Luiz Koodi Hotta
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Luiz Koodi Hotta has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Carlos Trucíos
Universidade Estadual de Campinas (UNICAMP)
8 shared publications Mauricio Zevallos
Department of Statistics, University of Campinas, Sao Paulo, Brazil
5 shared publications- Mauricio Zevallos
Departament of Statistics Universidade Estadual de Campinas (UNICAMP) Brazil
4 shared publications - Pedro Luiz Valls Pereira
Sao Paulo School of Economics - FGV and CEQEF- FGV
4 shared publications Luiz Candido de Souza Dias
Universidade Estadual de Campinas, Brasil
3 shared publications- Márcio Poletti Laurini
Faculty of Economics, Administration and Accounting of Ribeirão Preto, University of São Paulo, Ribeirão Preto 14040-905, SP, Brazil
3 shared publications
Latest Publications
- Open Access
- 28 July 2026
- Published by Springer Nature in Computational Statistics
- Vol. 41 (5) , 110
Robust estimation of Markov-switching GARCH models
The study of volatility is important in several areas of finance, and GARCH models have been widely used in the literature due to their ability to capture key stylised facts of financial time series. However, in some cases, financial time series exhibit structural changes in volatility dynamics, for which standard GARCH models may be inadequate; in such situations, Markov-switching GARCH models provide a more suitable framework. On the other hand, outliers are often present in empirical data. This paper shows that conventional estimators can be strongly affected by outliers and proposes an estimator that is more robust to their presence. The simulation results and empirical applications indicate that the proposed robust estimator is competitive and useful under contamination, but not uniformly superior to the QML-t estimator.
- 28 July 2026
- Published by Springer Nature in Computational Statistics
- Vol. 41 (5) , 110
Open AccessRobust estimation of Markov-switching GARCH models
The study of volatility is important in several areas of finance, and GARCH models have been widely used in the literature due to their ability to capture key stylised facts of financial time series. However, in some cases, financial time series exhibit structural changes in volatility dynamics, for which standard GARCH models may be inadequate; in such situations, Markov-switching GARCH models provide a more suitable framework. On the other hand, outliers are often present in empirical data. This paper shows that conventional estimators can be strongly affected by outliers and proposes an estimator that is more robust to their presence. The simulation results and empirical applications indicate that the proposed robust estimator is competitive and useful under contamination, but not uniformly superior to the QML-t estimator.
- Open Access
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Combining Covariance Forecasts for Large-Dimensional Portfolio Optimization
We study whether combining covariance matrix forecasts improves portfolio construction in largedimensional asset universes. Using daily U.S. equity data from 1980 to 2022, we evaluate ten covariance estimators-ranging from static shrinkage methods to dynamic multivariate volatility models-and five forecast combination strategies in minimum-variance portfolios of up to 1,000 assets. We assess performance under three portfolio specifications that progressively impose realistic implementation constraints, including long-only restrictions, turnover limits, and concentration bounds. In unconstrained portfolios, dynamic models achieve the lowest realized volatility but generate high turnover that substantially erodes net performance. Once realistic constraints are imposed, performance differences across estimators shrink considerably. Forecast combinations, particularly simple ones such as equal weighting, deliver robust performance consistently close to the best individual models. We formalize these findings through analytical results linking forecast dispersion to a diversification effect across models, including an MSE decomposition that connects pairwise Frobenius distances between forecasts to combination accuracy, and a minimax optimality result for equal weights. Our analysis of portfolio weight similarity across estimators reveals that implementation constraints compress differences in both portfolio composition and performance, helping explain why forecast combination is most valuable in realistic settings.
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCombining Covariance Forecasts for Large-Dimensional Portfolio Optimization
We study whether combining covariance matrix forecasts improves portfolio construction in largedimensional asset universes. Using daily U.S. equity data from 1980 to 2022, we evaluate ten covariance estimators-ranging from static shrinkage methods to dynamic multivariate volatility models-and five forecast combination strategies in minimum-variance portfolios of up to 1,000 assets. We assess performance under three portfolio specifications that progressively impose realistic implementation constraints, including long-only restrictions, turnover limits, and concentration bounds. In unconstrained portfolios, dynamic models achieve the lowest realized volatility but generate high turnover that substantially erodes net performance. Once realistic constraints are imposed, performance differences across estimators shrink considerably. Forecast combinations, particularly simple ones such as equal weighting, deliver robust performance consistently close to the best individual models. We formalize these findings through analytical results linking forecast dispersion to a diversification effect across models, including an MSE decomposition that connects pairwise Frobenius distances between forecasts to combination accuracy, and a minimax optimality result for equal weights. Our analysis of portfolio weight similarity across estimators reveals that implementation constraints compress differences in both portfolio composition and performance, helping explain why forecast combination is most valuable in realistic settings.
- Open Access
- 17 April 2026
- Published by Elsevier in SSRN Electronic Journal
Regime Dependence, Not Dimensionality: Forecasting Cryptocurrency Value-at-Risk
We study regime dependence in tail-risk forecasting for cryptocurrency markets. Using the large-dimensional Markov-switching factor model of Barigozzi and Massacci (2025), we produceone-step-ahead Value-at-Risk forecasts as quantiles of a predictive mixture of regime-specificempirical distributions, weighted by propagated smoothed regime probabilities. In a rolling out-of-sample exercise on 37 cryptocurrencies from 2022 to 2026, our estimator attains violationrates close to nominal levels and is the only model included in the 90% Model Confidence Setat the 1% quantile, significantly outperforming Gaussian and Student-𝑡 GARCH, MS-GARCH,time-invariant factor quantile regression, and rolling historical quantiles in pooled Diebold–Mariano tests. Our results indicate that the main limitation of existing high-dimensional methodsis not dimensionality, but the assumption of time-invariant factor–return relationships.
- 17 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRegime Dependence, Not Dimensionality: Forecasting Cryptocurrency Value-at-Risk
We study regime dependence in tail-risk forecasting for cryptocurrency markets. Using the large-dimensional Markov-switching factor model of Barigozzi and Massacci (2025), we produceone-step-ahead Value-at-Risk forecasts as quantiles of a predictive mixture of regime-specificempirical distributions, weighted by propagated smoothed regime probabilities. In a rolling out-of-sample exercise on 37 cryptocurrencies from 2022 to 2026, our estimator attains violationrates close to nominal levels and is the only model included in the 90% Model Confidence Setat the 1% quantile, significantly outperforming Gaussian and Student-𝑡 GARCH, MS-GARCH,time-invariant factor quantile regression, and rolling historical quantiles in pooled Diebold–Mariano tests. Our results indicate that the main limitation of existing high-dimensional methodsis not dimensionality, but the assumption of time-invariant factor–return relationships.
- Open Access
- 10 March 2025
- Published by Fundacao Getulio Vargas in Brazilian Review of Finance
- Vol. 23, e202503
Forecasting Bitcoin and Ethereum risk measures through MSGARCH models
Recent studies have suggested that more complex models than GARCH are better suited for forecasting cryptocurrency risk measures, such as Value-at-Risk and Expected Shortfall. Among these studies, some highlight the advantages of MSGARCH models over traditional GARCH models. While improvements over single-regime GARCH models have been observed by using MSGARCH, the literature has only focused on the MSGARCH specification proposed by Haas, Mittnik and Paolella (Journal of Financial Econometrics, 2004) overlooking several other well-established MSGARCH specification alternatives. In this paper, we illustrate that exploring alternative MSGARCH specifications can lead to improvements in risk measure performance, emphasizing the potential benefits of using several specifications.
- 10 March 2025
- Published by Fundacao Getulio Vargas in Brazilian Review of Finance
- Vol. 23, e202503
Open AccessForecasting Bitcoin and Ethereum risk measures through MSGARCH models
Recent studies have suggested that more complex models than GARCH are better suited for forecasting cryptocurrency risk measures, such as Value-at-Risk and Expected Shortfall. Among these studies, some highlight the advantages of MSGARCH models over traditional GARCH models. While improvements over single-regime GARCH models have been observed by using MSGARCH, the literature has only focused on the MSGARCH specification proposed by Haas, Mittnik and Paolella (Journal of Financial Econometrics, 2004) overlooking several other well-established MSGARCH specification alternatives. In this paper, we illustrate that exploring alternative MSGARCH specifications can lead to improvements in risk measure performance, emphasizing the potential benefits of using several specifications.
- Open Access
- 14 January 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (2) , 257
Out-of-Sample Predictability of the Equity Risk Premium
A large set of macroeconomic variables have been suggested as equity risk premium predictors in the literature. Acknowledging the different predictability of the equity premium in expansions and recessions, this paper proposes an approach that combines equity premium forecasts from two-state regression models using an agreement technical indicator as the observable state variable. A comprehensive out-of-sample forecast evaluation exercise based on statistical and economic loss functions demonstrates the superiority of the proposed approach versus combined forecasts from linear models or Markov switching models and forecasts from machine learning methods such as random forests and gradient boosting. The parsimonious state-dependent aspect of risk premium forecasts delivers large improvements in forecast accuracy. The results are robust to sub-period analyses and different investors’ risk aversion levels.
- 14 January 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (2) , 257
Open AccessOut-of-Sample Predictability of the Equity Risk Premium
A large set of macroeconomic variables have been suggested as equity risk premium predictors in the literature. Acknowledging the different predictability of the equity premium in expansions and recessions, this paper proposes an approach that combines equity premium forecasts from two-state regression models using an agreement technical indicator as the observable state variable. A comprehensive out-of-sample forecast evaluation exercise based on statistical and economic loss functions demonstrates the superiority of the proposed approach versus combined forecasts from linear models or Markov switching models and forecasts from machine learning methods such as random forests and gradient boosting. The parsimonious state-dependent aspect of risk premium forecasts delivers large improvements in forecast accuracy. The results are robust to sub-period analyses and different investors’ risk aversion levels.
- 23 December 2021
- Published by Taylor & Francis in Journal of Business & Economic Statistics
- Vol. 41 (1) , 40-52
Forecasting Conditional Covariance Matrices in High-Dimensional Time Series: A General Dynamic Factor Approach
Based on a General Dynamic Factor Model with infinite-dimensional factor space and MGARCH volatility models, we develop new estimation and forecasting procedures for conditional covariance matrices in high-dimensional time series. The finite-sample performance of our approach is evaluated via Monte Carlo experiments and outperforms the most alternative methods. This new approach is also used to construct minimum one-step-ahead variance portfolios for a high-dimensional panel of assets. The results are shown to match the results of recent proposals by Engle, Ledoit, and Wolf and achieve better out-of-sample portfolio performance than alternative procedures proposed in the literature.
- 23 December 2021
- Published by Taylor & Francis in Journal of Business & Economic Statistics
- Vol. 41 (1) , 40-52
Forecasting Conditional Covariance Matrices in High-Dimensional Time Series: A General Dynamic Factor Approach
Based on a General Dynamic Factor Model with infinite-dimensional factor space and MGARCH volatility models, we develop new estimation and forecasting procedures for conditional covariance matrices in high-dimensional time series. The finite-sample performance of our approach is evaluated via Monte Carlo experiments and outperforms the most alternative methods. This new approach is also used to construct minimum one-step-ahead variance portfolios for a high-dimensional panel of assets. The results are shown to match the results of recent proposals by Engle, Ledoit, and Wolf and achieve better out-of-sample portfolio performance than alternative procedures proposed in the literature.
- Open Access
- 30 November 2019
- Published by Instituto de Filosofia e Ciências Humanas - UNICAMP
- No. 27,p. 1
Teoria moderna de portfólio, paridade de risco e aplicações
Segundo a teoria moderna de portfólio, o melhor investimento é dado pelo portfólio de capitalização de mercado (PCM). Nessa pesquisa, é comparado o PCM com portfólios baseados em outras abordagens como a paridade de risco. Os resultados preliminares indicam que eles apresentam uma performance melhor. No entanto, é preciso levar em conta detalhes de implementação como custos transacionais, que tornam o PCM tão atrativo por sua simplicidade.
- 30 November 2019
- Published by Instituto de Filosofia e Ciências Humanas - UNICAMP
- No. 27,p. 1
Open AccessTeoria moderna de portfólio, paridade de risco e aplicações
Segundo a teoria moderna de portfólio, o melhor investimento é dado pelo portfólio de capitalização de mercado (PCM). Nessa pesquisa, é comparado o PCM com portfólios baseados em outras abordagens como a paridade de risco. Os resultados preliminares indicam que eles apresentam uma performance melhor. No entanto, é preciso levar em conta detalhes de implementação como custos transacionais, que tornam o PCM tão atrativo por sua simplicidade.
- Open Access
- 5 June 2019
- Published by Elsevier in SSRN Electronic Journal
Forecasting Conditional Covariance Matrices in High-Dimensional Time Series: A General Dynamic Factor Approach
Based on a General Dynamic Factor Model with infinite-dimensional factor space, we develop a new estimation and forecasting procedures for conditional covariance matrices in high-dimensional time series. The performance of our approach is evaluated via Monte Carlo experiments, outperforming many alternative methods. The new procedure is used to construct minimum variance portfolios for a high-dimensional panel of assets. The results are shown to achieve better out-of-sample portfolio performance than alternative existing procedures.
- 5 June 2019
- Published by Elsevier in SSRN Electronic Journal
Open AccessForecasting Conditional Covariance Matrices in High-Dimensional Time Series: A General Dynamic Factor Approach
Based on a General Dynamic Factor Model with infinite-dimensional factor space, we develop a new estimation and forecasting procedures for conditional covariance matrices in high-dimensional time series. The performance of our approach is evaluated via Monte Carlo experiments, outperforming many alternative methods. The new procedure is used to construct minimum variance portfolios for a high-dimensional panel of assets. The results are shown to achieve better out-of-sample portfolio performance than alternative existing procedures.
- Open Access
- 9 May 2019
- Published by MDPI AG in Econometrics
- Vol. 7 (2) , 19
Covariance Prediction in Large Portfolio Allocation
Many financial decisions, such as portfolio allocation, risk management, option pricing and hedge strategies, are based on forecasts of the conditional variances, covariances and correlations of financial returns. The paper shows an empirical comparison of several methods to predict one-step-ahead conditional covariance matrices. These matrices are used as inputs to obtain out-of-sample minimum variance portfolios based on stocks belonging to the S&P500 index from 2000 to 2017 and sub-periods. The analysis is done through several metrics, including standard deviation, turnover, net average return, information ratio and Sortino’s ratio. We find that no method is the best in all scenarios and the performance depends on the criterion, the period of analysis and the rebalancing strategy.
- 9 May 2019
- Published by MDPI AG in Econometrics
- Vol. 7 (2) , 19
Open AccessCovariance Prediction in Large Portfolio Allocation
Many financial decisions, such as portfolio allocation, risk management, option pricing and hedge strategies, are based on forecasts of the conditional variances, covariances and correlations of financial returns. The paper shows an empirical comparison of several methods to predict one-step-ahead conditional covariance matrices. These matrices are used as inputs to obtain out-of-sample minimum variance portfolios based on stocks belonging to the S&P500 index from 2000 to 2017 and sub-periods. The analysis is done through several metrics, including standard deviation, turnover, net average return, information ratio and Sortino’s ratio. We find that no method is the best in all scenarios and the performance depends on the criterion, the period of analysis and the rebalancing strategy.