I use topic modeling and principle component analysis to construct an index of inflammatory language using Twitter data from Senators. I estimate the effect of difficult election years on language moderation using different treatment groups— Senators who faced closed general elections in 2016, and Senators who ran in states that voted for the opposite party’s presidential candidate in 2016— in line with the median voter theorem. I find Senators in close elections are more inflammatory on average, and Senators in opposite party states did not moderate in the election year.
This thesis estimates the current voting power of each of the voting blocs in the Electoral College. It concludes that the current distribution of power, while significantly improved from the last similar published paper, is still significantly unfair in terms of the one man one vote principle. Since the Electoral College elects the highest authority in the United States, the President, the election should reflect the principle of one man-one vote. An alternative to the Electoral College is presented, based on L.S. Penrose's conclusions on weighted voting, that comes close to solving the issue of one man-one vote while retaining some of the traditions of the Electoral College.