Job Market Paper
Evaluating Alternative Designs for Carbon Border Adjustment Mechanisms [Current Draft]
This paper evaluates the Carbon Border Adjustment Mechanism (CBAM) as a potential tool to mitigate carbon leakage, with its design varying based on the inclusion of export subsidies and discrimination across trading partners. To this end, I adopt a quantitative multi-country, multi-industry trade model with climate externalities and abatement. I provide a novel theoretical decomposition of the welfare effects associated with carbon pricing in open economies, underscoring the incidence of the home country's carbon tax on foreign residents as a vital welfare channel. The welfare decomposition reveals ambiguous welfare effects when export subsidies are incorporated in the CBAM, as they mitigate leakage but reduce the incidence of home's carbon tax on foreign residents. The model is then mapped to data to evaluate these trade-offs quantitatively for the European Union. I find that non-discriminatory EU border adjustments lead to a Pareto improvement only if they exclude export subsidies, resulting in a 36 million tonnes reduction in carbon leakage. On the other hand, discriminatory EU border adjustments are Pareto improving if they feature export subsidies in addition to import tariffs, yielding a 130 million tonnes reduction in leakage. These results provide a possible justification for the current design of the EU CBAM. Working Papers
Evaluating Asset Pricing Models Under Endogenous Regime Switching [Draft]
with Yoosoon Chang and Joon Y. Park
This paper investigates the time variation in the Captial Asset Pricing Model (CAPM) betas by introducing a new approach that models panel regressions with endogenous regime-switching using a latent autoregressive factor. For our estimation, we model the CAPM using portfolio returns sorted on book-to-market ratio, where the factor loadings, the pricing errors, and the volatility of the error terms can vary across high and low volatility states of the market. We find that the behavior of this asset pricing model significantly differs across different volatility regimes and its performance improves significantly, especially when it is evaluated during the times where the market is in the low volatility regime. Works in Progress
On the Effectiveness of LSTM Models in Predicting Inflation Rates
with Yoosoon Chang and Joon Y. Park
This paper investigates the Long-Short Term Memory (LSTM) neural networks for flexible nonlinear inflation forecasting as an alternative to traditional time series models like VARs that rely on restrictive assumptions. A key challenge with LSTMs is their sensitivity to initial values. In this paper, I develop LSTM architectures for predicting inflation and compare their performance against benchmarks like VARs, ARIMA, and random walk models. The goal of this paper is developing techniques to obtain near-optimal initial values for LSTMs to improve their predictive accuracy. The models is trained and tested on standard macroeconomic datasets with model selection and evaluation based on RMSE. Enhancing deep learning forecasts of inflation through refined LSTM initialization will provide an additional tool for policymakers. The insights on setting starting values for LSTMs can facilitate their adoption in other macroeconomic applications as well. Publications
Revenue Mobilization for a Resilient and Inclusive Recovery in the Middle East and Central Asia [Published version] [Slides]
with Fiscal Policy Group, Middle East and Central Asia Department, IMF
Domestic revenue mobilization has been a longstanding challenge for countries in the Middle East and Central Asia. Insufficient revenue has often constrained priority social and infrastructure spending, reducing countries’ ability to reach the Sustainable Development Goals, improve growth prospects, and address climate related challenges. Moreover, revenue shortfalls have often been compensated by large and sustained debt accumulatiosn, raising vulnerabilities in some countries, and limiting fiscal space to address future shocks. The COVID-19 pandemic and the war in Ukraine have compounded challenges to sustainable public finances, underscoring the need for revenue mobilization efforts. The recent global crises have also exacerbated existing societal inequalities and highlighted the importance of raising revenues in an efficient and equitable manner. This paper examines the scope for additional tax revenue mobilization and discusses policies to gradually raise tax revenue while supporting resilient growth and inclusion in the Middle East and Central Asia. The paper’s main findings are that excluding hydrocarbon revenues, the region’s average tax intake lags those of other regions; the region’s fragile and conflict-affected states (FCS) face particular challenges in mobilizing tax revenue; In general, there is considerable scope to raise additional tax revenue; countries have made efforts to raise tax collection, but challenges remain; tax policy design, notably low tax rates and pervasive tax exemptions, is an important factor driving tax revenue shortfalls; weak tax compliance, reflecting both structural features and challenges in revenue administration, also plays a role; and personal income tax systems in the region vary in their progressivity—the extent to which the average tax rate increases with income—and in their ability to redistribute income. These findings provide insights for policy action to raise revenue while supporting resilient growth and inclusion. The paper’s analysis points to these priorities for the region to improve both efficiency and equity of tax systems: improving tax policy design to broaden the tax base and increase progressivity and redistributive capacity; strengthening revenue administration to improve compliance; and implementing structural reforms to incentivize tax compliance, formalization, and economic diversification. 