Skip to content
All library documents

Choosing Event Dummies for Exchange Rate OLS Studies

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

Summary

The document asks how to test whether multiple news announcements affected exchange rates across countries. The proposed market model uses exchange rates as the dependent variable, stock market returns as a control, and event indicator variables. One approach assigns a single dummy value to each announcement date across the full sample; another uses a shorter sample and marks a small window around selected announcements. The question is whether either setup can identify an event effect.

The document provides no answer, regression results, or design guidance, so it does not establish which specification is appropriate. It also leaves important details unresolved, including whether announcements overlap, whether their effects are expected to persist beyond the announcement day, and how events should be defined. Readers can take away the core problem of translating announcement dates into regressors, but need additional sources to choose event windows, account for repeated events, and interpret estimated effects.

Key ideas

  • The analysis concerns announcement effects on exchange rates across multiple countries.
  • A proposed market model includes exchange rates, event indicators, and stock market returns.
  • The question contrasts full-sample announcement-day indicators with narrower event windows.
  • The document provides no answer or empirical evidence to evaluate either design.
  • Event overlap and the duration of announcement effects remain unspecified.

Tags

ForexEvent-drivenStatistics
Full text
# How to set up the dummy variable for OLS event study regression


# How to set up the dummy variable for OLS event study regression












I've been going back and forth with how I should work to find an event effect. would be so grateful for some clarification.

I have daily time series of exchange rates for different countries ( 1 for each country). the period is from jan1, 2018 to dec 31, 2019. I want to know if a certain event that had multiple announcements during this period had an effect on each of the countries. ( I am taking the news that happened during this period 65 in total) and I want to analyse if the event had an effect on each country's exchange rate. this is the market model and my variables would be ER= (dependent)exchange rate, D= dummy, SP= stock market return

my question is, should i have 1 dummy variable for the whole regression with all 1 when a news happen and zeros when it did not and analyse that, as in my data frame would look like

and so on for 546 observations( business days) which contains 65 dummys. so total would be 481 observations with 0 and 65observations with a 1 during this period of jan1, 2018 to dec312019 when there was a news announcement.

I should then just run an OLS regression of the market model and the dummy=1 would tell me the effect of the event on the FX.

or should I have had to reduce the time window, to, for example, 6 months and analysed each news announcement separately, of course now focusing on less news announcement ( 1 every 6 months), and adding a dummy=1 to days around the event date (-1,+1). so in a 6months daily observations, i would only have 3 dummys=1 everything else would be zero for example:

or am i completely wrong with my both approaches?

thank you so much!

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.