In the recent Supreme Court judgment in Syed Iftikhar Andrabi v. National Investigation Agency, the Supreme Court granted bail to accused in a narco-terror case, citing, among other grounds, the “abysmal” conviction rates under the Unlawful Activities (Prevention) Act, 1967 (UAPA). The Court relied on National Crime Records Bureau (NCRB) data for 2019–2023 to say that there is a 99% possibility of acquittal in such cases in J&K — and 94–98% nationally — effectively using low conviction percentages as a proxy for the weakness of the prosecution’s case and the injustice of prolonged detention. Here is the exact table reproduced in the judgment for J&K:
| YEAR |
PERSONS ARRESTED |
PERSONS CONVICTED |
PERCENTAGE OF CONVICTION |
| 2019 |
227 |
0 |
0 |
| 2020 |
346 |
2 |
0.58 |
| 2021 |
645 |
0 |
0 |
| 2022 |
1238 |
11 |
0.89 |
| 2023 |
1206 |
10 |
0.83 |
The Court concluded: “the annual rate of conviction is always less than 1%. It means that at the end of the trial, there is 99% possibility of acquittal in such cases.” This statistic was invoked to “fortify” the application of Union of India v. K.A. Najeeb (2021) and to justify bail despite serious charges involving narco-terror links under UAPA and the NDPS Act.This seems to be a textbook case of statistical misinterpretation.
The percentage is calculated by dividing convictions recorded in a given calendar year by arrests made in the same calendar year. This is not a conviction rate; it is a meaningless ratio that ignores the fundamental reality of criminal justice timelines. UAPA cases — especially those involving terrorism financing, narco-terror networks, or cross-border linkages like the one against Andrabi (alleging recovery of heroin, cash, and links to LeT/HM operatives) — are not disposed of quickly. Trials routinely take 5–10 years or more.
Convictions entered in 2023 almost certainly arose from arrests made in 2018, 2019, or even earlier — cases that had already aged through investigation, chargesheet filing, framing of charges, and full trial. Meanwhile, the 1,206 arrests in 2023 had barely begun their journey through the system by the time the data was compiled.
Treating 2023 convictions (10) as reflective of the fate of 2023 arrests (1,206) is like judging the success rate of a five-year medical treatment by comparing patients who started treatment this year to those who completed it last year. It guarantees an artificially depressed “rate” every single year, regardless of actual case quality.
This is not how conviction rates are computed. A genuine conviction rate is: Convictions ÷ (Convictions + Acquittals) for cases disposed of in a given period. Or, even better, a cohort-based analysis: track what percentage of arrests from a specific year ultimately result in conviction after all appeals. The judgment uses neither.
Some have compared this calculation to “case clearance rates” used in judicial statistics (disposals vs. institution of cases). That analogy fails here. Clearance rates measure system efficiency — how well courts are keeping pace with inflow. They do not pretend to predict the outcome probability for newly instituted cases. The Supreme Court, however, explicitly used the figure to infer a “99% possibility of acquittal” — a predictive claim about case merit that the data cannot support.