Skip to content

Bayesian Machine Learning with PYMC3. Data from the Kyiv School of Economics.

Notifications You must be signed in to change notification settings

CarolMoore19/Business-Self-Sanctioning-of-Russia

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 

Repository files navigation

Business-Self-Sanctioning-of-Russia

Class Project for Bayesian Machine Learning, University of Virginia, Summer 2022

Authors: Elina Ribakova, Diana Morris, Carol Moore

Data was compiled by the Kyiv School of Economics (KSE) Leave Russia project.

See https://leave-russia.org/ to learn more about Leave Russia and the data.

This analysis and the views expressed are independent from KSE and the Leave Russia Project. Any errors are the responsibility of the authors.

Abstract

Following Russia’s full-scale invasion of Ukraine on February 24, 2022 many governments imposed severe financial and trade sanctions on Russia, including its access to global payments systems and the U.S. dollar, sovereign debt trading, commodity exports, and access to critical technology. There is also an emerging phenomenon of “self-sanctioning”: multinational companies exiting the Russian market, whether partial or fully, temporarily, or permanently, despite not being explicitly directed to do so by sanctions. In this analysis, we estimate a Bayesian multinomial logistic regression model to identify the factors associated with a companies' decisions to curtail business with Russia, and to what extent.

The database covers 2,479 companies from 55 industries and headquartered in 75 countries. The response variable is a company’s operational status with Russia as of July 31, 2022. "Leave": explicit action to close branches in Russia or stop orders to Russia; "Stay": take no action; "Wait": hold off planned investment or scale back operations. Features included in the regression were industry (aggregated to 7 to 9 categories), presence of official sanctions by the company's home country, global revenues, and percentage of revenues from Russia. Due to missing values, the model with %revenue from Russia included 484 observations and the version of the model without this variable had 1,157 observations.

Markov Chain Monte Carlo estimates of the posterior slope and intercept distributions with priors ~N(0,1) and N=1,157 found that companies in the healthcare/pharma industry had lower log odds of leaving than the reference industry, consumer cyclicals (94% HDI -2.8 to -.46). Official sanctions imposed by the home country was correlated with increased log odds of leaving (.091 to 2.33 94% HDI) and reduced log odds of staying (-2.68 to -.50), while the association with waiting could be postive or negative (-.68 to 1.55).

About

Bayesian Machine Learning with PYMC3. Data from the Kyiv School of Economics.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published