Did Restricting Index Derivatives Calm the Spot Market? Evidence from India's 2024 SEBI Reform A dose-gradient natural experiment across fifteen NSE indices

Abstract

In October 2024 the Securities and Exchange Board of India (SEBI) imposed the most restrictive set of index-derivatives rules in that market's history: minimum contract value tripled, weekly expiries withdrawn from all but one benchmark index per exchange, option premium collected upfront, and an additional 2% Extreme Loss Margin levied on short option positions on expiry day. The measures cut index-derivatives turnover sharply. The regulator's own market-stability rationale — and a chorus of contemporaneous commentary — predicted that curbing speculative derivatives activity would dampen volatility in the underlying spot indices. We test that prediction and find no support for it.

Identification exploits an asymmetry in the reform that has gone unremarked: it did not treat all indices equally. The National Stock Exchange retained weekly expiry for the Nifty 50 while withdrawing it from Nifty Bank, Nifty Financial Services and Nifty Next 50, and eleven sectoral indices carry no index derivatives at all. This yields a dose gradient across fifteen indices that share one market, one trading calendar and one set of macroeconomic shocks — converting a question that a single-index study could not identify into a difference-in-differences design with a control group the reform built itself.

The difference-in-differences estimate is −4.6% on 21-day realised volatility (β = −0.047, SE = 0.103), with an exact randomisation-inference p-value of 0.591 obtained by enumerating all 364 estimable treatment assignments. The estimate is stable in sign and insignificant across every alternative sample, volatility measure and event window we examine. Chow tests find no break at the implementation date (Nifty 50: F = 0.04, p = 0.96); estimated endogenously, no treated index selects a break near the reform — the Nifty 50 selects April 2020, the month of the COVID crash. An expiry-day triple difference, aimed squarely at the sessions three of the six measures targeted, is likewise indistinguishable from zero (β₃ = 0.004, p = 0.945 on the cleanest sample).

Machine-learning models trained on pre-reform data do not degrade after the reform; on treated indices they improve, landing below the 10th percentile of a 200-draw placebo distribution. The one index where forecast accuracy significantly deteriorates is Nifty IT — an untreated control.

We are explicit about what a null of this design can and cannot support. The minimum detectable effect at 80% power is a 25% reduction in volatility: we can reject reductions of that magnitude but cannot separate smaller changes from zero. Within that resolution, a reform that removed a large share of India's index-derivatives turnover left spot volatility where it found it. The investor-protection case for the reform is untouched by this finding; the market-stability case is not supported.