I read a column in the Gazette newspaper online it was an editorial on the census written by Dan Grander. He wrote a great piece of information on the long term census. He talked about how in 1936 a poll that was taken by 2.3 million Americans on the election of that year. The poll showed that Alf Landon would win and not Roosevelt ,and would beat Roosevelt by a landslide. Mr. Grander mentions the poll was bias, because the people who did the poll where finding the address to send them to by car records. This meant that poorer people where less represented in this situation, which voted for Roosevelt. The result was that Rosevelt did win even though 2.3 million americans polled said that the other guy would win by a landslide. As a comment on the blog stated by Tom he thinks that if a poll that is not mandatory is reliable than how is it that a non mandatory census isn't. The thing is polls that are not mandatory can be wrong. In the 2008 primary New Hampshire was suppose to go to Obama every poll said even the exit pols , but It went to Hiliary. In 1992 the poll in the U.K had the conservatives just winning the election with the max 316 seats (minority Goverment) and would lose 3%of the popular vote (Camapared to the last election which gave them 43% of popular vote , and with 376 seats). The exit poll was wrong , and in fact the conservatives still had a majority , and only lost 0.3% of the popular vote. This goes to show how non madatory surveys can be unreliable, but to be fair as long as it is done radomly with no bias the margin of error can be minor 3%, but even with a margin error of only 3% it seems that you can make a mistake of majority goverments and minority. The long term census which gives very important informations has a response rate of 94% which is high , and shows strong data that is very accurate. With a non madatory census staticians believe that poorer people will not answer the census as much as the other , and this will create a in acurate reading in the census.
Read the editorial here