I just wanted to clarify one point, because I think I had my numbers mixed up, and I know it's going on record.
It was the U.S. in 2002. They did a test to see what would happen if they went to a voluntary American community survey, and they found that 43% of white households responded, as did 20% of Hispanic households and 22% of black households. There's other data out there that shows that exactly that happens.
It would be great to know why. I don't profess to know why, but we know it happens. Statistics Canada have said sampling was unlikely to produce accurate data for small populations, so the smaller the population within the big population, the less likely it's going to be accurate.
And when Statistics Canada did a test and looked at what Canada would look like, they found the same thing. They did a simulation, and they found that in the Toronto population, blacks were underrepresented by 13.2%, Chinese were overrepresented by 17.6%, and reported Indians were underrepresented by 11.7%. This is in the last year or so. This isn't 10- or 20-year-old data. And strangely enough--I have no idea why--construction workers were overrepresented by 9.4%, and bureaucrats in most of the major cities were massively overrepresented. So we see completely skewed results, not just by ethnic origin or by economic class, but by job. We know that's happening. That will happen with a voluntary survey.
So to me, the fundamental issue is voluntary versus mandatory. You can tweak questions. That's a separate issue. The issue is whether it's voluntary or mandatory. Right now the mandatory census is used to correct this sort of thing. So when you get results back and you see that blacks are underrepresented by 13%, you can adjust for that and weight it. You can't if you don't have that.