Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Statistical power calculations were undertaken to determine the appropriate sample size. The following conservative assumptions were made based upon existing program data, an existing survey and the implementing partner's experience conducting these trainings in other parts of PNG.
• Currently around 60% of households have access to a formal financial product
• Around 20% of households in a treatment village will attend the financial literacy training
• Of the households that attend the training, around half will gain access to a bank account who previously did not have access
• Therefore, on average, the estimated effect size is 10 percentage points
• An intra-cluster correlation of 0.08 is used
Based upon these assumptions, a sample size of around 2800 households in 80 villages (40 treatment and 40 control) would have statistical power of 0.8.