BBC NEWS Americas Africa Europe Middle East South Asia Asia Pacific
BBCi NEWS   SPORT   WEATHER   WORLD SERVICE   A-Z INDEX     

BBC News World Edition
 You are in: In Depth: Local elections  
News Front Page
Africa
Americas
Asia-Pacific
Europe
Middle East
South Asia
UK
Business
Entertainment
Science/Nature
Technology
Health
-------------
Talking Point
-------------
Country Profiles
In Depth
-------------
Programmes
-------------
BBC Sport
BBC Weather
SERVICES
-------------
EDITIONS
Local elections Tuesday, 2 May, 2000, 14:55 GMT 15:55 UK
Pollsters peer at the locals
David Cowling, editor of the's BBC Political Research team, looks at why opinion pollsters are having a closer look at this year's local elections.

The opinion polls hardly covered themselves with glory in the 1992 general election.

Their collective failure was judged sufficiently serious that an inquiry was established under the auspices of the Market Research Society.

Because of the enormous scale of Labour┐s victory three years ago, less attention was given to the fact that the overall accuracy of the polls was hardly any better than in 1992.

In the light of this uninspiring record it is not surprising that the search for reliable alternatives continues apace.

One of the more serious contenders for consideration is the results of local council by-elections.

Most Thursdays of the year in various wards throughout the land voting takes place to fill occasional vacancies caused by deaths or resignations. In 1998, there were 288 by-elections involving 382,000 voters.

Ten years earlier, in 1988, the total was of 721,000 votes cast in 393 by-elections.

Two key workers in the field of local elections are Professors Colin Rallings and Michael Thrasher of Plymouth University.

They have pioneered the collection and publication of council election results and as a result have extensive information on which to judge whether these weekly by-elections can tell us anything useful about the outcome of the next general election.

They have developed a model based upon the results of contests in wards where Conservative, Labour and Liberal Democrat candidates fought in the by-election and the last time the seat was fought in the annual May elections.

The change in each party┐s share of the vote in the by-election compared with the previous May election was measured and applied to the appropriate national equivalent vote share.

Applying this method to individual by-elections would often result in absurdities but they found that if they used aggregate figures for all the by-elections in a month or a quarter then the results were smoothed out and a more reliable estimate of national party support could be produced.

Interestingly, they also discovered that the generally low turn-outs in council by-elections did not appear to have much impact on the accuracy of the model they developed.

A better model?

So, how accurate has all this modelling been?

Well, it had a pretty grim baptism at the 1992 general election when compared with the final day opinion polls the by-election model was less accurate for the Conservative and Lib Dem share and more accurate when it came to the Labour share.

The record of forecasting subsequent annual local elections was much more impressive.

In 1993, it was within 2% for each party. In 1994 and 1995 the maximum error of the model was 1% and in 1996 it was 2%.

However, the real test was the 1997 general election.

In their final prediction in the 20 April 1997 Sunday Times, taking into account all council by-elections held during the general election campaign period, Professors Rallings and Thrasher were spot on for both the Conservative and Labour shares but 3% out on the Lib Dem vote.

Overall, they were more accurate in forecasting the Conservative and Labour vote than were the opinion polls.

Don't throw in the towel

But I do not think that the polling industry should throw in the towel quite yet. The by-election model has a tendency to overstate support for the Lib Dems and this needs to be remedied.

Also, crucially, what the by-election model cannot tell us and the polls can is why people voted the way they did.

But nonetheless I think the work of Professors Rallings and Thrasher in developing their model has delivered an alternative measure of assessing political support that has earned the right to serious consideration.

Links to more Local elections stories are at the foot of the page.


E-mail this story to a friend

Links to more Local elections stories

© BBC ^^ Back to top

News Front Page | Africa | Americas | Asia-Pacific | Europe | Middle East |
South Asia | UK | Business | Entertainment | Science/Nature |
Technology | Health | Talking Point | Country Profiles | In Depth |
Programmes