>For one thing, has the federal government ever harmonized its data gathering
>and analysis with all the different states' 'employment security' offices,
>let alone all those other places (which, if they did, would seem to make
>double counting quite possible, but my hunch is that federal estimates of
>unemployment in the US are way too low)?
The BLS works closely with state labor departments.
But I don't see how this is relevant really. The unemployment stats are the product of a monthly survey of 60,000 households. As I recall, the British stats are based on who registers for unemployment benefits, which produces pleasingly low figures, and there's only a yearly survey comparable to the U.S. household tally. You can criticize, rightly, the definition of unemployment in the U.S. as too strict, producing too low an estimate, but it's a reasonably accurate picture by its own definition. But remember that the jobless stats are not intended as a measure of human deprivation, but of labor market slack. On that conception, it's appropriate to exclude people who aren't looking for work, since they've been demobilized from the reserve army of labor.
The BLS also produces wider measures of unemployment, which includes discouraged workers and unwilling part-timers. These tend to be about 1.5 times the official number.
What's "way too low," anyway? The U.S. has one of the world's highest shares of the adult population working. We've got some of the longest workweeks in the world. You could say that overwork is at least a serious problem here as underwork.
Doug