The Unrecognised Problem with Voter Photo ID

Photo credit: Kyle Glenn (

Written by:

Dr Natalie Butcher: Senior Lecturer in Psychology at Teesside University, Centre for Applied Psychological Science, Cognition and Decision-Making theme. Email: 

“Voting is a right and taking part in his country’s government is the cornerstone of democracy” Article 21 of the Universal Declaration of Human Rights.

The Elections Bill 2021-22 was introduced on 5 July 2021. If enacted, it would require voters to show voter ID in polling stations for UK parliamentary elections, local elections in England, and police and crime commissioner elections in England and Wales. Opposition to the Bill has largely focussed on the fact that ‘personation’ (i.e. the act of pretending to be someone else when you vote) is rare and that this measure therefore seems disproportionate, given its potential to preclude marginalised groups from voting, because they are less likely to have valid photo ID. It is argued that young people, people of colour, disabled and trans individuals and those with no fixed address are amongst those most likely to be affected. These opposition arguments have been documented widely and are addressed in the Voter ID Research Briefing published on 9 July 2021 so further in-depth discussion of these issues is not warranted here.

Instead, I want to talk about an unrecognised (pardon the pun) problem with the use of Voter Photo ID – the potential for legitimate voters to be turned away at the polling station due to polling station staff not recognising that the photo on their ID is a representation of that voter, instead believing it is a photo of someone else. The potential for this problem can been seen in research published by the Cabinet Office in May 2021 on access to photo ID. A nationally representative survey of 8,500 respondents in England, Wales and Scotland estimated that 96% held a photo ID with a recognisable picture. This might seem like a large and acceptable number but if we assume this proportion is the same for the overall population of voters, it suggests approximately 1.9 million voters do not hold photo ID that is recognisable, a figure I hope you would agree is not acceptable!

Add to this the known fallibility of our ability to recognise the face of a person we are unfamiliar with and we have a potentially big problem with the use of voter photo ID. In psychology we have known for decades that people are poor at looking at two faces, for instance the voter and the photo on their ID, and deciding whether they are the same person. In 1997 Kemp, Towell and Pike conducted a study in a supermarket setting. The store was closed and cashiers were told the study was looking at how fast and accurately they could process credit cards with photo ID. Of relevance here was the finding that cashiers questioned those with cards with their own face approximately 10% of the time. More recently, David White and colleagues (2014) conducted a study with passport officers whose main responsibility is to assess the eligibility of passport applicants. That is, in their job they are  tasked with confirming identity either by checking people against their ID photos (when citizens apply for passports in person) or by making photo-to-photo comparisons in the case of passport renewals, and when checking for potentially fraudulent duplicate applications. Despite this professional experience, White and collaborators found that valid photos were wrongly rejected 6% of the time. Taken together, these findings suggest that deciding whether a face in a photo is a representation of the person standing in front of you is not as simple as we’d probably assume – it is difficult and error-prone even when it is a fundamental part of your day job. These findings also demonstrate the potential for genuine voters to be turned away by polling station staff who aren’t routinely tasked with confirming a person’s identity in this way in their everyday lives.

As others have argued in relation to citizens access to photo ID, this problem of recognition is also likely to disproportionately affect those in a racial minority group. Decades of face recognition research has identified a phenomenon known as the ‘other-race-effect’ wherein people are significantly better at recognising the faces of people who are the same race as themselves when compared to the faces of people from a different race. Several socio-cognitive factors have been proposed to explain this phenomenon (see Hugenberg, Young, Bernstein & Sacco, 2010), including social categorization, motivated individuation, and perceptual experience. Originally the other race effect was observed in tasks that require a person to remember a face, learning the face in one viewing session and being asked to identify it in another. However, in 2011 Megreya, White and Burton showed that the phenomenon extends to face-matching tasks that don’t involve memory. These tasks are like the task polling station staff will be faced with, as they will simultaneously see the person and their photo and must decide if it’s the same person. In their study Megreya and colleagues recruited participants from the UK and Egypt and gave them a simultaneous face-matching task wherein they were shown a target face (Egyptian or UK) and were told that the target face may or may not be one of the ten faces presented below the target. Participants were asked to decide whether the target was within the ten faces, and if so, which of the ten was the target. In both participant groups they found better overall performance with own race faces than with other race faces, with the UK groups accuracy (mean of hits and correct rejections) being 67.4% for own race faces but only 52.6% for other race faces. Similarly, the Egyptian groups accuracy score was 69.4% when they viewed own race faces compared to 64.1% when they viewed other race faces. These findings, along with more recent examples of the other-race-effect in matching tasks, demonstrate that face verification of people from another race is particularly error-prone. Thus, in settings like a polling station, where person identification is being conducted with photo ID, as the Bill proposes, we would expect more people from minority racial groups to be turned away by station staff due to an inability to accurately match their face to their photo on their ID.

Given the wealth of research on the fallibility of face-matching and the other-race effect it is surprising that this potential problem with the use of voter photo ID has gone unrecognised in discussions about the Elections Bill 2021-22. In writing this blog I hope to be at least one voice speaking to this potential problem that could affect voters if polling station staff are not given adequate guidance on these psychological findings.


Hugenberg, K., Young, S. G., Bernstein, M. J., & Sacco, D. F. (2010). The categorization-individuation model: an integrative account of the other-race recognition deficit. Psychological review117(4), 1168.

 Kemp, R., Towell, N., & Pike, G. (1997). When seeing should not be believing: Photographs, credit cards and fraud. Applied Cognitive Psychology: The Official Journal of the Society for Applied Research in Memory and Cognition11(3), 211-222.

Megreya, A. M., White, D., & Burton, A. M. (2011). The other-race effect does not rely on memory: Evidence from a matching task. Quarterly Journal of Experimental Psychology64(8), 1473-1483.

 White, D., Kemp, R. I., Jenkins, R., Matheson, M., & Burton, A. M. (2014). Passport Officers’ Errors in Face Matching. PLoS ONE 9(8): e103510.