1 April 2012

STV: What lessons from Baillieston?

I managed to pique my own interest with Friday’s scintillating post about the single transferable voting system (“STV”) and the approaching Council elections, specifically in Glasgow. At the end of that post, I speculated gently on the effect of multi-member SNP-Labour punch ups in the city’s twenty one wards in the coming poll. In 2007 – Scotland’s first local election under STV – the Nationalists stood only one candidate apiece in all of the city’s wards (save for Baillieston, where two SNP candidates stood and were elected). Come 2012, Glasgow SNP are standing 43 candidates in the city, a near match to Labour’s 45. In the previous post, I took the Pollokshields ward as my example, for the modest reason that I cast my vote there in 2007. A look at the figures showed that the levels of preferences stated dropped off sharply. However, my ward looked much like a traditional election under first past the post, with just one candidate from all of the prominent political parties. What effect might multiple candidates from the same party have on the vote – or more importantly for 2012, multiple candidates from vying parties?

A dip back into the archive may have the answer. In point of fact, the choice put in Pollokshields was not typical of Glasgow in 2007. Multiple Labour candidates stood and were elected in respect of wards elsewhere in the city. Remember, wards elect three or four councillors apiece, and Labour gained two councillors in seven wards – Linn, Craigton, Southside Central, Calton, Anderston/City, Maryhill/Kelvin and Canal – and three of four seats in six city wards - North East, Shettleston, East Centre, Drumchapel/Anniesland, Garscadden/Scotstounhill, Govan and Greater Pollok.

Since it is the only 2007 ward where multiple SNP candidates stood against multiple Labourites, Baillieston seems the obvious case to study. Baillieston elects four councillors. Moreover, a look at the detail beneath the final results – which elected two Labour and two SNP councillors (marked below with a *) – Baillieston is a much clearer example of preference voting (and the indeterminacy of first preferences) at work than Pollokshields. To start, the ward’s first choices for councillor...

Stage 1: First Preferences

John Mason (SNP)* 3,199 (30.8%) 
Jim Coleman (Lab)* 2,689 (25.9%) 
Douglas Hay (Lab)* 1,342 (12.9%) 
Robert MacBean (Lab) 739 (7.1%) 
George Clark (Tory) 678 (6.5%) 
Michael Kayes (Solidarity) 466 (4.5%) 
Marjory Watt (Lib) 392 (3.8%) 
David McDonald* 250 (2.4%) 
Jim McVicar (SSP) 224 (2.2%) 
Ian Dickie (Scottish Unionist) 216 (2.1%) 
Raymond Morrison (Green) 181 (1.7%)

The quota for election was 2,076 votes (how this is calculated is explained here), so after the first round, both John Mason and Jim Coleman were elected (with a surplus of 1,123 and 613 votes respectively). This surplus needs redistributing (ibid). As Mason achieved the highest number of votes, we start with him. Helpfully, the data published by the council is very clear. Of Mason’s 3,199 ballots, only 510 (15%) were non-transferable (i.e. ballots which recorded only one preference, for him). Of the remaining 2,689 ballots, 70.7% recorded the second SNP candidate – David McDonald – as their second preference. Since we are merely transferring Mason’s surplus rather than all of his votes, McDonald gained only 70.7% of Mason’s surplus rather than the full whack – a not inconsiderable increase in his vote from a wan 250 to 917.32704: an increase of 667 plus five percentage points, putting him ahead of the third Labour candidate in the ward. The rest were distributed in the low hundreds to other candidates.

How did things fare with Jim Coleman’s ballots? A smaller surplus to redistribute, 604 of Coleman’s 2,689 ballots recorded no other preference – 22.5%, some 7.5% higher than Mason’s. This reduced his transferable stock to 2,085. Of this, Douglas Hay took 1,331 (49.5%) and Robert MacBean 439 (21%). Looking across the rounds, Baillieston of 2007 is also a lesson in the importance of timing. I’ve graphicised the ten stages of the allocation above. Candidates crash to zero in the round they are eliminated, with the other candidates’ increases coming from the eliminated candidate. Where a candidate is elected (Mason, Coleman), their tally is reduced to the quota level. From stage One to Two, for example, you can see the difference Mason’s transferred surplus made to McDonald – increasing from a paltry 250 votes and sneaking in front of the third Labour candidate – remaining a nose ahead of him in every subsequent round, leading to MacBean’s elimination in round nine, and the elections of Hay (Lab) and McDonald as the last remaining candidate short of the quota in round 10.

That last point is particularly important, as a nudge the other way could have made Baillieston a ward electing three Labour councillors instead of two. In the ninth round, Hay had garnered 2,072 votes – four short of the quota – and promptly eliminated the candidate with the lowest level of support in the tenth round – in this case, his Labour colleague, MacBean. 586 – 55% - of MacBean’s ballots transferred to his remaining colleague, lobbing Hay handily over the threshold. If, by contrast, Hay had achieved an increase in his vote, and hit the quota in round nine, his surplus would first have redistributed amongst the remaining candidates: in this case, his Labour colleague and a perfidious Nat. It is easy to envisage who the primary beneficiary of that transferred surplus would have been.

Baillieston also serves as a reminder that it is helpful – necessary – to distinguish (a) voter ranking behaviour under STV – how many of their preferences do voters use? – and (b) the effect of that behaviour, when the ballots are fed into the allocation system. These may be quite distinct. For instance, across all ballot papers in the ward, MacBean accumulated 739 first preferences, 866 second preferences and 1,782 third preferences. McDonald’s cache, by contrast, numbered 250 first, 1974 second and 361 third preferences, but went on to take the seat.

Why? It’s all down to the elimination of candidates achieving modest support, and the variable might of surplus transfers from successful candidates. For example, if, like Hay, you are in a leading position in the contest but have not yet attained the quota, next preferences expressed on your ballot papers go nowhere. Assuming a good number of those would have gone MacBean’s way, Coleman having been already been elected in the first round, they avail MacBean not at all while Hay is still waiting to achieve quota. Secondly, while a candidate who achieves well over the quota can substantially assist his colleagues when his surplus is redistributed – e.g. Mason in this example – candidates who scrape the quota aid their party colleagues very little, even if the vast majority of voters who support them also support their party colleagues.

And what of levels and extent of preference-stating? In my last post, we saw the diminishing number of voters in Pollokshields who availed themselves of their preferences, albeit in a council race with only one candidate from all of the parties. So what differences might a multi-party multi-option race make? Here are the equivalent figures from Baillieston:

In Pollokshields the number of ballots expressing two preferences was 30% lower than the number expressing only one. Just half of all ballots in Pollokshields identified three. In Baillieston, 80% of ballots identified two preferences, while 60% of all ballots cast identified three. Only 30% went beyond that, identifying four preferred candidates in order in Baillieston: nigh identical to the percentage of papers in Pollokshields who did so.  Potential indications, then, that fielding more than one candidate from major parties may have an impact on whether voters elaborate their preferences, composing longer lists of preferences.

Lessons for the 2012 Local Elections?

What lessons from all this for Labour and the SNP in 2012? Firstly, while it is clearly important to encourage your voters to transfer their preferences to all your candidates in the field, you only see the substantial benefits of doing so when your candidates surge past the surplus, and can cascade that surplus back to your candidates. While the second SNP in Baillieston attracted 70% of Mason’s second preferences, contributing significantly to his haul of votes, the Labour transfers were divided 49 : 21%. While this did not prevent a second Labour councillor being elected in the ward, we also have to take timing into consideration. If Hay’s support had been higher, he may have achieved quota before round ten, and not done so at the expense of his third Labour colleague. Admittedly, this doesn’t seem likely in 2007: even if he had done a little better and crossed the threshold in round nine, his redistributed surplus would have been insufficiently large to overhaul McDonald’s 208 votes lead over MacBean – but it serves as a reminder that timing matters, and that a long-drawn out trudge to the quota can disadvantage parties fielding a larger number of candidates.

You’ll also notice that by the last round, the second SNP candidate did not actually achieve the quota, but was elected anyway as he was ahead of his nearest rival. A lesson then that while the quota is an important anchoring point, if STV council races prove close run things, it does to be ahead of your main competitor throughout the process, as that may well be enough to elect you in an exhausted field. It will vary across different constituencies in Glasgow, but there are a number of city wards that are essentially big Labour-SNP dust ups, surrounded by a crowd of smaller parties and no hopers, ripe for elimination. This logic particularly obtains there.

On Baillieston’s example, at the very least it seems sensible for the SNP to run a two-candidate strategy in all wards – and in most of them, it seems like the best option, maximising the likelihood of transfers across the party slate, and so concentrating the chances of victory in a closely contested field, while maximising the number of candidates potentially electable, given the party's level of support (arguably undershot in Glasgow in the 2007 local elections). On this front, interesting races to watch this year will be wards like Glasgow North East, currently electing three of four Labour councillors, where two SNP candidates are fielded to Labour’s three. A familiar worry about fielding an excessive number of candidates in an STV system is that you split your vote and lose out on any councillors by a dormouse’s whisker. Intuitively, this may be less of a problem in Glasgow where the competition is likely to be more or less evenly divided between the two parties, rather than three or four being in contention.  Instead, the difficulty for Labour in their three candidate wards may be those which disappointed MacBean's hopes in Baillieston 2007 - divided surplus transfers, allowing the more concentrated SNP vote to remain just a nose ahead of them.  


  1. I was gobsmacked the first time I checked how many first preference votes David McDonald got.

  2. Tony,

    Quite so, and a stark example of STV preferences at work it is too. Curiously, you don't see anything like the same pattern in 2007 in wards with multiple Labour candidates standing. It'll be interesting to see in 2012 whether the SNP vote follows the Baillieston pattern (which may say more about John Mason than anyone else), or whether it comes to resemble the Labour pattern.

  3. If you want more information about the importance of preferences and transfers, you might be interested to see the table and charts you'll find here:

    What the charts don't show are the third-placed candidates (in 3-member wards) who were kept in third place and elected only by transfers and the fourth-placed candidates in 4-member wards who were kept in fourth place and elected only by transfers.

    The message for voters is: mark preferences for all the choices you can make because you never know how close the contest will be in your ward.

    The message for candidates is: don't just canvass for first preference votes - those second, third, fourth, and fifth preferences could make all the difference.

  4. Many thanks for the link Edinburgh. As you say, that seems very much the lesson for voters and parties, keen the squeeze the most out of expressed preferences.