Lib Dems predicted to poll 18% and make 30 net gains in May elections

The annual council election predictions from Michael Thrasher and Colin Rallings are out, based on what has been happening in council by-elections.

Vote share predictions: May 2018

On vote share, they predict the Lib Dems to be static on last year, and hence up sharply on 2014 (the last time most of these seats were contested). For Labour and Conservatives, they predict a neck and neck tussle whilst for Ukip it's a picture of continued dramatic decline.

• Con 36% (-3% on last year, +6% on 2014)
  Lab 36% (+8% on last year, +5% on 2014)
  Lib Dem 18% (no change on last year, +7% on 2014)
  Ukip 2% (-2% on last year, -16% on 2014)

Note: these are the national equivalent vote share figures; i.e. adjusted to allow for which seats are up for election each year and so comparable across years.

Seat predictions: May 2018

The huge decline in Ukip and Labour's very strong likely performance in London combine to mean there's still scope for the Conservatives to come out of the contests with loses which are not that large and for the Lib Dems to score an all too rare year of net gains.

• Con -75
  Lab +200
  Lib Dem +30
  Ukip -125

How accurate are these predictions?

As I now comment pretty much every year, although many commentators and journalists like to talk about data journalism and evidence based blogging, almost no-one cites the track record of these forecasts when talking about them.

Almost no-one, but lucky reader you're at the exception. This is the only site that systematically documents how the predictions perform against reality:

• · Accuracy of Conservative predictions: on average the Conservatives slightly over-perform the forecasts, but the level of variation makes that a pattern of limited use for interpreting future forecasts (full data here).

• · Accuracy of Labour predictions: these predictions have a lot of variation in their accuracy, including a spectacular failure in 2013 where the Ukip surge hit Labour badly and less strikingly in 2017 when Labour outperformed the forecast even before the big Corbyn general election surge (full data here).

• · Accuracy of Liberal Democrat predictions: these predictions have a run of over-estimating the Liberal Democrats by two or three percentage points (full data here).

So the verdict for the Liberal Democrats? Based on the historic accuracy of these predictions, optimism about seat gains should be tempered by caution based on the historic record. An unchanged vote share on last year would also be a missed opportunity for a further step forward in the party's recovery. But there's one way to make that a seized opportunity: hard campaigning.