Gustavo A. Caffaro

View the Project on GitHub

Versión Español aquí

About me

I am a highly motivated Economics student in the pursuit of new solutions to the world’s (in particular my country’s) most challenging problems. My academic and professional interests lie at the intersection of artificial intelligence and public finance. My work experience includes the domains of debt management, portfolio risk management, and fiscal analysis and statistics.

CV

You can find the latest version of my CV here (Oct 2020)

Research & Publications

Text Mining Algorithms: an Application to the Sovereign Differential of the Dominican Republic (Master’s Thesis)

Abstract

This paper presents a text mining algorithm that analyzes the relationship between the sovereign differential of the Dominican Republic and emerging market (EM) financial news. This algorithm is based on sentiment analysis methods and the topic modeling algorithm known as Latent Dirichlet Allocation (LDA). The main finding of this research is that the indicator based on the sentiment of EM financial news has a -0.6214 correlation with the sovereign differential of the Dominican Republic. Additionally, the results show that by grouping news by topics, we are able to identify the importance of some topics during periods of greater changes. This finding suggests a special relationship between these topics and the sovereign differential of the Dominican Republic.

Optimal FAVAR Model for Short Term Forecasting: An application to the Dominican Republic Caffaro, G. & Pérez, J., Research Document Series, Nov 2018, Dominican Republic Ministry of Finance

Abstract

This research identifies a “best” Factor-Augmented Vector Autoregressive (FAVAR) model used to forecast inflation and economic activity (IMAE) in the Dominican Republic. An optimization procedure is used for finding the “best” variables which contain the most forecast information. The model is compared against a VAR and a base FAVAR using the criteria of Root Mean Squared Error (RMSE). When compared against the VAR, the results show that the “best” FAVAR model improves forecasts in about 12.9% for the IMAE and 7.2% for inflation, while it displays a 8.9% and 14.7% reduction in the RMSE when compared to the base FAVAR model.

Download here (spanish)

Articles