Friday 4 January 2013

4. How much effort does your opponent put in, and are they any good?

Moving on to my mostly hypothetical musings on the performance of parties. Oh this is definitely the fun bit! I spent 25 years analysing vast quantities of pupil performance data in various ways when I was a teacher, and I love the application of maths. I'm presently intrigued by the use of above or below average as a measure of a campaign's effectiveness. I've only seen percentage of votes won used as a measure of campaign effectiveness, but that doesn't tell me much.

Let's take a typical ward - St. Peter's,  2008, because I know very little about it, and see what we can find out just from the numbers. (source: Wikipedia; Sunderland Local Elections). I've deliberately left off the candidates names.

Party
Number of Votes
% vote
Average % vote for that party in whole of Sunderland, 2008
Difference of actual from average

1,501
47.8
32.0
+15.8

840
26.7
39.6
-12.9

443
14.1
11.6
+2.5

358
11.4
10.2
+1.2
Majority
661
21


3,142
37.8








Let's assume that the % of  votes in the whole borough is a fair measure of how a party did in that particular election. Therefore, one would expect a candidate performing averagely to get a percentage of votes similar to that of the way his/her party performed in the area.

What I'm going to do is compare the average % vote for a specific party for all of Sunderland against the actual % vote a candidate achieved, and remember, this has nothing to do with national or larger geographical factors. We're neutralising the factors associated with how a party is performing nationally; it's irrelevant to this calculation, as is the area, whether it be a Sunderland suburb or a posh southern constituency. This is just a measure of individual ward achievement.

The Conservative candidate here performed +15.8% higher than his party's average achievement in Sunderland, whilst the Labour candidate underperformed by -12.9%. Small gains on party performance were made by the LibDems, +2.5 and the BNP candidate at +1.2%

Lots of campaigning put in by the Conservatives in this ward then, at the expense of Labour. If you like, the Conservatives added 15.8% to their vote share by - no doubt - hard graft. There are all sorts of additional factors apart from campaigning that can make the difference to a ward/candidate's "value added" measure of performance. Here are some I thought of:-

1. The Boris factor - in other words, personality. If  people feel that a candidate is personable, charming, and possibly even a tough sod who'll get the job done, they'll be more likely to vote for them. Conversely, any appearance of being "not up to the job" (I'm being diplomatic) will lower their vote share. I should perhaps add that I've absolutely no idea who this data refers to; I just tell the tale the numbers give me.

2. The deprivation of a particular ward. It's a fact that richer areas, with more expensive houses, have a higher Conservative vote share (See psephology blog 2 on left-right voting). It's traditionally believed that more educated, richer people are more likely to vote Conservative. Also, the poorest areas have a higher BNP vote. In other words, the class of the voter. Controversial these days, but it can't be ruled out.

3. What we're really measuring is that for some reason, less (or more) people wanted to vote for a party in the ward, compared to the rest of the city. The electoral make up in a ward may be significant - a more educated, voter population tends to vote LibDem (or so I've heard, but I'm partisan). And indeed, voter education also tends to be related to higher turnout overall. I won't go into gender, marital status, ethnicity and occupation, which all may be significant, but which would require a high powered multivariate analysis to analyse properly. Hey, this is just a blog!

5. Is the candidate right for the ward. I observed one LibDem who performed poorly in one ward;  +0.1%  and very well in a different ward;  +17.8% in the next election. Now I don't suppose the candidate had a personality change overnight, or suddenly had the urge to go on hundreds of extra leafleting runs, but in that second ward, with those constituents and against those particular Labour opponents, it was a whole new ball game.

6. In smaller parties, in particular, fielding many candidates in 25 wards instead of targeting say half a dozen key wards spreads support too thinly, with consequent falling votes - we're told that targeted campaigning works, again and again; it's a good point. For example:- one ward had a value added/performance measure for the LibDems of  +7.5% above average in 2007 when 15 candidates were fielded, and +1.4 in 2008, when 22 candidates were fielded.

You have to at least ask the question, was the increased workload on the party significant? The BNP on the other hand, obviously over-reached themselves financially after fielding many candidates for so long; now they tend to target the wards where they have the most support. I'd better not say what I think of that, except that deluded doesn't necessarily mean stupid.

How does my hypothetical measure compare to the usual measure of a candidate's performance - their share of the vote on election day. Here's the data for the 2012 local elections:-


The red shows the normal way of looking at this sort of data - how much the candidate increased or decreased votes since the last election. Apparently all was rosy in the Labour camp, apart from in Hetton. It's difficult to tell where they're failing. However, when you compare the performance of each ward using my performance  measure, it becomes obvious which wards (green) fall below expected standards, zero being the average. So as a measure of where this party is underachieving, it is quite effective. So for instance, in St Michaels, Millfield and Copt Hill, Labour performed quite badly. In fact, these wards were all taken/held by other parties. Conversely, Labour achievements in Ryhope and Sandhill, in contrast, were excellent. So this way of looking at things can be useful.

You could, of course, construct tables using this value-added measure of  candidates, to see which individual rivals are weakest and most likely to be open to competitive campaigning, and which of our own candidates are particularly effective. Would I do that? Of course I would! Which brings up some interesting ethical problems. Is it meretricious to publish this data to advise or indeed to put the wind up one's opponents? Hmmm. That's a tough one, but I'm an ethical sort, so let's leave it for now.  I will say this: well done to the LibDem candidate who outperformed his party's average % vote by +20% in one election - yet he never ran again.  Now that's depressing! We can however use such "value added" predictors as a good indication of which seats are good targets, for whatever reason you choose to espouse.

Case Study 1 - The Houghton wards

I worked out the "value added" statistic for two of the Houghton wards and saw a significant trend in Labour performance in this area for reasons which will become clear.

Ward
2004
2006
2007
2008
2010
2011
2012
Copt Hill
19.6
9.8
-7.7
-5.3
-4.4
-16.8
-20.5
Houghton
26.4
9.9
7.3
-2.6
-2.3
-8.4
-15.9


Remember, zero represents the average Labour performance (different each election) in the whole city. As you can see, in this area, what was initially a healthy performance of +20% above the average city vote gradually declined in this period until the Labour candidates were significantly underperforming (see graph). Just for interest, here are the increases/decreases in % vote.
2007
2008
2010
2011
2012
Copt Hill
-15.1
-1.1
10.5
14.3
9.59
Houghton
0.1
-13.7
10.1
5.2
8.79


Things seem to be getting better and better, and are fairly positive. Not so. In fact Labour lost to Independents in Copt Hill in 2007, 2008, 2011 and 2012, and in Houghton in 2008 and 2012. The Conservatives didn't even bother to run a candidate some years. Why you might ask? The big issue here; the reports coming out at this time about the huge local landfill site contaminating the local water supply. For a flavour of their very public spats:-

Case Study 2 - Labour achievement against other parties

My ward, St. Chad's is a battleground between Tory and Labour. You can see the Labour candidates' fluctuating performance as the ward elects councillors(3) of different persuasions.In Fulwell, a Conservative stronghold, their achievements are never less than dire. In Millfield, the LibDem stronghold, even when they ousted the last LibDem in 2012, their performance was still way short of where the rest of the party in Sunderland  were. 16.6% below the average percentage achieved by other wards in the city. I like to think it's because we gave them a run for their money.

Ward
2004
2006
2007
2008
2010
2011
2012
St Chad's
-4.1
-9
-3
-3.1
8.9
-7.4
-6.4
Fulwell
-32.7
-15
-18.2
-13.4
-13.5
-18
-16.4
Millfield
-18.6
-9
-10.4
-15.4
-11
-6.9
-16.6



More on targeting and ward ranking next time.

No comments:

Post a Comment