Uncategorized

Still Swinging Both Ways

After the general election last year, I a piece on the blog which evaluated the respective performance of Uniform National Swing (UNS), and Ratio Swing (RS), in projecting the correct winner in each seat. We found that UNS had projected the correct party to win in 38 of the 40 Welsh seats, and RS in 37 of the 40.

http://blogs.cardiff.ac.uk/electionsinwales/2015/05/18/swinging-both-ways/

 

I thought it would be useful repeat that exercise for the National Assembly election. Before getting into the detail of that, it might be useful to re-iterate some comments I made in the relevant article last year. As I said at the time, I think many people misunderstand “how analysts generally deploy assumptions like UNS. Of course those using UNS to project the potential seat implications of an opinion poll don’t believe that any future election will really see a party’s support levels change by an exactly uniform percentage right across the country. (We might be stupid, but we’re not that stupid). As I have tried to explain on the blog previously, I don’t see UNS as any sort of infallible prediction tool (which was why I was always very careful to make clear here that I was making projections of the potential seat implications of polling numbers, and definitely not predictions). I think UNS provides us with a useful baseline, or benchmark, against which pre-election we can assess the tasks facing parties in individual seats, and post-election we can evaluate their achievements. But it is nothing much more than that.”

One other preliminary comment: a reminder on what UNS and RS do. UNS simply means projecting the change in national vote share for each party onto every constituency and region. So imagine, for instance, a party has seen its constituency vote share go up from 10% to 20% between two elections. Under UNS you would then project a ten percentage point increase for that party onto every constituency. Ratio swing models the change in the ratio of a party’s support. So in the case of our hypothetical party we would project a doubling of their vote share in each constituency.

So with that out of the way, let’s first look at the constituency seats for the Assembly. The table below shows all forty of them. We know what the national changes in constituency vote share from the 2011 election were; using the UNS and RS formulae, do we then project the correct winner in 2016 for each seat? The table highlights in bold any cases where the formula projects a different party to win the seat than was the actual result.

 

2016 – Constituency Results

Constituency UNS Ratio RESULT
Ynys Môn Plaid Plaid Plaid
Arfon Plaid Plaid Plaid
Aberconwy Conservative Conservative Conservative
Clwyd West Conservative Conservative Conservative
Vale of Clwyd Labour Labour Labour
Delyn Labour Labour Labour
Alyn & Deeside Labour Labour Labour
Wrexham Labour Labour Labour
Clwyd South Labour Labour Labour
Dwyfor Meirionydd Plaid Plaid Plaid
Montgomery Conservative Conservative Conservative
Ceredigion Plaid Plaid Plaid
Brecon & Radnor Lib-Dem Lib-Dem Lib-Dem
Preseli Pembroke Conservative Conservative Conservative
Carmarthen West & South Pembs Conservative Plaid Conservative
Carmarthen East & Dinefwr Plaid Plaid Plaid
Llanelli Plaid Plaid Labour
Gower Labour Labour Labour
Swansea West Labour Labour Labour
Swansea East Labour Labour Labour
Neath Labour Labour Labour
Aberavon Labour Labour Labour
Bridgend Labour Labour Labour
Ogmore Labour Labour Labour
Rhondda Labour Labour Plaid
Cynon Valley Labour Labour Labour
Pontypridd Labour Labour Labour
Vale of Glamorgan Labour Labour Labour
Cardiff West Labour Labour Labour
Cardiff North Labour Labour Labour
Cardiff Central Lib-Dem Labour Labour
Cardiff South & Penarth Labour Labour Labour
Merthyr Tydfil & Rhymney Labour Labour Labour
Caerphilly Labour Labour Labour
Blaenau Gwent Labour Labour Labour
Islwyn Labour Labour Labour
Torfaen Labour Labour Labour
Newport West Labour Labour Labour
Newport East Labour Labour Labour
Monmouth Conservative Conservative Conservative

 

As we can see, UNS and RS both project the correct winner in 37 out of 40 seats. There are two seats that both formulae get wrong: both project Plaid Cymru to win Llanelli, and Labour to win in the Rhondda. I think this just highlights the scale of the achievement by the respective winning campaigns in both those seats. (Labour achieved a swing from Plaid Cymru in Llanelli that was a full 5.0 percentage points better than the national average; in the Rhondda Plaid managed a swing from Labour that was more than 19.6 percentage points better than the national average). The other two seats projected incorrectly are Cardiff Central under UNS, and Carmarthen West & South Pembrokeshire under RS.

What about the regional seats? Projections for those have to take into account, of course, the projected outcomes in the constituency seats – not all of which, as we have just seen, were correct. Anyway, this second table shows the projected outcomes under UNS and RS, and the actual result in each region.

 

2016 Regional Results

Region UNS Ratio RESULT
North Wales 2 UKIP

1 Plaid

1 Conservative

 

2 UKIP

1 Plaid

1 Conservative

 

2 UKIP

1 Plaid

1 Conservative

 

Mid&West 3 Labour

1 UKIP

 

2 Labour

2 UKIP

 

2 Labour

1 Plaid

1 UKIP

 

South West 2 Plaid

1 Conservative

1 UKIP

 

2 Plaid

1 Conservative

1 UKIP

 

2 Plaid

1 Conservative

1 UKIP

 

South Central 2 Conservative

1 Plaid

1 UKIP

 

2 Conservative

1 Plaid

1 UKIP

 

2 Conservative

1 Plaid

1 UKIP

 

South East 2 Plaid

1 UKIP

1 Conservative

2 UKIP

1 Plaid

1 Conservative

2 UKIP

1 Plaid

1 Conservative

 

It would appear here that RS does a better job than UNS – it only gets the result wrong in one region, whereas UNS gets it wrong in two regions.

However, when we look at the overall projected results, there is no obvious superiority of one method or the other. Both methods project the exactly correct numbers of Labour and Plaid seats. UNS also gets the Conservative number correct, but projects the Liberal Democrats one seat too high and UKIP one seat too low. RS, by contrast, gets the Lib-Dems right, but projects UKIP one seat too high at the expense of the Conservatives.

 

Overall Seats (Constituency + Regional)

Party UNS Ratio RESULT
Labour 29 (26 + 3) 29 (27 + 2) 29 (27 +2)
Plaid 12 (6 + 6) 12 (7 + 5) 12 (6+6)
Conservative 11 (6 + 5) 10 (5 + 5) 11 (6 + 5)
UKIP 6 (0 + 6) 8 (0 + 8) 7 (0 + 7)
Lib-Dems 2 (2 + 0) 1 (1 + 0) 1 (1 + 0)

 

What, if any overall lessons can we draw from this? There are two, I think.

  • As in the general election last year, UNS and RS both worked pretty well in terms of providing projections for the overall result, when looking at the national vote figures. Both methods are thus also likely to work well in terms of projecting the seat implications of national opinion polls, provided those polls are accurate.
  • However, UNS and RS should both be understood as generally accurate tools for projecting the aggregate For both Westminster and, even more so, for the Assembly, they tend to get to the right overall outcomes, but not necessarily by the exactly correct route. Idiosyncratic local results tend to balance out – or, for the Assembly, lead to balancing consequences in terms of the regional list seats.

In short, national seat projections made by UNS or RS from national polls in Wales will generally give you a good idea of the likely overall outcome (except in an election where one party is particularly successful in the key marginal seats). But as I have tried to point out at various times, it is very perilous to draw strong implications for individual local contests from national figures. Now, if only I could find a way to explain that point successfully to some journalists and party activists…

Comments

  • Welshguy

    What I find interesting is how, instinctively, Ratio swing seems like it “should” produce a more accurate result than UNS, because it’s so much less crude. In practice though there seems to be little or nothing in it.

    • Roger Scully

      I’m not sure it is any less crude. Both are a simple arithmetical formula, applied blindly. What is perhaps surprising is the high % of correct results that both manage to ‘predict’.

Leave a Reply

Your email address will not be published. Required fields are marked *