What our polls got wrong and why

All polls conducted in the past weeks converged on an 8 to 10 point gap for Labour. What went wrong? Asks James Debono.

Surveys conducted by MaltaToday, it-Torca and the Sunday Times before and during the campaign using a similar methodology had suggested a narrowed gap for Labour from the 42,500 gap in 2019 to one of between 15,000 and 30,000.

But this was a far cry from a result in which the gap between the parties has fallen significantly below 10,000 votes and a percentage-point gap of just four points.

The final surveys of all three pollsters showed a gap of between 24,000 (MaltaToday) and 30,500 (The Sunday Times) with it-Torca somewhere in between, projecting a 27,500-vote difference.

MaltaToday’s surveys showed the Labour lead increasing from 15,000 in February to 29,000 in March and dropping again to 17,000 in the beginning of April.

Subsequently, the gap in Labour’s favour progressively increased to 20,000, 21,500 and finally to 24,000.

In percentage terms MaltaToday registered a gap of between 6.9 points and 10.2 points with the projected turnout stabilising around 70% during the electoral campaign.

All three surveys did get some things right. For example, both the Sunday Times and the MaltaToday surveys clearly showed Metsola emerging as a front runner and both put Cassola in the top tier and both suggested a low score for Lowell’s far right.  

The latest MT survey also got the ranking order of Labour’s candidates right with Daniel Attard emerging in second place and Thomas Bajada and Steve Ellul following closely behind. It also predicted that Peter Agius would get more first count votes than David Casa and David Agius.

But while the projected turnout was not so much off the mark from the actual one (72.8) especially after removing invalid voters, the gap between the parties cannot be explained by the margin of error. 

The stark reality is that all surveys went off mark by a substantial margin even if MaltaToday’s was the closest to the result gap. Here are four reasons which may explain the discrepancy between surveys and the actual result.

1.Voters did their own thing silently in the polling booth

One possibility is that a segment of voters actually voted differently from the intention declared in surveys. One possibility is that traditional labour voters found it hard to declare their intention to vote for another party in surveys but felt comfortable to do so in the voting booth.

2.Undecided voters were allotted to Labour but voted otherwise

Another possibility is that some voters were undecided and only made up their mind towards the very end.  Since all three surveys used the imputation method, the undecided were still allotted to different parties using predictor questions. 

Possibly while still following the patterns of Labour respondents, these voters ultimately ended up voting for other parties.

In fact, previous MaltaToday surveys carried out using another methodology in which the undecided are included in the result, had shown the PN overtaking Labour in July 2023.

3.Some young voters voted against Labour instead of abstaining

It may also have been the case, that a segment of younger voters who declared their intention to abstain, ultimately ended up voting for a party which was not Labour. Surveys had shown that a considerable 40% of under 35 year olds were intent on not voting.   

4.Voters were overwhelmed by events while being surveyed

But one major reason could well have been that the surveys were held during a tumultuous month dominated by revelations and controversy around the hospitals inquiry.

Normally it is advisable not to hold polls right in the middle of events which are still evolving and to wait a few days to be in a position to assess their impact. But this was impossible during an electoral campaign.

It is possible that people were still deliberating on who they were voting for right up to the very end, with the publication of the Vitals hospitals inquiry by MaltaToday giving politicians less control over the political narrative. This undermined Labour’s ability to press on with its mantra blaming the establishment.

It is possible that the polls captured the beginning of a trend; that of a narrowed gap which was completed in the privacy of the polling booth.

Conclusion

But while pollsters have a lot to ponder on, the silver lining is that this should be a cautionary tale for people who treat polls as some sort of magical prophecy. And crucially pollsters themselves should avoid projecting themselves as some kind of oracles of truth.

Polls are ultimately a useful tool to explore political trends providing a snapshot in a specific and sometimes elusive moment in time. But while this time around polls have clearly underestimated Labour’s decline, let’s not forget that polls conducted by all three polling agencies have also predicted most political outcomes of the past two decades with a degree of accuracy.