Monday, May 03, 2010

UK Election Preview III: The Numbers

Unlike the US election in which we are polled and advertised to near-death, the UK election is over in a flash (this year, in three flashes called televised debates), so there is no chance to measure much change in momentum. While there is some attempt to take larger national polling samples that measure regional trends, sampling error on those has meant they haven't shown much clear movement, as opposed to some statistically significant shifts on the national level.

The state of the art in analysis is trying to infer changes in the national support levels for the major parties at the district ("constituency") level (where there is no polling data available). The Uniform Shift approach, which is predominant in English electoral analysis, takes the same percentage movement seen nationally and applies it to each district's race, which allows them to project the results for each and then aggregate to total party members elected. That's what all those "swingometers" (see also first preview) are about.

Our own Nate Silver (i.e., American) has brought his formidable analytical talents to the challenge and made an incremental improvement. Instead of a shift of X% of the total, he uses a matrix of From/To % shifts of their previous votes which add up to the desired national shifts. It also allows for a shift in the turnout (though he needs to guess where it goes), on the assumption that this more vigorously contested three-party battle will draw more voter interest. His findings are that the swingometers stickiness "in favour of Labour" is not justified, and that there could be a larger swing to the Conservatives than those methods predict, if there is a sufficient loss of vote level to Labour. His latest forecast is for the Conservatives to win 308 seats, to 198 for Labour and 113 for the LibDems (326 needed for a clear majority), with the popular vote going 35-27-28 (Cons.-Labour-LibDem); his forecast on the popular vote is in line with others', but he has the Conservatives further ahead with that margin.

The LibDems' surge has upset expectations that their ceiling is a mere 100 seats (about 15% of the total), even if they end up with something like 30% of the popular vote and ahead of Labour. Still, I have seen no projections anywhere that they can exceed 150 seats short of something totally unthinkable like their getting 40% of the vote.

The real crux of the matter, by all accounts, is still the relative position of the Conservatives vs. Labour. With even percentages, Labour gets more seats, and possibly a clear majority. More likely, based on the polling, Conservatives will lead Labour by 4-8%, but that range of variation allows for all the difference: if the Conservatives lead Labour by only 4% (say, 33-29), Labour will still likely have more seats, though not a majority; if the Conservatives lead by 8% (say 34-26), they will be at or very close to an absolute majority.

David Cameron is believed to have won the third debate, and the Conservatives were viewed by the polls to have a lead of about 6% over Labour going into that debate. Given that, the expectation now is that the Conservatives will have the most seats, and that they may be close to a majority. The LibDems may end up a couple of percent above Labour, but would (according to projections) have a lot less seats than them. This result will put the LibDems in a ticklish, and rather unfavorable position, in which they only have enough seats to help the Conservatives into power, either actively or passively.

At this point, though, nothing is certain. There could be a late move of support for any party, there could be some shifts due to "tactical voting" between LibDems and Labour (based on voters' expectations of the electoral chances of their local candidates), and the degree of turnout and its effect is very much uncertain. Parallel to the US electoral system, we can expect the Conservatives' turnout to be a given, with an increase in turnout likely to help either the LibDems, or Labour, or both. It is possible that the Conservatives will be short of a majority, but so close that minor and regional parties (such as Northern Ireland's Ulster Unionist Party) could put them over. Finally, local personalities figure prominently in voters' decisions, particularly when incumbents choose not to run again, and the effect of these changes is rarely considered in projections for individual seats.

There is a lot of betting in Britain on elections, something I have avoided looking at while doing these posts. I will take a look and post as comments, but here are my odds, and my bets:

Macro Election outcome: Conservative majority 30%; Labour majority 5%; LibDem majority 0%; LibDems hold margin to help either party to majority 20%; LibDems can help Labour only 5%; LibDems can help Conservatives only 40%.
This translates to my probabilities of the next Prime Minister: David Cameron 75%; Gordon Brown 15%; some other Labour leader 5%; Nick Clegg 5%.

My bet: Conservatives will hold a sufficient plurality (something like 290-230 over Labour) that LibDems, while technically able to help either party to a majority, will feel morally compelled to work with the Conservatives; they will choose to do so passively (a "minority government", in which the LibDems promise to abstain from opposing them in no confidence votes, short of a resolution to go to war) with the Conservatives promising a referendum (on proportional representation, which will not pass, by the way) in four years.

Postscript: I should mention the site, which makes projections seat-by-seat. They aggregate these into an overall projection, too, one that is not too different from others (297-235-86--Cons.-Labour-Libdem--in seats is their latest), but they also allow you to look up seat-by-seat and see what they think. Mostly, it's just a matter of "this seat is safe" or "this seat is definitely in play", but, I would argue, that's about as much as anyone else knows (except possibly a local insider, but that's not comprehensive across all seats).

1 comment:

Chin Shih Tang said...

Odds (from, a consolidator which shows bets from all the different houses):
No majority--approx. 2-to-3 (60%)
Conservative majority--varies, averaging about 11-to-8 (35% or so)
Labour majority--25-to-1
LibDem majority--100-1

On number of seats, BetFred has it by 25-seat ranges for the Conservatives. Both 325-349 and 300-324 are at 5-2 (29% each), 250-274 are at 10-1, 350-374 is at 6-1.

As for next Prime Minister, that's sort of a sucker bet (because Gordon Brown isn't included because he's already PM). The relevant bet is who will be PM on June 6 (i.e. in one month), and the figures are: Cameron 1-5, Brown 9-2 (he would probably remain PM in the case of a deadlock), David Milibrand 10-1 (the current Labour #2--there's a lot of sentiment among all parties to throw Brown over, even if Labour is to remain in charge), Nick Clegg 12-1, and all others 25-1 or worse.

My bet of all those would be for conservatives getting 275-299 seats at 9-2.