Written evidence submitted by Ipsos MORI
(PVSCB 25)
SUMMARY
1. Ipsos MORI is one of the largest and
most respected companies conducting survey research in Britain.
In recent years we have been commissioned by the Electoral Commission
to conduct several pieces of research into the state of the electoral
register in Britain and public attitudes to electoral registration.
2. Our registration research for the Electoral
Commission falls into several distinct categories. These are:
— national survey research, in which we
interviewed a nationally representative sample, but did not attempt
to match their responses to their entries on the electoral register;
— detailed research into the accuracy and completeness
of electoral register in eight local authorities chosen as case
studies, which combines interviewing with a check of the register,
but which measures the situation in these case study areas and
not the national picture;
— data mining exercises in the same eight
local authority areas, identifying apparent anomalies in the register
and determining whether these indicate inaccuracies; and
— qualitative research (discussion groups
and in-depth interviews) into attitudes towards registration,
which provide insight into the way people think and feel about
the issues but which do not produce quantitative findings to show
how many people think in a particular way.
3. Our case study research into the completeness
and accuracy of eight local authorities' registers was conducted
in 2009. (These local authorities do not make up a nationally
representative sample.) In each case, this consisted of checking
the accuracy and completeness of the register entries relating
to a sample of addresses within the local authority area, using
face-to-face interviewing at the selected addresses.
4. The research indicates that completeness
of the register (the proportion of eligible people who are registered)
in our eight study areas ranges from 73% up to 94%. Accuracy (the
proportion of register entries that correctly refer to people
eligible to be registered) ranges from 77% to 91%. Note that in
one case study, Knowsley, the register used for comparison was
from February 2009, and the remaining areas the register was from
May 2009, a longer time since the previous annual canvass.
5. The most frequent type of inaccurate
entries found are those listing a name not matching anybody living
at the address in question, many of whom may be former residents.
Entries referring to derelict or unoccupied properties are also
a common cause of inaccurate entries. Inaccuracies from being
deceased, ineligible nationalities and date of birth errors are
much less common.
6. Both inaccuracy and incompleteness are
more common in reference to privately rented accommodation. Incompleteness
is also more common among recent movers, young people and those
whose ethnic group is not White British. The lowest levels of
completeness and accuracy that we found were in urban local authorities
(Lambeth and Glasgow).
7. Our data mining exercise, also conducted
in 2009 in the same eight local authority areas, found that some
apparent "anomalies" of repeated names and unusually
large numbers of people registered at an address could be identified
using automated computer look-ups on the register. Our follow-up
interviews showed that around half of the "anomalies"
identified by the computer look-ups in the eight case study areas
actually refer to inaccurate entries.
8. Other research conducted by Ipsos MORI
for the Electoral Commission across the UK between 2004 and 2010
shows that the public are broadly satisfied with the system of
registration and are confident they know how to register.
9. While almost all of the public say they
believe they are registered to vote, more in-depth research shows
that many people think they are registered when they are not.
This may be part-related to thinking that registration is automatic
in some way.
10. Other common typologies of those not
registered are being suspicious or nervous of "bureaucracy",
being politically disengaged or hostile, being a recent mover
or "not having got around" to registering.
ELECTORAL REGISTRATION
Research into accuracy and completeness of electoral
registers for the Electoral Commission
11. The Electoral Commission is undertaking
a rolling programme of research into the extent to which registers
are incomplete or inaccurate, and commissioned Ipsos MORI to conduct
research as part of this exercise.
METHODOLOGY
12. The research consisted of case studies
of eight local authority areas in Great Britain. These were not
intended to comprise a representative national sample, but to
ensure a sufficient spread of different types of areas; it did
not include Northern Ireland, where the system of registration
is different. The figures should not therefore be seen as nationally-representative.
However, we understand that our findings are generally in line
with the Electoral Commission's conclusions about the national
picture based on other sources and methodologies.
13. The research was in two phases, with Phase
One acting as a pilot study to test the methodology. In Phase
One, interviews were conducted at 449 households in Knowsley between
23 March and 27 May 2009. In Phase Two, interviewing took place
between 6 July and 17 September 2009 in seven local authority
areas, with a target of 500 interviews in each: Derby, Glasgow,
Hambleton, Lambeth, South Ayrshire, Swansea and West Somerset.
In Phase One, our adjusted response rate was 67%, and in Phase
Two 72% across the seven areas, varying from 56% in Lambeth to
82% in South Ayrshire.
14. Ipsos MORI and the Electoral Commission
designed a random pre-selected methodology for this survey, using
face-to-face interviewing, as we believe that this is the best
way to estimate accuracy and completeness. Addresses where the
research would be conducted were selected both from the electoral
register and the Postcode Address File (PAF): this ensured that
it was possible to check both for addresses at which nobody was
registered to vote and for addresses appearing on the register
which did not exist or were inaccurately recorded.
15. The sample was clustered, taking one
in every four wards for each case study area. The selection of
wards was stratified to ensure a broad mix of areas, and wards
were selected with probability proportional to the size of their
electorates.
16. Full sample breakdowns for each phase,
including the number of interviews achieved, are detailed in the
Electoral Commission's published report.
17. Weighting was necessary in our final
findings to correct for the fact that the sample design meant
that all adults in each area did not have an exactly equal probability
of selection. We considered also weighting to compensate for non-response
to the survey, but after analysis of data from addresses where
no interview was achieved, we concluded that this was unnecessary.
CONFIDENCE INTERVALS
18. Because a sample has been interviewed,
and not the entire population of addresses in each area, our findings
are subject to sampling error. As this was a random probability
sample, where the exact probability of each person being chosen
for interview is known, we can predict the likely variation between
the sample results and the "true" values. These are
expressed as "95% confidence intervals" (or, less formally,
the "margin of error").
19. We also have taken into account the design
effects caused by using a clustered sample and weighted data,
which affect the confidence intervals.
THE FINDINGS
20. The research aimed at estimating the
levels of accuracy and completeness of the electoral register
in each case study area.
21. Accuracy is a measure of the proportion of
register entries that correctly refer to people eligible to be
registered, and inaccuracy is therefore synonymous with over-registration;
for example:
— duplication of names and/or households;
— ineligible people/addresses wrongly included;
and
— various forms of outdatedness (eg dead,
moved).
22. Completeness is a measure of under-registration.
That is to say, incompleteness is the proportion of people who
are eligible to be on the register but not registered.
23. Table 1 shows our survey estimates of
accuracy and completeness in each area, together with the confidence
intervals associated with them. Fuller breakdowns for each area
are included in the Electoral Commission's published report. Note
that in Phase One, Knowsley, the register used for comparison
was from February 2009, and in the Phase Two areas the registers
were from May 2009. We may expect accuracy and completeness to
decline the longer the period of time since the previous canvass
in the Autumn.
Table 1
ESTIMATES OF COMPLETENESS AND ACCURACY, WITH
CONFIDENCE INTERVALS
| Completeness
| Accuracy |
| Area | Estimate
| CI | Estimate
| CI |
| Phase One | |
| | |
| Knowsley | 94% | +1.3
| 91% | +3.0 |
| Phase Two | |
| | |
| Derby | 86% |  6.3
| | 6.0 |
| Glasgow | 74% | 6.3
| 77% | 7.6 |
| Hambleton | 89% | 4.1
| 91% | 2.9 |
| Lambeth | 73% | 7.4
| 79% | 4.1 |
| South Ayrshire | 87% | 4.6
| 91% | 3.0 |
| Swansea | 89% | 4.1
| 83% | 11.9 |
| West Somerset | 86% | 5.6
| 89% | 3.3 |
| |
| | |
COMPLETENESS
24. We were able to make comparisons of the levels of
completeness and accuracy among different groups of the population
across the seven areas studied in Phase Two. Again, it must be
borne in mind that the sample was not designed to be nationally
representative, so conclusions relating to our case study areas
may not hold more generally.
25. Housing tenure is a key factor in completeness. Fewer
than half (44%) of eligible adults who rent from a private landlord
are on the register, while 95% of eligible adults who own their
homes outright are registered. Private renters make up around
two in five of all those eligible but not registered across the
case study areas, rising to almost three in five in Lambeth. More
generally, owner occupiers (with or without a mortgage) are significantly
more likely to be on the register than those who rent.
26. Recent movers make up a majority of those eligible but
not registered. Those who are both a recent mover and who rent
privately make up around one in three of those eligible but not
registered when all seven Phase Two areas are taken as a whole.
27. Age is also an important factor in completeness.
Fewer than half of those aged below 25 (and eligible) are on the
register, while two-thirds (66%) of eligible adults aged 25-34
are on the register. Completeness levels generally rise with age,
with 84% of those aged 35-44 on the register, rising to 97% of
those aged 60-64.
28. There is also variation by ethnicity, with those
who are "White British" and eligible more likely to
be on the register than those who are eligible and from another
ethnic group (86% versus 69% respectively).
ACCURACY
29. The most frequent type of inaccurate register entries
at addresses where we completed interviews is entries listing
a person not living at that address. Inaccurate entries listing
names at properties our interviewers reported as derelict or unoccupied
are also a common cause of inaccurate entries.
30. The highest proportion of errors in register entries are
in those relating to properties rented from a private landlord,
though many of these are minor errors such as mis-spellings rather
than inaccuracies.
DATA MINING
AND FOLLOW-UP
INTERVIEWS
31. In the same eight local authority areas, we also
conducted a "data mining" exercise, searching for apparent
anomalies in the registers and then investigating a sample of
these with follow-up in-home interviews to determine which were
in fact inaccurate entries and which, despite appearing anomalous,
were nevertheless accurate. These interviews were conducted over
the same period as the Phase One completeness and accuracy research.
This exercise was in part intended to understand the ways in which
local authorities could undertake "cleaning" of their
own registers.
32. Anomalies were identified by computer look-ups on the
eight registers, refined by "eyeball" checking. We searched
for several types of anomaly, but only two proved sufficiently
frequent to merit investigation. These were identical names appearing
twice (or more) in the same area, and addresses with more than
the average number of names registered for that postcode (taking
those addresses with three or more names over the average as potentially
anomalous).
33. In total we found across the eight authorities 13,499
"repeated name" anomalous entries (about 1% of the total
number of entries on the register) and 23,845 entries at addresses
with an anomalously large number of names registered (about 1.8%).
These figures varied substantially between the different authorities,
as would be expected given the differing characters of their populations.
(The number of "large number of name" anomalies, for
example, will be affected by the degree to which there is genuine
variation in household size within single postcodes.) The inaccuracies
so identified account for only a small proportion of the total
number of inaccuracies in the register, as estimated in the completeness
and accuracy research.
34. We then conducted in-house interviews at a sample
of these addresses; in total, 610 interviews were achieved. Of
the potentially anomalous cases we were able to investigate, we
found in just over half of the cases (53%) of unusually large
number of names that some or all of the entries were inaccurate,
and that in two-fifths (41%) of the repeated names cases some
or all were inaccurate.
35. In these "anomalous" cases, most of the
inaccuracies we found consisted of ineligible entries on the register
(including incorrect multiple entries for a single eligible individual),
with far fewer related to nationality or date of birth inaccuracies.
OTHER IPSOS
MORI RESEARCH ON
ELECTORAL REGISTRATION
36. In a survey for the Electoral Commission just after
the 2010 General Election, 97% of UK adults (aged 18+) claimed
to be registered to vote. However, this probably over-estimates
registration levels as those who are not registered may prefer
not to admit it or may be less likely to participate in surveys.
Respondents may also assume that they are registered when they
in fact are not, as our wider research implies is often the case.
37. Young people, members of ethnic minorities, students and
those in privately-rented accommodation are less likely to say
they are registered (reflecting the patterns in our case study
research and in our post-election research going back several
decades).
38. The most common explanation for not being registered
is not being eligible. The next most frequent are a lack of interest
in voting and not knowing how to register. This reflects the findings
of eight discussion groups we conducted in 2004-05 to explore
public views on registration, which identified a number of typologies
of unregistered people:
— false positives (people who believe that they're registered
when they are not);
— "Big Brother" phobics (unwilling to provide
their details because of suspicion of "bureaucracy"
or "government");
— those nervous of bureaucracy who find the actual process
of completing the forms daunting;
— the politically disengaged who see no point in registering
or voting;
- the politically hostile who actively dislike politicians
and politics;
— recent movers who have not yet registered. They tend
not to have heard of the rolling registration process; and
— some others who have not got around to filling in
the form.
39. Those who are registered to vote give a wide variety
of reasons for registering. In the 2010 post-election survey,
the most commonly-cited reason is a desire to safeguard their
vote or to have their "voices heard" at elections. In
the 2004-05 discussion groups, we divided those who were registered
into three typologies: the politically engaged, the duty bound,
and those who were also aware that registering is mandatory (though
in our 2010 post-election survey only 12% of the public were aware
that you can get fined for not registering).
40. Four in five of the public (82%) said they were satisfied
with the system of registering to vote in December 2009. However,
young people, many BME groups and those renting privately are
less likely than average to be satisfied, as are people who are
not on the register (only 44% of those who say they are not registered
are satisfied with the system).
41. Common reasons expressed for dissatisfaction include
the process being seen as inconvenient, difficult or time-consuming,
lack of confidence in the Council/authorities not to make mistakes,
and a belief that the post is unreliable.
42. Most of the public (89% in December 2009) are very
or fairly confident that they know how to register. Again, the
young, many BME groups, students, private renters and those not
registered are less likely to feel confident.
43. In the 2004-05 discussion groups, common top-of-mind
associations with registering included it being "old fashioned",
"time consuming", and a "chore". These perceptions
were particularly common among unregistered voters. However, participants
also did not have a great depth of knowledge about processes of
registering to vote.
44. However, more recent Ipsos MORI qualitative research
for the Electoral Commission in 2009 about the collection of Personal
Identifiers under a system of individual electoral registration,
showed that many valued the current canvass process as both painless
and reliable, and many participants spoke of an "routine"
process of completing the annual form.
45. Both our quantitative and qualitative work show low
levels of knowledge about many aspects of registration, and many
misconceptions. A common misconception is that registering to
vote is in some way "automatic". In the 2010 post-election
survey, 43% of UK adults said they believed that registering to
vote is automatic if you are aged 18 or over, and 31% that it
is automatic if you pay council tax.
46. Many also falsely believe that everyone has to provide
their National Insurance number, their date of birth or their
signature when registering to vote.
3 September 2010
| 
