Skip to content

Useful numbers


This very simple poll aggregator (a relic of covering past elections) does a few things:

  • Each morning, it wakes up and scrapes the Wikipedia page listing the latest polling data for the next UK general election. Then, it cleans and sorts the data into the charts below.  
  • The average for each day includes, in principle, any poll released in the prior weeks. But it only permits each organisation's latest poll into each average. The idea here is that we do not want to over-weight evidence from one pollster.
  • It also applies a gentle decay, such that newer polls are upweighted over older polls. A day-old poll is worth roughly 50 per cent more than a week-old poll. This means high-frequency pollsters are more likely to matter more any given day - but only because they are more likely to have a newer poll.

The parameters are disputable - but this seems to work.

The following graphs use the same simple averages to benchmark how Starmer is doing. First, a phase plot showing how well he is doing at closing the gap on the Tories. Each dot is a snapshot in time, a week apart. The light dots are the oldest, the dark ones most recent.

For each week, it shows the current gap between the Tories and Labour (on the x-axis) and the speed at which Labour increased its vote share over the prior 6 weeks (on the y-axis).

Labour will want to get into the top left quadrant: moving ahead of the Tories by building up their own support base. The bottom right quadrant is the danger zone: behind the Tories and losing vote-share.

Second, a histogram set. These frequency plots are of weeks since 1980. They show how many percentage points Labour's voter support moved by in the preceding 6 weeks. So you can see how rare it is to have a single six-week spell where the polls move by, say, 5 points in Labour's favour.

For comparison's sake, I have put Starmer's performance (thus far) in alongside - and the comparable time series for the other recent Labour opposition leaders.

For a slightly longer time-series on the polling:

General economic indicators

Real wage growth.

...and the OECD's consumer confidence indicator: