How to mobilise the electorate: lessons from the University of Manchester ‘Get Out the Vote’ experiment

Abstract

This paper reviews the practical experience of running the UK’s first experimental evaluation of a Get Out the Vote (GOTV) campaign, which was run by the University of Manchester and took place during the 2005 General Election in the constituency of Wythenshawe and Sale East. The campaign encouraged voters to go to the polls by a telephone call or a visit to their homes, replicatingmethods deployed inthe USA(Gerber and Green 2004). This article reports on the challenges of selecting a sample, running a GOTV campaign, and checking whether the intervention was successful by accessing electoral registers. The conclusion is that researchers can successfully replicate GOTV evaluations outside the US, but that the UK electoral registration system presents extra challenges.
Policy-makers allocateconsiderable funds to schemes intendedto improve voter knowledge and turnout, partly as a means to try to stem the decline in voting in contemporary elections. The idea is that a contact from a non-partisan source, either by providing information or seeking to persuade someone of the desirability of voting, can encourage voters to go to the polls.An example is the Time for Democracy project in Sweden, which spent 35m Kronoron community-based democracy development, funding142 projects providing different kinds of support, mainly focusedon young people, women, people from foreign backgrounds, disabled people and the unemployed.[1] These projects, however, are often very difficult to evaluate, especially when seeking to isolate the independent influence of the intervention because of the many linked factors that affectthe voting decision. The contribution of a unique series of studies,conductedin the United States of America since the mid 1990s(Gerber and Green 2000a, b, 2001; Gerber et al. 2003; Green et al. 2003; Green 2004; Nickerson 2006), and summarised in Gerber and Green’s Get Out the Vote (2004), is that the experimental method offersan authoritative evaluation of these interventions. By randomising the sample into control and treatment groupsthat differ only on the basis that one receives the treatment and the other does not, it is possible to make an inference about the efficacy of Get Out the Vote (GOTV) campaigns. After contacting the treatment group, it is possible to use validated turnout data to find out whether voter turnoutdiffers between the groups.

As well as summarising these experimental studies, one of the aims of Get Out The Vote is to illustrate the practical issues involved in running such campaigns, partly as an invitation for others to do the same, which will enable evidence on these interventions to be cumulative. With these explorations of the instrument, the question arises as to whether the challenges and opportunities of running GOTV campaignsare universal or whether their features are moderated by the contexts and institutional rules operating in different voting and electoral systems. Equipped with this knowledge, researchers and policy-makers across the world will be able to examine GOTV proposals with a more informed appreciation of the task, being able to assess the nature of the environment within which these campaigns can be carried out, and whether there are limits to the application of this novel intervention.

We draw the lessons from a GOTV experiment we carried out at theUniversity of Manchesterduring the May 2005 General Election. We considered a number of interventions – leafleting, e-mail contact, telephoning, canvassing, and mailshots–all of which are popular methods of contacting voters and encouraging voting and are summarised in Green and Gerber (2004). We decided to evaluate two – canvassing and telephoning – on the grounds that the period prior to the 2005 election campaign experienced a massive increase in telephone canvassing, almost to the point of overload, whereas the political parties were increasingly moving away from door-to-door approaches except in the marginal seats.

Selecting the sample

We selected a constituency in which to carry out the research. We opted for Wythenshawe and Sale East on the grounds that, firstly, it is local but not proximate to the University of Manchester, therefore easily accessible but without the potential disadvantages of working in a constituency with a large number of students (e.g. hostility to student canvassers from local residents, a transient population more likely to be registered to vote elsewhere and many households without landline telephone numbers); secondly, it is largely residential and has well-laid out streets, which make it easier to cover the ground; thirdly,it had a turnout in the 2001 General Election of 48.6 percent, which was lower than the national average of 59.4 percent; and fourthly, and most importantly, it had a very safe majority for the sitting Labour Member of Parliament, which protected us from any allegation of seeking to influence the outcome – as well as the turnout – of the election in that constituency.

This last factor is especially important given that Labour voters are traditionally less likely to turn out and so the party would, at least in the eyes of our potential critics, could be the beneficiaries of a GOTV campaign.[2] In a safe seat, we did not expect the political parties to object to our study because politicians from all hues could not contest the aim of raising turnout.[3] The choice of a safe constituency also meant that our study would not be likely to encounter the campaigns of the political parties as these tend not to be extensively carried out in safe seats. Political campaigns are interested in turnout as well as votes for a particular party, which could confound the study or make it hard to separate our efforts from those of other groups. On the other hand, a safe seat presents its own problems because,from a short-term instrumental perspective, it is not rational for voters to go the polls if the outcome appears to be predetermined, thereby privileging justifications for voting based on civic duty rather than those appealing to the likely impact on the outcome. In practice, however, we did not find any voters in the constituency who had thought about this problem directly, or at least mentioned it to us.

We commissioned Vision 21, a Manchester-based company that carries out social and political survey research, to source the names, addresses and telephone numbers of individuals on the electoral register in the constituency. These were supplied by a data verification company, originating from the rolling electoral register and matched with BT OCIS, a central repository database which is updated daily and matches contacts with telephone numbers. To ensure random selection of subjects into treatment and control groups whilst still enabling us to carry out the telephone part of the experiment effectively, we confined the sample to those on the electoral register for whom we were able to obtain landline telephone numbers. We randomly selected three groups of 2,300 from the 9,976 available for the treatment and control groups, using Microsoft Excel. We did include the remaining 3,076 individuals in the study.[4]

Carrying out the interventions

Having separated our sample into three equal groups, we sent letters to everyone in the treatment groups: the group selected to receive the telephone call (the telephone group); and the group selected to receive the visit (the canvassing group), to alert them to the imminent contact. We had no contact with the control group. We badged ourselves as the University of Manchester ‘Get Out the Vote Campaign’, a non-party political group supported by the McDougall Trust interested in increasing electoral turnout. The letters advised recipients that we would be contacting them to discuss voting and provided contact details to enable recipients to tell us of any concerns. We also provided details of the campaign website,[5] which listed possible reasons to vote and indicated sources of further information and support. The objectives of sending weretomake the job of the callers and canvassers easier and to allow the respondents to contact us to advise us that the intended recipient was deceased or did not live at that address, thus saving us time and resources incurred by calling or canvassing them, and inconvenience or distress on the part of the residents.

The door-to-door canvassing was coordinated by the university’s Institute for Political and Economic Governance (IPEG). The canvassers were predominantly postgraduate students from the School of Social Sciences, who were enthusiastic about raising electoral turnout, had a good knowledge of the research topic and had an interest in the objectives of the project.[6] We also recruited three canvassers from Vision 21’s workforce, who had been fully trained in interview techniques, thus ensuring a combination of skills and knowledge in the campaign team as a whole. Vision 21 administered the telephone canvassing.

We devised a script for the canvassers and callers to work from; this was intended as a guide to be used in a fairly informal conversation rather than a text to which they should rigidly adhere. The workers on the door-to-door intervention developed their own strategies to engage the voters, which partly reflected the impossibility of supervisingwhat happens on the door step, but also the recognition across the whole team that what is likely to work is an engagement with the voters rather than a mechanical reading out of the script. The practice was not just to have a conversation, but to introduce the questions and the script in a natural way, that did not involve putting the interviewer in the dominant and active position, but yet where the canvassers got their message across.

In the course of the conversation, which was planned to last five minutes, the callers and canvassers were instructed to ask three questions:

  • Generally speaking, do you think voting is important?
  • Do you intend to vote?
  • Will you be voting by post?

However, the main purpose of the conversation was to persuade the contact to vote, both by providing reasons why voting is important and by attempting to respond to any concerns about the voting process. The reasons we provided for the importance of voting were:

  • It keeps our democratic system working. If not many people voted it could threaten our democracy. Turnout has been falling in recent elections and was only 59 percent in the last general election.
  • Earlier generations fought for the right to vote and in many countries people are still fighting for that right.
  • Voting gives you a voice and a chance to express your views about issues which affect your life. You can influence the outcome and politicians have to listen to communities where more people turnout to vote as their position depends on those people.
  • Voting is easy to do. It doesn’t take much time or effort but it is your chance to make a difference.

We also invited respondents to add their own reasons. The long list of options reflects the pilot nature of this study: we did not have a large enough sample size to vary the responses by civil duty or instrumentality so as to compare them, as was done in the US studies, though in practice US studies have not found differences on this dimension (Gerber and Green 2000a: 658). Instead, we adopted a ‘scatter gun’ approach, designed to find the reason that would work with the voter.

We supplied suggested responses to likely questions or comments, at all times avoiding discussion of individual parties or policies. In addition to suggesting reasons why respondents should vote, the campaign equipped canvassers to provide information and advice about the practicalities of voting. They were able to provide details of the voting process, such as the opening hours of polling stations and the alternative methods of voting, and refer contacts to sources of further advice and support. By so doing, the campaign aimed to overcome both psychological and practical obstacles to voting.

Both canvassers and callers received a brief training session during which we explained the objectives of the exercise and the process for carrying out the work. In particular, we emphasised the importance of conducting ‘interviews’ only with the named contact and of keeping accurate records. As part of the training process, we implemented a pilot of the script outside the target constituency and used the feedback to make minor amendments.

Each evening of the canvassing period,[7] meeting in a central location (the café of a local sports centre), we provided the workers with lists of the names and addresses of individuals to contact, together with maps detailing walking routes. The complexity of this exercise should not be underestimated as routes needed to be feasible, have enough addresses to call at in the planned period, be accessible by the mode of transport the canvasser had, and then be updated each day to include call backs. Each day we encountered a series of glitches and coordination problems that had to be resolved immediately. The canvassers also wore campaign badges and carried identification. They were required to complete recording sheets detailingthe date and time of contact; whether the interview was successful or unsuccessful; whether contact was with someone other than the intended respondent; any reasons for an unsuccessful contact (such as whether the respondent wasdeceased or had moved away); and the responses to the three questions posed. The daily meetings were also an opportunity for the GOTV coordinators to find out about canvassing and to keep up the morale of the group, as well as respond to immediate logistical problems.

We were concerned that an economically deprived constituency, with high proportions of public housing and some acute social problems, would be hard to canvass. We anticipated that some residents would be hostile to our student canvassers,given the reputation for social disorder and lack of civility of some English urban areas. For these reasons we were extra-vigilant about security: providing alarms; ensuring that the canvassers worked in pairs; checking they had mobile phones; giving them a list of contacts to call; making sure one of the co-ordinating team was ‘on call’; and trying to ensure that men canvassed with women. Bar a few low-key encounters with young adolescents, there were no problems working in the constituency and the canvassing proceeded smoothly. There was a positive spirit among the canvassers, who believed in the aims of the project, and who took pleasure in encouraging waverers to vote.

For twelve days over the two weeks prior to the General Election, canvassers knocked on doors and conducted brief conversations with their named contacts in which they attempted to persuade them of the merits of voting. The results were recorded on the sheets we provided and collected at the end of each day. We carried out repeat visits when the initial attempted contact had been unsuccessful;[8] a process helped by the information provided on the recording sheets, though time and resources limited the number of call backs we would have liked to have made.

The telephone calls were conducted by Vision 21 staff and took place between 20th and 27th April.[9] As a result of an agreement with the GOTV campaign, the company deployed their best interviewers and stressed the importance of this project. These workers also received a briefing from one of the GOTV coordinators, explaining the objectives ofthe campaign and how different it was to a normal survey.The interviewers used the same script as the canvassers, thus enabling a comparison of the impact of each method.oTh Though no records were kept of the length of the calls, we believe the telephone calls were shorter because of the nature of the medium and the training of the interviewers to complete calls efficiently, whereas the morepersonal nature of face-to-face interviews meant that these took longer. The telephone interviewers recorded their results electronically. They made up to two repeat calls for each contact. In the middle of the interviewing period, one of the GOTV coordinators visited the premises of the company and listened in on several of the conversationsto monitor the performance of the interviewers and to assess the extent to which they had achieved their brief. Our vigilance and continual contact ensured that the company had no doubt about the core objectives of a GOTV campaign and the imperativeof getting the vote out.

Collecting and categorising the responses

Following completion of the canvassing and calling, we coded the contacts into categories to reflect the diverse range of responses we encountered. While the most important distinction for the purposes of the experiment was whether contact had been made or not, there was also a variety of other responses which may prove pertinent, both for our analysis and for coordinating future experiments.

During the canvassing we encountered families or others who reported the person we were seeking had died. Even with electoral registers updated recently, we were concerned about the quality of this data, and raised the issue with the company that supplied it. Whilst assuring us that the data were up to date, in the end the company supplied a list of people who were registered deceased, admitting that it should have done so in any case. However, as the list of responses shows (see below), this strategy did not remove all the people who the canvassers found to be dead. We can only speculate why this is case. Do people lie to the canvassers as an easy (if macabre) way of getting rid of them? More likely, the version of the electoral roll that commercial companies use to generate samples like this may be inaccurate or out of date? If these rolling registers are inaccurate, then data quality is a real problem for this kind of research because it affects the response rate. The only way to overcome this would be to carry out a pre-survey of all the respondents, and weed out the dead and others from both the treatment and control groups. But this strategy would risk affecting the control group thus biasing the experiment. In research where it is not necessary to tie together voters with landline telephones, it should be possible to use a more accurate version of the electoral register.