national party closely monitored campaigns being conducted there (senior Conservative grassroots figure and commentator, September 16, 2015).

These tendencies are exacerbated by the UK party system and the psephology of contemporary elections. In terms of overall vote share, the two major British political parties have been in decline for several decades, and are now having to exist in a more complex, multiparty system (McKibbin 2016). The last two majority governments in the UK have been achieved with the lowest vote share in modern history—Labour in 2005 with 35.2 percent and the Conservatives in 2015 with 36.9 percent (Audickas et al. 2016). While it did not come up explicitly in our interviews, it is not hard to see how this new political environment would encourage data-driven campaigning. Parties now need to navigate a more complex topography, where constructing a majority becomes a matter of what might be referred to as the aggregation of marginal gains.

Data-driven targeting techniques are also changing the geo-institutional form of UK election campaigns, especially how we define national and local campaigning efforts. The interviews provided some evidence of growing central control over the periphery of campaigns, matching predictions made by political scientists (Norris 2000). This process was not uniform or one-sided, however. There was also some conflict between the national and local parties, with the latter often critiquing the actions of the former (Conservative Party scholar, July 22, 2015; constituency Conservative activist, August 2, 2015). In addition, as has been noted elsewhere (Barwell 2016), some non-target seats ran vigorous, locally based campaigns. Nevertheless, national control of local campaigns was clearly occurring, especially in marginal seats targeted by political parties (senior Conservative grassroots figure and commentator, September 16, 2015; senior Labour Party campaign official, September 16, 2015).

Significantly, these developments provide a fundamental challenge to the way UK campaign finance law works. This is because the system is based on two sets of spending limits, one national and one local (Anstead 2008). The logic behind this system assumes a neat institutional separation between the two. Our interviews uncovered a great deal of anger about how parties, especially the Conservatives, had used this system to distribute national resources in local contexts. The Liberal Democrats were particularly angry about this, as they clearly felt they had been targeted with these techniques (senior Liberal Democrat activist, August 17, 2015; constituency Liberal Democrat activist, August 20, 2015; research seminar, September 15, 2015). In addition, representatives of the Electoral Commission (interviewed September 23, 2015) noted that they had received several requests for definitions on local and national spending, both from parties keen to ensure their own practice was appropriate and from parties worried about the activities of their opponents. In the months following our interviews, several high-profile investigations were launched into spending practices during the election, focusing on exactly these questions (Belam 2016). The rights and wrongs of the issue are beyond the scope of this article, but what is clear is that there is ambiguity in how to regulate these newly emerging, centrally dictated, local campaigns.

Finally, longstanding cultures within political parties played a significant role in shaping their use of data and local campaigns. Prior to the election, Labour was very confident in its ability to mobilize its support base. This seems to have been because of a belief that the party had fought a strong election campaign in this regard in 2010, avoiding a much more damaging defeat (senior Labour Party campaign official, August 27, 2015; senior Labour Party campaign official, September 16, 2015). This reading of the events of 2010 may well be true. Certainly, some of Labour’s opponents believed it (e.g., senior campaign official from a political party, September 14, 2015). However, it is also important to note that this reading of events is also a comforting one for a party founded as the political wing of the trade union movement and which lauds community-based activism (McKenzie 1963). It is therefore not surprising that a narrative focused on grassroots communication became central to Labour’s election campaign, when the party heavily publicized its idea of having four (later five) million conversations with members of the British electorate (senior Labour Party campaign official, August 4, 2015; see also Hope 2015).

This approach was not universally popular with Labour’s election team. Some argued that it prioritized quantity of interaction over quality (regional Labour Party campaign official, August 17, 2015), while others saw it as an attempt to paper over the party’s lack of a compelling overarching policy narrative (senior Labour Party campaign official, September 16, 2015). It also contrasts sharply with the Conservatives. The Tories spent the years before the election constructing powerful political messages, especially on the economy and Labour’s lack of fiscal rectitude (Clarke et al. 2015). To this, in the weeks before the election, was added the message of the potential dangers of a Labour/SNP coalition, targeted at the voters who were most likely to be responsive (Conservative Party scholar, July 22, 2015; senior national Liberal Democrat official, August 5, 2015). This is a precise inversion of Labour’s approach—for Labour, the campaign was the narrative; for the Conservatives, the campaign was a mechanism to distribute their narrative.