How do you figure "it does not include army of unemployed people"? Census data should reflect members of the population regardless of employment status.
From a quick read of the census website linked, it appears that what unemployment data there is comes from the 2000 Census, whereas income data is much more current and numerous. That's rather concerning.
It's also the case that unemployment numbers are always underreported, something which gets worse during periods of extended unemployment. The government bases its estimates primarily on unemployment benefits filings. People who are unemployed beyond the period of available benefits don't get counted.
So, the unemployed are somewhat factored into the study, but it doesn't seem like a very clear picture of the situation.
That's my prima facie judgement at least.