Can we see a direction of travel from polling on Remaining or Reunifying?

Updated to add graphs excluding NILT/Liverpool datasets.

More opinion poll data from LucidTalk in the Belfast Telegraph (February 24th 2025), suggesting that 48% of their NI sample wish to remain in the UK (“Remain”) and 41% wish for a reunified Ireland (“Reunify”). This follows hard on the heels of a recent ARINS poll in the Irish Times (Ipsos) (48% remain in UK; 34% reunify).

When we exclude the don’t knows or won’t vote percentages, we find that the Remain – Reunify percentages (as a % of those who expressed an opinion either way) are 59-41 for the ARINS poll and 54-46 for the LucidTalk poll.

It is a brave person who would extrapolate from just one poll. If you did this from these two polls, it suggests a majority for reunification by Summer this year, unlikely to say the least. Looking at all the reunification polls since Brexit (polling data tabulated on Wikipedia), can we draw any firm conclusions?

Some concern has been raised about the reliability of both Northern Ireland Life and Times (NILT) and University of Liverpool polling, in relation to the deviation between their election polling and the actual results: see Professor John Doyle’s RIA/ARINS blog and Professor Katy Hayward and Ben Rosher’s comments on the 2020 NILT survey. For this reason, each graph is followed by another similar one excluding the University of Liverpool and NILT data.

********

Figure 1 shows all of the survey data. We can see that the number of polls increased in the immediate aftermath of the negotiation of the Brexit withdrawal agreement, and these show a narrowing of the gap between Remain and Reunify. Since then, that gap seems to have widened and the polls have become less frequent. The best-fit line is a second-order curve (which means it is of the form ax² + bx + c, so not a straight line). The R² values are relatively low, suggesting a not particularly good fit, as the datapoints are well spread out. (If all of the datapoints lay on the best-fit line, the R² value would be 1.0.) The best-fit curve in Fig.1 suggests that reunification is unlikely to happen, as the best-fit gap is widening.

********

Figure 2 shows just the polling data from the past 4 years. The two datasets are essentially flatlining, and we consequently get R² values of 0, implying no movement of any consequence over this time period. Excluding the NILT/Liverpool datasets, support for Remain seems to have very slightly increased, with concomitant decrease for Reunify.

********

Another approach to looking for trends is to get the average Reunify and Remain figures in a fixed time period. Figure 3 calculates the average poll numbers within a six-monthly period. However, the number of polls within a six-monthly period varies: four between September 2023 and February 2024, but only one between March-August 2023. This graph does show some slight recent narrowing in the gap.

********

Figure 4 shows yet another approach: to try to minimise fluctuations by calculating the smoothed average values for Remain and Reunify over three neighbouring opinion polls. The date is also averaged: so keep in mind that the ‘date’ of a poll in this case does not represent an actual opinion poll, but the average date of the three most recent polls (which is why the 27/8/24, 7/2/25, and 24/2/25 dates average to 20/12/24). This graph suggests a narrowing of the gap since October 2023. If this recent rate of narrowing were to continue, Reunify would have over 50% by Summer 2027. (But remember that the most recent 3-poll smoothing ‘date’ of 20/12/24 includes the two February 2025 polls, so a Summer 2027 average ‘date’ might include a poll from over a year later, if the frequency of polls drops.) And notice that Remain is higher now than it was 2018-2020.

********

Figure 5 tweaks the last graph: this time averaging the polls over three neighbouring opinion polls. We can see that the datapoints for the 5-poll smoothing do trace a smoother path across the graph than those for the 3-vote smoothing. One cannot be as definite that there is a downward trend compared to the 3-poll graph, but something might be happening gap-wise since April 2024. Extrapolating this rate of narrowing suggests Reunify would have over 50% by the end of 2025, extremely unlikely I would imagine. (And we can only say December 2025, assuming the same rate of polls up to then.) And, remember that the most recent 5-poll smoothing ‘date’ of 3/8/24 includes the two February 2025 polls, so a December 2025 average ‘date’ might include a poll from over 2-3 years later, if the frequency of polls drops.) Fig.5, minus Liverpool/NILT data, suggests parity is over two years away.

********

So. Five graphs:

1. including Liverpool/NILT datasets: one showing the Remain-Reunify gap widening, three showing a narrowing gap, and one showing no change.
2. excluding Liverpool/NILT datasets: two showing the Remain-Reunify gap widening, three showing a narrowing gap.

Can we draw a definitive conclusion? It’s important to note that the smoothed polling trends also show Remaining in the UK rising 2020 – spring 2023. Opinion poll interpretation, like beauty, lies in the eyes of the beholder.