Taxa de demissão na pandemia: uma análise de países europeus via modelos lineares generalizados inflacionados de zeros

Abstract

This study aimed to understand which variables are associated with the employee layoff rate during the COVID-19 pandemic in European countries. We analyzed microdata from the World Bank's 2020 and 2021 Business Surveys, based on questionnaires completed by owners or senior leaders of 3,779 companies across ten European countries, which included specific questions related to the pandemic. The response variable in this study was the number of employees laid off during the pandemic. Explanatory variables included the country, the economic sector of the company, annual revenue, business age, and binary variables indicating whether the company received government assistance, increased its online presence, or implemented innovations during the period. We used the total number of employees as an offset to model the layoff rate. The employed methods were Poisson and Negative Binomial regression models, both with and without zero-inflation, since 87.6% of the surveyed companies reported no layoffs during the period. The model with the lowest AIC (Akaike Information Criterion) was the Zero-Inflated Negative Binomial Regression, including the variables country, industry, and log(revenue), with the latter two being significant in all fitted models. The estimated coefficients indicated that an increase in company revenue is associated with a decrease in layoffs, showing an inverse relationship. Furthermore, companies in the retail sector tended to have higher layoff rates compared to those in the manufacturing sector, suggesting higher layoff vulnerability during the pandemic. The application presented in this study highlights the potential of the Poisson and Negative Binomial regression model classes in analyzing count data and estimating associated rates.