Note: Mainly this question pertains to predictions from a model.
If the unit of analysis of a regression (or any predictive model really) is the individual retail store and these stores are organized into geographic regions, when does it make sense to include variables that describe the geographic region (i.e. a coarser grain than the unit of the analysis) versus just adding a factor (nominal variable) for each region?
For example, adding a factor for each region would literally be a variable that takes on values 'A', 'B'......etc.
Other descriptions of the region might be total number of households, per capita income. etc
Given this set-up, there are two cases I can think of and I would like to know it it makes any sense to add the additional descriptions of the region (versus just the factor):
1) We don't expect the descriptions to change over the time we will use the model for prediction (e.g. We have 2013 total household information per region and we wont get an update until 2016).
2) We do expect the descriptions to change over the time we will use the model for prediction (e.g. We have 2013 total household information per region and we will get an update next month - or whenever we will next use the prediction).