My dissertation consists of three unrelated chapters. The first chapter studies the presence and effect of a serious labor-market distortion in the Surgical Instrument (S.I.) industrial network of Sialkot. The problem arises due to the non-perfect substitutability between family and hired managers. In particular, the firm owners are hesitant to hire managers who can potentially steal their business. As a result, the owners prefer to engage mostly close family relatives into management positions i.e. those who are considered trustworthy. This restriction prevents the firms from optimally choosing the management size. The distortion is manifested by a significant positive correlation between family size of the founder and firm output; firm founders who have more brothers end up with higher output.
The second chapter evaluates a teacher incentives program conducted in Kenyan primary schools in 1998 and 1999. Out of 100 randomly selected schools, 50 were chosen to participate in the program, while the remaining 50 were treated as the control group. The impact of the program on various teacher and student outcomes is studied. We find a positive and significant effect of the program on extra (out of class) coaching by the teachers, but no effect on teacher attendance, homework assignment, and pedagogical practices. We also find a positive impact of the program on student test scores.
The third chapter uses a human capital approach to model juvenile participation in criminal activities and/or legitimate labor market activities. In a two-period setting, the individual decides how to allocate time to crime and labor market in each period. We endogenize skill formation by assuming that the time spent in criminal and labor market activities in the first period determines the investments in the corresponding stocks of human capital. The investments lead to a larger stock of human capital for these activities in the subsequent period, and therefore affect the second-period returns from these activities. The model is tested using data from the National Longitudinal Survey of Youth 1997 (NLSY 97). We implement the model using a multinomial logit framework and find that the data match the predictions fairly well.
