Census Block Group ("CBG") study areas are notorious for changing over time. Even from 2009 to 2013, about 10-25% of the CBG's in South Carolina have not remained consistent over that time.
Seeing that CBG's provide the most geographic "resolution" for comparing data, I vastly prefer them over any other group. Trying to measure trends over time using them, however, can be difficult.
One way I'm trying to do this is by finding the CBG's that have not changed between 2009 and 2013 and measuring changes in just these areas. This way, even though I won't be able to get a complete picture of demographic changes over time, I will still be able to get a supposedly decent sample of the data changing over time within these tightly defined areas.
My concern is that maybe by only looking in these areas, I'm only looking at a certain type of area, namely, one that isn't changing enough to justify a "split" into multiple, smaller areas. Should I be worried about the statistical bias of only measuring the changes of CBG-defined areas that are not being redefined over time? If so, how much bias should I expect?