In 1999 the media projected from a safe Conservative seat, this time they use safe Labour seats
The Preston, Glasgow Anniesland and West Bromwich West by-elections
Labour should fear last weeks election results
By Quentin Langley
Dateline 24 November 2000
By-election results always face the media and academics with a problem. How do you project a result from one constituency, which may be hopelessly unrepresentative, to the country as a whole? The traditional answer is the “swing”. This assumes that if the Conservative vote declines from 43% to 32%, as it did in 1997, the proportion of the electors backing the Conservatives would be 11% less in all constituencies. This would mean the Conservative vote falling from 60% to 49% in some safe seats and falling into negative figures in some inner city areas. Plainly this is absurd, and Dr Gordon Reece, a mathematical engineer at Bristol University has shown that his Proportional Loss Hypothesis has been more accurate at projecting election results than the “swing” in all cases.
Broadly, Dr Reece assumes that a party will lose twice as many votes in seats where it has 50% of the vote as in seats where it has 25%. So why does the mainstream media persist with the swing? Because it tends to get the results of marginal constituencies right. At the last election the swing analysis overestimated Labour support in safe Conservative seats and underestimated Labour support in its own heartlands. But this had no effect on Labour’s majority, because in the Labour-Tory marginals it got the results right.
Where the swing analysis goes wildly wrong is when the information comes from atypical seats, like Kensington and Chelsea, where there was a by-election last year. The swing would assume that Labour’s vote would decline in marginal seats, and even in safe Labour seats, would be the same as in K&C. This is hugely unlikely. Surely a by-election in a safe Labour seat would produce a bigger fall in Labour’s vote than in a safe Tory seat?
Had the only by-election last week been in Preston, the headlines would have screamed a disaster for Labour. The Labour vote fell from over 60% to not much more than 45%. If Labour’s vote fell from 44% to 29% nationally – applying the uniform swing - it would be the party’s worst result since 1983. Labour’s performance in Glasgow Anniesland – losing 10% was not much better. Labour was saved from these headlines by the result in West Bromwich West, where its vote appeared to go up. Averaging the three results lead the Financial Times to declare that Labour’s majority in a general election would have been 80 seats.
But Labour’s vote in West Bromwich West only appeared to go up. There was no comparable result in 1997, as the Speaker is traditionally unopposed. Labour’s vote was not up compared with an actual result in 1997, but compared with a notional result. And if the swing underestimated Labour’s support in its safe seats – and it did – then the notional Labour result in West Bromwich must be far lower than Labour would have actually scored if the seat had been contested. Far from being up on 1997, Labour’s vote in West Bromwich West was certainly down, fully in line with the other two by-elections.
So if we ignore West Bromwich West, is Labour heading for 29%? No. It is not that bad. In Preston Labour’s vote was down by a quarter – this would represent a fall from 44% to 33% nationally. The Conservative vote was up by 3.1% of the electorate or 15% of the Conservative vote. The LibDem vote was up by 2% of the electorate, also about 15% of the party’s vote. This would represent an increase in the Conservative vote nationally from 32% to 38% and an increase for the LibDems from 17% to 20%. In Glasgow Anniesland Labour lost “only” a fifth of its vote, but the disposition of these votes cannot be relevant to English constituencies because of the complicating presence of the SNP.
Both these results were way out of line with national opinion polls but show a Conservative lead not much larger than in the local elections of May 2000 and slightly less than in the Euro-elections of 1999.
While a national projection shows Labour’s vote falling by 11% not all those votes went to the Conservatives and LibDems. Being a by-election some votes went to minor parties. In a general election this would be unlikely. If Labour held those votes then the Conservative lead would be very similar to May 2000, but if those votes went to the Conservatives it would represent a Conservative victory almost as big as the 1987 election.
Quentin Langley is a freelance commentator and consultant.
Copyright © Quentin Langley 24 November 2000