Pricing behavior of exchange traded funds

This dissertation examines the pricing behavior of exchange traded funds (ETFs) in three essays. (1) The Overreaction of International ETFs, (2) Fragmentation of Night Markets, and (3) The Impact of the Creation of the QQQ on the Underlying Securities. The overreaction study examines the role of information in global overreaction. Univariate analysis reveals that overreaction associated with informed events is less pronounced than with uninformed events following extreme price decreases. Further, positive firm-specific announcements are met with investor overreaction while negative firm-specific announcements are not. Finally, significant reversals of winners during bull markets relative to bear markets support the hypothesis that bull markets contribute to investor overconfidence and overreaction. The fragmentation study examines the cost of market fragmentation across day and night trading sessions. Using a sample of intraday transaction data for three ETFs, I show that night markets do not impound information available in net order flow to the same degree as day markets. Bid-ask spreads are wider at night and these costs are due to higher order processing costs, market maker rents and higher inventory holding costs. Furthermore, market concentration costs at night are associated with significantly higher spreads. The QQQ creation study investigates whether the creation of tradable baskets of securities affects the pricing efficiency and risk of the underlying securities. The results show that extreme price movements in the post-QQQ period are met with less pronounced corrections than in the pre-QQQ period, and that this pricing pattern does not hold true for the control sample. A decomposition of spreads finds that quoted spreads widen and effective spreads tighten in the post-QQQ period. Furthermore, though more heavily weighted components of the QQQ experience tighter spreads, this benefit is less pronounced in the post-QQQ period implying relative benefits to the less heavily weighted components. Cross-sectional analysis reveals that liquidity is directly related to pricing efficiency, but this relationship lessens in the post-QQQ period. The results also show that systematic risk for the underlying securities declines while total risk rises, though the control sample experiences a similar rise in total risk.