Measurement, outcomes and decision support
Two failures sit at either end of an evaluation. At the front, measuring an outcome nobody cares much about, precisely. At the back, producing a good estimate that never reaches the person choosing.
Measurement
Outcome measures inherit assumptions from whoever built them. If a measure was developed on one population and one condition, it does not automatically capture what matters in another. This strand works on outcome measurement and economic evaluation with that inheritance in view, particularly where a measure is being carried into a setting it was not designed for.
Decision support
Read more
The second failure is the more common one, and the more fixable. Estimates from a choice experiment can be packaged so that someone with a decision can change the assumptions and watch the answer move. That is a different object from a report. It exposes the sensitivity of the conclusion instead of asserting the conclusion.
Three are public:
- eMANDEVAL, mandate design against public preference estimates.
- STEPS, costing and planning for field-epidemiology training scale-up.
- Farming benefit-cost tool, appraisal for soil and farming interventions.
Building them has its own lesson. A tool forces every assumption into the open, because it has to be entered somewhere. Assumptions that survive a paper’s methods section do not always survive that.