Essays on the impact of politicians' characteristics and speeches on policy outcomes

This thesis consists of three self-contained essays. The first substantive chapter, entitled "Should you want an educated mayor? Evidence from close elections in Italy", studies the impact of politicians' education on policy choices and public finance. A probabilistic voting model with candidates' education level and public expenditure decomposition is used to explain how a politician's education matters for policy decisions over public budget allocation. Empirically, I use Italian municipality data on electoral results, balance sheets, and mayor candidates' educational attainment from 2000 to 2015. To estimate the causal effect of mayors' education on public finance I rely on regression discontinuity design focusing on close elections. Overall, I find that educated mayors boost public investment, especially in the education sector, without compromising the fiscal stability of the municipalities.

The second substantive chapter, entitled "The impact of women's political leadership on institutional quality", studies the impact of female political leadership on future women's participation in politics and hiring in public administration, as well as on different measurements of institutional quality. Empirically, I use Italian municipal and provincial data on elections, politicians, and institutional characteristics from 2000 to 2015. To estimate the causal effect of electing female mayors I rely on regression discontinuity design, and instrumental variable approach, focusing on close elections. Overall, I find that women's political leadership boosts women's participation as politicians and public servants while improving a wide set of institutional quality metrics, reducing corruption, increasing government effectiveness, the rule of law, and political accountability. Furthermore, the positive impact of female leadership on institution quality is stronger in areas with more established gender norms, and if the female politician is holding a tertiary education degree.

The third substantive chapter, entitled "Do speeches at the UN General Assembly affect international aid allocation?", studies whether countries' speeches at the United Nations General Assembly matter for international aid allocation. I use a supervised machine learning algorithm called wordscore and 7,533 different statements from 198 countries made in the period from 1975 to 2018 to measure countries' preferences from text. I find considerable evidence that countries' preferences derived from speeches affect aid allocation. Both the US and Russia provide more aid to those countries whose speeches please their current agenda and preferences the most.