Research
2024
- Automatic Identification under Polynomial Moment Restrictions with an Application to Social NetworksLuan Borelli, Guilherme Exel, and Marcelo MoreiraWork in Progress, 2024
We provide necessary and sufficient conditions for the global identification of parameters in models defined by polynomial-in-parameters unconditional moment restrictions. This broad class of models encompasses most simultaneous equations models, panel data models, time series models, and even certain nonparametric models under discrete instruments or regressors. Our conditions are purely algebraic, enabling the automatic, exact, and analytical characterization of identifying restrictions through computer algebra. The results we present provide a practical device for algorithmically proving (or disproving) identification and uncovering potentially interpretable identifying restrictions. We illustrate the practical relevance of our results with an application to social network models under unknown network structures. When network structures are unknown, social network models become highly nonlinear polynomials in parameters, posing significant challenges for characterizing identification—an issue that has remained underexplored in the social networks literature until recently. We show how our results can be applied to analyze identification in these contexts by proving identification for specific cases of the symmetric connections model of Jackson and Woinsky (2003) and the geometrically-distanced-in-a-circle model of Blume et al. (2011), which had previously only been demonstrated to be set identified up to a finite set of values.
- Global Optimization for Continuously Updating EstimatorsLuan Borelli, Guilherme Exel, Marcelo Moreira, Whitney Newey, and Mahrad SharifvaghefiWork in Progress, 2024
The Continuously Updating GMM Estimator (CUE) presents a challenge for numerical optimization algorithms due to the structure of its objective function. In weakly-identified models, it often presents multiple local minima and flat sections, rendering conventional methods such as gradient descent and grid-search inadequate. In this paper, we establish an equivalence between the problem of finding all critical points of the CUE objective and a generalized form of eigenvalue problem. For models defined by linear moment equalities, this result leads to a practical method for simultaneously computing all critical points. The only numerical step required in this approach is the computation of eigenvalues, which is accomplished by performant algorithms from numerical linear algebra. This makes the method fast, precise, and easy to implement.
- Global Optimization for Nonlinear Continuously Updating EstimatorsLuan Borelli, Guilherme Exel, Marcelo Moreira, and Mahrad SharifvaghefiWork in Progress, 2024
- Optimal Identifying Instrument LearningLuan Borelli, Guilherme Exel, and Marcelo MoreiraWork in Progress, 2024
2022
- Forecasting Brazilian GDP under Fiscal Foresight with a Noncausal Fiscal VARLuan Borelli, and Christian VonbunForthcoming at Brazilian Review of Econometrics, 2022
This paper stems from my time at the Institute for Applied Economics Research (Ipea) during my undergraduate years.
Due to fiscal foresight, standard fiscal VAR models are inherently susceptible to issues of nonfundamentalness and noncausality, which can result in invalid estimates. While these problems have been extensively addressed in the fiscal literature, they have largely been overlooked in Brazilian fiscal VAR studies. To address this gap, we estimate a noncausal fiscal VAR model for Brazil—an alternative specification that may correct these issues—and use it to forecast Brazilian GDP. The results show that the noncausal VAR model outperforms the standard purely causal VAR in terms of forecasting performance, particularly when considering the typical Brazilian fiscal VAR dataset. This suggests that fiscal expectations may play a crucial role in shaping the dynamics of Brazilian GDP.
- Do Macroeconomic Laws Exist?Luan BorelliValor Econômico, 2022
A book review of Kevin Hoover’s The Methodology of Empirical Macroeconomics, published in Valor Econômico, Brazil’s leading financial newspaper. The review is written in Portuguese, and the title provided here is a free translation of the original.
2021
- The Macroeconomics of Epidemics: Interstate Heterogeneity in BrazilLuan Borelli, and Geraldo Sandoval GóesEconomiA, 2021
This paper stems from my time at the Institute for Applied Economics Research (Ipea) during my undergraduate years. A VoxEU column presenting its results to a broader audience is available here. This paper previously circulated as a working paper in CEPR’s Covid Economics Series and, in Portuguese, as an Ipea Discussion Paper.
We applied the SIR-macro model proposed by Eichenbaum et al. (2020) in its complete version to comparatively study the interaction between economic decisions and COVID-19 epidemics in five different Brazilian states: São Paulo (SP), Amazonas (AM), Ceará (CE), Rio de Janeiro (RJ), and Pernambuco (PE). Our goal was to analyze qualitatively how the main intrinsic differences of each of these states could affect the epidemic dynamics and its consequences. We computed and compared the model for each of the states, both in competitive equilibrium and under optimal containment policy adoption, and analyzed the implications of optimal policy adoption. We concluded that the intrinsic characteristics of the five different states could imply relevant differences in the general dynamics of the epidemic, in the optimal containment policies, in the effect of the adoption of these policies, and the severity of the economic recessions. One year after the original ex-ante calibration, we evaluated the death toll and economic recession predicted by the model comparing it against real data. The model predictions showed to be qualitatively sufficient to anticipate the size of the pandemic risk that later materialized in Brazil.
- Lack of Coordination Between Federal and State Levels in Combatting COVID-19 in BrazilLuan Borelli, and Geraldo Sandoval GóesCadernos ENAP, 2021National School of Public Administration (ENAP)
This paper stems from my time at the Institute for Applied Economics Research (Ipea) during my undergraduate years. It is the result of the project Implications of the Lack of Coordination Between Federal and State Levels in Managing COVID-19 Pandemic Response Policies in Brazil, awarded with a R$ 15,000.00 research grant by the National School of Public Administration (ENAP). The project was ranked 5th out of 228 research projects submitted. The paper is written in Portuguese, and the title provided here is a free translation of the original.
2020
- Interstate Heterogeneity and Combatting COVID-19 in BrazilLuan Borelli, and Geraldo Sandoval GóesCEPR VoxEU, 2020
A VoxEU column presenting the results of my work The Macroeconomics of Epidemics: Interstate Heterogeneity in Brazil to a broader audience. This column was written prior to the paper’s publication, while it circulated as a working paper in CEPR’s Covid Economics Series and, in Portuguese, as an Ipea Discussion Paper. This column stems from my time at the Institute for Applied Economics Research (Ipea) during my undergraduate years.
Brazil has faced great difficulties in controlling the COVID-19 epidemic, having become the world’s epicentre of the coronavirus pandemic and recently reaching 50,000 fatalities. This column argues that the great heterogeneities between states in Brazil, together with difficulties in political coordination, may have shaped these consequences. Looking at five states, it investigates whether certain differences in the states’ intrinsic characteristics may have influenced the dynamics of the local epidemic. Governments may need to consider local conditions and adopt heterogeneous containment policies.