I don't want to speculate too much on who's undercounted. We do know that people with language difficulties and the poor will probably be undercounted. But the only way of knowing for sure is to do a properly designed study comparing the two.
On July 27th, 2010. See this statement in context.