Looks like there's enough struggling with this one that it would pay for me to weigh in. Let's give it the red carpet treatment.

The U.S. census is not perfect: thousands of Americans probably go uncounted. However, the basic statistical portrait of the nation painted by the census is accurate. Certainly some of the poor go uncounted, particularly the homeless; but some of the rich go uncounted as well, because they are often abroad or traveling between one residence and another.
Which of the following is an assumption on which the argument above depends?

Conclusion: Basic stat portrait of census is accurate
Premise: Some poor uncounted, but also some rich
Assumption: The uncounted rich and poor are roughly equal. Rich and poor are emblematic of the portrait of other groups.

Notice I was able to come up with a couple of assumptions here. This is a pretty bad argument, so there might even be more. It's always helpful to think ahead as much as possible. I wrote both of these before looking at the answer choices, so I have an idea of what I'm looking for.

(A) Both the rich and the poor have personal and economic reasons to avoid being counted by the census.
PROBLEM: We don't care why people avoid the census, only how many of them do.

(B) All Americans may reasonably be classified as either poor or rich.
PROBLEM: This relates to my second assumption, but the language goes too far. We don't need all Americans to be classifiable as poor or rich, because even if there were Americans who could not be classified as poor or rich, the argument would not fall apart. It would become a little muddy (what about those middle class folks?), but we wouldn't have definitive proof that the census was bad.

(C) The percentage of poor Americans uncounted by the census is close to the percentage of rich Americans uncounted.
ANSWER: This relates to my first assumption, and is tough to figure out, so it helps to actually take apart the math a bit. Let's imagine that there are 100 people in the country, but we only end up counting 80. If 50 poor, 50 rich is reality and 50 poor and 30 rich are actually counted, we'd have missed 66% of the rich and 0% of the poor, and now our count doesn't provide an accurate portrait of the break down in America (which should have been 50/50, but is no 5/8ths versus 3/8ths). On the flip side, if we miss 10 poor and 10 rich (The same percentage of 50), our breakdown of the country's statistics is still accurate (40 and 40 out of 80, versus 50 and 50 out of 100), even though the hard numbers are wrong.

(D) The number of homeless Americans is approximately equal to the number of rich Americans.
PROBLEM: How many actual people there are in each category is immaterial here. We only care about percentage breakdowns.

(E) The primary purpose of the census is to analyze the economic status of the American population.
PROBLEM: The purpose of the census is immaterial, only its statistical results.

Answer: C