Research methods are inflexible because the instruments cannot be modified once the study begins.
Reduction of data to numbers results in lost information.
The correlations produced (e.g., between costs and benefits, gender, and access to services or benefits) may mask or ignore underlying causes or realities.
Untested variables may account for program impacts.
Errors in the hypotheses tested may yield misimpressions of program quality or influential factors.
Errors in the selection of procedures for determining statistical significance can result in erroneous findings regarding impact.