In rural areas we are used to saying that once we’ve identified the first house (pen toss, numbering the houses, etc.), to continue to the next house/compound you take the first and closest door on the right (or left).

Now imagine you’re working in a city with a grid pattern (e.g. Manhattan). Doesn’t this tend to make you go straight all the way up or down the same street ? Is there a way to make sure you move around within your identified neighbourhood/cluster?

Thank you Reza for the question.

Unfortunately, no guidelines are available yet for sampling in urban settings. However, I don't know if you were aware that we're trying to move away from the EPI (spinning the pen) method now because it's a statistically biased method (each sampling unit is not independant from the other.) We use it if there is really no choice and we justify its use in the final report. We now recommend the enumeration method. It consists of counting the number of households in the cluster (or getting the list, if available) and then applying the systematic random sampling to find the households to visit. This might be done in urban settings. In a very large city, you will have to do segmentation in order to have smaller equal sections of the city, and then you choose randomly one of them to do your simple or systematic random sampling. If you cannot divide the city into equal parts, you'll have to choose your segment using the probability proportionate to population size (PPS) method in order to give equal chances to all households.