Publications
Despite their high predictive performance, random forest and gradient boosting are often considered as black boxes which has raised concerns from practitioners and regulators. As an alternative, we suggest using partial linear models that are inherently interpretable. Specifically, we propose to combine parametric and non-parametric functions to accurately capture linearities and non-linearities prevailing between dependent and explanatory variables, and a variable selection procedure to control for overfitting issues. Estimation relies on a two-step procedure building upon the double residual method. We illustrate the predictive performance and interpretability of our approach on a regression problem.
What is the role of income polarization for explaining differentials in public funding of education? To answer this question, we provide a new theoretical modelling for the income distribution that can directly monitor income polarization. It leads to a new income polarization index where the middle class is represented by an interval. We implement this distribution in a political economy model with endogenous fertility and public/private educational choices. We show that when households vote on public schooling expenditures, polarization matters for explaining disparities in public education funding across communities. Using micro-data covering two groups of school districts, we find that both income polarization and income inequality affect public school funding with opposite signs whether there exist a Tax Limitation Expenditure (TLE) or not.
This paper introduces an autoregressive conditional beta (ACB) model that allows regressions with dynamic betas (or slope coefficients) and residuals with GARCH conditional volatility. The model fits in the (quasi) score-driven approach recently proposed in the literature, and it is semi-parametric in the sense that the distributions of the innovations are not necessarily specified. The time-varying betas are allowed to depend on past shocks and exogenous variables. We establish the existence of a stationary solution for the ACB model, the invertibility of the score-driven filter for the time-varying betas, and the asymptotic properties of one-step and multistep QMLEs for the new ACB model. The finite sample properties of these estimators are studied by means of an extensive Monte Carlo study. Finally, we also propose a strategy to test for the constancy of the conditional betas. In a financial application, we find evidence for time-varying conditional betas and highlight the empirical relevance of the ACB model in a portfolio and risk management empirical exercise.
Lower tariffs typically raise productivity, production, and trade, increasing the benefits from building infrastructure. Infrastructure spending by governments should therefore increase after countries open up to trade. I test this hypothesis empirically using a trade reform in India and find that a 1 percentage point reduction in tariffs increased states’ infrastructure spending by 0.5% between 1991 and 2001. To understand the mechanisms behind my empirical findings, I develop and calibrate a multi-region model of international trade, private capital accumulation, and infrastructure spending, in which each government chooses such spending to maximize their state’s welfare. I find if governments choose infrastructure following the reform optimally, infrastructure would have increased by 60% on average. The actual increase, based on my empirical findings, was about 29%. Counterfactual exercises show that raising aggregate infrastructure towards its optimal following the trade reform will result in state GDP to increase by 7% points on average.
Following Bergeaud et al. (2022), we construct a new measure of proximity between industrial sectors and public research laboratories. Using this measure, we explore the underlying network of knowledge linkages between scientific fields and industrial sectors in France. We show empirically that there exists a significant negative correlation between the geographical distance between firms and laboratories and their scientific proximity, suggesting strongly localized spillovers. Moreover, we uncover some important differences by field, stronger than when using standard patent-based measures of proximity.
This article synthesizes the insights gained through presentations and discussions at the 2023 IEEE Workshop on Norbert Wiener in the 21st Century (21CW2023), which focused on “The Future of Work in the Age of Automation.” Hosted at Purdue University, this interdisciplinary convening of technologists, social scientists, and humanists explored the impacts of automation on labor, drawing on Wiener’s legacy of insights as a backdrop to examine the technologically mediated future we face in coming decades. The workshop presented a rare opportunity to reflect critically on these issues at a pivotal moment in human and technological history, and to elicit underappreciated dimensions. Areas of focus include: the qualitative and quantitative losses associated with automation and AI, the impacts automation has for questions about the meaningfulness of work, the challenges we face related to uncertainty and lack of predictability in technological advancement, and the opportunities that exist for centering human values and agency in these conversations. While acknowledging many items for concern in the context of automation in the future of work, such as the domination of economic narratives, a potential loss of qualitative texture, and the neglect of certain issues key to human identity, the authors conclude by offering optimistic visions—or calls—for redefining value and labor, preserving human agency, and embracing creative problem-solving.
Mexican cities along the US-Mexico border, especially Cd. Juarez, became notorious due to high femicide rates supposedly associated with maquiladora industries and the NAFTA. Nonetheless, statistical evaluation of data from 1990 to 2012 shows that their rates are consistent with other Mexican cities’ rates and tend to fall with increased employment opportunities in maquiladoras. Femicide rates in Cd. Juarez are in most years like rates in Cd. Chihuahua and Ensenada and, as a share of overall homicide rates, are lower than in most cities evaluated. These results challenge conventional wisdom and most of the literature on the subject.
This paper highlights how technology can contribute to reaching the 2015 Paris Agreement goals of net zero carbon dioxide (CO2) emissions and global warming below 2°C in 2100. It uses the Advanced Climate Change Long-term model (ACCL), particularly adapted to quantify the consequences of energy price and technology shocks on CO2 emissions, temperature, climate damage and Gross Domestic Product (GDP). The simulations show that without climate policies the warming may be +5°C in 2100, with considerable climate damage. An acceleration in ‘usual’ technical progress not targeted at reducing CO2- even worsens global warming and climate damage. According to our estimates, the world does not achieve climate goals in 2100 without ‘green’ technologies. Intervening only via energy prices, e.g. a carbon tax, requires challenging hypotheses of international coordination and price increase for polluting energies. We assess a multi-lever climate strategy combining energy efficiency gains, carbon sequestration, and a decrease of 3% per year in the relative price of ‘clean’ electricity with a 1 to 1.5% annual rise in the relative price of polluting energy sources. None of these components alone is sufficient to reach climate objectives. Our last and most important finding is that our composite scenario achieves the climate goals.
Wine experts' ratings provide quality information and reduce the information asymmetry for the consumer. We hypothesize that consumers' ratings will come to dominate expert ratings in the wine expertise market. We employ a hedonic regression framework on the attributes of 36,970 French red wines to determine the relative impacts of expert and Vivino community ratings on wine prices. Average consumer ratings are found to have a larger effect on price than expert scores. These results are found to be robust to outliers and the general conclusion that peers matter more than experts holds when we exclude the top-end wines.
