Just under 90 days to go.  Do the polls give us any pointers from the past?

There is an interesting exercise we can do.  Take the last 10 general elections, and take the polling averages for all the parties with just under 90 days to go to the final vote.

In other words, take the point we are at now, and take the same points for the previous 10 elections.

Did the closing 90 days make any difference to the eventual outcome?

For the Conservatives, no difference. The final vote share has been, on average, exactly what polls said it would be with 90 days still to go.

For Labour, the run-up means lost ground.  On average, the party’s poll share was 3.5 percentage points lower on polling day than it was with 90 days to go.

For the Liberal Democrats (and all previous permutations), the actual vote share was 3.2 percentage points better than implied by polls with 90 days to go.

Now, the statistical small print.

The Conservative performance is remarkable.  The variation between T -90 and polling day across the last 10 elections is consistently tiny – under one percentage point.  There is only one outlier to the trend, and that was the run-up to the 1997 poll when the Conservative position slipped by 6 percentage points between T – 90 and polling day.

The Labour performance is more varied.  In only one lection did Labour manage to increase its share between T -90 and polling day, and even then it was by just one percentage point (February 1974). In one election there was no change at all (1987).  All others show a loss of support between T -90 and polling day. There are two outliers – 1997 and 2001, when Labour’s stratospheric pre-election lead dropped back slightly between T -90 and polling day.  Strip out the distortion of those two, and the average dip between T -90 and polling day is 2.7 percentage points.

The trends for the two major parties are not affected by whether they were in government or in opposition in the pre-election run-up.

The Liberal Democrats  -and their previous incarnations – gained support between T -90 and polling day in seven elections, remained stable in two, and slipped in only one.

What if we apply these precedents to today?

The current poll average gives us:

Labour                  33

Conservative         32

Liberal Democrat   8


Apply the average movement seen over the last 10 elections (adjusting for the outliers) and we will lend up with:


Labour                   30

Conservative         32

Liberal Democrat  11


What would that give us in terms of seats?  Hard to say with precision because of the pentagonal nature of this election, and the unknown impact of UKIP in Con/Lab marginal, and of the SNP on Labour’s base in Scotland.

But making cautious assumptions on these new variables, we would get a Parliament roughly like this:


Conservative   288

Labour              285

Lib Dem              27

SNP                     30

Others                20


If the polling precedent holds, we are heading for a Parliament where no combination of one large party and one small party can command a majority.

But what about that ‘if’?

The past is not necessarily a guide to the future.  And some factors about election 2015 are new:

  • There is no clear precedent for a force like UKIP;
  • The Liberal Democrats are entering the contest with a record in government, for the first time;
  • There are contests with very different profiles taking place in the constituent countries of the United Kingdom;
  • The percentage of voters indicating, at T -90, that they may still change their mind is unprecedented.

Those points alone suggest that this General Election may write new history, rather than repeat the patterns of the past. But then again, it may not.

Either way, for party leaders planning the final push, polling precedent might be worth taking into consideration. If only because it shows what tens to happen in the minds of around 30 million voters as that clock ticks down.

James Plaskitt


Polling Station” by secretlondon123 is licensed under CC BY-SA 2.0.