My old advanced statistics tutor would probably love this article. The entire year we were with him was spent learning examples of the sort of maths involved that make 100% sample sizes completely unnecessary. And yes, this included the National Census, an institution that he had no trouble tearing into and making various references to throughout the year.
There's plenty of voluntary sampling methods that will provide you with oodles of market research data (sorry, 'information about the populace') without forcing everyone to provide an ever more detailed analysis of their personal life to the government every 10 years, under penalty of fines and imprisonment. At least, it'll be to the 95th or 99th percentile certain, which is probably just as accurate as the census after you filter out the comedians and refuseniks playing silly buggers with the forms.