Cooper and Schindler (2008) said that a population is the total collection of elements about which we wish to make inferences." What is the risk of making inferences instead of analyzing all of the population? Your thoughts?
Ohhh now we get to discuss sampling and type I and type II! Goody, but not on Sunday. The basic issue is sampling error of type I or II See: Type I and type II errors - Wikipedia, the free encyclopedia for an explanation. But this does not really discuss the problem, whihc is how big is the population? If you are talking about a small number (say less than 100,000) you could reasonably study all of it (if you have enough time). But realistically, resources are ususally constrained, so you need to sample and then make inferences (eg. how many democates are ther in the US: sample; how many democrates in Richville, Vt you could probably test the whole population (if they were all home and would talk to you)).
The risk with sampling is that the sample is not valid so the resulting inferences could be wrong. Cooper and Schindler do a good job discussing this (at least in the 6th Ed/1998).