I am compelled to address the discrepancies between the recent MEP election results in Malta and the opinion polls leading up to it.

This article aims to shed light on why this may have occurred and what can be amended to improve future polls’ accuracy.

The role of 'Don’t Knows' and 'Refusals' in Surveys

High numbers of “don’t know” responses and “refusals” in opinion polls are quite normal. In our most recent survey, one in every five respondents claimed they either do not know or refused to answer who they would vote for in the MEP election.

It is for this reason that Esprimi has adopted machine learning to address this issue in the last general election and this MEP election.

Without this adoption, the discrepancy between opinion polls and election results would be much higher.  What we have learned from this election was that our methodology did not accurately capture the protest votes that took place during these MEP elections which appear to be less likely to occur during a general election.

It is also essential to remind readers that all surveys come with a degree of margin of error. In our case, with a sample size of 650 people, the opinion poll comes with a 95% confidence level and a 3.84% margin of error.

What does this mean? The 95% confidence level indicates that if we conducted the same poll 100 times, we would expect the results to fall within a certain range 95 of those times.

The margin of error tells us how much the poll results might differ from the true opinion of the entire population. A margin of error of 3.84% means the actual opinion could be 3.84% higher or lower than the poll result.  Therefore, all surveys can only give an estimate and not a certainty of the result. 

The results

MEP elections	Turnout %	PL %	PN %	Other %
Esprimi poll	72.40	50.80	40.40	8.80
Actual results	72.98	45.26	42.02	12.72
Discrepancy	-0.58	5.54	-1.62	-3.92

 Overall analysis

The poll results for voter turnout and the Nationalist Party are within the margin of error, indicating a high level of accuracy and reliability.

The actual support for the Labour Party is slightly below the lower margin of error range, indicating a slight overestimation by the poll.

The support for other individuals/parties is slightly above the upper margin of error range, indicating a slight underestimation by the poll.  

External political influences

External political events and sentiments also definitely played a role. It cannot be overlooked that voters might have used this election as an opportunity to send a strong message to the main parties about the status quo.

Throughout fieldwork, respondents were asked what issues would have determined their vote. In the wave closest to the election, the matter of corruption spiked as a concern among voters when compared to the previous two waves.

Additionally, the issues of cost of living and immigration were also mentioned.

Methodological considerations and limitations

Three waves of surveys were held during this electoral campaign with samples of approximately 600 responses per wave.

In comparison, other polls with larger samples and more frequent surveys, reaped similar results which suggests that the discrepancies were therefore not due to sample size or frequency but more likely influenced by broader, exogenous factors as mentioned above.

Moving forward

As we move forward, it is essential to treat opinion polls as tools that provide a snapshot of public sentiment as well as indicative predictions of election outcomes.

While the recent MEP election results may not have fully aligned with poll predictions, they offer valuable lessons for future polling efforts. We need to investigate further the effect of votes inherited from independent candidates to the main parties and votes that were invalidated.

As pioneers in using machine learning in our methodologies, this was the first time we applied this technique for MEP elections. At Esprimi, we remain committed to refining our methodologies and understanding the dynamic factors that influence respondent behaviour.  

Morgan Parnis is the CEO of Esprimi, commissioned by Times of Malta for opinion polls