The Ten Per Cent Solution
Amending the Parliamentary Voting Systems and Constituencies Bill: Why a 10 per cent permitted variation is superior to the government’s 5 per cent rule
Much less crossing of county boundaries
A 5 per cent rule involves violating the boundaries of well-established local units in a way that does not take place in apparently comparable systems of equalisation in Australia and the United States. For a county to avoid sharing one or more seats with another county, it needs to meet two criteria.
• Its electorate size permits a whole number of seats to be given within the allowed variation. For instance, a county with an entitlement to 5.3 constituencies cannot be given a whole number of seats because its average seat would be 106 per cent of the standard national size.
• Even if a county is technically entitled to a whole number, it might be practically impossible – for instance, if a county has 5.7 times the national quota of electors, it could have six seats all at 95 per cent of standard size. In practice it will be impossible to find a sensible division of the county to permit such exact slicing.
• Even if its electorate is compatible with a whole number of seats, it may still need to have a cross-county seat because a neighbouring county is out of balance. For instance Suffolk, of itself, could have 7 seats quite easily under the government plan. But because Norfolk is a long way off a whole number entitlement, Suffolk ends up having to share.
Very few counties meet these criteria in England with a 5 per cent limit. In the Democratic Audit model boundaries using a 5 per cent rule, only Cumbria, Staffordshire, North Yorkshire, Humberside, Lincolnshire, Gloucestershire, Berkshire, Oxfordshire and Buckinghamshire escaped – 9 out of 46 counties, accounting for 67 of the 503 seats proposed for England. Relatively small future changes in electorate size would lead to disruptive change to the county groupings every parliament.
A 10 per cent tolerance of variation would transform this chaotic picture. No counties fail outright (other than the Isle of Wight) but in practice a few are close enough to the edge to make pairing perhaps necessary. Wiltshire and Dorset, and West Yorkshire and South Yorkshire, would be the only pairings required under a revised plan based on 10 per cent.
Much less splitting of wards
It is probably impossible to implement a 5 per cent rule without splitting wards between constituencies, something which the Boundary Commissions currently avoid doing because of the potential for voter confusion and highly artificial constituency boundaries, not to mention causing headaches for the organisation of all political parties. The model distribution using 5 per cent run for Democratic Audit attempted to minimise ward-splitting, but it proved unavoidable in some areas. The worst-affected areas are those where wards have large electorates, such as the English metropolitan boroughs, most of Scotland and some unitary authorities and London boroughs.
A rigid 10 per cent rule might still involve a few isolated cases of ward-splitting, but it is likely to be very uncommon in comparison with a 5 per cent rule.
Fewer and less disruptive boundary changes in future
The government’s Bill proposes that the boundaries will change every election, which disrupts the relationship between MP and constituency and will no doubt lead to confusion. Because the 5 per cent limit is so tight, many constituencies that were the right size in one boundary review will be too big or too small by the next. This will happen because of growth and decline in population. It will also happen because of variations in electoral registration from year to year, which are likely to be larger under the forthcoming Individual Electoral Registration system. It is quite possible that radical changes in boundaries will be made for no better reason than fluctuating registers, which as we know have become much less stable, complete and accurate.
With a wider permitted variation, fewer constituencies will go above or below the limit because of population or register change even if the boundaries are reviewed every election. It would make for a more sensible and stable system of boundaries, and better representation for constituents, if the frequency of reviews were to be 10 rather than 5 years anyway.
Closer concordance with community identities
The government’s proposals will involve areas, particularly rural areas, being moved into constituencies to make up numbers at the cost of making the constituency less cohesive. It is not clear, because the government has not bothered asking, whether people would prefer to be in a slightly larger seat that makes up a coherent whole, than in an appendage to a constituency based on different communities that is the ‘right’ size.
Taking the borough of Doncaster as an example, it would be entitled to 2.87 seats under the new rules. In practice, drawing three seats of just over 95 per cent of the standard size would involve splitting wards and creating illogical boundaries, so it is more likely that a ward of a neighbouring borough would be drawn into a seat. With a 10 per cent limit, Doncaster would have three seats of its own without a problem. Coventry would be festooned with rural wards bringing up numbers with 5 per cent, but again capable of having three city seats under 10 per cent.
The hard cases
The government’s Bill proposes to tolerate an extremely small group of anomalous seats – two island groups in Scotland plus perhaps a seat with a large land area in the Highlands. Most other systems of districting (including the US and Australia) allow some latitude for small states, difficult territory, or both. The Bill insists that 99.5 per cent of seats are within 5 per cent of the national quota, while 86 per cent is good enough for the United States and 67 per cent is fine in Australia. There is ample room for a few more constituencies to depart from the national average in the interests of geography, local identity or simple common sense.
Among the most deserving cases are:
• Isle of Wight. The Bill proposes to have at least one seat that straddles the channel between the Isle and Hampshire, something that would create an absurd constituency with no coherent identity. In past boundary and local government reviews local opinion has opposed the division of the island.
• Cornwall. Cornwall has something of a special status among English counties – in some ways it is only administratively rather than culturally ‘England’ at all. Local opinion in Cornwall has been against a ‘Devonwall’ seat crossing the county boundary. It would not affect overall equality much to give it (plus the Isles of Scilly) a specific exemption, although the decision whether to give it 5 or 6 seats of its own is a fine one.
• Anglesey. The island is, unlike Wight or the Scottish island groups, linked to the mainland by bridges, but then again it is not as far from the national quota at around 50,000 electors. Equalisation would mean putting Bangor into the same constituency.
• Argyll & Bute. This is in some ways an even sparser constituency than the Highland seats because of its extremely long coastline, islands and the difficulty of internal communications.
• North west Wales. Two factors come into play in this area. The difficult and mountainous geography has caused the Boundary Commission to exercise a bit of leniency in the past. The representation of indigenous national minorities is also a recognised criterion for drawing up electoral districts, and the government’s plan would probably involve creating fewer Welsh-speaking majority seats than their proportion in the population.
• Estuaries. It would make sense to ban constituencies straddling wide estuaries such as the Mersey, Humber, Clyde, Forth and Thames.
• Welsh valleys. There might be a case for allowing small departure from the usual rules if following them could lead to an absurd seat with a small part of one valley attached to a seat based on another valley.
The population/ register factor
There is scope for a fuller discussion of the appropriate way of measuring constituency ‘size’, particularly given the deterioration in the completeness of electoral registers in some seats in inner urban areas in particular. However, the government’s haste to act has not permitted this.
In the absence of a thorough look at the issue, it would be possible and desirable to create a class of seats that were allowed to be a bit smaller than the norm because they are grossly overpopulated. If it is unacceptable for an MP to have a constituency with more than 13,000 square kilometres in area because it makes it impractical to represent, it is also unacceptable for a constituency to have a vast resident population. It is at least as difficult for an MP to represent an abnormally large number of constituents as it is to navigate around a large seat.
Overall, the electorate is about 75 per cent of the population – there are around 61 million people in the UK population and a bit over 45 million in the electorate. However, there are a number of constituencies where there are vastly more constituents than there are registered electors. Often, the factors that produce low levels of electoral registration or large ineligible populations coincide – young people, including foreign citizens, living for short periods in private rented accommodation comprise a large proportion of the population in inner London in particular.
The table below shows, on the basis of Office of National Statistics research compiled in 2007 (for the old constituencies), the English constituencies in which the population represented by the MP is much larger than normal compared to the registered electorate. Several Northern Ireland constituencies would also feature in this table, but none in Scotland or Wales.
An amendment to the Bill, analogous to the provision that already exists for land area, could limit the population allowed in any constituency. If no constituency were permitted to have more than, say, 125,000 in population the number affected would be relatively small but the impact on the service to constituents in some hard-pressed areas could be considerable.
Partisan effects and electoral bias
The results of a – rougher – Democratic Audit simulation using 10 per cent toleration are actually more favourable to the Conservatives than those under strict 5 per cent equalisation.
This may well be an effect of the model’s methodology, but the conclusion has to be that there are no significant differences between 5 per cent and 10 per cent equalisation as regards their partisan effect. The differences are in the ability of 10 per cent equalisation to better accommodate natural communities and administrative boundaries.
It is worth noting that whatever set of boundary changes are proposed, the effect on partisan outcomes is relatively small. This is because, as demonstrated by repeated academic research, constituency size is only a small element of the electoral system’s current bias in favour of Labour. The illustrious psephological team of Ron Johnston, Colin Rallings and Michael Thrasher, and Galina Borisyuk, has analysed the 2010 election (http://www.essex.ac.uk/government/epop/Papers/Panel16/P16_Borisyuk_EPOP2010.pdf) and the general issues (Borisyuk, G., R. Johnston, C. Rallings, and M. Thrasher. “Parliamentary Constituency Boundary Reviews and Electoral Bias: How Important Are Variations in Constituency Size?” Parliamentary Affairs 63, no. 1 (2010): 4-21) conclusively. Their estimate of Labour’s gain from the size factor is a net 13 seats over the Conservatives; using a different methodology I calculated it as being a net 15, a lot of which was caused by the over-representation of Wales.
The real major sources of electoral system bias are not to do with constituency size, but with differential turnout and the uneven way parties’ votes are distributed across the country. Some model examples are given in the Appendix.
There are significant new issues to consider in the sphere of electoral registration and how constituency boundaries are drawn, which would benefit from a period of consideration and consensus-forming which the government has prevented by the haste of its Bill and its unwillingness so far to compromise. It would be better to take a little longer and devise a system that will last. I am not alone in thinking that there will be only two boundary reviews under these rules – one reporting by 2013 and in force from 2015, and another reporting in 2018. At that point, MPs will revolt at the prospect of repeated disruptive boundary reviews, as they did in similar circumstances in 1958.
The Bill is capable, though, of being sensibly amended to produce a system that will last a bit longer and work with the grain of practicality and local identity. The key measures to ameliorate the Bill would be, in summary:
• A 10 per cent permitted variation in seat size, replacing the 5 per cent rule
• Either a list of further permitted geographical special cases, or the restoration of the Boundary Commission’s ability to use its discretion for special geographical circumstances.
• Creation of a new category of anomalous seats to deal with the new problem of constituencies with grossly oversized population compared to electorate.
• Boundary reviews every 10 years rather than every 5 years.
I have already argued elsewhere for the restoration of the right to a public inquiry.
Amendments of this nature will not prevent the achievement of a level of equality between constituencies that is towards the top of the international league table, and closer to equality than the Australian House of Representatives (and possibly the American House of Representatives as well). They will enable the new constituencies to be more identifiable by their constituents, more stable, more sensible with respect to the physical and administrative geography, and therefore more likely to be sustainable – while still meeting the government’s aim of greater equality.
It would be in the best traditions of the House of Lords in improving legislation if the Upper House were to pass amendments such as these to the Parliamentary Voting System and Constituencies Bill.
ELECTORAL SYSTEM BIAS: SOME ILLUSTRATIVE EXAMPLES
In this example there are three seats, one marginal and one safe seat for each of the two big parties. Parties win their safe seats by exactly the same share of the vote – 68 per cent to 32 per cent. In an election where turnout is uniform this produces the ‘right’ result – see ‘1964’ below.
However, if there is a fall in turnout which is unevenly distributed this disrupts the relationship between seats and votes. Even if there is no swing in any constituency – the winners still hold their safe seats 68:32 – but turnout drops in one party’s safe seats this causes an apparent national swing. In the example below there has been a fall in turnout in the safe Labour seat.
We now have a strong bias to Labour deriving simply from turnout – Labour has ‘won’ this election despite being 4 points behind the Conservatives in the popular vote.
In 2010 the turnout in Labour’s hundred safest seats was 58.9 per cent, and in the Conservatives’ hundred safest seats it was 68.5 per cent.
If one did away with differential turnout, the Conservatives’ lead over Labour would have shrunk from 7.1 per cent to 5.4 per cent – the latter figure is the one the electoral system ‘thought’ there was.
There are only three methods of correcting turnout bias. One is to eliminate the turnout differential by making voting compulsory. Another is to – unlike any other country in the world other than Belarus – use the numbers voting as the basis for drawing parliamentary seats. Another is to have an element of national proportional representation in the electoral system.
VOTE DISTRIBUTION BIAS
In this example of electoral system bias Labour has come third in votes and first in seats, even though every seat is the same size and turnout is equal too. The bias arises because Labour’s vote is enough to win in the party’s stronger seats but the party has few votes in its weaker areas. By contrast, the Lib Dem and particularly the Conservative vote is relatively high even in the seats that party is not winning. Labour’s vote was distributed in this way in 2005, while in the 1950s the Conservatives tended to benefit from this form of electoral bias.
This example also demonstrates the ‘third party seats’ effect. In the three seats won by the Conservatives or Labour, the two parties’ votes are equal at 80,000 (Labour’s is more efficiently distributed, hence the 2:1 advantage in seats). In the Lib Dem seat, there is a substantial Conservative vote but a low Labour vote. This Tory advantage in ‘Berwick’ is reflected in the vote total but does not affect the number of seats.
Another form of vote distribution effect on bias is when a party wins its safe seats by an enormous margin but the other party wins more narrowly. This was the source of anti-Labour electoral bias in the 1950s because, crudely, Labour was winning working class strongholds 80-20 and the Conservatives’ rural and suburban seats were being won 65-35 or so.
Lewis Baston All are equal, but some are more equal than others. Is ‘equalisation’ of constituency electorates an international norm from which Britain departs? Democratic Audit, January 2011 http://www.democraticaudit.org/download/Equalisation%20-%20international%20experience.pdf
Lewis Baston How pressing is the need for further equalisation of constituency electorates? Democratic Audit, December 2010 http://www.democraticaudit.org/download/How%20pressing%20is%20the%20case%20for%20further%20equalisation.pdf
Lewis Baston Written evidence to the Political and Constitutional Reform Committee of the House of Commons August 2010 http://www.democraticaudit.org/download/Evidence%20on%20boundaries%20bill.pdf
Lewis Baston and Stuart Wilks-Heeg How strong is the case for having fewer MPs? Democratic Audit, November 2010 http://www.democraticaudit.org/download/How%20strong%20is%20the%20case%20for%20reducing%20the%20number%20of%20MPs.pdf
Lewis Baston Written evidence to the Welsh Affairs Select Committee of the House of Commons August 2010 http://www.democraticaudit.org/download/Wales.pdf
Stuart Wilks-Heeg Written evidence to the Political and Constitutional Reform Committee of the House of Commons September 2010 http://www.democraticaudit.org/download/Political%20and%20constitutional%20reform%20select%20committeee%20%28S%20Wilks-Heeg%29.pdf
House of Lords Constitution Committee report http://www.publications.parliament.uk/pa/ld201011/ldselect/ldconst/58/5803.htm#a4
House of Commons Political and Constitutional Reform Committee report http://www.parliament.uk/business/committees/committees-a-z/commons-select/political-and-constitutional-reform-committee/news/pvsc-bill-report-/
House of Commons Welsh Affairs Select Committee report http://www.publications.parliament.uk/pa/cm201011/cmselect/cmwelaf/495/49502.htm
Borisyuk, G., R. Johnston, C. Rallings, and M. Thrasher “Parliamentary Constituency
Boundary Reviews and Electoral Bias: How Important Are Variations in Constituency Size?”
Parliamentary Affairs 63, no. 1 (2010): 4-21.
Borisyuk, G., R. Johnston, C. Rallings, and M. Thrasher “Electoral bias in 2010: evaluating its extent in a three party system” EPOP paper September 2010 http://www.essex.ac.uk/government/epop/Papers/Panel16/P16_Borisyuk_EPOP2010.pdf