12.1 Event-study design
Motivation
"Markets fell because of the war" is a claim about a counterfactual: what prices would have done without the war. An event study makes that counterfactual explicit, estimates it, and tells you whether the gap is larger than noise. Without one, you will fit a story to every wiggle. If you have read M06A, its 6A.2 primer (Proposition 6A.4) already gave the market model, abnormal returns and the single-firm CAR -test for results days; 12.1–12.2 restate them in full generality, and 12.3 onward is new: multi-event tests, event-induced variance and events that move the market itself.
Setup
Index trading days by event time , with the event day. That is the first trading session in which the market could react.
- The estimation window is , of length , for example with .
- The event window is , of length , for example .
- A gap between the windows keeps pre-event leakage out of the normal-return estimate.
Returns are for asset and for a market or factor proxy.
Normal-return models
- Constant mean: .
- Market model:
- Factor models: add regressors. For Indian assets around oil events, natural factors are Brent and USDINR.
Model choice matters most when the event moves the regressor itself. A war moves the "market". To ask "how much did OMC stocks fall beyond the Nifty", use the Nifty as . To ask "how much did the Nifty fall", the Nifty cannot be its own benchmark. Use a constant-mean model, or a world index such as MSCI World or the S&P 500 lagged one day, being aware that the world index also reacts.
Choosing
- Time zones. The NSE cash market closes at 15:30 IST. A US announcement at 14:00 ET (23:30 or 00:30 IST) is day 0 for the US but the next Indian session is India's day 0.
- Overnight news. Russia's invasion began in the early hours of 24 Feb 2022 (European time), so 24 Feb is day 0 for the Nifty. Operation Sindoor began at 01:05–01:30 IST on 7 May 2025, so 7 May is day 01.
- Weekend events. An event on Saturday maps to Monday's session.