The Scholarship List That Leaned One Way
A provincial government runs a PHP 6,000,000 annual college scholarship program for 400 students, selected by a committee using a composite score: 60% household income rank, 30% academic average, 10% interview. Applications this year: 1,240. The committee's shortlist of 400 shows that 212 come from three barangays in the capital town, which together hold 9% of the province's poor households. The committee chair says the interview is where "we see who really needs it." An internal review finds that interview scores correlate strongly with the interviewer's own barangay. The governor asks you, as planning officer, to assess the selection process and propose a defensible alternative before the next cycle. You have the applicant database, the poverty incidence per barangay from the national statistics office, and three years of selection records.
1. Diagnose the problem using the data described, and state what evidence would confirm or refute each suspected cause.
Read the worked answer
Two findings stand out.
Geographic concentration. 212 of 400 slots (53%) go to barangays holding 9% of poor households. If selection were proportional to need, those barangays would receive roughly 36 slots. The gap is large enough that chance is an implausible explanation.
Interview as a leak. The interview is only 10% of the score, but if it correlates with the interviewer's barangay, it is not measuring need; it is measuring familiarity. A 10% weight can swing close calls when many applicants cluster near the cutoff.
Evidence to confirm or refute each cause:
- Weight sensitivity: recompute the shortlist with the interview set to zero and with the income weight raised. If the list changes substantially, the interview is doing hidden work.
- Cutoff clustering: plot composite scores around the 400th rank. If dozens of applicants sit within a point of the cutoff, small components decide outcomes.
- Interviewer assignment: cross-tabulate interview score against interviewer and applicant barangay. A consistent within-interviewer pattern supports bias.
- Data quality: check whether income documents were verified or self-declared; unverified income would distort the 60% component.
- Access: compare application rates per barangay against poverty incidence. Low application from poor barangays may indicate information or travel barriers rather than merit differences.
2. Propose a revised selection design that is both defensible and administrable with existing staff.
Read the worked answer
1. Make need the gate, not just a weight. Use a verified poverty proxy (barangay poverty incidence plus household income documents validated by the social welfare office) to create an eligible pool. Rank within the pool on academic average. This removes the situation where a well-documented middle-income applicant outranks a poor applicant on presentation.
2. Replace the scored interview with a structured needs assessment. Keep a conversation, but use it to verify circumstances (dependents, distance, disability, orphanhood) against a fixed checklist with defined points. Two assessors, neither from the applicant's barangay, score independently; disagreements above a threshold go to a third reviewer.
3. Publish the rubric and the weights before applications open, and publish the anonymized distribution of awards by barangay after.
4. Add an equity floor and ceiling. Guarantee each municipality a minimum share of slots proportional to its poor households, with a cap so no barangay dominates. Any unused slots revert to the provincial pool.
5. Extend reach. Coordinate with barangay officials and schools in low-application areas to distribute forms and hold application drives, since access, not merit, may explain low uptake.
6. Audit annually. Recompute the weight-sensitivity test each cycle and report to the governor.
3. The committee chair resists, arguing that the new design ignores "character" and that his staff cannot handle two assessors. How do you respond?
Read the worked answer
Address the substance first, then the workload.
On character. Character matters, but a single interviewer's impression is a weak and unverifiable measure of it, and the data show it tracking geography rather than need. The structured checklist keeps the qualities the chair values - responsibility, hardship, commitment - while making them observable and reviewable. Offer to keep a short written essay or a referee letter as a qualitative input that is read but not scored, so no signal is lost.
On capacity. Two assessors do not double the work if the interview is shortened to 15 minutes per applicant and scheduled in blocks at the municipal hall, with assessors drawn from the social welfare, education, and planning offices. Estimate the load openly: 1,240 applicants times 15 minutes is about 310 assessor-hours, or roughly 20 staff-days across the province if spread over three weeks. That is feasible with scheduling.
On trust. Propose a one-cycle pilot: run the old and new methods in parallel on paper, compare the resulting lists, and let the committee see the difference before committing. Frame the change as protecting the program from a challenge it would likely lose, not as a verdict on the chair.