Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)

Graduate Group


First Advisor

Itay Goldstein

Second Advisor

Amir Yaron


In the first chapter "Shock Spillover and Financial Response in Supply Chain Networks: Evidence from Firm-Level Data", using machine learning methods on firm-level textual disclosures, I construct a large-scale dataset featuring firm-specific shocks to production. I map these shocks into a unique, hand-built network of firm-level supply chain connections to empirically quantify how these localized shocks affect remote firms along the chains. Surprisingly, contrary to prediction by typical network theories, these firm-specific shocks impact the revenue of firms even up to 4 connections away from the origins. This pronounced spillover effect is explained by three features--uneven distribution of monopolistic power, variations in supplier substitutability, and different inventory levels--that are salient in the data but usually not present in existing models. In addition, firms seem to respond to these spillovers by increasing their working capital and financial leverage. Moreover, the stock market reacts to shock spillovers from distant connections with slower speeds: post-shock abnormal returns are persistently negative for up to 40 days.

In the second chapter "Deciphering Fedspeak: The Information Content of FOMC Meetings", we present a new approach to quantify the economic and policy content of Federal Reserve communications by dissecting the Federal Open Market Committee (FOMC) meeting minutes into distinct economic topics, and simultaneously extract the tone and uncertainty level of each topic. We use market reaction to objectively assess the relative informativeness of each topic, and we find significant incremental informational value from the topic contents, despite that the minutes are released several weeks after the original meetings. Furthermore, we find evidence consistent of the Fed possessing superior information, which is then transmitted to the market through the language of the minutes.