Two essays in applied econometrics and finance

My Ph.D. dissertation consists of two chapters. In the first chapter, I conduct a thorough analysis of execution costs comparison between the NYSE and the Electronic Markets, adopting a variety of the econometric techniques to correct for the selection bias problem, and using the most recent Dash-5 reports data on a large sample of stocks. I find that after controlling for the selection bias, the Electronic Markets offer lower execution costs during my sample period. I carry out my analysis at different levels of order difficulty, instead of just controlling for the selection bias at the sample average level of order difficulty, measured by a vector of controlling variables. My results confirm the superiority of the Electronic Markets' execution quality and the results are robust. My results can have implications for listing decisions by public firms. Since the Electronic Markets' trades of NYSE-listed stocks are less costly than in the NYSE, it may make sense for some NYSE firms to switch to the NASDAQ, especially for those stocks whose trading volume is mostly made up of easy orders. Moreover, my results are consistent with Chemmanur and Fulghieri's (2006) theoretical model: trading costs alone do not drive firms' listing decisions.;In the second chapter, I propose a simple method for estimating betas (factor loadings) when factors are measured with error: Ordinary Least-squares Instrumental Variable Estimator (OLIVE). OLIVE is intuitive, easy to implement, and achieves better performance in simulations than other instrumental variable estimators, especially when the number of instruments is large and the sample size is small. OLIVE can be viewed as a one-step GMM estimator using the identity weighting matrix. I also derive a two-step GMM estimator by choosing an optimal weighting matrix. As an empirical application, I replicate Lettau and Ludvigson's (2001b) test of the (C)CAPM using OLIVE instead of OLS to estimate betas. Macroeconomic variables usually contain large measurement error. I find that in regressions where macroeconomic factors are included, using OLIVE instead of OLS beta estimates improves the R-squared significantly. More interestingly, my results based on OLIVE beta estimates help to resolve two puzzling findings by Lettau and Ludvigson (2001b) and Jagannathan and Wang (1996).