Currently in Kenya, secondary school government bursaries are administered through committees set up at the level of parliamentary constituencies. However, there is widespread consensus that this system is not functioning adequately, as the process is often haphazard and funds are often spread too thinly across students. The Ministry of Education has thus expressed an interest in introducing a more standardized system to target scholarships to the neediest children, either through proxy-means testing or through participatory community meetings. This study builds upon current research literature in this area by experimenting with different forms of proxy-means testing, complemented by a community assessment of relative student neediness. The research question addressed in this paper is the following: How can the Kenyan Ministry of Education best target scholarships to 8th graders at the highest risk of dropping out? This study provides evidence on the effectiveness of various tools in capturing the ability of each of these methods to predict secondary school enrollment. Specifically, the targeting methods under examination are: (1) government proxy-means surveys (filled by teachers and guardians); (2) school community "participatory rural appraisal" sessions (PRA) (both with and without parents); and (3) a comprehensive household survey. In the case of the community-based sessions, 36 schools were randomly assigned to two groups: (1) sessions with teachers and the school management committee only; and (2) sessions with teachers, the school management committee, and parents. This was done in order to test to if the addition of parents made the relative rankings more accurate due to the additional information and level of transparency, or more biased, as parents may have a highly vested interest in their student's receipt of scholarship aid. Results show that teacher-filled government surveys, household survey data, and Community PRA meetings held with only teachers and the school administration (and not parents) seem to be the best methods of identifying students who may be vulnerable to dropout. Of these, teacher-filled government surveys may be the most cost-effective way to gather this data. Regarding schooling determinants, from the analysis thus far (employing the variables currently available), it seems that in addition to student performance, student orphan-hood status, household assets, and household access to credit of some kind (formal or informal) all tend to have a strong and very significant impact on the probability of enrollment. Tables are appended.