Gian Paulo Soave.

This project extends the methodology proposed by Fritche, Klein, and Reith (2021) to a panel context. The method allows to recover the fiscal shock from the second moments of the data. The project applies the methodology to a panel of 60 contries at quarterly frequency. Estimates of the fiscal mutipliers are in line with the previous literature for advanced economies in normal times, but not in uncertain times. The new results for developing and emerging countires suggest that, in uncertain times, multipliers can be close to zero.

This paper investigates the relationship between financial conditions, macroeconomic uncertainty, and business cycles in emerging markets. For a set of 15 emerging economies, using a model that allows shocks to affect second moments, we show that macroeconomic uncertainty, interpreted as a deterioration in the accuracy of forecasts of key macroeconomic variables, increases after some macroeconomic shocks under financial stress regime but not in normal times. We then show that a model for a sudden stop-prone small open-economy featuring a Fisherian debt-deflation mechanism and Epstein-Zin preferences generates responses similar to what we found in our empirical analysis. When the economy is sufficiently close to a binding constraint, the risk of a regime change induces a deterioration in the forecast accuracy. It happens because the economy behaves very differently across regimes, and regime-switching is affected by random shocks. Thus, in such regions, uncertainty responds endogenously to the state of the economy. However, macroeconomic volatility is mainly driven by exogenous shocks when the economy is sufficiently away from a binding constraint.

This project analyzes some algorithms for solving the decentralized equilibrium in multidimensional, occasionally binding Fisherian Models in which the collateral constraint is based on the value of some stock asset. This kind of model is numerically challenging because an extrapolation step is (usually) needed to compute the next period quantity of the asset. When there are available closed-form solutions, both Carrol's Endogenous Grid Method and Maliar and Maliar's Envelope Condition Method perform faster, yet with comparable accuracy as standard Euler Equation Methods. In models with complex collateral constraints, the set of available methods is substantially reduced, and the project presents a hybrid version of the Envelope Condition Method that is accurate and computationally cheaper than standard Euler Equation Methods

Empirical evidence has shown that fiscal multipliers - particularly spending multipliers - tend to be smaller in emerging markets compared to those estimated for advanced economies. This project aims to add to the literature in two dimensions: first, by using a Panel Bayesian Thresold VAR model for a set of 17 Emerging Economies, we complement the previous evidence showing that multipliers are lower (or even negative) whenever a measure of sovereign risk is high enough - according to the data. To better understand this result, we aim to evaluate the contribution of expansionary government spending shocks on the economy's volatility in bad times, comparing the results with those from situations in which the sovereign risk is relatively stable. 
Second, we argue that this seemingly puzzling piece of evidence obtained from the empirical analysis can be rationalized in a theoretical model for a small open economy in which the sovereign authority has an option to default on its debt. As a defaulting country is excluded from the international capital markets, the private sector faces an occasionally binding constraint in the financial markets that create systemic risk. As a consequence, rational expectation agents may increase precautionary savings under certain levels of default risk. In such an economy, Keynesian multipliers may arise whenever the sovereign default risk is low, but an expansionary fiscal policy in times of high indebtedness potentially leads to a crowding-out effect. The project, then, intends to analyze the costs and benefits of the stimulus policies in times of fiscal fragility.

Nonlinear DSGE models with endogenous state-dependent regime switching are quite costly to be solved due to the computation burdens associated with the grid methods commonly used. This paper presents a set of tools developed within the Julia Language that aims to reduce such costs. Julia Language has similar learning costs to Matlab Language and can be as fast as FORTRAN, or even C++. The tools include N-Dimensional interpolation and extrapolation functions - both linear and Delaunay methods -, Newton-type with dogleg strategy solvers for constrained and unconstrained nonlinear system of equations, and functions for integrating out the uncertainty associated with expected variables. Examples with applications to simple and more complex models are provided