Esprimi CEO Morgan Parnis announced today that his market research company will stop trying to predict the vote gap between PL and PN in future political surveys and focus solely on percentage forecasts.

“In this election, Esprimi projected a PL-PN gap of 10.6 per cent; the actual gap was 7.1% per cent,” he wrote in the Times of Malta.

“This 3.5 per cent gap difference is due to the prediction error of the two major parties by 1.7 per cent and 1.8 per cent respectively, both of which are within the margin of error.”

“This is why, across the major democracies, polls are reported and judged on percentage share, not absolute vote counts. It is the international standard, and it is the standard Esprimi will subscribe to going forward.”

Mr Parnis’ declaration cuts to the heart of a debate that has arisen amongst pollsters in the wake of the 2026 general election.

For years now, Maltese media headlines and subsequent public discussion have revolved around the forecast gap between the two parties.

Morgan Parnis and Vincent Marmara on WhosWho Talks

This is understandable – speaking in terms of vote gaps can be more easily visualised than cold, hard percentage numbers.

The problem is that small percentage swings, well within the margin of error, are equivalent to thousands of votes.

Vincent Marmara’s final pre-election survey projected the PL at 53.3 per cent and the PN at 42.8 cent, with a standard margin of error of 2.8 per cent and a forecast PL victory of around 30,000 votes.

PL ultimately secured 51.8 per cent of the vote, while the PN obtained 44.6 per cent, both comfortably within the survey’s margin of error.

In percentage terms, forecasting a PL result of 53.3 per cent only for it to end up at 51.8 per cent appears completely regular.

However, when framed as a projected 30,000-vote lead that eventually resulted in a 22,000-vote gap, the exact same survey appears disastrous.

Prof. Marmara tells WhosWho.mt that he took a larger voter sample than would be deemed acceptable by international standards so as to keep the margin of error as low as possible.

“A margin of error of 4 to 5 per cent is considered acceptable by international standards, but I took on greater risk by using a larger sample size to reduce that margin of error to 2.8 per cent.”

Vincent Marmara

A swing of 8,000 votes from a 30,000 to a 22,000 gap is politically significant.

It means PN has halved the gap seen at the last election some and allows the party to head into the next legislature with degree of optimism for the first time since 2013. It will motivate MPs, party officials and volunteers and likely secure the immediate political future of PN leader Alex Borg.

However, statistically it is a different story.

“A margin of error of 1.5 per cent is equivalent to around 4,500 votes,” Prof. Marmara explains. “If those people all switch to the other side, it becomes 9,000 votes. While a 1.5 per cent difference is statistically insignificant, these figures are leading to misinterpretation in our society.”

While MaltaToday’s final pre-election survey forecast was very close to the mark, it may be premature to claim that “one survey is better than others”. MaltaToday’s 2022 survey predicted a PL victory by 29,000 votes, while the final gap was closer to 40,000.

As Mr Parnis pointed out, all three pollsters correctly called a PL victory and all three landed within their stated margin of error.

“That is the headline, and it is worth stating plainly before the numbers, because much of the post-election commentary has focused on a single figure, the gap in absolute votes, that is the least reliable way to measure a poll's accuracy,” he said.

Prof. Marmara points out that focusing on the percentage share of votes is common practice in foreign political surveys, and used to be standard in Malta until a few years ago.

“Before 2017, Maltese surveys mainly focused on vote percentages,” he said. “But over time, people and media houses became more interested in the projected vote gap.”

While hearing about vote gaps is way more appealing than hearing about percentage shares, opinion polls were never designed to predict exact gaps down to the last 1,000 ballots.

And the problem is that pollsters are constantly attacked for “getting it wrong” when they are actually getting it right. They often end up resorting to mathematical reasoning when, for many people, the story stops at ‘You predicted X, the final result was Y, you aren’t reliable anymore’.

The question facing Malta’s pollsters now is whether they are ready to shift the focus back to percentage shares and statistical accuracy, even if it may come at the cost of tempering public debate surrounding the surveys themselves.