Avaliação do uso de modelos GARCH para criptomoedas : uma análise das dez criptomoedas mais líquidas entre 2018-2022
Resumo
Resumo: Este trabalho tem como objetivo identificar o modelo GARCH que melhor se aplica para cada um dos dez criptoativos estudados: Bitcoin, Ethereum, Theter, BNB, XRP, Cardano, Dogecoin, Tron, Litecoin e ChainLink, no período de janeiro de 2018 a dezembro de 2022. Para isso, foram utilizados modelos ARIMA para a média condicional e modelos GARCH de primeira ordem para a variância condicional. Diferentes modelos GARCH foram analisados, incluindo GARCH, EGARCH, GJRGARCH, Power ARCH (APARCH), Component GARCH (CSGARCH) e IGARCH, e o modelo ótimo foi escolhido com base nos critérios de informação AIC, BIC e RMSE. Os dados utilizados foram obtidos publicamente online, através da biblioteca crypto2 do software gratuito R, com base no site CoinMarketCap. Foram investigadas as estatísticas descritivas de cada um dos criptoativos analisados e testada a estacionariedade dos seus log-retornos. O modelo ARIMA que melhor se aplicou para cada um dos criptoativos foi definido e foram testados efeitos ARCH para cada um desses modelos ARIMA. Finalmente, foi determinado o modelo GARCH com melhor goodness-of-fit in-sample e out-of-sample. Os resultados mostraram que os modelos GARCH testados se aplicaram de forma diferente para cada um dos criptoativos estudados, indicando que é necessário modelar o comportamento dessas criptomoedas de forma individual. Abstract: This work aims to identify the GARCH model that best fits each of the ten crypto assets studied: Bitcoin, Ethereum, Theter, BNB, XRP, Cardano, Dogecoin, Tron, Litecoin, and ChainLink, in the period from January 2018 to December 2022. ARIMA models were used for conditional mean and first order GARCH models for conditional variance. Different GARCH models were analyzed, including GARCH, EGARCH, GJRGARCH, Power ARCH (APARCH), Component GARCH (CSGARCH), and IGARCH, and the optimal model was chosen based on the information criteria AIC, BIC and RMSE. The data used was obtained publicly online through the crypto2 library of the free software R, based on the CoinMarketCap website. The descriptive statistics of each crypto were investigated, and the stationarity of their log returns was tested. The ARIMA model that best applied to each crypto asset was defined, and ARCH effects were tested for each of these ARIMA models. Finally, the GARCH model with the best goodness-of-fit was determined for in-sample and out-of-sample. The results showed that the GARCH models tested applied differently to each of the crypto assets studied, indicating that it is necessary to model the behavior of these cryptocurrencies individually.
Collections
- Dissertações [63]