An X against your preference

Voters will soon be asked to choose the next Government. For the political parties, finding where those voters are and understanding their intentions has probably never been harder. David Reed finds out why data on the electorate may not be fit for purpose.

There are seven registered voters in Richard Webber’s house in Highgate, North London – himself, his wife and three children of voting age. The Webbers also rent out part of their house and the two tenants are on the Electoral Register.

This could be seen as the very definition of a Chattering Classes home. (Webber did after all create Mosaic, the geodemographic profiling system that describes one population segment that way.) It should certainly be a prime target for electioneering by the parties contesting both the local elections on 6th May and the general election taking part the same day.

But there is a problem. “One of our children is in Bristol, one is in Oxford and one is in Germany. Getting postal votes to them in time is a challenge. Our two renters have also moved on and we don’t have a forwarding address for them. So out of seven people on the ER, there are five who may not get to vote,” says Webber.

At the last local election in his borough, postal voting forms for his children actually arrived after the date when postal votes needed to be cast. “The growing number of students moving around is greater than ever and is one reason for low registration rates and a decline in voter turnout,” says Webber.

Disenfranchisement seems almost to have become built in to the voter registration process. In a report published by the Electoral Commission in March into the completeness of the ER, a close correlation was found between densely populated areas (the report looked at Glasgow and Lambeth) and low rates of ER completeness. It also found very high levels of non-registration among 17 to 24 year olds (56 per cent not registered) private sector tenants (49 per cent) and black and minority ethnic British residents (31 per cent).

CORE may very well lead to a decline in the number of people on the electoral register.

The Commission estimates that the voter’s roll is generally over 90 per cent complete, but that a growing minority of local registers have less than 85 per cent of eligible voters on them. Moving house is the main cause of this inaccuracy – among people who have lived at an address for less than 12 months, just 21 per cent are registered to vote. Students are typical of this group, although there are other highly transient people, such as migrant workers who may actually be eligible to vote in the UK.

Based on these omissions, the Electoral Commission believes as many as 3.5 million eligible voters will not be able to participate in May’s elections. Given the apparently close nature of the contest, ensuring those people gain their franchise would go a long way to giving a clear mandate to the incoming government.

The irony is that it has never been easier to get the vote. “Changes to electoral law mean that people can now register to vote much closer to this General Election than ever before. People can now register up to 11 days before the election date,” points out Douglas Stewart, senior manager, public sector, at Deloitte’s.

WIth 6th May as the date for the national poll, that could give eligible citizens up to April 21st to get on the register. “This has big implications for electoral registration officers who need to have robust and efficient data management processes in place to collect, process and publish a register in time for election day,” says Stewart.

Amendments to the Political Parties and Elections Act 2009 opened up individual registration, improved the accuracy, security and reliability of the ER and will create a platform for potentially innovative methods of voting in the future. It also makes provisions for a new Co-ordinated Online Record of Electors Keeper.

“This is a significant and challenging project for local authority EROs and may very well lead to a possible decline in the number of people on the electoral register. This apparent drop is due to the cleansing effect that removing duplicates and ’dormant’ registrations will cause,” says Stewart .

He points out that, “these reforms to the way electoral registration data is collected and maintained is perhaps the most significant and challenging change to the electoral franchise since it was extended to include women.” The goal is to maintain the integrity of the democratic process by ensuring more complete data, but which is held securely.

For those parties campaigning in 2010, however, much of this improvement will come too late. Their campaign war maps will have been drawn up using ER data without those improvements and missing large swathes of potential voters. This has two important implications – at a local level, seats may look more marginal than they should, and at a national level, assumptions about the total number of votes available could be out by as much as 10 per cent.

Stewart says: “Incomplete and inaccurate electoral register data can be a particularly difficult problem for the political community to manage. Not only does it make targeting voters a more time consuming and arduous process, it also makes checking the permissibility of donors more challenging, given that individuals must generally be on the electoral register to donate money to political parties.” government needs is a fresh scandal about having raised campaigning funds from somebody who is not eligible to vote. The revelations about Lord Ashcroft, who has donated much of the Conservative Party’s election fund, have made the electorate very sensitive to this issue.

A late surge in registrations is possible, especially with Labour planning to use Obama-style local campaign techniques that also centred around optimising voter numbers and turnout. They may discover too late that other changes in data have loosened their grip on power.
“I am surprised the Labour party – and the Liberal Democrats – have not commented before on the criteria used by the Boundaries Commission to redistribute seats,” says Webber, who is also managing director of Origins Information and visiting professor of geography at King’s College, London.

He notes that in Harringay, Islington and Camden, nine MPs were returned by the three boroughs 30 years ago, whereas the new boundaries will see just 5.5 this time. “At this election, London is losing seats, predominently because of one reason – the Boundaries Commission is working on the basis of registered voters and not population.”

High levels of non-registration by eligible citizens, as identified by the Electoral Commission, is having a big impact on reducing the number of MPs returned in inner-city boroughs, especially in London, where the transient or student population is large. To further harm the chances of the incumbent government, Webber says that non-registered citizens tend not to be Labour voters. So any registration drive by the party could end up being counter-productive.

There is a further data quality issue with ER that Webber has identified. “We have been going through the registers looking for odd names, especially where people may have put down their name back to front because they don’t understand the difference between first and last name,” he says.

The Boundaries Commission is working on the basis of registered voters and not population.

Many new arrivals in the UK can be eligible to vote, but come from a cultural background where name structures are very different. The problem is that, once a paper record has been completed and submitted for registration, EROs are not allowed to make any changes.

“The people capturing that data often can’t decipher unusual names, or mistake ’l’ for ’i’. That has a knock-on effect for any identity checking against the ER,” say Webber. If knowledge about that sort of problem is getting shared anecdotally within sub-groups, then it could act as a major inhibitor on registration, further embedding disenfranchisement.

It will also make back-checking of donations harder where funds have been solicited from communities with a strong ethnic profile that may contain many new immigrants. Gearing up party machinery at a local level to cope with the difficulties and absences of voter data is just one of the challenges facing the main parties.

Seeing the deficiencies in public sector information about voters, the three big players have all turned to private sector sources. Their experience with commercial data may not prove to be much happier. In March, the Information Commissioner, Christopher Graham, put all parties and candidates on notice that encouraging individuals to vote for a particular party or candidate, appealling for party funds or campaign support constitutes marketing.

As such, parties must abideby the same rules as any other brand. The ICO noted that previous election campaigns had been marred by some parties flouting the rules, with hundreds of thousands of people being targeted unlawfully. Automatic dialling of pre-recorded campaign messages to people registered on the Telephone Preference Service was a common breach.

“For good reason, there are strict rules concerning the way in which political parties and candidates can contact unsuspecting members of the public with campaign and promotional materials,” says Graham. “They are the same rules that apply to anyone else marketing a product. This guidance will help political parties respect people’s privacy rights. I strongly urge the parties to adhere to the ICO guidance, especially as their collective track record to date has been disappointing.” If the run-up to the General Election has been characterised by any one thing, it is uncertainty about the outcome. Media coverage has been asuming a hung parliament, based on opinion poll results suggesting no one party will have an overall majority. This has conditioned the tone of much of the early debate.

Yet even here, there are reasons to believe that the underlying data and the way it has been modelled may be faulty. By applying an entirely new approach, Resolver Systems is predicting a very different outcome giving the Conservatives a majority of 24, close to the 21 MPs who gave John Major the premiership in 1992.

“Everyone has realised that the forecasting model used – uniform national swing – has produced terrible results,” says Robert Smithson, director of Resolver Systems and son of Mike Smithson, the man behind the influential blog In 1992, the polls were suggesting an increased Tory majority of 100, for example, while Neil Kinnock had based his famous “we’re alright” speech in Sheffield on private polls showing Labour would win. In 1997, the scale of the Labour landslide was missed.

What all of those forecasts assumed was that any shift in voting intention picked up during polling would be replicated uniformly at a national level. But according to Smithson: “The big move is from voting to not voting. In 97, voters did not move from Conservative to Labour but from Conservative to not voting. Standard analysis fails to spot that shift.”

He has developed a new model, Voting Intentions Predictive Analysis, to understand how tactical voting is likely to affect each constituency. This has been made possible by ICM releasing its poll results for each seat in the country. One of the key differences this has picked up is that many LibDem voters will switch to the Tories. “The result looks a lot like the election of 1987,” says Smithson.

A key factor is the results for the “forced choice” question in polls, which ask how an individual would vote if their preferred party was not an option. This unlocks a shift among voters who say their primary choice is for Other parties, such as UKIP, BNP or Green. The largest of these is likely to contest fewer than half the seats in the UK, leaving supporters either not to vote or to choose an alternative.

“A lot of voters intending to vote for them will find they don’t have that choice. Who will they vote for instead? Only two out of ten say Labour – Conservatives get the bulk of the rest,” he says. VIPA has been validated by testing its model against historical election data and was found to be out by 10 seats for 2001 and 15 for 2005, considerably better than the conventional pollsters.

What this model reflects is the fragmentation of national voting patterns that have been enshrined in everything from the Electoral Register to polling day forecasts. With a more mobile electorate and more parties contesting seats, finding voters and predicting what they will do has never been harder.

But whichever party gets in, don’t expect a dramatic move to improve matters. Politicians are always grateful for a mandate, but superstitious about messing with its source.


    Leave a comment