When the government collects and reports data on "Household Income" it includes the income of all the people living in a particular household. Two people living separately will be counted as two household units and their income counted separately.
However, if they marry (or co-habitat) then they will have one combined household income. Half the households, double the income on a per household basis (no more total income, however).
As households are formed over time their characteristics based on several factors will affect total household income.
Education is one such variable that has been identified as one of the primary drivers of income inequality. Not the only one but significant enough to note
New data provided by the Census Bureau and interpreted by the Pew Research Center (graph below) shows that the rate of household formation by marriage has decreased overall since 2008 by an average of 11.4%. However, the decrease is not evenly spread over educational attainment levels.
As you can see, those with a college degree get married at a higher rate than people with lower levels of educational attainment, and even though they have decreased they have decreased at a slower rate than the other categories (7.8% is significantly below the average).
Additionally, those with a degree have seen an uptick from 2011 to 2012 (brown bars to the extreme right) in family formation by marriage.
If income gains are going to those with more education and educated people are pairing up at a higher rate than other groups, then income inequality as measured by "Household Income" can only get wider.
I don't know how to fix this. I just thought it was interesting and I don't think it is something that is talked enough about at a policy making level.