In a survey of $N=5000$ households, a cluster sampling design selects $k=50$ clusters, each with $m=20$ households. The intra-class correlation coefficient is $\rho = 0.15$. What is the Design Effect ($Deff$) for the sample mean, and how does this affect the required sample size compared to a simple random sample to achieve the same precision?