Essays in Macroeconomic Analysis: Oil Shocks, Textual Forecast, and Bubble Cycles
This dissertation consists of three essays in macroeconomic analysis.
The first essay introduces a new method for studying the macroeconomic effects of oil price shocks by incorporating textual analysis of news articles on OPEC announcements. Text features extracted from major newspapers serve as external instruments in a proxy SVAR model to capture shifts in oil price expectations. The results show that 91.3% of oil price surprises are linked to OPEC-related factors, 85.6% from supply decisions, and 5.7% from sentiment and demand expectations. Historical decomposition reveals supply news dominates during geopolitical crises, while demand sentiment is more influential during downturns like the Global Financial Crisis and COVID-19. Experiments using synthetic news confirm that real narratives significantly affect market behavior, underscoring the importance of news-driven expectations.
The second essay proposes a method to improve text-based forecasting of crude oil prices. Using advanced techniques such as pattern validation and attention mechanisms, the study finds substantial improvements in predictive accuracy. Notably, incorporating the full text of articles, instead of just headlines, enhances forecasts. A model leveraging verb-noun and noun-verb collocation pattern validation (e.g., “prices tumbled”) outperforms benchmarks and headline-based models across forecast horizons. Integrating macroeconomic indicators with textual features further improves prediction accuracy, showing the value of combining structured and unstructured data.
The third essay develops a continuous-time heterogeneous agent model to examine the relationship between bubbles, financial frictions, and wealth inequality. The model features entrepreneurs facing productivity shocks governed by a Cox-Ingersoll-Ross (CIR) process. These agents allocate wealth across capital, credit markets, and a speculative bubble asset, subject to borrowing constraints. Firms operate in perfectly competitive markets with endogenous capital returns. The equilibrium is defined by a coupled system of Hamilton-Jacobi-Bellman (HJB) equations and Kolmogorov Forward Equations (KFEs), which together characterize individual behavior and the aggregate evolution of productivity, capital, wealth, and bubble holdings.