Who is going to win the Senedd election? National polls can only tell us so much

Dafydd Trystan

Today has seen the publication of BBC/ICM’s yearly St. David’s Day poll for Wales – which shows Labour strengthening its position somewhat and the Conservatives and Plaid vying for second place but largely sharing the regional spoils between them.

Based on the BBC graphic used in their report, which shows a projection based on the 2016 results, one could be tempted to believe that Labour are locked on for 30 seats.

But the truth is far more complex and suggests we shouldn’t rush to judgement based on any single poll (or for that matter groups of polls).

This is because even the best national polls will have a margin of error of around 3%; and the way these votes are then distributed across the country can have a very significant impact on how many seats the parties finally end up with.

Let’s take a real-world example from 2016 – the Rhondda and Llanelli constituencies – contested by four significant and well-regarded figures in Welsh politics.

The last polls before the election were pretty close to the actual result – and in the actual national result, Labour’s vote share fell by 7.6% and Plaid’s increased by 1.3%. A swing of some 4.5% Labour to Plaid.

Applying that national swing to those seats would have seen Helen Mary Jones comfortably returned in Llanelli, while Leanne Wood would still have fallen significantly short of her aim of capturing the Rhondda.

On the same day, as we soon found out as the count progressed, Lee Waters secured a small swing to Labour in Llanelli, while Leanne Wood secured a pretty mammoth 25% swing to Plaid.

Likelihood

The point here is that while we can say with some confidence that the most recent poll would deliver Labour between 27 and 31 seats we cannot be totally confident how any of the individual 40 constituencies might vote.

This situation is further complicated by the semi-proportional regional list system that might then lead to seat gains or losses as a consequence of very marginal constituency results.

My plea therefore to journalists, commentators and political activists alike is to consider this poll (and any subsequent poll) as a national snapshot but not to read too much into the likely results in individual seats.

In an ideal world, someone might develop a ‘Nate Silver-esque’ prediction model. The FiveThirtyEight website in the USA, rather than predicting with certainty that party A or B will win a State, Senate Seat or Congressional District, offers a percentage likelihood.

My back of an envelope calculation based on the latest poll would suggest that the incumbents in both Llanelli and the Rhondda have around a 60-70% chance of retaining their seats.

Such a nuanced understanding might help us all consider and analyse the polls for what they are – a useful and illustrative national snapshot – but not predictions!