If we make it
possible for
local leaders to pilot their initiatives
on a limited
scale first — say, by launching an intervention in 100 treatment classrooms and tracking results relative to a set of 100 comparison classrooms — we would provide both more upside potential and less downside risk.
Another
possible objection could be that the performance of GCMs
on the point basis is expected to be low because of the «noise» of
local (spatial or temporal) weather conditions, whereas the performance should be higher at large spatial
scales.