Análise do prêmio de risco implícito e realizado brasileiro: duas abordagens forwardlooking
Autores
Costa, Maurício Fiori Gomes da
Orientador
Co-orientadores
Silva, André Castro
Citações na Scopus
Tipo de documento
Dissertação
Data
2025
Resumo
Este trabalho tem como objetivo principal analisar a dinâmica do prêmio de risco implícito do
mercado acionário brasileiro entre os anos de 1998 e 2024, comparando duas metodologias
forward-looking de estimação baseadas no modelo de Gordon: uma abordagem tradicional de
dividendos descontados (DDM) e uma abordagem baseada em lucros (EB), que assume, como
hipótese central, a equivalência entre o retorno sobre o patrimônio líquido (ROE) e o custo de
capital (k). A relevância do estudo reside na tentativa de identificar qual das duas metodologias
consegue capturar de forma mais eficiente a percepção de risco dos agentes econômicos.
Ambas as metodologias são aplicadas ao índice Ibovespa com dados trimestrais extraídos da
Bloomberg e do Banco Central do Brasil. A estimativa do prêmio implícito é obtida por meio da
diferença entre a taxa de retorno exigida (k), calculada conforme cada modelo, e a taxa livre de
risco representada pelo CDI. A taxa de crescimento de único estágio é estimada com base na
relação ROE ∗ (1 − payout), o que pressupõe reinvestimento total dos lucros retidos à mesma
taxa histórica de retorno. No modelo earnings-based, adota-se a hipótese simplificadora de que o
ROE é igual ao custo de capital ROE = k, o que permite inferir diretamente o earning yield
(E/P) como medida do retorno exigido.
Os resultados revelam importantes diferenças entre os dois modelos. O prêmio implícito estimado
via DDM apresentou média de 5,95% e volatilidade σ = 6, 64%, com comportamento
marcadamente sensível a crises e mudanças nas expectativas de crescimento. Já o IRP calculado
pelo modelo earnings-based mostrou-se sistematicamente próximo de zero, com média de -1,09%
e menor dispersão σ = 2, 11%, o que indica uma possível subestimação do prêmio de risco
exigido pelos investidores. O prêmio de risco realizado (retorno ex post do Ibovespa em relação
à taxa livre de risco) apresentou média praticamente nula (0,23%) e comportamento errático,
com desvio-padrão de 15,36%.
A discrepância entre os prêmios implícitos estimados pelas duas abordagens pode ser explicada
pelas premissas estruturais adotadas. A suposição de que ROE = k impõe uma restrição severa
ao modelo earnings-based, desconsiderando a possibilidade de retornos econômicos excedentes
ou vantagens competitivas. Isso é especialmente problemático em mercados emergentes, onde
a heterogeneidade entre empresas, os choques macroeconômicos e a volatilidade institucional
comprometem a validade de hipóteses de equilíbrio. Em contrapartida, o modelo DDM, mostrouse
mais sensível às mudanças na percepção de risco e nas expectativas dos agentes, revelando
maior aderência empírica às flutuações do mercado brasileiro.
Do ponto de vista preditivo, nenhuma das abordagens apresentou capacidade significativa de
antecipar o prêmio de risco realizado. As regressões lineares simples entre IRPs e retornos subsequentes apresentaram coeficientes baixos e sem significância estatística, com R² inferiores
a 4%. No entanto, a análise de janelas móveis indicou que, em períodos de crise — como 1999,
2008, 2015 e 2020 —, a relação entre os prêmios implícitos e o retorno realizado se intensificou,
o que sugere que o IRP pode funcionar como termômetro da aversão ao risco em momentos de
estresse sistêmico, mesmo que não seja um previsor robusto de retornos futuros.
A principal contribuição desta dissertação está na demonstração de que diferentes metodologias
de estimação do prêmio de risco implícito produzem resultados significativamente distintos
e que, em particular, o modelo de dividendos apresenta maior sensibilidade às variações nas
expectativas de mercado. Em contraste, a abordagem baseada em lucros, ao adotar premissas
mais rígidas, tende a suavizar excessivamente as estimativas, comprometendo sua utilidade
como indicador dinâmico. A pesquisa reforça, portanto, a importância de considerar as hipóteses
e limitações inerentes a cada modelo na escolha da metodologia a ser utilizada em análises
financeiras, principalmente em mercados instáveis e menos eficientes.
Como propostas de extensões futuras, sugere-se a incorporação de modelos de crescimento em
múltiplos estágios, a utilização de projeções de analistas de mercado para estimar o crescimento
dos fluxos, bem como o aperfeiçoamento do tratamento de outliers nos dados contábeis e
a inclusão de recompra de ações como componente do retorno ao acionista. Tais melhorias
metodológicas são essenciais para o desenvolvimento de modelos mais robustos e aderentes à
realidade dos mercados emergentes, permitindo estimativas mais precisas e úteis para gestores,
reguladores e tomadores de decisão.
This study aims to analyze the dynamics of the implied equity risk premium in the Brazilian stock market between 1998 and 2024, comparing two forward-looking estimation methodologies based on the Gordon model: a traditional dividend discount approach (DDM) and an earnings-based (EB) approach, which assumes as a central hypothesis the equivalence between return on equity (ROE) and the cost of capital (k). The relevance of the study lies in identifying which of the two methodologies more efficiently captures the risk perception of economic agents. Both methodologies are applied to the Ibovespa index using quarterly data extracted from Bloomberg and the Central Bank of Brazil. The implied premium is estimated as the difference between the required rate of return (k), calculated according to each model, and the risk-free rate represented by the CDI. The single-stage growth rate is estimated based on the relation ROE ×(1−payout), which assumes full reinvestment of retained earnings at the historical rate of return. In the earnings-based model, a simplifying assumption is adopted: that ROE equals the cost of capital (ROE = k), which allows direct inference of the earning yield (E/P) as a proxy for the required return. The results reveal important differences between the two models. The implied premium estimated via the DDM presented an average of 5.95% and volatility σ = 6.64%, with behavior notably sensitive to crises and shifts in growth expectations. On the other hand, the IRP estimated using the earnings-based model remained systematically close to zero, with an average of -1.09% and lower dispersion (σ = 2.11%), suggesting a possible underestimation of the risk premium demanded by investors. The realized risk premium (ex post return of the Ibovespa relative to the risk-free rate) had an almost null average (0.23%) and erratic behavior, with a standard deviation of 15.36%. The discrepancy between the implied premiums estimated by the two approaches can be explained by their structural assumptions. The assumption that ROE = k imposes a strict constraint on the earnings-based model, disregarding the possibility of excess economic returns or competitive advantages. This is especially problematic in emerging markets, where firm heterogeneity, macroeconomic shocks, and institutional volatility undermine the validity of equilibrium-based assumptions. In contrast, the DDM model proved to be more responsive to changes in risk perception and expectations, showing greater empirical adherence to fluctuations in the Brazilian market. From a predictive standpoint, neither approach showed significant ability to anticipate the realized risk premium. Simple linear regressions between IRPs and subsequent returns yielded low and statistically insignificant coefficients, with R² values below 4%. However, the rolling window analysis indicated that during crisis periods — such as in 1999, 2008, 2015, and 2020 — the relationship between implied premiums and realized returns intensified, suggesting that the IRP may serve as a barometer of risk aversion during systemic stress, even if it is not a robust predictor of future returns. The main contribution of this dissertation is to demonstrate that different methodologies for estimating the implied equity risk premium yield significantly different results and that, in particular, the dividend-based model exhibits greater sensitivity to changes in market expectations. In contrast, the earnings-based approach, by adopting more rigid assumptions, tends to overly smooth the estimates, compromising its usefulness as a dynamic indicator. The research thus reinforces the importance of considering the assumptions and limitations inherent to each model when choosing the methodology for financial analysis, especially in unstable and less efficient markets. As suggestions for future extensions, the study proposes the incorporation of multi-stage growth models, the use of market analysts’ projections to estimate future cash flow growth, improved treatment of accounting outliers, and the inclusion of share buybacks as a component of shareholder return. These methodological improvements are essential for the development of more robust models aligned with the realities of emerging markets, enabling more accurate and useful estimates for managers, regulators, and decision-makers.
This study aims to analyze the dynamics of the implied equity risk premium in the Brazilian stock market between 1998 and 2024, comparing two forward-looking estimation methodologies based on the Gordon model: a traditional dividend discount approach (DDM) and an earnings-based (EB) approach, which assumes as a central hypothesis the equivalence between return on equity (ROE) and the cost of capital (k). The relevance of the study lies in identifying which of the two methodologies more efficiently captures the risk perception of economic agents. Both methodologies are applied to the Ibovespa index using quarterly data extracted from Bloomberg and the Central Bank of Brazil. The implied premium is estimated as the difference between the required rate of return (k), calculated according to each model, and the risk-free rate represented by the CDI. The single-stage growth rate is estimated based on the relation ROE ×(1−payout), which assumes full reinvestment of retained earnings at the historical rate of return. In the earnings-based model, a simplifying assumption is adopted: that ROE equals the cost of capital (ROE = k), which allows direct inference of the earning yield (E/P) as a proxy for the required return. The results reveal important differences between the two models. The implied premium estimated via the DDM presented an average of 5.95% and volatility σ = 6.64%, with behavior notably sensitive to crises and shifts in growth expectations. On the other hand, the IRP estimated using the earnings-based model remained systematically close to zero, with an average of -1.09% and lower dispersion (σ = 2.11%), suggesting a possible underestimation of the risk premium demanded by investors. The realized risk premium (ex post return of the Ibovespa relative to the risk-free rate) had an almost null average (0.23%) and erratic behavior, with a standard deviation of 15.36%. The discrepancy between the implied premiums estimated by the two approaches can be explained by their structural assumptions. The assumption that ROE = k imposes a strict constraint on the earnings-based model, disregarding the possibility of excess economic returns or competitive advantages. This is especially problematic in emerging markets, where firm heterogeneity, macroeconomic shocks, and institutional volatility undermine the validity of equilibrium-based assumptions. In contrast, the DDM model proved to be more responsive to changes in risk perception and expectations, showing greater empirical adherence to fluctuations in the Brazilian market. From a predictive standpoint, neither approach showed significant ability to anticipate the realized risk premium. Simple linear regressions between IRPs and subsequent returns yielded low and statistically insignificant coefficients, with R² values below 4%. However, the rolling window analysis indicated that during crisis periods — such as in 1999, 2008, 2015, and 2020 — the relationship between implied premiums and realized returns intensified, suggesting that the IRP may serve as a barometer of risk aversion during systemic stress, even if it is not a robust predictor of future returns. The main contribution of this dissertation is to demonstrate that different methodologies for estimating the implied equity risk premium yield significantly different results and that, in particular, the dividend-based model exhibits greater sensitivity to changes in market expectations. In contrast, the earnings-based approach, by adopting more rigid assumptions, tends to overly smooth the estimates, compromising its usefulness as a dynamic indicator. The research thus reinforces the importance of considering the assumptions and limitations inherent to each model when choosing the methodology for financial analysis, especially in unstable and less efficient markets. As suggestions for future extensions, the study proposes the incorporation of multi-stage growth models, the use of market analysts’ projections to estimate future cash flow growth, improved treatment of accounting outliers, and the inclusion of share buybacks as a component of shareholder return. These methodological improvements are essential for the development of more robust models aligned with the realities of emerging markets, enabling more accurate and useful estimates for managers, regulators, and decision-makers.
Palavras-chave
Titulo de periódico
Texto completo
Título de Livro
URL na Scopus
Sinopse
Objetivos de aprendizagem
Idioma
Português
Notas
Membros da banca
Silva, André Castro
Área do Conhecimento CNPQ
CIENCIAS SOCIAIS APLICADAS::ECONOMIA::METODOS QUANTITATIVOS EM ECONOMIA::METODOS E MODELOS MATEMATICOS, ECONOMETRICOS E ESTATISTICOS
CIENCIAS SOCIAIS APLICADAS::ECONOMIA::ECONOMIA MONETARIA E FISCAL::INSTITUICOES MONETARIAS E FINANCEIRAS DO BRASIL
CIENCIAS SOCIAIS APLICADAS::ECONOMIA::ECONOMIA MONETARIA E FISCAL::INSTITUICOES MONETARIAS E FINANCEIRAS DO BRASIL
