Crewe and Numbskulls - The Snow In The Summer or So-So

Voting here and there

Election season continues, with two more chunks of psephology passing our radar.

First, the by-election in Crewe and Nantwich, following the death of Mrs. Dunwoody. Though there will be boundary changes between to-day's by-election and the next general election, they amount to about 1500 people, and slightly favour Labour.

Crewe and Nantwich was narrowly held by Mrs. Dunwoody in 1983 and 1987, becoming a little more secure following the last boundary revisions in 1997. It's the sort of seat where the Conservatives need to be there or thereabouts if they're to form an overall majority. It's not unfair to argue that, as goes Crewe and Nantwich, so goes the rest of the semi-rural north. If the Conservatives can win here, they can win a swathe of seats in semi-rural Cheshire, Lancashire, Yorkshire, Derbyshire, into north Wales... On the other hand, if Labour hold the seat, they're stronger in those marginal seats than might be the impression.

The incumbency effect means that Labour will be advantaged if they hold the seat at the next elections, and the Conservatives will be advantaged if they hold it. We should also consider a traditional protest vote at by-elections against the government.

At the last election, the result was Lab 48.8%, C 32.6%, LD 18.6%. An 8% swing in their favour will give the Conservatives the seat. Here's the matrix of outcomes for the major parties; we define clear win as a 5% margin over the runner-up:

Crewe and Nantwich results
PartyClear winNarrow winNarrow lossClear loss, secondThird place
ConservativesVery pleasedPleasedUnhappyVery unhappyUnthinkable
LabourEcstaticVery pleasedNeutralVery unhappyDisaster
Liberal DemocratsEcstaticEcstaticVery pleasedPleasedNeutral

Our gut instinct is that the Conservatives need that 5% cushion to have any realistic hope of retaining the seat at the next election. Anything less gives Labour a very good crack at winning it back. For Labour, retaining the seat would be a significant achievement, and a narrow defeat is not terminal. For the Lib Dems, third place is a given, anything better - even running Labour close for second - would be a feather in their cap, particularly after being squeezed from both sides.

The minor candidates are from UIP, G, OMRL, ED, and independents standing on platforms of Former Conservative, Cut Tax on Diesel and Petrol, and Miss Great Britain. For any of them, retaining their deposit would be a major achievement.

How will this by-election work? It's really very simple. Voters go to a designated site, authenticate themselves, and express their first preference on a form. At the end of polling, all the papers are counted and verified, and the single candidate with the most first preferences is declared the winner.

Over in St. Francisville, they can't come up with anything so simple. In fact, they seem to take great delight in proposing one system, and then using something similar, but different. Yes, it's another bite at the proposed rules for Livejournal's internal election. We pointed out last month that the rules, as published, would have led to the appointment of Jon Cruddas as the Labour party's deputy leader.

Again, the fact is that we contacted Livejournal on 25 April asking them to comment on this matter. At the time of writing, no response - substantive or holding - has been received. One candidate in the election, I. M. C., has been reviewing the relevant code, and finds that the counting algorithm will work in a slightly more sensible manner. We can thus revise our description of the original for accuracy, using strikethrough for removed text, and emphasis for new additions:

References to 51% throughout should be read as strictly more than half the valid vote. The threshold to exceed is not defined as valid_vote * 0.51 but valid_vote /2. This is a far more sensible implementation.

Furthermore, the non-effective non-transferable votes are not counted to the valid vote total. Non-effective non-transferable votes are ignored for the purposes of determining a winner: in formal terms, the Quota can reduce at each round. Again, far more sensible.

Taken in combination, these changes would (generally) return the same result as any other AV system used in the wild. Mrs. Harman, and not Mr. Cruddas, would be the Labour deputy leader.

Some of the more marginal arrangements do still cause us concern. In the event of a tie for elimination, all the candidates concerned are eliminated at once. This goes against the Electoral Reform Society's model rules, which require a review of earlier rounds, to eliminate the candidate securing the fewest votes early in the process. If that's still a tie, procedure differs: some will remove one candidate at random, others will conditionally exclude candidates and retrace to see if this affects the result. In the short time available to us, we've not found any other implementation that eliminates all the tied candidates and hopes for the best.

The final clause pertains to a tie at the end of the process: two or more candidates having equal numbers of votes. In this case, it's not entirely clear what happens - it could be that all active votes are transferred to their next preference, or the process halts, or something else happens. Nor is it clear if the first-preference tie-break is between the tied candidates, or all candidates; this case is not handled by the published code.

The code explicitly allows a voter to specify the same candidate in more than one preference; it is perfectly legal to vote 1) IMC 2) RM 3) RM. Again, we cannot find a previous implementation that allows a single candidate to appear twice on the preference list.

But these are minor details, and unlikely to alter the result. There is a far more fundamental question regarding the election process: is it governed by the code used to count the votes, or the description published to voters? There is, quite clearly, a discrepancy between the two rulesets. Followed to the letter, one of them would elect Harriet Harman, one of them would elect Jon Cruddas. For there to be any confidence in the process, the rules actually used in counting must be the same as the rules given to the electorate. Yet the algorithms give different results. They cannot both be right.

We have no particular preference for which set of rules should be used. We do have a position on votes that publish one set of rules and operate by a different set of rules. Words such as Kafkaesque, surreal, and phantasmagorical spring to mind, as do words such as Florida and Ohio. We find votes under such conditions to be fundamentally and terminally flawed, a violation of the standards of democracy and fair representation we hold dear. This confused method is not fit for anything other than meaningless, nugatory sinecures.

That'll be why such a vote is used for this position.

Addendum, 23 May

The Conservatives won the Crewe and Nantwich by-election by a margin of 19%.

When Livejournal finally opened the poll, early in the morning of the 23rd, the counting process had been rewritten, with all instances of 51% replaced by more than half. This addresses the substance of our criticism, though Mr. Pokery has documented other potential irregularities.

| Permanent link