How Might Differential Constituency Turnout Affect a NI Referendum?

Recently I speculated on how differential turnout could  impact on any future referendum on a United Ireland. Having done a bit of number-crunching since then on Assembly election results since the seventies, focusing on the post-GFA elections in particular, the answer seems to be: six years.

Differential constituency turnout means that the turnout varies between constituencies. Stating the bleedin’ obvious, I know, but there seems to be a political edge to it in NI. Border (mostly nationalist) constituencies have greater turnout than east-coast (mostly unionist) constituencies. This is probably a legacy of contested Westminster seats in Border areas since Partition, whereas many east-coast seats were uncontested between 1925 and 1970, being safe unionist seats.

I am not investigating differential party turnout within constituencies as I don’t have data for that. If party vote turnout is broadly similar within a constituency then – in a PR-STV election – the number of elected candidates will be broadly proportional to the party vote within that constituency. But a low turnout east and north of the Bann means that overall unionist and middle-ground bloc share of the vote is lower than it would be if turnout were equal in all constituencies, because these constituencies are, generally, overwhelmingly non-nationalist.

Consider two constituencies, both of whom have the same electorate. Suppose one constituency had a vote share of 70% for PartyA, 20% for PartyB and 10% for PartyC, and a valid poll of 80%. And suppose the other had a vote share of 70% for PartyB, 20% for PartyA and 10% for PartyC, and a valid poll of 50%. If turnout had been equal in both constituencies, that would give overall vote share of 45% for PartyA, 45% for PartyB, and 10% for PartyC. However, the effect of differential constituency turnout gives the overall vote share of 50.8% for PartyA, 39.2% for PartyB, and 10% for PartyC. If PartyA were to agitate for a referendum based on their 50.8% of the actual vote, and a huge turnout everywhere meant that there was no differential constituency turnout of any significance, their referendum proposal might be defeated 55%-45% despite PartyA’s vote share of 50.8%.

So how has differential constituency turnout impacted on bloc vote share in Northern Ireland? I am looking at Assembly elections only. (Westminster results under the first-past-the-post electoral system often have tactical voting and probably differential turnout within a constituency as very few Westminster seats are inter-bloc marginals.) My approach is to apply the bloc vote percentage within a constituency to the entire electorate within that constituency and then to add up these adjusted figures for all of the constituencies. This can then be compared to the actual (or unadjusted) bloc share, which is based on simply adding the votes.

The first image shows the effect of differential turnout since PR-STV was reintroduced for Stormont elections in 1973.

Impact of Differential Turnout on Bloc Share

If we take 2007 as an example, the unadjusted bloc shares of the vote were 48.62% U, 43.55% N and 7.83% MG. The adjusted bloc shares become 49.93% U, 41.69% N and 8.39% MG. The adjusted nationalist bloc is 1.86% less, whereas both adjusted unionist and middle-ground blocs increase, respectively, by 1.31% and 0.56%.

This graph suggests that the arrival of Sinn Féin into electoral politics for the 1982 ‘rolling devolution’ assembly election had an energising effect on turnout for a significant number of nationalist voters who had previously stayed at home. As Sinn Féin became part of the political establishment in the past decade, the voting surge in nationalist constituencies has abated. In 2016 and 2017, the nationalist bloc bonus due to differential constituency turnout was at its smallest since Sinn Féin entered electoral politics following Bobby Sands’ election as MP for Fermanagh & South Tyrone in 1981.

The second image below shows how average valid poll percentage has varied by constituency since 1998. Constituencies are colour-coded based on what bloc predominated in the 2019 Westminster constituency. It is striking that all the above-average valid poll constituencies are nationalist-leaning ones; that Alliance’s only victory happened in the constituency with the lowest valid poll; and that unionist-leaning constituencies all have a valid poll percentage lower than the overall valid poll percentage.

Constituency Average Valid Poll Since 1998

The third image, below, shows how the valid poll percentage has changed since 1998.

Constituency Valid Poll Change Since 1998

Notice how the biggest percentage decreases in the valid vote have occurred in mostly nationalist-leaning constituencies. The effect of this is to reduce the gap between the adjusted and unadjusted bloc shares.

Some boundary adjustments since 1998 have occurred, but they are probably unlikely to have significantly impacted on these figures.

How might differential constituency turnout impact future results and referenda? The fourth image, below, shows the unadjusted and adjusted gap between the unionist and nationalist blocs in Assembly elections since 1998.

Extrapolation of Bloc Vote Difference: Unadjusted and Adjusted

While there can be sizeable inter-bloc gap variation between successive elections, there is a clear downward trend. Note that the gap between the adjusted unionist and nationalist bloc vote shares is always larger than the gap for the unadjusted (i.e. actual) bloc vote shares.

If a best-fit linear regression is fitted to both datasets and extrapolated to when the gap closes (i.e. equal share for both unionist and nationalist blocs), both graphs suggest that a nationalist bloc vote plurality (largest vote, not necessarily a majority) is still quite a while away. The ‘adjusted’ best-fit line suggests that actual plurality could be six years later than ‘unadjusted’ plurality: 2042 rather than 2036.

These extrapolations suggest that an Assembly vote in favour of asking the NI Secretary of State to call a referendum on a united Ireland is a long way away: one would imagine there would have to be at least a nationalist plurality – or probably a majority – of the Assembly seats for that to happen.

This projection doesn’t take into account the rise of the middle-ground Alliance and Green parties. Both seem to be eating equally into the two tribal behemoths: Catholic demographic growth has failed to translate into a nationalist bloc share increase since 2003 (see graph 2); and the unionist bloc has dropped sharply since the RHI and Brexit débacles (graph 1 of same link).

It is extremely likely that any referendum would see nationalist, unionist and middle-ground voters flocking to the polls in equal numbers in most, if not all, constituencies. In the 1921 NI Parliament election, unionist constituencies tended to have higher turnout than nationalist constituencies.

Every electoral system produces results that don’t align with the popular will. In Scotland, the May 6th election returned a SNP/Green MSP majority for independence (72 of 130 seats) even though their combined constituency vote share was 49%, their combined regional vote share was 48.4%, and opinion polls at the same time were showing more Scots preferring to remain within the UK. In 2015, pro-independence Catalan parties gained a majority of the seats even though they gained less than 50% of the vote: rural (pro-independence) seats have smaller electorates than urban (remain Spanish) seats.

Those who desire, and those who organise, referenda need to factor in the vagaries of their electoral system to minimise any political instability that might occur due to the difference between what the elected politicians are saying and what the people are saying.

The current disarray in unionism post-Brexit/Protocol is likely to result in the election of a SF First Minister after 2022’s Assembly election, even though the unionist bloc vote is still likely to be 3-6% greater than the nationalist bloc next year (and even though SF’s vote share may drop a bit given the continuing Alliance surge and Colm Eastwood’s strong performances in Westminster). If this were to happen, and if SF were able to definitively – in the eyes of a significant section of pro-Union voters – disassociate themselves from commemorating IRA actions and combatants while governing responsibly in NI, they have a 14- to 20-year opportunity (2022 to 2036/2042) to make a desire for, or an acceptance of, a United Ireland become the settled will of a substantial majority of voters. And, of course, governing responsibly in Stormont will reap dividends at Dáil elections.

The stakes are enormous for the republican movement: is support for 20th-century physical force republicanism compatible with a united Ireland voted for by the Catholics, Protestants and Dissenters of Northern Ireland? If the answer is no, which will they choose?

What if a fudged Protocol deal between Brussels and London restores stability, and unionism coalesces into one party – not in thrall to religious fundamentalism – under a media-savvy leader? (We’ve waited ages and now, like buses, two have come along.) Such a civic unionism might look more attractive to many Alliance, Green and SDLP voters in the absence of a SF move to unambiguously future-oriented entirely-democratic policies and activities.

So who will the middle-ground voters go for in a future referendum: the pro-London vision of civic non-fundamentalist unionism or the pro-Dublin vision of ‘non-commemorative past-definitively-put-behind-us has-gone-away-you-know’ nationalism working in tandem with a transformed constitution in the South? Elections since 2017, and current opinion polls, have shown us that parties dominated by either religious fundamentalism or trying to justify their movement’s violent past are losing votes to middle-ground parties.

When neither bloc has a majority, zero-sum politics doesn’t work because it produces lose-lose outcomes for both blocs. If republicans transform themselves away from the treadmill of constant commemoration of violent acts and if unionists stay divided, a united Ireland seems likely. On the other hand, if civic unionism predominates over religious-fundamentalist unionism while SF stays as hagiographer-in-chief of the armed struggle ‘carving tomorrow from a tombstone’ (as Paul Brady put it), NI will probably be in the UK for the foreseeable future.