Skip to content
All library documents

Choosing Return Data for Long-Run BHAR Event Studies

Article Quant Q&A · Author: dasanicola
Read the original

Summary

The document asks what return observations are needed to calculate buy-and-hold abnormal returns (BHAR) over a long event window. The example concerns an event in mid-December 2017 and a window extending from just before the event through twelve months afterward. The question is whether monthly non-log returns alone are sufficient or whether daily data are needed to align returns to the event date and build the monthly holding periods.

The response points to an event-study reference and says the appropriate data frequency depends on price volatility. It does not explain a specific calculation procedure, compare daily and monthly sampling, or report evidence from the cited paper. Thus, it signals that frequency choice can affect BHAR measurement, but leaves implementation and the implications of volatility to further reading. The brief exchange is a pointer to methodology rather than a complete guide.

Key ideas

  • The question concerns return frequency for long-run buy-and-hold abnormal return studies.
  • Event-date alignment can complicate using standard calendar-month returns.
  • The response indicates that price volatility affects the choice of data frequency.
  • The exchange refers readers to event-study literature but does not give a calculation recipe.

Tags

EquitiesEvent-drivenStatistics
Full text
# BHAR Event Study Data


# BHAR Event Study Data












I am about to run a long-run event study on certain events. For a short-term event study, I previously have used daily log returns. My question is now, what data I need for the BHAR one. Just monthly non-log returns or also daily returns?

Example: My event is on 15th December 2017 with a window of (-1;+12) months. Do I have to go back 30 days, then calculate the monthly return until 15th December and so forth for 12 months or how can I implement that in real life?

Many thanks for any hints.

## Answer by Vitomir (score 1)

https://quant.stackexchange.com/a/46016

This paper illustrates your problem. Basically, it depends on the volatility of prices.

S. P Khotari and Jerold B. Warner : “Econometrics of Event Studies” (2006)

https://www.bu.edu/econ/files/2011/01/KothariWarner2.pdf

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.