|dc.description.abstracteng||The dissertation at hand is divided into three chapters. Chapter 1 presents an impact evaluation of a widely spread conditional cash transfer in Brazil called “Bolsa Família”. We analyze the impacts of the program on two educational outcomes: enrollment and attendance. In our analysis we consider the heterogeneous effects of the program. Chapter 2 and 3 study the influence of leaders’ identity on the group dynamics. Chapter 2 focuses on the effect of leaders’ identity on cooperation while Chapter 3 studies the dynamics of embezzlement or what we call elite capture.
Chapter 1 conducts an impact evaluation of Brazil’s Conditional Cash Transfer Program “Bolsa Família”. The program provides grants to households below the poverty line conditioned to school attendance and medical care. Reaching more than 13 million families, this is one of the biggest programs of its kind in the world. Yet research on its impact is limited. To deal with the non-experimental nature of the program we use Propensity Score Matching (PSM). The use of an ample Brazilian household survey - Pesquisa Nacional de Amostra de Domicílio (PNAD) - makes it possible to match beneficiaries and non beneficiaries on a large set of observable characteristics. In this way we are able to present convincing causal inference. Moreover, thanks to the richness of the data set, we can study the impacts of the Program by age, gender and regions. Our results point to a significant increase in Enrollment rates and Attendance among recipients. Furthermore, when different groups and regions are compared, we find that recipients in less developed regions benefited more from program participation than groups in less developed regions, indicating that the program was able to close the gap in education between more and less developed areas.
Chapter 2 studies the effect of leader's identity on cooperation. Due to the increasing demands of more competitive markets traditional producer cooperative leaders often need to be replaced by managers with specialized skills. While managers can bring technical expertise that leads to better economic outcomes, the loss in representativeness of members’ interest might risk the sustainability of the organization.
Using a laboratory experiment, we test the impact of leader's skills and identity on groups’ cooperation. Our design uses a three stage procedure in groups formed by 4 subjects. In the first stage we generate identity in the lab allowing participants to interact for 10 minutes while solving a joint task and competing with other groups. The second stage is instrumental and is designed to measure participant's ability in a real effort task. The last stage consists on a modified public good game in which participants have to decide which proportion of their endowment they allocate to the group account and the individual account. Investments in the individual account are multiplied by 1 while investments in the group account are multiplied by a factor M<1 that is an increasing function of the leader's outcome of a real-effort game. The more the Leader works, the higher the multiplier, M. In the experiment we use a 2X2 design that combines two levels of skills of the leaders and shared or non-shared identity of the leader and the group members. Our hypothesis is that leaders who share the same identity as the group members are more motivated and work harder in the real effort task than out-side leaders. We expect that group members will anticipate it and contribute less when led by an out-side leader than by an inside leader. Leaders with higher skills are expected to perform better in the real effort task and hence attract more cooperation. However it is not clear whether the gains of higher skills are negatively affected by the lack of identity of the out-side leader.
Our findings show that the lack of identity decreases leaders’ productivity, while members successfully anticipate it and cooperate significantly less. When leaders’ skills and social identity are directly compared, we find that the gains in cooperation levels due to leaders’ skills, is just enough to make up for the loss in identity. We do find, however, that out-group skilled leaders were able to sustain cooperation over time, while groups led by random in-group presented a continuous cooperation decrease over time.
Chapter 3 analyses a negative aspect of social identity. Small communities in developing countries (such as tribes, smallholders’ association and cooperatives for example), who share language, ethnicity and culture, are often inclined to suffer from an abusive behavior from their local leaders. This abusive behavior often occurs during long periods of time becoming even close to be an accepted political norm. In Chapter 3 we question if the high social identity condition in the community can be the cause of abusive behavior being hence the trigger to a perpetuation mechanism of power abuse in the community. We investigate two channels of how such a perpetuation mechanism can occur: trust and willingness to monitor. We hypothesize that groups who share the same identity with the leader trust him more, and would hence be less willing to pay for a monitoring mechanism. As trust decreases scrutiny, leaders would feel free to increase embezzlement.
To do that, we present a laboratory experiment where groups formed by three subjects playing a three-stages game. In stage one subjects solved a group task similar to the group task presented in Chapter 2. In stage two, one subject was randomly assigned as group leader and other two subjects as group members. To allow members’ payoff to be dependent on leaders’ performance, we implemented the effort task followed by a trust game. The effort task was to sum 5 two digit numbers during 30 seconds. A minimal performance of 1 correct sum by the leader generated a multiplier A which would take a value between 0.8 and 1.9. The non-achievement of the minimal performance would set A=0.8. Members received 20 points as endowment, which they could decide to keep, where it would be multiplied by 1, or to pass it to the leader, where it would be multiplied by A. Leaders received a fixed amount of 20 points plus the amount of points leaders decided to take from each group member. In stage three subjects played a lottery game. We use a 2X2 design combining in- and out-group treatments with monitoring and without monitoring possibility.
Our findings show that members led by in-group leaders were less willing to monitor while this lack of scrutiny significantly increased amount captured. We did not however find that trust by itself is significantly higher in homogenous groups, which questions the identity as a cause of perpetuation mechanism.||de