Comment by c-riq

To get a hard to manipulate, representative sample of popular opinion it's not necessary to include everyone in the results. Only if the sub-population sharing IP's is large enough and has systematically different opinions than the people who don't share an IP, then this method would skew the results in a certain direction and may not be useful for certain polls in those regions. But it may also be possible to adjust the results, if it is known which IP's are shared and which are not.