First, I evaluated the existing literature about the relationship between gender, crime and politics. The majority of existing studies are focused on developing countries and (classical) political institutions, while developed countries and judiciary systems are mostly ignored. I also interacted with different experts on this topic, thanks to (on-line) conferences, seminars and meetings.
Second, I collected the data. In spite of some issues in data collection and COVID-19 pandemic, I managed to collect data from Italy and the USA. The Italian dataset includes information about media coverage of corruption at municipal level and about mayoral races between 1993-1998. The USA dataset includes corruption data and information about female politicians at state level between 2000-2011. It also includes some information about the judiciary branch for North Carolina (Jury Sunshine Project) between 2010-2012.
Third, I acquired the necessary empirical knowledge to implement my analyses, by studying previous works and interacting with experts.
Forth, I Implemented my empirical analyses:
1. Using the Italian dataset, I investigated the effect of mayors’ gender on corruption scandals. I used a methodology called regression discontinuity design on mixed-gender electoral races (Lee (2008)). In broad terms, this technique takes advantage of the similarities between municipalities where a woman lost (control group) or won (treatment group) for a very small margin. My findings indicate that there is no statistically significant effect of mayors’ gender. I developed a (soon-to-be discussion) paper, called “Female Mayors and Corruption Scandals: an RDD Approach with Italian Data”.
2. I also explored the relationship between politicians’ gender and corruption in the USA. As suggested by the precedent literature (Lee (2008)), I first evaluated the relationship using graphical analysis and preliminary regressions. Overall, the results indicate that no statistically significant effect was present, although these findings are not extremely robust.
3. Using the data provided by Jury Sunshine Project, I investigated the role of female judges on the probability of guilty verdicts in jury trials. My identification strategy is based on judge’s rotation and the results indicate that female judges increase the probability of guilty verdict in jury trials. I created a working paper, called “Lady Justice: The impact of female judges on trials' verdicts in US”.
4. Using the data provided by Jury Sunshine Project, I explored the effect of jury gender composition and jury political composition on guilty verdicts. I, first, evaluated the impact of jurors’ gender compositions on the probability of guilty verdicts. My identification strategy is based on the selection of the jury pools (Anwar et al. (2012)). Jurors’ gender does not play a statistically significant impact. However, jurors’ political affiliation composition has a significant role and my findings indicate that politically independent jurors reduce the possibility of a guilty verdict. I created a working paper, called “Beyond Reasonable Doubt: The impact of jurors' political affiliation on jury trials in US”.
5. I attempted to evaluate the impact of gender in media coverage of crime. However, the lack of crime data for Italy and the low quality of US data made it impossible to implement the analyses with the necessary rigor.
6. Using the Italian data, I investigated the effect of gender quotas on corruption, using a difference-in-difference strategy. In broad terms, this technique is a quasi-experimental approach that compares the changes in outcomes over time between a group of municipalities affected by a reform (treatment group) and a group of municipalities that is not (control group). Following the approach of Baltrunaite et al (2014), I used the introduction (1993) and removal (1995) of gender quotas in candidate lists in mayoral elections. The identification strategy is based on the comparisons between those municipalities voting between June 1993 and November 1995 (treatment group) and those voting after November 1995 (control group). My findings are non-statistically significant and not particularly reliable.
Finally, I disseminated my findings. In spite of COVID-19 limitations, I managed to participate and to present in many (on-line) conferences and workshops.
