Election Essay 5: Transfers: Myths and Reality

(Written for Stratagem, 04 May 2011)

One of the joys of the PR-STV system for us analysts is the wealth of detail revealed about voter preferences when one looks at transfers. This became particularly glorious in the election for the Dáil in February, when Fianna Fáil’s difficulty in attracting transfers meant that they won only 19 of the 165 contested seats, rather than the 29 one might have expected from their vote share.

Transfers tend to be more predictable in Northern Ireland’s elections. Nationalists will transfer to nationalists first, then to centre parties or moderate unionists if there is no remaining nationalist candidate. Unionists will transfer amongst themselves, then to centre parties if they are available, then to the SDLP ahead of Sinn Féin.

The most spectacular example of transfers making a difference in 2007 was the Upper Bann count, where George Savage, in eighth place with only 2,167 first preference votes, was pulled ahead of the trailing DUP candidate on transfers from his own party colleagues, and then was pulled ahead of the second SF candidate once the DUP transfers trickled to him, taking the last seat by a comfortable four-figure margin.

There were four other constituencies in 2007 where one of the top six candidates on first preferences failed to get elected, in all four cases losing out to the SDLP whose candidate had placed seventh: in West Belfast and South Antrim, the losers were the DUP, and in Fermanagh & South Tyrone and Foyle, Sinn Féin candidates similarly lost. The SDLP deserve good marks in general for their vote management, which however went awry in West Tyrone where three SDLP candidates with a quota of votes between them failed to transfer between each other and allowed Independent candidate Kieran Deeny to retain his seat. Sinn Féin, of course, did even more impressively, getting five candidates elected in West Belfast, with less than five hundred votes separating Gerry Adams’ four running-mates.

But it is terribly difficult to make concrete generalisations about how voters use transfers. The Assembly itself reported that in the 2007 election, “79% of transfers from unionist voters went to other unionist party candidates; 12% went to nationalist party candidates. 64% of transfers from nationalist voters went to other nationalist party candidates; 13% went to unionist party candidates.” But this is probably no more than a mathematical statement, only loosely related to voters’ actual behaviour.

To explain this, it may be worth exploring the subtleties of the three different circumstances in which votes actually get transferred. The first, and most obvious, is when a candidate is deemed elected on the first count – when their number of first preferences exceeds the electoral quota (which in a six-seat Assembly election is a seventh of the total valid vote, plus one). In this case all of their votes are checked to see if they have a later preference, and only the ones that do are taken into account – so in other words, if you get 1,000 votes, and the quota is 800, you have a surplus of 200 votes; but if half of your 1,000 votes are from ‘plumpers’ who only voted for you and nobody else, that means that the other half are transferred at a value of 0.40, the surplus (200) divided by the number of transferable votes (500). The ‘non-transferable’ votes which appear in the results sheets at this point are a fictional accounting factor, and are very indirectly related to the number of votes which actually were non-transferable.

A real-life example from the last election: in North Antrim, Sinn Féin’s Daithí McKay received 7,065 first preferences, 731 more than the 6,334 quota. He then transferred 300.74 and 280.39 votes to the two SDLP candidates, 99.44 and 0.88 to the two Independent candidates, 10.34 to Alliance, 2.64 and 0.33 to the DUP candidates, 2.09 and 0.44 to the UUP candidates, and 0.11 to the UKUP, with a reported non-transferable tally of 33.60. The full details are not reported, but there is enough information here to work it out: if the UKUP received only one transferred ballot paper, then all McKay’s transferred first preferences went at a value of 0.11, so we have reported destinations for 6,340 of them, and the other 725 must have had no further preferences (or else voted only for McKay and for either or both of the two Ian Paisleys, who had been elected by the time McKay’s votes were transferred).

The logic of this is simple enough – voters who plump for a candidate get their full value for their vote with that candidate; voters who transfer get a sufficient fraction of their vote remaining with their first choice to keep them just over the quota, and the rest is transferred. The purist in me would like to see Northern Ireland elections use another decimal place in their calculations – if McKay’s votes had been transferred at a ratio of 0.115 rather than 0.11, the figure added to the ‘non-transferable’ tally would have been 1.900 rather than 33.60, which strikes me as neater. On the other hand I’m not aware of any case where this would have made a difference to an election result in the last 40 years.

There are two other circumstances where votes are transferred. If a candidate reaches the quota at a later stage in the process, then again their surplus votes are transferred – but only from the last batch of votes that came in. Again looking at North Antrim in 2007, on the second count Ian Paisley Jr was elected when he received a transfer of 1158.04 votes from his father, taking him from 6,106 first preferences to 7264.04 total votes, well over the 6,334 quota. The votes that then got transferred from Ian Paisley Jr’s surplus were not his own first preferences, but those he had received from his father (and they might also, though we may consider it improbable, have had second or third preferences for Daithí McKay).

This also means that we have to sometimes be cautious in describing where transfers come from. In the last count in Foyle in 2007, for instance, a Sinn Féin candidate’s surplus of 109.04 transferred fairly evenly between three SDLP candidates (one got 42.93, another 36.45 and the third 28.35) with 1.31 added to the non-transferable tally. But this bare account of the numbers leaves out interesting detail: we were actually looking at 361 ballot papers, of which only 133 (37%) transferred to the SDLP at a value of 0.81, the rest going nowhere; and the source of these ballot papers was not in fact Sinn Féin, but a transfer to Sinn Féin from Eamonn McCann, then of the Socialist Environmental Alliance. (Similarly the distribution of the surplus of Alliance’s Kieran McCarthy in Strangford, which heavily favoured the SDLP over the DUP, actually consisted of first preference votes for the Green Party.)

The final circumstance of votes being transferred is when candidates are eliminated. This happens when a candidate, or a group of candidates, has fewer votes than anyone else and no possible arrangement of spare votes can save them. In general this works as you would expect, the votes being examined to see if there are preferences for any other candidates remaining in the race, and if so, allocated to them. There is, however, one little-known extra subtlety. If an eliminated candidate has acquired votes with a fractional value, directly or indirectly from an elected candidate’s surplus, those votes are transferred last, and whole votes are transferred first; and if any other candidate is taken over the quota by the transferred whole votes, they are not considered for the distribution of the fractions. So it is not at all unusual to see a column of transferred votes in which only one or two are whole numbers, and those go to candidates who get elected on that count.

All of this frustrates those of us who are trying to calculate definitive transfer ratios and work out where voters believe that the parties are in relation to one another. Votes are not simple; voters operate the system with subtlelty. But we psephologists should remember that the system was not designed for our mathematical convenience, but to give voters fair representation, while not over-complicating the task of the counting staff; and it has performed very well for the last four decades.