Skip to content
All library documents

Choosing OLS or Panel Models for Monetary Policy Event Studies

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

Summary

The document considers how to estimate an event study of unconventional central bank measures using euro-zone government bond yields. It compares separate country-level ordinary least squares regressions with panel estimation, where observations across countries are pooled to represent effects thought to be shared. The suggested starting point is a simple country-by-country OLS model with a small set of controls. This offers a benchmark before adopting a more complex specification.

The choice of a panel model depends on the researcher's assumptions about which effects should be common across countries; separate regressions allow more country-specific freedom. The discussion also raises a confounding concern: sovereign debt stress and turbulent markets coincided with policy announcements, potentially contributing to unexpected positive estimated coefficients. It does not provide a method for resolving that issue or empirical results. Its main guidance is to make the assumptions behind each regression explicit and interpret the model as a representation of a particular economic story.

Key ideas

  • Begin with country-level OLS and a small number of controls as a benchmark.
  • Use panel estimation when the research question supports pooling information across countries.
  • Separate country regressions allow more country-specific variation than a pooled model.
  • State the economic assumptions behind the estimator before choosing the regression specification.
  • Market turbulence coinciding with policy announcements can complicate interpretation of estimated effects.

Tags

Fixed incomeEvent-drivenStatistics
Full text
# What estimation method is best to conduct event study on unconventional monetary policy


# What estimation method is best to conduct event study on unconventional monetary policy












I have collected bond yield data from 01/01/2008:31/12/2019 for several euro-zone countries. I would like to conduct an event study analysis of the main Non standard measures announced by central banks, but I am not sure which model is best from a theoretical point of view. I basically have dummy variables and control variables as regressors, that are the same for each countries, except for the first lags of the dependent variable. Would you estimate the model for each country seperately via OLS, or should I use a Panel estimation method? One of the main problem I face is that, bond yields were rising because of the sovereign debt crisis and many policies were announced on a period of turbolent markets so that I get for instance positive loadings on many non standard monetary policy announcements. How would you handle this issue?

## Answer by Stéphane (score 1)

https://quant.stackexchange.com/a/53862

As part of your analysis, it is always a good idea to do something simple before pulling out the big guns. So, OLS by country with perhaps a handful of controls would be a good benchmark, if only to tell later if your complicated ideas don't amount to squashing a fly with a sledge hammer.

As for the panel idea, you have to think about how you'd be pooling your information. Country by country allows a maximal degree of freedom, but if you think some effects are shared or should be shared, you'd need a panel to capture this effect. The point here is that each estimator and regression equation tells a STORY. Think about it using words or pictures -- THEN, do the math.

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.