Skip to main content Skip to navigation

Livia Silva Paranhos

About me

Hello, I am a PhD candidate in Economics and I work in the intersection of macroeconomics and econometrics. I am particularly interested in using machine learning techniques to analyse macroeconomic data, where my current work covers applications to inflation forecasting and the transmission of monetary policy. Prior to my PhD, I pursued Engineering at the École Centrale de Lyon, France, and the Federal University of Rio Grande do Sul, Brazil.


Job Market Paper

How Do Firms' Financial Conditions Influence the Transmission of Monetary Policy? A Nonparametric PerspectiveLink opens in a new window

[slidesLink opens in a new window]

Winner of the Dwyer Ramsey award for best paper presented at the SNDE Symposium by a graduate student

Abstract: How do monetary policy shocks affect firm investment? This paper provides new evidence on US non-financial firms and a novel non-parametric framework based on random forests. The key advantage of the methodology is that it does not impose any assumptions on how the effect of shocks varies across firms thereby allowing for general forms of heterogeneity in the transmission of shocks. My estimates suggest that there exists a threshold in the level of firm risk above which monetary policy is much less effective. Additionally, there is no evidence that the effect of policy varies with firm risk for the 75% of firms in the sample with higher risk. The proposed methodology is a generalization of local projections and nests many common local projection specifications, including linear and nonlinear.

Working Papers

Predicting Inflation with Recurrent Neural NetworksLink opens in a new window, R&R at the International Journal of Forecasting

Media coverage: Francis X. Diebold's blog No HesitationsLink opens in a new window

Abstract: This paper applies a recurrent neural network, the LSTM, to forecast inflation. This is an appealing model for time series as it processes each time step sequentially and explicitly learns dynamic dependencies. The paper also explores the dimension reduction capability of the model to uncover economically meaningful factors that can explain the inflation process. Results from an exercise with US data indicate that the estimated neural nets presents competitive, but not outstanding, performance against common benchmarks (including other machine learning models). The LSTM in particular is found to perform well at long horizons and during periods of uncertainty. Interestingly, LSTM-implied factors present high correlation with business cycle indicators, informing about the usefulness of such signals as inflation predictors. The paper also sheds light on the impact of network initialization and architecture on forecast performance.