The electoral register is the definitive record of who can or can’t vote in a particular election. Missing people means people aren’t able to exercise their democratic rights. Erroneous entries open up possibilities for fraud and for people who shouldn’t vote getting to cast a ballot. Statistics derived from the register are widely used to inform and shape other decisions. So having accurate registers is important.

Knowing how accurate our registers is a tough question to answer. Estimates as to the theoretical electorate if everyone entitled to vote registered – and no-one else – can be derived from population statistics. But those statistics are not perfect and the margins of error on the final calculations make it hard to judge whether our current rates of electoral registration fall at the good or bad end of the fairly narrow band that separates one from another. 98% registration rate would be good; 89% would be bad.

The Electoral Commission has therefore recently been carrying out some in depth research using a series of local case studies scattered across the country. An interim report on them has just been published. What does it say?

The study has looked at Derby, Glasgow City, Hambleton, Knowsley, Lambeth, South Ayrshire, Swansea and West Somerset:

Once the eight areas had been selected, data-mining of their registers was carried out. This involved checking the registers for visible, potential anomalies. A number of house-to-house interviews were then undertaken in each local authority in order to check whether apparent anomalies were an indicator of an inaccuracy on the register … A house-to-house survey was also undertaken in Knowsley, which was designed to produce estimates of completeness and accuracy of the local register. This entailed drawing a random, preselected sample which was subsequently used to approach interviewees…

The most common anomalies identified through data-mining were the duplication of names and the inclusion of an unusually large number of people in a household, compared to the average for that postcode…

Lambeth’s register had the highest rate of an above average number of entries (compared to the postcode average) per household: at 4.6%, over twice the rate in Derby, Swansea and Knowsley and three to six times the rate in the remaining four areas. It is not surprising that Lambeth should exhibit this pattern. As an inner-London borough, Lambeth has a significantly higher proportion of single-person households (40%, compared to an average in England of 30%), as well as a slightly higher proportion of very large households (three per cent of Lambeth households have six people or more, compared to two per cent nationally). There are also clear variations in repeated names as an anomaly on the local registers. Repeated names are more frequent in Glasgow and Knowsley (1.8% and 1.5% of the register respectively) than in the other case study areas.

However, when anomalies were followed up by door-to-door surveys, it turned out that only around a fifth to a quarter of the anomalies were due to an error on the register. The other anomalies were false alarms.

Therefore, this sort of data analysis may have a role to play in identifying areas of high possible risk with inaccurate electoral registers and to help focus investigations, but it is a relatively crude tool that cannot be used on its own.

In addition, this sort of data analysis is not going to pick up the other, more significant, sources of error – such as people not being registered at all.

The door-to-door survey in Knowsley therefore gives a better idea of the overall state of the register:

Completeness: among the eligible population, the survey found that 93.6% of those surveyed appeared on the electoral register – a figure broadly in line with available national estimates. Registration levels were found to be significantly higher among those aged 20 and over, and among interviewees who had lived at an address for at least two years and who live in properties which are owner occupied.

Accuracy: 91.4% of register entries were confirmed as being accurate. The great majority of inaccurate entries represented cases where electors are registered at addresses at which they are no longer living, or less commonly, have never been resident. Inaccuracies arising from incorrect information being entered in relation to nationality or date of birth are negligible.

This is though only one survey in one local authority. The next stage of the Electoral Commission’s research will involve carrying out similar research across the other authorities.