Averaging CARs Across Multiple Events and Countries

Summary

The document addresses how to organize cumulative abnormal returns (CARs) in an event study that includes several countries and multiple events within some countries. The proposed procedure is to calculate a CAR for each country-event observation, then average the event-level CARs within each country. Country averages can subsequently be combined with equal country weights or weights based on each country’s number of events.

The response also notes that multiple events can be grouped when they are independent, and cautions that overlap in estimation periods can complicate inference. It does not provide a full statistical testing procedure or explain how to handle dependence across countries, event windows, or common shocks. The weighting choice changes the estimand: equal country weighting gives each country the same influence, while weighting by event count gives countries with more observations more influence.

Key ideas

• Compute CAR separately for each country-event observation before aggregating results.
• Average event-level CARs within each country to form country-level estimates.
• Combine country averages with equal weights or weights based on event counts, depending on the research question.
• Independence between events supports grouping without a cross-correlation adjustment, subject to the stated caveat.
• Overlapping estimation periods and dependence structures can affect inference and are not fully addressed.

Tags

# Event study : multi-country and multi-events study (CARs - cumulated abnormal returns)


# Event study : multi-country and multi-events study (CARs - cumulated abnormal returns)












I'm trying to cnduct an event study accross countries, where sometimes there is more than one event per country.

I'm following the methodology explained here to compute $CAR_i$.

From what I understand, I can use one event per country if I'm following the formulas.

What should I do if I have more than one event per country, should I compute the $CAR_i$ for each event and then create an average ?

Thank for your help !

## Answer by Igor Pozdeev (score 2)

https://quant.stackexchange.com/a/41128

We do exactly that in our paper on FX returns before monetary policy announcements (link): there are several countries, and in each country there are several announcements per year, so the event matrix (in the time/country dimensions) looks like a punched card. In short, you do compute the CARs for each country-event, then compute the average CAR across events for each country, then you could average those weighing them equally or by the number of events for each country.

Details on the inference in the paper.

## Answer by Mads (score 2)

https://quant.stackexchange.com/a/42514

I can not see why you can't use more than one event per country. As long as they are independent of each other, you can easily group them without adjusting for cross correlation. Also, if they don't overlap on the estimation period as well.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.