← PureCar News

Why DVA withheld Ballymena examiner pass rates: the evidence and the limits

Published
FEATURE
Why DVA withheld Ballymena examiner pass rates: the evidence and the limits

A Freedom of Information request sought anonymised driving-examiner pass rates and complaints data for Ballymena. DVA refused the requested breakdown, citing the risk of identifying individual examiners. What can candidates actually verify?

Anonymised was not the same as anonymous

The relevant document is FOI reference DFI/2025-0274. The request arrived on 19 June 2025 and DVA responded on 8 July 2025. The requester asked for three examiner-level measures at Ballymena Driving Test Centre: individual pass/fail rates, male-versus-female candidate outcomes for each examiner, and the number of complaints about each. Rather than ask for names, the requester proposed numbered labels such as Examiner 1 and Examiner 2.

DVA said it held information relevant to the request but would not release the requested breakdown in that format. Its stated reason was that Ballymena had a small number of examiners and the data could permit their indirect identification. It invoked section 40(2) of the Freedom of Information Act 2000, the exemption dealing with personal information.

This was not a statement that examiners are all statistically identical. Nor was it an admission of inconsistent marking or discriminatory treatment. It was a disclosure decision about whether particular data could be released. Those are fundamentally different propositions, and a credible investigation must keep them separate.

Why even a numbered table can identify someone

Consider a hypothetical centre with only a handful of examiners. If a person knows when one examiner worked, which shifts they covered, or information publicly associated with particular instructors, a numbered row containing pass rates and complaints might be linkable to a real person. Removing names does not guarantee that such links disappear. This example illustrates the concern; it is not a claim about the actual staffing or local knowledge at Ballymena.

The Information Commissioner’s Office explains that information can still be personal data where someone is identifiable indirectly. Its FOI guidance also shows that small numbers do not invariably justify withholding; authorities must evaluate the context and the actual likelihood of identification. The DVA release does not publish a detailed risk analysis, so an outside reader cannot independently test every step of the decision.

An additional complication is that complaint counts can include allegations that were not upheld. Publishing a small examiner-level complaint total without context could be misleading and potentially unfair. Equally, treating every form of aggregate scrutiny as impossible would risk leaving the public unable to examine whether systems produce consistent outcomes. Neither extreme is justified by the disclosed letter.

What DVA says is already publicly available

The refusal points to official quarterly driver, vehicle, operator and enforcement statistics. In particular, it identifies table 4.7 as a source for driving-test pass and fail rates by test centre and candidate gender. That is an important distinction: the absence of examiner-by-examiner data does not mean the entire driving-testing system is a statistical black box.

Published test-centre figures can support questions about how local pass rates change over time and whether overall outcomes differ by gender, but they cannot directly show variation between examiners in the same centre. An average can conceal spread in a dataset, yet the spread cannot be presumed large merely because it has not been published.

DVA also referred to its annual complaints reporting. An agency-wide complaint count is valuable for accountability, but it cannot be directly translated into a probability that a particular examiner has behaved improperly. To assess complaints responsibly, researchers need the type of complaint, its outcome, the number of tests delivered and whether repeat contacts relate to one case.

Pass rates alone cannot prove unfair marking

Imagine two examiners each conducting 100 tests: one passes 40 candidates and the other passes 60. That apparent 20-point gap is not by itself evidence of bias. The groups might differ in prior experience, test time, route complexity, training background, weather conditions or random variation. These are illustrative possibilities, not claims about Ballymena.

An analysis of consistency would ideally use a sufficiently large sample and investigate assignment of candidates and routes, case mix, complaint resolution, quality-assurance observation and any internal moderation. Some of those records could potentially be disclosed in broader anonymised bands without exposing individuals, but this FOI does not demonstrate what DVA could safely release.

There is a basic statistical trap in choosing the lowest pass-rate examiner and calling them too strict: if you sort enough examiners, one must be last even when all are applying the same standard. The relevant question is whether any variation exceeds what might reasonably be expected after taking account of meaningful differences.

A better transparency test

The strongest accountability question is not “Which examiner is the easiest?” Publishing rankings without context could encourage candidates to seek a particular examiner rather than improve safety. The better questions are whether DVA has a structured process for monitoring examiner consistency, whether significant outliers are reviewed, and what information it can publish about the process without exposing employees.

Test-centre-wide figures, time trends, documented quality assurance and complaint-outcome categories could provide independent reassurance. It is also reasonable to ask whether broader time windows or statistical thresholds could reduce the identification risk. DVA’s July 2025 response answered the particular request it received; it did not settle all conceivable forms of aggregate publication.

Candidates with concerns should preserve the date, centre, test details and any contemporaneous evidence, and use the appropriate official complaint or appeal route. An unpopular result, however, is not automatically an administrative error. This investigation is about transparency and the evidence necessary to judge fairness, not a claim that Ballymena examiners are marking improperly.

What a robust statistical review would require

The first step in comparing examiners would be to fix a common period and define the sample. For each examiner, a researcher would need the number of eligible tests conducted and the number passed; pass percentages without the count of tests are unstable, particularly when caseloads are small. Then they would need to understand whether candidates are assigned randomly or whether some staff disproportionately handle particular test types, sessions or circumstances.

Suppose an examiner has conducted only 20 tests. One extra pass changes the rate by five percentage points. For an examiner conducting 200 tests, one extra pass changes the rate by half a percentage point. A ranking that puts those two percentages beside each other without uncertainty creates an illusion of precision. This arithmetic explains why published raw examiner tables could be misleading even if privacy were not a concern.

Complaints also need denominators and definitions. The number of complaints received is not the number upheld, and an examiner conducting more tests has more opportunities to receive a complaint. A useful public summary could potentially include the total received, the categories, the share substantiated and any system-level learning, without publishing enough detail to identify individual staff. The appropriate disclosure balance is a matter for the authority and information law; PureCar has not determined that an alternative anonymisation method would necessarily be lawful.

The FOI is valuable precisely because it identifies the boundary between what the agency says it holds and what it publishes. That boundary deserves scrutiny, but proving unfairness would require much more than a rejected request.

Sources and methodology

PureCar examined the original published government response and cross-checked the interpretations against the official guidance linked below. Calculations are derived from the published figures; they are not new data supplied privately to PureCar. The response dates and coverage are stated in the article because historical disclosures must not be represented as newly released information.

SOURCES

Sources & further reading

www.infrastructure-ni.gov.uk www.infrastructure-ni.gov.uk www.infrastructure-ni.gov.uk www.infrastructure-ni.gov.uk ico.org.uk ico.org.uk ico.org.uk
CARS FOR SALE

Find Your Next Motor

Explore NI
Find Your Next Car Search Northern Ireland Cars Browse cars for sale across Northern Ireland