This would be better if the criteria didn't change every time. Are we supposed to be surprised that the results change when the things being measured are changing every time too? If you're going to measure one parent households, measure one parent households for all 3. If you're going to include no parent households, include no parent households for all 3. If you're going to measure unmarried mothers, measure unmarried mothers for all 3.

When you change your criteria every time, it makes it look like you're manipulating the data to achieve a specific result, i.e. making the percentage go up over time.