Ensaios sobre seleção de variáveis e impulso-resposta em estruturas a termo de spreads de crédito soberanos

• Autor(a) principal: Candido, Guilherme Amaral
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: https://hdl.handle.net/10438/32906
• topic: Term structure modeling; Credit spreads; Nelson Siegel decomposition; Selection of variables; Structural shocks; Modelagem de estrutura a termo; Spreads de crédito; Decomposição de Nelson Siegel; Seleção de variáveis; Choques estruturais; Economia; Créditos - Avaliação de riscos; Dívida pública; Risco país; Taxas de juros

This thesis seeks to identify the determinants of variations in the term structures of Latin American offshore sovereign debt spreads, differentiating between global and idiosyncratic movements. In particular, I seek to answer what are the determinants of these term structures and what are the expected effects of structural shocks. In the first chapter, I analyze different models, with Nelson Siegel latent factors driven by macro-finance state variables, to describe the offshore term structure of the Brazilian sovereign credit spreads, from 2004 to 2020. The models rely on combinations of different indicators, which I select using three methods: the sorting by the explanatory capacity of each one in individual regressions; Lasso selection; and genetic algorithm. Models that combine external risk appetite variables with idiosyncratic variables in Brazil offer the best explanatory power. In the second chapter, I extend the discussion to identify the effects of structural shocks on macro variables in the offshore term structures of other Latin American sovereign issuers, including: Chile, Peru, Colombia and Mexico. This time I extract credit spreads from Credit Default Swaps (CDS) prices, in order to ensure data uniformity and compatibility. For comparison sake, I revisit the Brazilian term structure as well. Among the term structure drivers, I consider three common external and three individual local indicators, and identify shocks using structural VAR models and Local Projections. In addition, I compare the results with narrative shocks. I find that Latin American term structures are strongly driven by external factors, responding in similar ways to structural shocks, irrespectively of the identification method that I use.