A Village Latrine Program That Did Not Reduce Diarrhea

The program cannot be credited on this evidence. The decline in program villages (30% to 26%, a 4-point drop) is nearly matched by the comparison district (29% to 25%, also a 4-point drop). The difference-in-differences is essentially zero (program change minus comparison change ≈ 0), which suggests secular trends - such as seasonality, a regional health campaign, improved water supply, or economic changes - explain the decline rather than the latrines. This is the central threat: secular trend/co-intervention.

Other threats:

• No randomization or matching: the comparison district may differ in baseline characteristics, and villages were not randomly assigned, so selection bias is possible.
• Measurement bias: diarrhea is often self-reported and subject to recall and social-desirability bias; reporting may have changed because the program increased attention to diarrhea.
• Implementation failure (not effectiveness failure): only 70% of households built latrines, only 40% had functional water seals, many were used for storage, and handwashing stations were rare. Even a real effect would be diluted by low and incomplete use. This is a process/implementation gap, not proof that latrines do not work.

The honest conclusion: the observed decline is consistent with background trends, and the program's implementation was too incomplete to expect a measurable effect.

Program redesign (address the implementation gap):

• Shift from construction subsidies alone to behavior-change and maintenance components: community-led total sanitation approaches, follow-up visits, repair support, and promotion of consistent use.
• Ensure functional water seals and provide handwashing stations with soap, since handwashing is a major transmission barrier and was largely absent.
• Address storage use of latrines through community engagement, norms, and practical design (size, location, privacy).
• Consider sanitation plus water and hygiene as a package rather than latrines alone.

Evaluation redesign:

• Use a stronger design: cluster-randomized trial or a matched comparison with baseline and endline measurement in both arms, or a stepped-wedge design.
• Measure objective outcomes where feasible (e.g., stool studies, anthropometry, structured observation of handwashing) alongside reported diarrhea, to reduce reporting bias.
• Track implementation fidelity: proportion of functional latrines, consistent use, handwashing with soap, and maintenance over time.
• Pre-specify the analysis (difference-in-differences with clustering accounted for) and the sample size.

Communication: Do not claim effectiveness now. Report that the observed decline paralleled the comparison area and that implementation was incomplete, then propose the redesigned pilot with rigorous evaluation before scaling.

Acknowledge the plausibility but insist on evidence and stewardship. Epidemiologically, sanitation is a proven intervention in principle, but the observed data do not show this program worked, and the implementation data show why it might not have: non-functional seals, storage use, and no handwashing. Scaling an incomplete model would waste resources and could entrench the belief that latrines do not help. The relevant question is not "are latrines good?" but "does this delivery model, at this fidelity, reduce disease?"

Ethically, resources are finite, and scaling without evaluation risks diverting funds from effective interventions and creating opportunity costs. There is also an obligation to beneficiaries not to promise health benefits that have not been demonstrated. A reasonable compromise: fix the implementation (functionality, handwashing, use), run a rigorous pilot with a credible comparison, and scale only if the pilot shows an effect. If the office wants to proceed on grounds of plausibility, it should at minimum build in monitoring and evaluation so that the decision can be corrected. This balances the precautionary appeal of sanitation with accountability and evidence-based stewardship.