It's important to put it in context.
It's true that the response rate has dropped, going from 93.5% to 69%. And the problem that gives rise to is this. If we do nothing and keep the sample size as is, we will lose statistical reliability. But we have offset that effect. Instead of having a sample size of one in five households, we established a sample size of one in three households. In the end, more people responded in 2011 than in 2006, with the mandatory long-form census. So we rectified that problem.
A lower response rate can skew the data. Some segments of society happily respond, whereas others refuse altogether. We invested a considerable number of resources and made sure we had a tremendous amount of information at our disposal to make every possible effort to correct that kind of bias. We'll never be able to fix it 100%, but our analysis shows that we've been able to largely offset that response bias. There has always been a certain measure of response bias, even in the census data.