Authority, Leadership and Social Skills: Evidence from Police Academies

In this project, we aims to investigate the following research questions: Do communities recognize leadership abilities in better leaders? What is the relationship between perceptions of leadership and preferences for authority? What are the roles of social networks and gender in shaping these perceptions? We plan to explore the relationship between leadership skills, social network, and preferences for authority, with a particular focus on gender dynamics and differences, by combining surveys, online activities and lab-in-the-field experiments.

The importance of leaders is recognized in all aspects of life—social, political, economic. (Jones & Olken, 2005; Funke et al., 2023; Ottinger & Voigtlander, 2020). Managers for instance are known to play a key role in driving firms’ productivity and economic success (Bloom et al. 2013; Lazear et al. 2014; Weidmann et al. 2023). A large literature has documented how political leaders can lead by example (Hermalin 1998), shape identity and norms (Acemoglu & Jackson 2015; Akerlof & Holden 2016; Giuliano et al. 2022) and affect social movements (Dippel & Heblich, 2021; Ferrara et al. 2025; Assouad 2025). Recently, the literature has also highlighted the critical role played by lower-level organizational leaders (Ferrara et al. 2025).
The relationship between leadership and authority is a canonical question in social sciences. Weber (1922)’s tripartite classification of authority identified that for most recent forms of authorities (rational-legal authority), the legitimacy of authority comes from powers that are bureaucratically and legally attached to certain positions. To ensure its legitimacy, intensive control mechanisms are usually in place, and public servants are recruited on the basis of a high degree of loyalty. Recent studies have investigated these hypotheses in the context of the police. Ang et al. 2025 study abuse of delegated authority in the context of police violence, which affects community engagement. Friebel et al. 2019 provides evidence of self-selection of intrinsically motivated agents into police. To our knowledge, there is still limited understanding of the relationship between perceptions of leadership and preferences for authority, in particular in the context of the police.

Research Questions

In this project, we aims to investigate the following research questions:
Do communities recognize leadership abilities in better leaders?
What is the relationship between perceptions of leadership and preferences for authority?
What are the roles of social networks and gender in shaping these perceptions?

We plan to explore the relationship between leadership skills, social network, and preferences for authority, with a particular focus on gender dynamics and differences, by combining surveys, online activities and lab-in-the-field experiments.

2025-10-20

2025-10-31

Perceptions of leadership skills by self and peers, leadership nominations, rate of delegation of authority (authority game), leadership nominations after group nominations.

We plan to collect several outcomes in each activity. For the social network survey, we will collect perceptions of leadership by self and peers, and (random) general nominations or gender-specific nominations. For the authority game, our outcome of interest is the rate of delegation of authority. For the group activities, our main outcomes are performance and leadership nominations. In particular, we want to measure the causal impact of revealing name effects on leadership nominations.

Position in social network, baseline individual performance measures.

For the social network survey, we will collect measures of social interactions. For the individual activity, we will collect baseline performance measures in different subjects (math, language or civic skills). Using student information provided in the informed consent, we plan to match these outcomes to students’ administrative data. Our goal is to correlate leadership skills with students’ position in their social network, and with their preference for authority as measured in the authority game.

For the group sessions, we first randomized participants within each section (classroom) into teams of three. The second randomization—assigning the name-revealing treatment, designed to study who benefits from this condition—was conducted at the team level. We also introduce other simple randomizations such as roles in the authority-delegation game, the payment structure in the authority-delegation game, the peer rating in leadership skills where we will have variation in the gender of the rater and the person who gets rated. Finally, we randomize the gender-specific vs. non-gender specific nominations at the individual level.

For the group sessions, we first randomized participants within each section (classroom) into teams of three. This randomization was repeated 100 times, and we selected the draw that maximized the minimum p-value for balance in the proportion of males across teams (i.e., using a max–min p-value criterion). The second randomization—assigning the name-revealing treatment, designed to study who benefits from this condition—was conducted at the team level. Similarly, we performed 100 randomizations and selected the one that achieved the best balance across treatment arms according to the same max–min p-value criterion.
We also introduce other simple randomizations such as roles in the authority-delegation game, the payment structure in the authority-delegation game, the peer rating in leadership skills where we will have variation in the gender of the rater and the person who gets rated. Finally, we randomize the gender-specific vs. non-gender specific nominations at the individual level.

computer

Most randomizations are performed at the individual level. The name-revealing treatment for the group activities is at the team level. (2 sessions of 81 teams, 40 team in Treatment and 40 in Control). 160 teams in total.

Yes

160 teams in total.

For activities 1-3, 243 individuals per activity. For activities 4-5, 243 individuals divided in 80 teams.

80 teams per treatment arm.

MDE for impact of name-revealing on the total number of leadership nominations is 0.496 based on previous pilot studies, with a baseline mean of 0.76.