Democracy Volunteers deployed 4 observers, in two teams of two, following the international best practice of the so-called ‘four eyes’ principle, to observe the Gorton and Denton by-election on 26th February 2026.
The data collected after having observed 22 polling stations across the course of the day, was readily available at the close of poll, which allowed us to make an immediate statement of our findings at 10pm, before the result of the election was declared.
We don’t usually issue our findings at the close of the poll because the quantity of data outside of the context of a byelection is usually too large to facilitate this. Following us issuing the data, it was disappointingly misused by political actors, with some stating it was evidence of certain parts of the community being more likely to commit electoral fraud, which the data in no way stated.
The by-election has provided us with an opportunity to reflect on our own practices, and to implement steps to ensure that our data is not misused in the future.
Our purpose is to assess the extent to which elections meet domestic law as well as international standards and law, to increase public confidence in the electoral process and offer recommendations for future improvements. This by-election highlighted the importance of clarity alongside transparency in our communications.
The term “Family Voting”
Democracy Volunteers has been collecting data on ‘family voting’ since we began observing elections in 2016. We, and other election observation groups define ‘family voting’ in an advanced democracy, like that in the UK, where the secret ballot is well established, as any breach of the secret ballot. We have observed ‘family voting’ in the 15 countries and territories we have observed around the world.
However, it is clear from the reaction to our release on 26th February 2026, that a technical term understood by election observers around the world, is easily misunderstood by others less familiar.
The term covers a broad range of behaviour. While any violation of the secret ballot is a problem, the nature of that problem differs. On one hand, there are people unambiguously coerced, or subtly pressured to cast a vote they do not want to cast. On the other end of the scale, there are people who may, without prior approval from election officials, request assistance from a family member or friend, or fill out ballots together without considering the lack of secrecy a problem.
That’s why we are clarifying our definitions to make it clear exactly what we are talking about. We will publish this detail before our next data release.
Evidence we supplied to those investigating the issue in Gorton and Denton has been adjudged not to reach an evidential threshold for prosecution, we would not suggest it does. However, any decisions taken by the relevant authorities involve a different evidential threshold; we stand by our methodology and have offered to work constructively with Manchester City Council and others to support improvements going forward.
Lessons Learned
We have long prioritised transparency by publishing detailed observation findings. In addition, we must improve the clarity of the information we provide.
The misuse of data by third parties is not unique to this by-election and can arise across a wide range of actors, including media commentators, political campaigners, and members of the public. The reporting of our data reflected wider trends of the means through which data is disseminated across various platforms for differing purposes.
As part of this review process we have consulted with interlocutors, interested parties, and our observers, through a series of focus groups and conversations to learn appropriate lessons from the Gorton and Denton by-election.
1. It is very important to clarify the context of the data collected during our observations:
We found that when presenting election data, clearer, more specific, terminology is essential to promote full understanding.
2. It is important to anticipate how the data will be interpreted:
We recognise that election observation occurs within a pressured, fast-paced and highly networked political environment. Once information is made public, it may be interpreted, shared, or reframed in a variety of ways.
Therefore, we must consider how information might be understood as part of our reporting process from the outset. Data can be used to identify or infer evidence from our findings that are not stated by us, such as our observations being associated with one demographic group or another, without us evidencing or stating that.
Our data, however neutral in our statements, can be weaponised to malign individual communities leading to characterisation of certain behaviours associated with that community. Unintended by us, this can lead to potential assertions of bias on our part, demonisation of individuals, groups or communities, which is not the objective of our evidence. It can also lead to reputational risk and damage to our organisation, which is proud of our independence, non-partisan nature and transparency with our findings.
3. The timing of the release of data is important:
We also learned that the timing of the publication can influence how findings are received and understood. Releasing detailed information shortly after the close of the poll, without a fully developed contextual framework surrounding the data, led to incorrect interpretations.
Changes Implemented by Democracy Volunteers
In response to these lessons, we have refined our approach to data handling and reporting.
We have introduced structured internal review processes to assess:
- Clarity of language and potential for misinterpretation
- Whether sufficient context has been provided
- The overall balance and proportionality of findings
In addition to these changes, we are also placing an increased emphasis on communicating our findings directly with key interlocutors including the Electoral Commission, and the Association of Electoral Administrators, policymakers, and academics. This ensures our findings are clearly understood by those directly involved in electoral processes.
Finally, we are refining how we communicate publicly to highlight the extent to which the electoral process meets domestic and international standards, to avoid the misrepresentation of our work, and to present findings in a clear and balanced manner.
Although it has rarely been our practice to issue data so quickly, as by-elections allow us to do, we will revert to our general previous practice of issuing reports no sooner than 4 weeks after an election. Bearing in mind the large datasets for large-scale elections, this is also a practical as well as policy decision.
