# Methodology

HTML version: [https://datalibertarian.in/babuwatch/methodology](https://datalibertarian.in/babuwatch/methodology)

# Methodology & disclaimer

Babuwatch publishes only what a court decided, attributed to that court, and says exactly how far each record has been checked.

Babuwatch documents court-adjudicated findings against public servants in India, drawn from court judgments and from the official conviction records that anti-corruption and vigilance agencies publish. Every finding is the court's, not ours. Inclusion records what a court decided on the date shown; it does not establish anything beyond that finding, and a trial-court conviction may since have been appealed. Records are corrected or withdrawn as facts develop. We withhold names by default and never publish the identities of complainants, victims, witnesses or minors.

## One register, one method

Babuwatch lists every published record in one register — police officers and civil servants alike. Every record shows the verification level it was published at, and the appendix below sets out what each level means and how each source programme’s rules apply. Police records began life as Copwatch India, and that provenance is kept in their sources. Further registers (for example, for a single service) will follow the same method.

**Official sources only.** A record is published only when one of its sources is the court order itself or an official agency record (a court or government website, a judgment repository, or an archived copy of one). A record that rests on news reports alone is held back until an official copy is found, and its page says so. High Court and Supreme Court records publish at V2 and above.

**Who is covered.** Babuwatch covers the core civil services first: the All India Services (IAS, IPS and the Indian Forest Service), the Indian Foreign Service and central services such as the IRS, and state and central government employees from Group A (formerly Class I) to Group C (Class III). Employees of public-sector enterprises, government banks among them, will be added after that.

## Civil-servant convictions

These rules govern records of civil servants. Police records follow Copwatch India's methodology; the verification levels, the naming gate and the treatment of appeals are the same in both.

**Dataset size.** Babuwatch publishes 2,443 court-adjudicated records against public servants — 522 civil servants and 1,921 police records, from 1168 Supreme Court / High Court judgments and 1275 trial-court convictions. A further 22 trial-court convictions later set aside on appeal are kept out of the counts.

### Inclusion rules

**Who.** An employee of a Union, state or local government department, office or body — for example a revenue official, clerk, engineer, registrar, or municipal or panchayat employee — convicted for conduct in that office. Elected representatives, employees of public-sector banks and companies, cooperative-society staff, and private persons (co-accused, middlemen) are not included in this register; public-sector company and bank staff and cooperative-society staff are planned for a register of their own.

**What.** A criminal conviction entered by a trial court — a Special Judge under the Prevention of Corruption Act, a CBI court, a Lokayukta court or a Sessions court — recorded either in the judgment itself or in an official conviction list, press release or annual report of the agency that prosecuted the case (a Lokayukta, Anti-Corruption Bureau, Vigilance department, the Directorate of Vigilance and Anti-Corruption, or the CBI). Acquittals, discharges, pending trials and departmental action alone are not included.

### Verification

A civil-servant record publishes at **V2** or **V1** — the same standard as the police trial-court records, which also publish at V1. At V2 the official source was retrieved (an archived copy where the agency has since removed the page), and the official's post, the court, the conviction date or case number, and the conviction itself were matched in the same passage of that source. At V1 the source entry shows the conviction, but one element is weaker — the date or case number was not matched in the cited passage, the employing office is not stated, or the entry also contains acquittal wording — and every V1 record says on its page exactly what kept it from V2. A record where the conviction itself is not established — no conviction wording in the source entry, or an acquittal nearer the person's name than the conviction — is not published at all. A record where the source reports an acquittal for that person is rejected. V2 means the record is accurately attributed to an official source; it does not mean Babuwatch has independently corroborated the underlying facts (that is V3).

### Appeals

Agency conviction lists rarely report what happened on appeal. “No appeal known” means only that the cited source does not mention an appeal; it is not a statement that none was filed. Trial convictions are appealable and some are set aside; a conviction known to have been set aside leaves the counts. If you know a conviction was overturned, tell us and we will correct the record.

### Naming

Officials are shown by post and department. A name appears only after the two-key naming gate used across Babuwatch: the source must show that person convicted (first key), an independent check must confirm it (second key), a subsequent-history check must find no later reversal, and a qualified lawyer must review the record. No civil-servant name has yet cleared this gate, so none is shown. Complainants and other private persons are never named.

### Categories

Each record carries one category: **Bribery & extortion** (demanding or accepting a bribe, Prevention of Corruption Act section 7 and related provisions); **Disproportionate assets** (assets beyond known income, section 13(1)(e), section 13(1)(b) after the 2018 amendment); **Fraud & misappropriation** (criminal breach of trust, cheating, forgery, misappropriation of public money); **Violence & unlawful detention** (almost entirely police records); and **Other misconduct**.

### Duplicates

The same conviction often appears in several lists (an annual report and a press release, or two years of reports). These are merged into one record, keeping the best-evidenced source.

### Limitations

Coverage reflects which agencies publish conviction records and how far back their archives reach. Several states publish none, so absence from Babuwatch is not evidence of absence. These records are not a representative sample of corruption or of prosecutions; counts show what was found and verified, not how common misconduct is. Dates are as precise as the source allows.

### Corrections

If a record is wrong, incomplete or out of date, write to us with the record's link and the source that shows the correction. A record we cannot stand behind is corrected or withdrawn, and the change is logged.