P.S. Jud, I'm not so sure I agree with your premise about more=homogenous.
For example,
If we polled everyone in the U.S. as to what thier religion is...I'm guessing we would have a fairly homogenous result that was mostly Christianity.
But if we polled everyone in the world, which is a much bigger sample size, the result would become far more heterogeneos in that now there would be Christianity, Judeaism, Islam, Hinduism, Buddhism, etc, etc.
Kalen, Jud is correct. It is pretty basic statistics. More samples you have, more homogenous the result is.
You are using the example in the wrong way. If you polled 10 people in US, there is a chance that all 10 people say they are Wiccans - which would not be very representative of the overall population. More you sample, more reflective of the true representation the result will be.
In your example, polling just US for the world-wide spread of religion would result in a skewed result - because your sample was too isolated. Once you opened up the sample size to the rest of the world, more representative result was found.