I would add several things to that, but first to compliment the United States on the mechanism. They were looking at the issue of moving away from a mandatory census, or portion of the census, so they tested it. They took several years and got very serious about testing.
Their conclusions were as follows.
They would lose approximately 30%. There were certain groups—ethnic and racial minorities, poor people, and the wealthiest people—who would be heavily under-represented and there would be no way to benchmark the results.
The second finding was that the costs would increase dramatically, and to push response rates beyond the 70% range, which is still unacceptably low, became impossibly expensive.
The third and very interesting finding was counterintuitive. One of the things they tested was a question raised earlier by Mr. Lake, which is whether you get better quality answers item by item by having it voluntary, because people could opt out of questions. They looked very carefully at analyses of item responses and what is called “item non-response”. I think it would have been to my surprise—I don't know what their researchers thought going in—that they could find no difference. It was counterintuitive, but that's why we do research, to find out whether the things that seem intuitively appealing or clear in fact are proven out.