Date of Award

August 2020

Degree Type


Degree Name

Doctor of Philosophy


Management Science

First Advisor

Valeriy Sibilkov

Committee Members

John R Huck, Richard D Marcus, Jangsu Yoon, Donghyun Kim


This dissertation consists of two chapters about the momentum and idiosyncratic volatility anomalies, respectively, and one chapter about estimating clustered standard errors.

Chapter 1, Flights to Quality and Momentum Crashes, relates crashes of momentum strategies in stock markets around the world to investor behavior called flight to quality phenomena. The momentum crashes, defined as extremely negative returns of momentum portfolios, occur in most developed stock markets and are centered in economic recovery periods after recessions. I find that their negative returns and negative market betas are associated with investor behavior known as flights to quality (FTQ). Low quality—i.e., high default risk—stocks experience larger investor withdrawals and consequential stock price plunges at financial market collapses, featuring higher market betas particularly during recessions. So the momentum strategies, which tend to sell these plunging stocks, exhibit negative market betas before their crashes and underperform once those stocks bounce back to an economic recovery phase. Worldwide momentum returns and two FTQ proxies, US institutional ownership changes and stock market-bond market disagreements, show consistent results.

Chapter 2, Which Volatility Drives the Anomaly? Cash Flow Versus Discount Rate, examines whether the cross-sectional idiosyncratic volatility anomaly is because of the volatility’s cash flow news part or its discount rate news counterpart. In detail, I reexamine the idiosyncratic volatility anomaly of Ang et al. (2006) and investigate the relative importance of cash flow news and discount rate counterpart in driving this anomaly using the news decomposition of Vuolteenaho (2002). The results from idiosyncratic volatility-sorted portfolios show that the arbitrage portfolio with two extreme portfolios earns about 1.3 (1.2) percent quarterly alpha return after the market factor (the Fama–French factors). I also create two decile portfolios sorted on discount rate news volatilities and cash flow news counterparts. While the average return of the arbitrage portfolio from discount rate news volatilities is insignificant, the counterpart from cash flow news volatilities exhibits about 1.5 (1.2) percent quarterly alpha return after the market factor (the Fama–French factors). These findings indicate that cash flow news rather than discount rate counterpart governs most of the anomaly. The results suggest that investors prefer cash flow news volatilities to discount rate news counterparts, and hence not all idiosyncratic volatilities are equally priced in the cross-section.

Chapter 3, Multiway Clustered Standard Errors in Finite Samples, proposes new clustered standard errors less biased than existing clustered standard error estimators in finite samples. Specifically, I demonstrate the downward bias of existing one-way and two-way clustered standard error estimators (Petersen, 2009; Thompson, 2011) in finite samples using Monte Carlo simulations. When there exist both firm and time effects in a panel regression with N≫T, a firm clustered standard error is always the worst. A clustered standard error estimator by time is the third best, but worsens as T increases. A clustered standard error estimator by both firm and time is the second best, but is biased downward in finite samples. I suggest two first best standard error estimators that always outperform the other competitors.