Belatedly, I can at last say that my spreadsheet model is now available to download from Google Docs. The instructions are now in the adjacent post on PB2 and in that post is a link to the model.
Looking at the effects identified in my original analysis, we can now expand on the details â€“ thanks largely to the in depth and constructive discussion on pb.com.
Firstly, the â€œConstituency Effectâ€. I now think that this is a poor name, but I can’t think of another. To expand upon what it is:
Labour, when they win power, typically have a vote share composed of their core vote and the further vote they’ve won to get over the line (non-core vote). The core vote tends to turn out noticeably less and be concentrated in a couple of hundred safe seats, which are (on the whole) smaller than the average constituency size in terms of population and overall turnout. When they are on the up, they capture a segment of the vote that is non-core to any party. This segment turns out to vote at a noticeably higher rate than the Labour core. So when they are on the way up, the increase in their core vote (concentrated in the smaller seats) is dwarfed by the increase in their non-core vote (which, would therefore be disproportionately represented in slightly larger seats). This is reversed when they are on the way down â€“ the lower turn-out core vote will drop less than the higher turnout non-core vote.
Therefore, on average, each constituency Labour vote will increase by more than the national swing, as the smaller constituencies will increase by slightly less (higher proportion of core voters) and the larger ones by slightly more (larger proportion of core voters).
Reports from pbc’s Labour MP that, unusually, the non-core vote is holding up whilst the core vote is dropping strongly (largely due to â€œstay-at-homesâ€) should therefore be considered very important. If this is so, and it is replicated across the country, Labour could even prevent the expected unwind â€“ or even wind it further.
The second effect, the Tactical Vote. From thinking about the Marginals Boost Effect, I’ve concluded that I have split the tactical voting into two components of which this is one (and the other is merged in with the marginals boost). This is the average effect across all Con/Lab battles, regardless of how marginal. It should be concentrated in marginal constituencies (and those within reach of a big swing). There will be some effect even in safe seats â€“ voters are often not fully in the frame of how marginal a seat is, and who is the main competitor (which is the entire reason for the existence of bar-charts â€¦). However, this should be viewed as the average tactical vote nationwide, with the concentration of tactical votes in marginal consistencies being contained within the marginal boost effect (positive in marginals/semi marginals, negative in safe seats).
The third effect, the Marginal Boost. This consists of many components:
– Concentration of tactical vote
– Increased attraction to C1/C2s (Mondeo Man and Worcester Woman). This is a crucial component, as they are highly concentrated in METTHs (see Blair Freedman’s excellent article )
– Increased attraction/turnout amongst â€œGuardianistasâ€ (usually AB demographic).
– Number of activists in marginal seats. This effect has been well explained by â€œBunnco’s excellent article on pb2 .
– Net effect of targetting. Money spent and concentration of effort on the identified target seats. All parties will be doing this, the net effect is a factor of which party wins the â€œGround Warâ€
– Other. Effects not mentioned above, beyond those identified. One effect that might factor in the coming election was identified by Bunnco on pb2.
My â€œwet finger in the airâ€ view is that the above components will weight as follows:
– Tactical vote concentration: 1/5th
– Attraction towards C1/C2: 1/5th
– Guardianista attraction: 1/10th (smaller demographic segment and could be subsumed into one or other of the tactical voting effects)
– Relative activist concentration: 1/5th
– Net targetting effect: 1/5th
– Others: 1/10th
(I emphasise that this is a guesstimate of relative weights).
One element I picked up from the discussion which lodged firmly with me is that the Marginals Effect largely built up over 2 elections (1997 and 2005, not dissipating significantly in 2001) â€“ the full unwind in one election could be rather sudden. Further, the precision implied by the figures given is rather artificial. Iâ€™ve therefore reduced the effect slightly and rounded to half-percents rather than tenths.
The above factors will act as â€œdistortionsâ€ on the output provided by a probabilistic UNS (which is, itself, a variation from â€œstraightâ€ UNS). My original articles summed up the accumulated effect over the Blair Years, and gave the forecast of what would happen if the pressure exerted on the UNS pendulum by the Blair Effect was released. My model is available from Google Docs now, so I’d encourage people to make their own assumptions on the basis of their own judgement of what’s going to happen. In the model, a base level of unwind on Average Tactical Vote and Marginal Boost are assumed, and the full effect is assumed as linear from there (at zero swing) to the point of return to 1992 (and max out at that point). Please feel free to enter your own assumptions (with all assumptions set at zero, the model will act as a probabilistic UNS swing calculator). Iâ€™ve also provided a boost for Lib Dem defence in their seats, as the much-vaunted Lib Dem incumbency effect does seem to have some grounds in the data