Intraclass Correlations Values in International Development
Knowledge of comparative analysis of intraclass correlations (ICCs) forms an important tool to properly design impact evaluations of development interventions, especially cluster randomized control trials, by ensuring that an evaluation has sufficient statistical power.
This paper takes an important step towards helping researchers to accurately estimate appropriate sample sizes for their clustered RCTs by presenting ICCs for a wide range of domains common for development research in sub-Saharan Africa.
AIR used its deep pool of cash transfer experimental evaluations to perform a comparative analysis of ICCs for common development outcomes. Unfortunately, this parameter is rarely available to evaluation practitioners prior to commencing their study.
AIR’s article on ICCs in Sub-Saharan Africa provides a valuable contribution to study design by providing findings from rich datasets from Kenya, Malawi, Zambia, and Zimbabwe. These findings yield a set of ICC estimates for many common domains for development research. AIR’s research found that ICCs may be lower than researchers regularly assume while conducting power calculations.
The lowest ICCs tend to be in child nutrition and food security while the highest ICCs tend to be for community-determined outcomes, such as housing conditions. These findings suggest that cluster randomized control trials may be viable in certain settings even when evaluations face limited budgets.