Skip to main content

AHS Metro Areas

HUD.GOV HUDUser.gov
eList
AHS

I would appreciate any thoughts you might have about how to define metropolitan areas in the American Housing Survey, particularly the metro surveys. Should we maintain a fixed geography, change to match whatever OMB definition is current, or try to find a compromise between these?

Since the AHS is a longitudinal survey, one point of view would be to freeze the geography at the time we draw the sample. Subsequent surveys would then apply only to that region. It would be easy to use a series of surveys to examine how that area changed over time. If the original geography matched the then-current OMB definition, it would include the core of the metropolitan area in any case. We wouldn't have to worry about confidentiality problems caused by changing the geographic definitions, and users would be less confused about exactly what bits of ground were covered in each year. The disadvantage is that the survey would not cover any fringe counties that were added over the years. Tabulations from the AHS would reflect only the core of the metro area. Thus, the survey would be good at answering questions about a specified patch of ground but less useful to answer them about currently defined metropolitan areas.

The alternative is to modify the AHS geography to track changes in the official metro area definitions. This would involve drawing new samples in any areas added to the definitions. Since new areas are typically sparsely populated at the time they are added, we will not usually have enough new population (100,000 or more persons) to define new zones. Thus, the new areas will have to be added to existing zones. Because of longitudinality and confidentiality, zones assignments cannot be changed later. In some situations (such as now, when OMB is changing the way it defines metropolitan areas), parts of the old metro area may have to be deleted. In order to do this, all cases in the affected zones will have to be dropped, and a new sample drawn for the new, modified zones. Thus, longitudinality will suffer, even in those parts of the zones that remain in the metro area. Changes from one AHS to the next will reflect changes in geography as well as changes within the previous geography, and users will not be able to isolate the two kinds of changes from one another. Users interested in geography will have to carefully trace the modifications over the years. On the other hand, tabulations from each survey will accurately reflect the metro area as currently defined.

A side issue that also affects longitudinality is that, for budgetary reasons, we are more-or-less freezing the sample sizes in the metro surveys. The metro surveys have always been under pressure to keep costs down, and we feel that the total sample sizes are currently large enough for most purposes. Since we do need to add cases to reflect new construction, we have to eliminate some existing cases to balance the new ones. We can compensate for this by changes in the weighting, and the full sample will reflect the proportional representation of newer and older units. However, this cost-saving strategy does gradually erode longitudinality. At some point we may have to draw entirely new samples. This is an expensive proposition, and I don't expect that we will do this before the 2010 census results are available.

A more radical strategy would be to dispense with longitudinality for the metro surveys and simply draw a new sample, based on current OMB definitions, each time. We are not entertaining this idea at present, but it does have a certain methodological purity.

Our current strategy is based on the belief that users are more interested in having accurate tabulations for OMB-defined metropolitan areas than in following changes to units in specified areas over time. Thus, we are following the second method described above and working to conform the AHS survey areas to the new Core Base Statistical Areas (CBSA). We'll be giving you the details as they become relevant for each metro survey, beginning with the 2006 survey.

Dav Vandenbroucke
Economist
U.S. Dept. HUD
david_a._vandenbroucke@hud.gov
202-708-1060 ext. 5890