4
Geographic Adjustments to Thresholds
The thresholds used in the current poverty measure do not vary by where people live. That is, there is a U.S. standard for determining whether a family is poor, regardless of whether the family lives in the North or South or in a large metropolitan area or on a farm. As described above, the official poverty thresholds initially had a farm/nonfarm distinction, where poverty thresholds were lower for people living on farms (who were assumed to grow rather than purchase some of the food they consumed), but this distinction was eliminated in 1981.
The 1995 National Research Council (NRC) report recommended that poverty thresholds should be adjusted for differences in the cost of living across areas, though it was also recognized that implementing such adjustments would be complicated by the lack of reliable data and methods for doing so. Given the dearth of knowledge, the report recommended that adjustments be limited to the housing component of the threshold, noting that this element is the item for which prices vary most across the country and for which the methods for estimating adjustments are the most advanced. The report also recommended further research on methods for improving and updating geographic adjustments.
The report presented a set of indexes to adjust poverty thresholds on the basis of six metropolitan area (or nonmetropolitan territory) population size categories and residence for nine detailed regions (or Census Bureau “divisions”). This resulted in a set of indexes for 41 geographic areas (rather than 54, as some categories had no members). The report used a
modified version of a method used by the Department of Housing and Urban Development (HUD) to create annual fair market rents (FMRs), which are used in the administration of Section 8 rental housing subsidies. Essentially, data from the 1990 census on rents for two-bedroom apartments were used to gauge differences in the cost of housing across areas (see National Research Council, 1995:194-200; Nelson, 2004).
U.S. Census Bureau reports and papers have implemented experimental poverty measures with and without geographic adjustments and have also investigated the effect of these adjustments on state-level poverty rates (Short et al., 1999; Short, 2001b). This research has shown some of the limitations of the NRC-recommended approach, in which, for example, all metropolitan areas in New England have the same index value, though housing costs in Maine are lower than those in other New England states.
The second Census Bureau report devoted to experimental poverty measures (Short, 2001a) presented geographic indexes based explicitly on the HUD FMRs. FMRs are available annually for all U.S. metropolitan areas and nonmetropolitan counties. They represent the gross rent, including utilities, at the 40th percentile (with some exceptions) of the rent distribution of standard-quality rental housing. This adjustment is applied to the housing portion of the poverty threshold. The advantage of this method over the NRC-recommended approach is it provides a finer level of geographic detail and allows for fairly simple annual updating.
Applying these geographic adjustments has a considerable effect on many state poverty rates. The poverty rates decline in low-cost areas and increase in high-cost areas (when compared with rates for which no geographic adjustments to thresholds are used). For example, the poverty rate in Alabama drops from 14.8 percent to 10.2 percent, while the poverty rate in California rises from 13.1 percent to 18.4 percent (Nelson, 2004).
The presenters at the workshop noted that the geographic adjustment method recommended in the NRC report and the refined approach implemented by the Census Bureau—both of which are based on FMR information from HUD—have several limitations. First, it would be advantageous to adjust the thresholds for regional differences in costs of other basic items rather than just housing. Second, FMR data, by design, incorporate only rental costs and not owner-occupied housing. Third, rents reflect amenities and disamenities of geographic areas. John Ruser (Bureau of Economic Analysis) raised the question of whether, for example, people who live in low-cost areas should have a lower poverty threshold (which makes them less likely to be counted as poor) if they live in an undesirable place.
A number of other technical limitations of these methods were discussed during the workshop. For example, FMRs were developed to run HUD’s Section 8 certificate and voucher program and not for poverty measurement purposes. FMRs measure the gross rents of recent movers, not the entire rental stock. Rental markets can be volatile. Methods for determining FMRs sometimes vary across areas. Charles Nelson’s (Census Bureau) presentation of the FMR method for making geographic adjustments listed 12 limitations of the methodology (Nelson, 2004). Discussants John Ruser and Mark Shroder (Department of Housing and Urban Development) were also highly critical of the FMR approach to adjusting poverty thresholds. They both suggested that further research on other approaches to making geographic adjustments was necessary. One possible avenue for future research mentioned involves using Consumer Price Index data to construct interarea price indexes. Such indexes do not yet exist for the entire country.
In the open discussion period, Timothy Smeeding (Syracuse University) and Rebecca Blank (University of Michigan) argued that while incorporating geographic adjustments to poverty thresholds in a poverty measure was appropriate in principle, the methods currently available to make these adjustments were simply too crude, especially in light of the fact that these adjustments have a substantial effect on state-level poverty rates—a politically sensitive issue. Others argued that the methods for geographical adjustment are sound. Many workshop participants argued that regardless of whether the methods were technically acceptable or not, it would not be worthwhile to spend significant resources improving the methods because it is very unlikely that geographical adjustments to the official measure would ever be adopted because of the political infeasibility.
Many—though not all—workshop participants indicated agreement with these views. In looking at the need for further research on improving methods for making geographic adjustments to thresholds and including more than just the variation in the housing costs in possible future adjustments, Rebecca Blank said: “At present we should set aside putting geographical price adjustments into the poverty calculation, but … continu[e] to improve our methodology on how to do that, including research on improving the housing price issue, which is how the geographic adjustments are largely done right now, and its interarea price distribution, as well as work on geographic variation and other prices that might add to our information about housing.”