Small sample sizes are much more likely to produce extreme results, because it's easier to get an extreme result by chance. In the US, the highest rates of cancer occur in the most rural counties. So do the lowest rates of cancer.

Think about it like grabbing from a giant bag of M&M's. The fewer you select, the higher the chances your results will be skewed towards one color.