The Fifth Elephant 2026 Winter Edition

Call for Data Problems and Proposals

Rajat Gururaj

More Than Paper Leaks: 110 Incidents of Exam Fraud in India, by Mechanism

Submitted Oct 10, 2026

Submission type

Showcase (20 mins)

Title

More Than Paper Leaks: 110 Incidents of Exam Fraud in India, by Mechanism

Session description

India talks about exam fraud as paper leaks. A dataset of 110 incidents from 2004 to 2026 shows that only 25% are classic pre-exam physical leaks. The rest are insider collusion, WhatsApp and Telegram circulation, impersonation, hoaxes, and tampering with answer sheets after the exam. About 85% of incidents are outside entrance exams, mostly in recruitment for government jobs. Police and teacher recruitment stand out, with 18 and 15 incidents, and each is hit in a different way. Incidents a year have more than tripled since 2020, and Madhya Pradesh, Uttar Pradesh and Rajasthan together hold a third of them.

The session covers the findings that surprise people most: post-exam tampering that leaves no trace until someone audits the marks, and the gap between arrest and consequence, where 91 of 110 incidents stall at an arrest or a probe and 10 reach a conviction. It also covers what the data cannot tell us, and what happened when this analysis became an article in a national newspaper, Mint.

Takeaways

  1. A quarter of incidents are classic paper leaks. Securing the paper alone leaves the other three quarters untouched.
  2. An arrest is not accountability. Of 110 incidents, 91 stall after an arrest or a probe and 10 reach a conviction.

Audiences

Data journalists, analysts and civic tech teams who work with public-interest data. People who work on fraud, governance or education policy will also find the findings useful.

Bio

Rajat Gururaj is a data scientist and writer based in Bangalore, and a Senior Product Owner at Swiss Re. He writes about data on his Substack, and his first newspaper article, on this exam fraud analysis, ran in Mint in September 2026.

Slides

[Add link to PDF or PPT. Comments need to be enabled.]

What I don’t know yet

  • I have no count of exams held, so I cannot tell whether a state’s share of incidents reflects more fraud, more exams, or more press coverage.
  • Posts and applicant counts are known for only 33 of the 84 recruitment incidents.
  • I do not know how many of the 91 stalled incidents reach a conviction later, and I would like court outcome data to find out.

Match-making tags

  • I can help with: experience (taking a dataset to a national newspaper), dataset
  • I need help with: dataset (applicant counts, court outcomes), collaborator
  • I’d like to meet: researchers, domain experts, dataset owners

Topic tags

fraud, governance, public sector

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hosted by

Jumpstart better data engineering and AI futures