Why do official road-death figures differ?

A difference between official figures deserves examination. Start by checking whether the sources actually count the same thing.

This example comes from Rajya Sabha Unstarred Question 2820, answered on 12 August 2026 by the Ministry of Road Transport and Highways. The reference year is 2024, not 2026. Both pages of the answer were inspected for this article. 1 2

The figures given in the answer

These numbers are attributed to the Ministry's parliamentary answer. This article does not claim to have independently reconstructed all three underlying reports. 1

What the Ministry says explains the differences

The answer describes differences in state reporting channels and distinguishes accidental-death records from deaths recorded as penal-law offences. It also notes the accidental-death series' treatment of self-caused road accidents. These are the government's stated explanations, not a reason to stop asking methodological questions. 2

The criminal negligence category should not be relabelled as an alternative count of every road fatality. A scope difference is not repaired by averaging the numbers.

Question and answer are different evidence

A parliamentary question may contain a premise, an estimate or an allegation. Its appearance in an official document does not turn that premise into a finding by the government or Parliament.

Read the reply and check what it accepts, rejects, qualifies or leaves unanswered. This article uses the figures stated in the answer rather than repeating the question's framing as an established fact.

What you can conclude

The answer records three distinct reported figures and offers an explanation of their differing scope or collection. You can reproduce the difference between the first two figures: 177,175 − 175,142 = 2,033.

That calculation does not establish which series captures every real-world death, prove intentional manipulation or show that reconciliation has been completed. Those are additional questions requiring additional evidence.

This is a model for reading conflicting data: preserve the units and categories, attribute explanations and make unresolved questions visible.