Spoilt Ballots in the PCC Elections: What Do the Numbers Tell Us?

Much has been said about spoilt ballots in the Police and Crime Commissioner elections, but no one has thoroughly analysed the evidence.  In this post, Alan Renwick, Reader in Comparative Politics at the University of Reading, goes through the data that are available.  He finds that there is evidence of more deliberate spoiling of ballots than usual, but the extent of this was limited.  The main story of the election is low turnout, not high spoiling.

One of the talking points surrounding the Police and Crime Commissioner elections has been the number of spoilt ballot papers.  Many observers at the 41 counts around England and Wales saw ballots with mini-essays on them, rather than votes.  The perception is that some voters expressed their disagreement with the idea of politicizing the police by deliberately casting an invalid vote.

But there are always a few voters who purposefully spoil their ballots and others who do so by mistake.  To see whether there is anything unusual about this election, we need precise numbers.  Unfortunately, these have not been easy to come by.  There is no official agency that gathers information on results across the country.  The closest we have to that is the BBC, but the BBC has published results that exclude rejected votes.  The only way the ordinary citizen can find out about spoilt ballots is to go to the website of each of the 41 local authorities responsible for organizing the counts and check their numbers.

The information available is summarized in the table below.  Unfortunately, even this information is not complete: ten of the responsible local authorities have (so far as I can see) not published the relevant information.  Some of them have given no information on rejected votes.  Others have given detailed information only on the second round of counting, whereas it is the number of votes rejected at the first round that matters here.  Nevertheless, the figures that are available cover most of the country and give us a good impression of what has been going on.

Police   area

Total   votes cast

Votes   rejected at 1st count

%   spoilt ballots

No.   candidates

Source   of data

Humberside

136,071

2303

1.69

7

East Riding
Kent

208,848

3931

1.88

6

Kent PARO
Essex

171,686

3452

2.01

6

Chelmsford
Durham

71,586

1445

2.02

4

Durham
Nottinghamshire

135,742

2769

2.04

4

Rushcliffe
Northumbria

182,694

3887

2.13

4

Sunderland
Cheshire

111,335

2415

2.17

5

Halton
Merseyside

126,171

2915

2.31

6

Liverpool
Gwent

60,921

1555

2.55

4

Newport
Dorset

98,679

2527

2.56

4

Poole
Hampshire

217,481

5595

2.57

6

Southampton
Gloucestershire

80,618

2115

2.62

4

Stroud
Derbyshire

115,974

3049

2.63

4

Amber Valley
Leicestershire

127,081

3371

2.65

3

NW Leics
North Wales

77,753

2150

2.77

5

Flintshire
Staffordshire

98,826

2843

2.88

2

Stoke
West Midlands

245,447

7063

2.88

7

Birmingham
Northamptonshire

120,581

3474

2.88

5

Kettering
South Wales

146,899

4465

3.04

4

Neath
West Mercia

139,123

4273

3.07

3

Shropshire
Cumbria

64,263

2014

3.13

4

S Lakeland
Cambridgeshire

91,495

2892

3.16

7

East Cambs
Devon and Cornwall

196,987

6339

3.22

10

Cornwall
Norfolk

100,408

3251

3.24

5

Norwich
Thames Valley

226,516

7445

3.29

6

Oxford
Wiltshire

81,477

2683

3.29

6

Wiltshire
West Yorkshire

223,005

8277

3.71

4

Wakefield
Suffolk

88,498

3330

3.76

4

Suffolk Coastal
Avon and Somerset

244,042

9190

3.77

4

Bristol
Dyfed–Powys

67,572

2912

4.31

2

Pembrokeshire
North Yorkshire

88,619

6406

7.23

2

York
Total of these 31 areas

4,146,398

120,336

2.90

 

 

Areas for which information on rejected ballots is not available from official sources: Bedfordshire, Cleveland, Greater Manchester, Hertfordshire, Lancashire, Lincolnshire, South Yorkshire, Surrey, Sussex, Warwickshire.

 

The headline figure for the country as a whole (or at least for the 31 areas for which we have information) is that 2.90 per cent of all votes cast were rejected.  Is this high or low?  Compared to general elections, it is very high: spoilt ballots have made up about 0.3 per cent of the votes cast at recent general elections, and the figure seen this week is about ten times that.

So what might explain the difference?  Three obvious possible explanations immediately spring to mind: confusion over the electoral system; dissatisfaction with the candidates; and disapproval of the elections themselves.

In attempting to distinguish between these, we might begin by taking advantage of the fact that different electoral systems were used in different areas.  The basic system was the supplementary vote: voters could state their first and second choices in separate columns on the ballot paper; where no candidate won 50 per cent of first preferences, all but the top two candidates were eliminated and votes cast for eliminated candidates were redistributed according to second preferences.  If there are only two candidates, however, the second round is redundant; in such cases, simple first past the post was therefore used instead.  We might expect, if confusion over how to express two preferences was a problem, that the number of rejected votes would be lower where first past the post rather than supplementary vote was used.

In fact, the opposite pattern holds: of the 31 areas to have published the relevant information, three had only two candidates and used first past the post; two of these – North Yorkshire and Dyfed–Powys – were the two areas where the proportion of spoilt ballots was highest.

But that does not show that the supplementary vote is less confusing than first past the past.  The more likely explanation is that voters faced with a choice of only two candidates were more likely to find that choice inadequate and therefore cast a spoilt ballot.  Dyfed–Powys, for example, is an area in which the Liberal Democrats and Plaid Cymru both typically do well, but neither party ran a candidate.  Some of their supporters may have chosen to invalidate their vote rather than back either Labour or the Conservatives.

Two other ways of differentiating the three probable main sources of invalid voting are available.  One is to compare these elections with others using the supplementary vote.  London mayoral elections (and, indeed, all other mayoral elections) have been using the supplementary vote since 2000.  These elections have attracted many candidates and the value of holding them has generally been accepted, so we would expect the number of deliberate spoilings in them to be no higher than usual.  The proportion of rejected votes at these elections therefore gives us a rough indication of the level of invalid voting generated by the electoral system alone.

The proportion of rejected ballots in London was 2.2 per cent in 2000, before rising to 3.0 per cent in 2004 and then falling back to 1.7 per cent in 2008 and 1.8 per cent in 2012.  The 2000 figure might roughly indicate the proportion of spoilt ballots that we should expect to arise through voters’ confusion when confronted with the supplementary vote system for the first time.  The figure of 2.9 per cent for the PCC elections (2.8 per cent in the areas using supplementary vote) is only slightly higher than this.  On this evidence, it seems likely that most spoilt ballots were produced by voter confusion over the electoral system.

The final way of investigating the effect of the electoral system utilizes the fact that returning officers categorize rejected ballots according to the reason for their rejection.  For this election, they employed five categories:

  • want of an official mark;
  • voting for more than one candidate as to the first preference vote;
  • writing or mark by which the voter could be identified;
  • unmarked as to the first preference vote;
  • void for uncertainty as to the first preference vote.

Of the 41 police areas, 24 have (so far) reported the breakdown of rejected ballots.  The following table shows the total numbers across these areas.  It also shows the results separately for the 21 of these areas that had more than two candidates and therefore used the supplementary vote system and for the three areas with only two candidates, which used first past the post.

 

All   24 police areas

21   areas using SV

3   areas using FPP

 

Votes

As   % rejected ballots

As   % all ballots

Votes

As   % rejected ballots

As   % all ballots

Votes

As   % rejected ballots

As   % all ballots

No official mark

68

0.07

0.00

68

0.08

0.00

0

0.00

0.00

>1 first preference

28746

30.39

0.91

26585

32.25

0.91

2161

17.77

0.85

Identifying mark

1546

1.63

0.05

1477

1.79

0.05

69

0.57

0.03

No first preference

33043

34.93

1.04

26407

32.03

0.91

6636

54.57

2.60

Uncertain first pref

31194

32.98

0.99

27899

33.84

0.96

3295

27.09

1.29

Total

120336

 

100

2.90

 

108175

 

100

2.84

 

12161

 

100

4.77

 

 

The first of these categories of spoilt ballot is technical and tiny.  Of the remaining four, it is tempting to posit that the first is likely primarily to reflect voter confusion – confused voters might mark both of their choices in one column – and that the remaining three primarily come from deliberate spoiling – voters casting a deliberate protest vote are unlikely to express any preference and might sometimes write on their ballot paper in ways that generate uncertainty; some might sign their anti-election statements.  If these assumptions are accurate, then it appears that around two thirds of the rejected votes were spoilt deliberately, while one third were accidental.

But are these assumptions justified?  On the one hand, a comparison with London suggests that they capture a useful part of the truth.  The proportion of all ballots that were rejected because more than one first preference was marked was 0.91 per cent in the PCC elections and 0.97 per cent in the 2012 London mayoral election (the breakdown of rejected ballots in London is available only for 2012).  The proportion of all ballots rejected for the third, fourth and fifth reasons was 1.92 per cent in the PCC elections but only 0.81 per cent for London mayor.  The similarity of the first two numbers fits the suggestion that they reflect broadly constant levels of accidental spoiling.  The difference between the second two numbers fits the idea that they capture deliberate spoiling, which was higher in the PCC elections than the London elections.

On the other hand, that 0.85 per cent of voters expressed more than one preference even in the three police areas using first past the post casts doubt on the notion that all of these votes were accidentally spoilt: there is no reason to think such accidental spoiling would be higher in this election than any other.  Some voters may deliberately have spoilt their ballots by voting for everyone.  It is also likely that some voters accidentally end up in the fourth and fifth categories.

So what can we conclude from all of this?  First, the proportion of spoilt ballots in the PCC elections was slightly higher than in the most comparable other election: the first London mayoral election, in 2000.  Second, there is some evidence from patterns in the reasons for rejection that the difference was due to higher levels of deliberate spoiling.  On the other hand, the evidence is not overwhelming and the differences are small.  Deliberately spoilt ballots were perhaps cast by something in the order of 1 per cent of those who voted.  Third, there is strong evidence that some deliberate spoiling reflected dissatisfaction with the choice of candidates.  But heightened spoiling was found in most areas, so a protest against the system itself very likely also played a part.

These conclusions are rather more measured than might be expected given the talk that has surrounded the results.  It seems likely that the extent of deliberate spoiling has been exaggerated.  What might account for that?  One explanation is of course just that high levels of invalid voting make for good headlines.  Another may be that the two areas that declared their results first – Wiltshire and Dyfed–Powys – both had unusually high numbers of spoilt ballots.  A third possibility is that different voters were spoiling their ballots this time from in most elections: this time, spoilt ballots may have been cast by the commentariat and their friends, whereas normally they may be cast by voters who are more marginalized.

In any case, some voters clearly chose actively to express their rejection of the new system of Police and Crime Commissioners, but their numbers were low.  The main story of the election is low turnout, not high spoiling.

  1. cer’s avatar

    Thanks for doing this, I’ve been looking for these figures.

    Reply

  2. Iain Simpson’s avatar

    Very interesting article. As a voter who deliberately spoilt his vote for the first time in 50 years, I have no hard evidence. I spoke to two or three people about my intention to spoil my vote as a protest about the election and the new system and received some encouragement with a suggestions that they might do likewise. Also others indicated that they had considered the option before casting a valid ballot.
    It may be that as a member of the ‘commentariat’ I was just unwilling to be associated with those who were too lazy or uninvolved to vote.

    (Staffordshire)

    Reply