We could have done that. It was the first time there were two different forms, so it meant that it was up to the statisticians to decide what should be in the short form, in which they get 100% surveyed, and then what can be in the shorter form. The questions in the longer form, rather, as Mr. Fellegi quite clearly pointed out, are questions for which that sample size is accurate, whereas on the short form there are things you really need to survey 100%, because it's a head count. There's statistical evidence that a smaller survey of that long form, mandatory, can give you the information you need.
On March 8th, 2011. See this statement in context.