For the Journal of Liberal History‘s special issue on the Liberal Democrats and coalition (Issue 88, Autumn 2015, available here), I wrote the following piece about why the Liberal Democrats won only 8 seats in May 2015, which I’ve lightly updated for use online and to add in some extra cross-references.
It is a sequel of sorts to my earlier history of Liberal Democrat campaigning, The Liberal Democrat approach to campaigning: the history and debunking some myths. Together with What went wrong with the Liberal Democrat polling and key seat intelligence? these pieces form a painful trio. Unfortunately, whichever order you read them in, the outcome is the same but there is also a treasure trove of lessons to be learnt.
8 per cent but also only 8 MPs: the death of the fabled Liberal Democrat grassroots campaigning machine?
There were two elements in the disaster that was the 2015 general election result for the Liberal Democrats: just 8% of the vote and also just eight MPs. Many of the other articles in the special edition of the Journal of Liberal History explain the 8% vote share. However, the fact that even such a low vote as 8% turned into only eight MPs also needs explanation, both because it was well below prior expectations – inside and outside the party – and also because in the past the party had consistently won more seats than the percentage of the vote it secured.
Indeed, up until the 2005 general election, the Liberal Democrats had been starting to learn to live with the bias that first past the post (FPTP) imposes on smaller parties who do not have a very strong geographic concentration in one part of the country.
As I set out in ‘The Liberal Democrat approach to campaigning’,[i] from the 1970s through to 2005 the Liberal Democrats became progressively better at turning votes into seats at Westminster elections. This is best illustrated by the party’s seats to vote share ratio.
With about 650 seats in parliament[ii] but only a maximum of 100 per cent of voters to be won, a proportional result would mean a ratio of around 6.5:1. The party has never got close to that, but from the early 1970s its predecessors’ and then the Liberal Democrats’ results consistently improved, rising from a low of 0.7 up to 2.9 by 2005. To put flesh on that ratio, had the party still been at 0.7 in 2005, it would have won fifteen seats, not the sixty-two it actually secured. This was not trivial progress.
|Table 1: Liberal, Alliance and Liberal Democrat general election performance|
|Election||Seats won||% share of the vote||Seats:votes ratio|
However, after the 2005 peak things slipped back in 2010 before plummeting in 2015, returning the ratio to its pre-merger levels.
Lessons of the 2010 general election
As I wrote of the 2010 slippage for the Journal in 2014,[iii] the 2010 seats to votes ratio made it the then worst for the party since 1992: a poor reflection on the campaign machine’s ability to turn national vote share into actual seats. For pessimists this was the result of the Conservatives in 2005 having largely cottoned on to how to do intensive target seat campaigning, and by 2010 Labour doing so too, leaving the Liberal Democrats’ ability to outperform the national picture in selected constituencies hugely reduced. For optimists, there were specific mistakes in 2010 which could be remedied in the future. One was the weakening of the focus on target seats in the heady wake of Nick Clegg’s first TV debate victory and the resulting poll surge. Another was that the party called several seats wrongly in the last few days before polling day, misdirecting resources as a result. For example, a great deal of effort was directed to Oxford East on polling day, which Labour held on to by a significant margin – 4,581 votes – whereas, had the effort been directed instead to neighbouring Oxford West & Abingdon, Evan Harris would not have ended up losing by just 176 votes.
The party’s own official review into the 2010 general election, chaired by James Gurling,[iv] was a relatively low-key affair. Some of the causes of the party’s failure to win more seats in 2010 it ascribed to specific one-off factors such as the failure of the party’s immigration message and the old ‘you can’t win’ argument (both of these factors came through strongly in the party’s post-election private polling). It made many detailed recommendations, and some significant organisational ones – particularly that the party should change its computer database software for fighting elections. In addition, the increasing emphasis in the Labour Party on the virtues of canvassing rubbed off on the Liberal Democrats, with a switch from viewing canvassing as a data-gathering opportunity, where a virtue is made of talking to each person for as little time a possible, to an attempt to get into longer conversations about issues.[v]
It was hoped that these organisational improvements, plus the fact of being in coalition government giving a completely different spin to the ‘you can’t win’ argument, would allow the party to regain its local campaigning edge.
However, given the 2015 result, those hopes were not only not met but the seats to votes ratio crashed catastrophically. What went wrong?
Targeting decisions post-2010
During the 2010–15 parliament, the Liberal Democrats certainly put in a considerable effort to target campaigning activity at winnable Westminster constituencies. This included a ‘Dragon’s Den’ style application process for support, whereby constituency teams had to make the case that their seat was winnable and their applications were assessed with the assistance of an extensive constituency polling programme.
The process was far more ruthless than in previous parliaments, with the party willing to withdraw support even from seats held by long-standing Liberal Democrat MPs[vi] – something that had been a bone of contention in some previous parliaments where, regardless of the constituency campaigning performance of an MP, the party would always end up putting in outside support to sort things out.
Those seats that made it through the process received, despite the party’s low poll ratings throughout the parliament, generously funded support thanks to the efficacy of the party’s fundraising operation. It was a regular feature of the quarterly donation figures published by the Electoral Commission for the party’s fundraising from individuals to be more successful than that of Labour.
They also received more effective targeting of volunteer resources than in previous parliaments.[vii] This was partly due to a widespread understanding in the party of how few seats were truly winnable in 2015 and therefore a greater willingness on the part of volunteers to travel to help elsewhere.[viii] It was also due to the increasing use of telephone canvassing via an easy-to-use web system (called VPBs, or Virtual Phone Banks). VPBs made it easy for people to help a seat without having to travel to it, and replaced the previous reliance on printing and posting back and forth paper canvass sheets, which had been a rather cumbersome mechanism even when the arrival of the fax machine and later digital scanning/photography brought a little IT relief.
The introduction of Connect
VPBs were possible due to the party’s migration, as recommended by the Gurling review, to a new web-based electoral database, called Connect. Supplied to the party by the American firm NGPVAN, Connect was based on the same technology as used by the 2008 and 2012 Obama presidential campaigns (and the Canadian Liberals).[ix] In addition to investing in Connect, the party also commissioned micro-targeting research to score uncanvassed voters on their likelihood of being Liberal Democrat, Conservative or Labour in order to prioritise canvassing and to allow the more accurate targeting of campaigning such as direct mail to otherwise uncanvassed voters.
Although Connect had a severe slow down on polling day in 2015, resulting in some features being scaled back, even in the midst of the post-election dismay there has been no call to abandon it or blame it from the result. Rather VPBs, scoring and Connect were to varying degrees successful.
Problems with constituency data
Unsuccessful, however, was the party’s constituency polling. This had been inaccurate in places in 2010 (see above) and was even more so in 2015, leading the party to believe it would return far more than eight MPs and hence to Paddy Ashdown’s promise on live TV on election night to eat a hat if the exit poll prediction of the party winning just ten seats turned out to be right.
That so many public pollsters got their polls wrong too provides some cover for the party’s error, and indeed some of the criticisms of the party’s polling methodology (such as question order) do not stack up as a similar methodology was followed by other, successful pollsters.[x]
More likely, the problem with the polling was that it was asking about one sort of election but the public decided to vote in a different sort. That is, the polling accurately captured how people would vote if they were not thinking about who would be prime minister (such as if it was a foregone conclusion); however once worrying about the prime minister came into consideration, they switched away from the Liberal Democrats to other parties.
That would fit with a broader pattern of the party doing best (1997, 2001, 2005) when the name of the prime minister after polling day is little doubted and doing worst (1992, 2010, 2015) when there is real doubt over who will be prime minister.
This change in voter perspective also helps explain why the hoped for Liberal Democrat incumbency boost was muted: the more people worried about who the prime minister was, the less their love of their MP mattered. Hence the strong polling results for the party’s MPs during the parliament[xi] were not enough to save most of them.[xii]This polling problem was exacerbated by a breakdown in accurate analysis of canvassing information from target seats. Partly that was caused by the party getting to grips with a new system and not always looking at the right numbers. Partly too it was about who was willing to pay attention to what within the decision making processes.
Add to the mix an emphasis on KPIs which focused on volume of activity rather than outcomes – a focus which had worked in the party’s heydays but didn’t work in 2015 – and the party was caught out by what was really happening.
Geography or skills?
Three other factors, however, contributed to the party’s dreadful seats to votes ratio aside from the strategic political landscape. One was the abortive attempted major restructuring of staff at party HQ in 2012. It produced some positive results, with the turnaround in the party’s membership figures starting following the renewed focus on membership services.
However the attempt to change the campaign staffing structure was very controversial. The idea was to move from primarily geographically-based staff to skills-based staff, but the handling of the axing of the geographically-based posts led to widespread protests through the party.
The eventual structure that emerged was very similar to the pre-restructure one, but with a greater emphasis on monitoring performance standards[xiii] than on collaboration with seats, and with those staff with geographic responsibilities covering huge areas and so spending very considerable amounts of time travelling. Moreover, there was widespread bad feeling – and some rather complicated wrinkles, to cater for particular personality clashes and differences. This contributed to a significant cadre of highly skilled and experienced staff deciding to move on from party employment, often also dropping out of voluntary party activity too.
An absence of testing
A second factor was that, despite the attempt to move to a more skills-based structure, the party did very little in the way of testing out alternative campaign tactics in the field, such as by splitting voters into two different groups and trying a different direct mail design on each.
In the US, such A/B split-test field experiments have been the norm for political campaigners for many years now[xiv] and are spreading to other parties, but the Liberal Democrats almost never carried out A/B split testing except for online campaigning and the party’s campaigning tactics changed little from 10 years previously.
More broadly, the party’s development of campaign tactics had in many areas stalled and where testing was done the results were rarely widely implemented.
The death of constituency expense limitsThe final factor is one, however, also outside the party’s direct control: the death of constituency election expense limits. Although campaigning in constituencies in the months running up to polling day is nominally tightly controlled by constituency expense limits, there is very large scope for campaigning to be done that is targeted at swing voters in marginal seats but which does not count against the local limit. A mailshot from David Cameron, for example, to soft Lib Dem voters in a Lib Dem MP’s seat did not need to count against the constituency limit.
As a result, the Tories were able to spend millions of pounds extra on ‘national’ campaigning in Liberal Democrat-held seats, outgunning the Liberal Democrat campaigns and undermining the party’s traditional incumbency advantage.[xv]
This regulatory death played into the hands of the Conservatives not only because of the parties’ relative finances but also because the key messages to such swing voters for Tories were national ones (be afraid of Ed Miliband in hock to the SNP), yet for the Lib Dems were local ones (praise for the local MP). Therefore even when the Liberal Democrats spent money on ‘national’ campaigning in key seats, it was not as effective for the party as the equivalent Conservative campaigning was.
The lessons for the future, then, are twofold. Wider political circumstances – not only the party’s overall popularity but also the degree to which the election result is seen as a forgone conclusion – matter, as do regulatory issues the party cannot unilaterally influence.
Nevertheless, in addition to pushing for the revival of meaningful constituency expense limits, there are other factors under the party’s control which can be altered ahead of 2020, including a revised approach to polling and a reinvigoration of campaigning tactics, fuelled by a belief in testing and experimentation.
Dr Mark Pack worked at party HQ from 2000 to 2009, heading up the party’s online operation for the 2001 and 2005 general elections. He is author of 101 Ways To Win An Election and the party’s election law manual, as well as co-author of the party’s general election agents’ handbook.
Acknowledgments: Thank you to Neil Fawcett and Ed Maxfield for their comments on an earlier draft of this piece.
[i] Journal of Liberal History, 83 (Summer 2014).
[ii] The number has varied with boundary reviews and devolution. During the period in Table 1, the Liberals, then Alliance and subsequently Liberal Democrats contested nearly every seat, with a few exceptions such as the Speaker’s constituency and, in 1997, Tatton.
[iii] Journal of Liberal History, 83 (Summer 2014).
[iv] In his role as chair of the party’s Campaigns and Communications Committee (CCC). He discussed his review at the Liberal Democrat History Group’s fringe meeting at the autumn 2010 party conference. See the meeting report at https://www.markpack.org.uk/16887/the-2010-election-in-historical-perspective/.
[v] One cause for this renewed emphasis on canvassing, and canvassing moreover as more than just brief data gathering conversations, was the Obama 2008 election campaign which was a high profile example, widely reported in Britain, of the resurgence of doorstep canvassing in the United States.
[vi] Although there was some controversy over whether such decisions were made soon enough.
[vii] With lower membership and fewer councillors in 2015 than in 2010, the total volunteer base available was smaller. Hard data is not yet available as to whether the net result of fewer volunteers used better was more or less activity in key seats.
[viii] The contrast with 2010 was especially notable, when the Cleggmania poll surge resulted in more activists thinking their own seat was winnable and so damaging the party’s attempt to target efforts.
[ix] A small number of local parties continued to use the previous database system EARS, which had bid for the new party contract but lost out to Connect. This number of local parties declined steadily during the parliament, and the central party’s influence on the target seats operation helped push any slow movers over to Connect.
[x] For details on this, see Mark Pack, ‘What went wrong with the Liberal Democrat polling and key seat intelligence?’, revised August 2015: https://www.markpack.org.uk/132249/what-went-wrong-with-the-liberal-democrat-polling-and-key-seat-intelligence/.
[xi] For example, see the research by Phil Cowley and Rosie Campbell, using polling data from July 2013: http://www.totalpolitics.com/opinion/416802/polling-not-love-actually.thtml.
[xii] Another factor was the number of Liberal Democrat MPs retiring in 2015. See http://nottspolitics.org/2014/07/29/lib-dem-incumbent-mp-retirements-could-cost-the-party-four-seats-in-2015-before-any-votes-are-cast/.
[xiii] The value of the key performance indicator (KPI) framework used to monitor seat progress and help determine allocation of resources is thrown into severe doubt by the 2015 results as seats which were consistently rated as performing in the top tier on the KPIs ended up being lost heavily.
[xiv] On which the seminal work is Donald P. Green and Alan S. Gerber, Get Out the Vote: How to Increase Voter Turnout (Brookings Institution, 2004).
[xv] For more details on this see Mark Pack, ‘Constituency expense limits are dying off in the UK, but neither politicians nor the regulator will act’, March 2015: https://www.markpack.org.uk/130283/internet-speeds-up-the-killing-off-of-expense-controls-in-marginal-seats/.