This paper provides logit estimates of the probability that students will cheat in a specific class using randomized response and direct question data in two logit models. The results predict that there are several indicators of the probability of cheating occurring in a class. These factors include both student and instructor characteristics. They suggest several steps that can be taken to reduce the incidence of cheating which are relatively inexpensive yet potentially very successful. Further, this study explores the usefulness of the randomized response survey technique in obtaining information about sensitive behavior. Estimates indicate that there are steps that instructors can take to reduce the amount of cheating that takes place in their classes. This study suggests that using multiple versions of each exam, non-multiple choice exams and reducing the weight of each exam score toward the final course grade are all measures which will lower the incidence of academic dishonesty in a class. By allowing a respondent more anonymity the randomized response method encourages more truthful answers than direct questioning. In both models studied here, randomized response yields higher estimates of cheating. The randomized response estimates also appear to be more consistent with previous estimates of cheating than do the direct question estimates. This lends confidence to the conclusion that when surveying respondents about potentially sensitive or threatening information the direct question method yields inaccurate predictions of actual behavior and randomized response is a more appropriate methodology. / Graduation date: 1994

This three-article dissertation aims to apply Bayesian data analysis to improve the methodologies that process effectiveness findings, cost information and subjective judgments with the purpose of providing clear, localized guidance for decision makers in educational resource allocation. The first article shows how to use a Bayesian hierarchical model to capture the uncertainty of the effectiveness-cost ratio. The uncertainty information produced by the model may inform the decision makers of the best- and worst-case scenarios of the program efficiency if it is replicated. The second article introduces Bayesian decision theory to address a subset of methodological barriers that hamper the influence of research on educational decision-making, including how to generalize or extrapolate effectiveness and cost information from the evaluation site(s) to a specific context, how to incorporate information from multiple sources, and how to aggregate multiple consequences of an intervention into one framework. The purpose of this article is to generate evidence of program comparison that applies to a specific school facing a decision problem by incorporating the decision-makers' subjective judgements and modeling their specific preference on multiple consequences. The third article proposes a randomized control trial to detect whether principals and practitioners update their beliefs on the effectiveness and cost of educational programs in the light of uncertainty information and localized evidence. Supplemented by a pilot qualitative study that guides decision makers to work on self-defined decision problems, the pilot testing of the experiment provides some evidence on the plausibility of using an experiment to identify the causal impact of research evidence on decision-making.

The thesis explores two main topics, Catholic schools in a school choice setting and decentralization of education decision making. The first chapter uses Chilean data to assess the impact of Catholic versus secular voucher or public school attendance on student outcomes. I address admission selection bias by leveraging exogenous variation from school admission lotteries and controls for students' probability of securing a seat in each type of school. My causal estimates reveal that students attending Catholic schools have a 17 percent higher probability of taking the college entry exam (CEE) than those who attend secular public schools. Additionally, Catholic school attendance raises students' chances of scoring above the national mean by 33 percent in math and 45 percent in reading. Catholic school attendance also increases students' probability of applying and being accepted to college. Nevertheless, attending Catholic schools raises dropout rates for boys with low baseline ability. Notably, the positive CEE effects are driven by girl students; however, attending a Catholic school appears to dissuade girls from pursuing STEM majors. Survey evidence reveals that Catholic schools have stricter disciplinary measures and foster higher levels of parent involvement than other public and secular institutions---characteristics commonly associated with high-achieving charter schools. The second chapters explore how student selection changed in terms of socioeconomic characteristics and baseline ability after a centralized school admission system (CAS) was mandated to all publicly funded schools in Chile. That includes private Catholic and secular runned institutions as well as public schools. Estimating a difference in difference model with multiple time periods, leveraging the regional staggered implementation of the CAS. Results indicate that after the centralized admission system (CAS) started, private Catholic schools enrolled a higher proportion of lower-income and lower-ability students than before, this effect is also positive for private-secular and public-schools, but in a lower magnitude. Finally, the third chapter evaluates the effects of the administrative decentralization of education on teacher quality and student outcomes in Colombia. In 2001, the government established an arbitrary rule that granted municipalities with a 2002 population greater than 100,000 almost complete autonomy to provide education services (certification). This analysis takes advantage of this rule to evaluate, using difference-in-differences and regression discontinuity methodologies, the effect of municipal autonomy on teacher quality and student outcomes, including achievement and enrollment. The control group is made up of municipalities for which the provision of education was centralized and managed by the departmental authorities. The results indicate that administrative decentralization (being certified) improves both school enrollment and student achievement as well as the quality of teachers, as measured by teachers’ education level and scores on teachers’ entry competency exams. Using a mediation analysis, the paper finds that higher-quality teachers hired by the certified municipalities partially explained the improvement in student achievement. This analysis also shows that “certified” municipalities invest more local resources in education which also contributes to explain to a much lesser extent their superior educational outcomes. Finally, the results suggest that achieving better student outcomes is less related to the amount of resources that decentralized municipalities managed and more associated with the fact that those resources seem to have been better allocated, generating significant efficiency gains. These gains may be the consequence of lower transaction costs of matching local preferences with local educational interventions.