The responding sample would increase, which would mean that the sampling error would actually be less in the national household survey than we saw in the 2006 census. It wouldn't remove the risk of non-response bias—
On March 8th, 2011. See this statement in context.