|Abstract:||This dissertation consists of three independent essays on applied econometrics. The first chapter extends the model averaging estimator (Hansen, 2007) to the time series scenario. To this end, I propose a Mallows model averaging autocorrelation corrected (MAC) estimator, in which the tapered estimator (McMurry and Politis, 2010) is employed as the working covariance matrix. The optimality of the proposed criterion for MAC is proved under mild regularity conditions. Furthermore, the estimator works for a general family of covariance matrix estimators. Monte Carlo simulations demonstrate the effectiveness of MAC in comparison with the existing counterparts. A significantly improved paper based on this chapter (Cheng, Ing, Yu, 2015) has been accepted for publication, where model averaging weights residing in a discrete set are extended to continuous ones. Additionally, the published paper proposes the use of a modified Cholesky decomposition method, instead of the tapered estimator, for the inverse autocovariance matrix.
The second chapter examines the causality between the odds of winning an election and vote-buying. In Taiwan in 2010, eight cities and counties accounting for 60% of the entire population were merged and upgraded to five special municipalities. Along with the merger event, the number of district seats was changed, exogenously shifting the winning chances of candidates in the 2010 municipal elections. This policy intervention allows us to investigate whether a candidate is incentivized to buy votes when winning odds deteriorate. For every additional seat cut in a district, the difference-in-differences estimate suggests that approximately 4% additional incumbents seeking reelection decided to buy votes.
The third chapter estimates determinants of adolescent romantic matching using a maximum score estimator. Sorting patterns are suggested from the results. Findings are consistent with the literature, but more detailed results are revealed. In line with the literature, I find evidence of positive assortative matching along physical traits and grade. In addition, race is suggested to be the most important determinant of match value in the context of teen romantic relationships. Finally, adolescent matching seems pleasure-driven, in that past sexual experience plays an active role.