I was reading the survey below (it's been referenced in Washington Post, NBC, Salon, VICE, etc), and the numbers appear to be off. So I need someone to point out what I'm doing wrong.

When I look at each group's percent of the total N, and multiple that by the group's percent of each answer, I can't get each answer's numbers to add up to 1.
My work: [ibb.co]

So, if the "af" group are 27.6982% of the total N, and 60% of them picked the answer "hc" that gives me 16.1619%. If I do that for each group, and add them up, I should get something pretty close to the percent for N for that answer.

G4 should match G14, G5 should match G15, etc. They're off by multiple percent points.

Few possible things I thought I might be getting wrong:
- I realize they round the percents for displaying. Even if I subtract 0.5 from each number that is too high, and add 0.49 to each number that is too low, I can't get close enough. That will at most give you a total of 1% change. It's not enough. These percents are off 5-6%.
- The (other) group is not actually recorded. They have the groups, and the total. I subtracted all the groups from N to get (other) number. Because this only 2% of the total N, it makes no practical change.

Can anyone figure out what I'm doing wrong?

[genforwardsurvey.com]

- Paul



Edited 3 time(s). Last edit at 12/29/2017 07:06AM by PaulO.