MARCH Outcome Evaluation
IV. EVALUATION METHODOLOGY
4.1. Study design
This evaluation study employed two approaches: (1) a comparison of the exposed and non-exposed respondents of the outcome survey and (2) a comparison of the baseline and the outcome surveys results. In both of the cases the designs were non-experimental. The pretest-posttest (baseline-outcome) design lacks a control group, and as a result subjected to several threats of validity, including history, maturation and instrumentation. Therefore, in order to minimize these threats, this evaluation heavily relied on a Static-Group Comparison of the exposed and non-exposed groups on the posttest survey data (outcome survey data). This approach contrasts survey respondents that were exposed to the MARCH intervention against those respondents who were not exposed to the intervention. Since the selection of the two groups (exposed and non-exposed) did not involve a random process, the primary problem with this design is therefore the confounding factors of selection. To somehow address this problem, multivariate analyses were employed by adjusting for differences in selected respondents’ characteristics.
The baseline and outcome surveys, respectively, were fielded in September 2003 and May-June 2005.
4.2. Sampling methods and sample size
A multistage random sampling technique was employed in which each household had an equal chance of being selected. All the 14 kebeles in Woreda 15 were included in the sampling frame for selection. For the purpose of this evaluation 7 Kebeles were selected using probability proportional to size (PPS) (see Annex I). A kebele or peasant association (PA) represents the smallest administrative unit.
Addis Ababa and West Hararghe