Sampling in Randomized Evaluation

I am designing an experiment in which the treatment is administered through information priming. I plan to conduct the experiment in a large city town. The town comprises of several blocks which are geographically contiguous. I plan to randomly pick four blocks and assign treatment (information priming) to two and control (no information priming) to other two.
Within each block I plan to use systematic sampling for administering treatment / no treatment at household level. The outcome variable is dichotomous (yes/No).
My question is that do I need to correct for cluster sampling when calculating sample size or doing power calculations such as using ICC or not.
i would appreciate if you can point out any other flaw/issue in this scheme. Thanks