Political Figures: How Data Works

Once a month we are going to hand over to our friends at Medway Elects, who are going to dig into the Medway data and, using wizardry, work out where are right now.

As any opinion pollster will tell you, trying to predict how people will vote at an election is notoriously difficult. You can ask a sample of the population and then attempt to extrapolate that out as a representation of how the population as a whole will vote. You can study trends on social and traditional media to make educational guesses. Or, you could simply make it up as you go along. Whichever option you choose, as history has shown time and time and time again, the end result will always be unreliable, for a number of long and complex reasons (but that’s a thought for another day). 

However, it is with the unreliability of such predictions in mind that I am today writing the first in a series of posts for The Political Medway on my own data model for the upcoming local elections, prefaced with a number of caveats. Most important among them is this: this is not a prediction. This is a data-based projection, based on local and national polling data, to forecast how much support each party has in each ward. Retaining that support, or gaining additional support, and ensuring those supporters go out and vote is the responsibility of the parties themselves.

I have done this before. In the run up to the 2010 general election, via Democracy in Practice (the precursor to Medway Elects), I ran projections for the three parliamentary constituencies in Medway, again using a combination of local and national polling data. The eve-of-poll projections were close to the final results, in most cases within a reasonable margin of error (generally getting the Conservatives’ vote share about right, overestimating the Labour vote and underestimating the other parties):

However, these projections worked, in part, by amalgamating data from the various wards in each constituency and merging these together for a single set of projections. For local elections, the constituency data must be split between individual wards, which is a more difficult process. One cannot assume there will be a uniform swing, as that rarely happens, and one cannot simply guess what swing there will be in each ward. The algorithm takes this into consideration, but the accuracy of data for individual wards is still a little shaky. For this reason, I am not publishing projections for individual wards, only a Medway-wide vote share and an estimated number of seats won by each party. 

The projection model is also built on the assumption that parties will be standing candidates in the same wards as they did in 2015. As the list of candidates becomes clearer, the model can be refined. However, this does naturally introduce some unknown quantities to the equation: if a party stood in 2015 but are not standing in 2019, where will those voters go? If a party did not stand in 2015 but are standing in 2019, what would their notional base be in 2019? For the latter, I can use earlier election results and tweak the model to provide an (admittedly slightly less reliable) projection, but, again, the process for the former is educated, but imprecise.

Labour and the Conservatives fielding candidates for all 55 seats is pretty much a given (even if they have so far only announced (officially or otherwise) 33 and 32 candidates respectively), but the Liberal Democrats, Greens and UKIP will likely not have 55 candidates across Medway. Even where they are not standing, their absence will have an effect on the outcome of the election, as those people who would have voted for them will be seeking other candidates to vote for instead. A party standing in fewer wards will also, obviously, have a lower share of the Medway-wide vote.

And what of independent candidates, particularly Medway People’s Voice, the upstart new party launched earlier this month promising at least 23 candidates? With no previous polling data available, an accurate projection is impossible. Theoretically, they could end up with 23 seats on the council, or they could end up with 3% of the vote and without a single seat. Either way, their (potential) success cannot be reflected in the projections as previous performance cannot be analysed. They are, for this projection, an unknown quantity which could either upset the results dramatically on election day, or barely make a dent. Only time will tell. 

Finally, it is worth noting that a computer model such as this cannot take into consideration the local issues which may affect how voters in individual wards will vote, nor the capacity for parties to target one or two wards heavily and win a small number of seats in those couple of wards, without necessarily affecting the overall vote share to a large extent. As such, once again, this is not a prediction. Between now and 2 May, the figures are likely to change and even after the final projection is published, people will still be changing their minds and deciding whether to vote or not. This projection is intended as a rough guide as to voter support across Medway, with a seat estimate provided as an indicative figure for illustrative purposes only. 

Over the next couple of months, I will be taking a closer look at some of the trends we have seen in past elections in Medway, and trying to analyse some of the factors which have an effect on the final election result. All of the data I will be using is freely available on Medway Elects (I’m sure Jennings & Keevil will not mind my shameless plug) and I will be on hand on Twitter to answer any questions you may have about the analyses I produce. 

With each article, I will provide an update on the projection. The first projection is below, together with a one-off look at which wards could be key to deciding which party runs Medway Council for the next four years. 

Battleground Wards 

The wards to watch are those in which the projected majority is 10% or less. These are key battleground wards where two or more parties may be in with a shout of taking some, or all, of the available seats – and could be the key to deciding who gets the keys to Gun Wharf in May:

Rochester South & Horsted
Rochester West
Strood North
Strood South

That’s 18 seats which could (although it appears to be unlikely based on the current projections) dramatically alter the composition of the Council in May. 

2019 Medway Election Projection 

So now for the “headline” figures. If you’ve read all the way to here, well done! If you have just scrolled down here, I would urge you to note the various caveats above before trumpeting (or rubbishing) the figures on the social interwebs.

The projected vote share and estimated seat distribution as at today’s date is:

Check back in February for an updated projection – and a fun game of ballot roulette…

Alan Collins is the creator of Medway Elects, which is committed to building a complete electoral history for Medway.

Leave a Reply