Sample vs. Census in the Election

In FE8B Math we have been doing Data Management. We have been learning about different kinds of graphs, how to infer things from a set of data, measures of Central Tendancy (mean, median, mode), if data is biased or reliable (which I talked about in my last blog), and data collected from a sample versus a census.
I was thinking about where you would see samples instead of a census in the real world and one obvious thing came to mind - polls that were predicting the Canadian Election and the actual results. On nodice.ca (http://www.nodice.ca/elections/canada/polls.php), the last poll posted before the actual results was a sample taken from a company called "Angus Reid Strategies". Compared to election results, they were pretty close. Here are what they said (ordered Bloc Quebecios, Conservative, Green Party, Liberal, NDP):
Agus Reid Strategies: 9%, 37%, 7%, 27%, 20%
Canada: 10%, 37.6%, 6.8%, 26.2%, 18.2%
As you can see, the results are very similar. However, if you look at a sample taken from three and a half weeks ago (Sept. 25th) from the same company, the results are very different. The Liberals and NDP are tied, and the Conservatives are much higher. Therefore, some of the samples are close, but some aren't, so you have to be careful as the sample could be almost exact, or way off.
One last thought: Even the election results aren't a census. Only 59% of Canadian citizens elidgable to vote did, so therefore the results are still only a sample. I wonder how the results would change if ALL Canadians 18+ voted...
Until Next Wednesday,
-K