Adjusting Thresholds for Geographic Area

Robert Blaine and Marya Sosulski

A common criticism of the current official poverty measure is that it fails to account for regional and metropolitan-area variations in prices (Ruggles, 1990; Citro and Michael, 1995). Cities tend to be more expensive to live in than rural areas, and regions such as the Northeast and West tend to have higher living costs than the Midwest or South. Thus, individuals and families living in these areas will, on average, need more resources to meet their minimal consumption needs.

Such adjustments have never been made in the current poverty measure in part because policymakers lack a sufficient data source with which to estimate price differences across all of these areas. Moreover, assessing differences in prices often results in numerous subjective judgments that rarely survive criticism. Nevertheless, the Panel on Poverty and Family Assistance of the National Academy of Sciences (NAS) asserts that any improved poverty measure must include such price variation adjustments and that data exist to construct a roughly accurate index.

Approaches to Measuring Variation in Prices

To adjust poverty thresholds to account for differences in prices requires choosing among several methods to estimate costs of living. A common method to adjust for price differentiation is to compare the price of representative baskets of goods in defined regions and metropolitan areas. The Bureau of Labor Statistics (BLS) uses a similar process to calculate the Consumer Price Index (CPI). Shortcomings in this approach, however, have prompted the Panel to favor other measures of differences in prices across regions. For instance, the BLS compares inflation in baskets of goods thought to reflect the purchases of typical consumers, not just lower-income consumers closer to the poverty line. Also, the market baskets are region-specific and reflect differential regional tastes as well as differential needs. The Panel suggests that a fixed-weight interarea price index, currently under development by the BLS, could address these issues (Citro and Michael, 1995).

Choosing the representative market basket relies on several judgments as to what consumers need or use; and the representative basket in, for example, Los Angeles, California, likely differs from one in Nelson, Wisconsin. There seems to be universal agreement, however, that housing is a relevant market basket element, and researchers have found that housing costs tend to vary geographically more than other items such as food and clothing. Based on this idea, the NAS Panel recommends adjusting poverty thresholds for differences in housing costs (including utilities) to represent variations in prices.

Various methodologies exist for constructing housing-cost indexes. Many economists have used hedonic regression models to control for various characteristics of housing such as location, housing quality, and amenities that affect price levels. The Panel, on the other hand, supports a process used by the Department of Housing and Urban Development (HUD) to estimate a "fair market" rent in a given area. The Panel recommends examining each of the nine Census regions to find housing costs across six metropolitan areas (see Appendix A). In looking at the distribution of rents in each area, a fair market rent would be roughly equal to the 45th percentile of the distribution. The Panel also recommends making adjustments for the fraction of the poverty budget accounted for by housing.

To construct an accurate index, the Panel recommends using the Decennial Census, for neither the Current Population Survey (CPS) nor Survey of Income and Program Participation (SIPP) has a large enough data set to estimate rent distributions across all areas. Though the Census does not account for as many variables as the CPS and SIPP, its sheer volume of observations make it the best source of data for a geographic index. The Panel acknowledges the need for more updating in years between Census reporting to account for variations in housing costs. The Panel recommends the use of data from the American Housing Survey, local area CPI shelter cost indexes, and random digit dialing surveys to update housing cost information and adjust thresholds accordingly (Citro and Michael, 1995).

Effects of Adjusting Thresholds for Geographic Variations in Housing Costs

Adjusting poverty thresholds for region and metropolitan size will have impacts on the composition of the poor within these areas. Assuming that overall poverty rates are held constant, the Panel predicts that the poverty rates for the Northeast and West will increase by 2.0 and 1.7 percent respectively, while poverty rates for the Midwest and South will fall by 0.8 and 1.1 percent. Moreover, thresholds will tend to be higher as the metropolitan size of an area increases, corresponding with the costs of living. Thus, one may expect poverty rates in larger metro areas to rise while rates in rural and other non-metropolitan areas are expected to decline.

Several issues remain if the Panels proposal becomes the practice in poverty measurement. First, the Panel suggests that more work should be done to understand how housing costs vary by region, as well as to construct more accurate indexes. The Panel suggests that the current HUD methodology for updating the decennial census housing costs in intercensal years can be improved through more use of hedonic regression. Thus, a commitment to geographic adjustment requires a commitment to further research into developing index methodologies.

Second, the publics understanding of the poverty measure will be affected by making such adjustments. For instance, one advantage of the current measure is that it is reasonably easy to understand. However, the Panels proposal, if enacted, would essentially create 50 different sets of poverty thresholds, one for each metropolitan-area size category within each region. This differentiation may increase the complexity of comparing poverty rates among different areas and subgroups.

Third, critics of housing cost adjustments--and of price adjustments in general--argue that a correlation exists between prices in a given area and income. That is, since people in areas with higher housing costs tend to have higher average incomes, their larger incomes will offset the larger prices. The NAS Panel argues, however, that this correlation may not be relevant at lower income levels at which poverty thresholds are significant. The poverty measure is intended to account for minimum levels of need, and should presumably reflect geographic areas in which low-income families face higher costs. Therefore, in places where prices are greater the thresholds should be higher to correspond with them, even if peoples average income is also greater.


A majority of experts agree that poverty thresholds should account for variations in prices, and measuring differences in housing costs appears to offer the best method for making such threshold adjustments. The NAS Panel does not provide an absolute methodology or plan for making such adjustments. Rather, the Panel recommends that social scientists should continue researching ways to improve housing-cost indexes and develop general price indexes. The lack of a definitive methodology in the Panels recommendations and the problem of finding a sufficient data set may generate the most difficulty for this aspect of the proposal.


Citro, Constance F. and Michael, Robert T. (eds.) 1995. Measuring Poverty: A New Approach.Washington D.C.: National Academy Press.

Ruggles, Patricia. 1990. Drawing the Line: Alternative Poverty Measures and Their Implications for Public Policy. Washington, D.C.: The Urban Institute Press.

Appendix A -- Metropolitan Size Categories

Nonmetropolitan Areas

Metropolitan areas under 250,000

Metropolitan areas 250,000-500,000

Metropolitan areas 500,000-1,000,000

Metropolitan areas 1,000,000-2,500,000

Metropolitan areas 2,500,000 or more

Appendix B -- Census Regions

Region States
1. New England Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont
2. Middle Atlantic New Jersey, New York, Pennsylvania
3. E North Central Illinois, Indiana, Michigan, Ohio, Wisconsin
4. W North Central Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota
5. South Atlantic Delaware, District of Columbia, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia
6. E South Central Alabama, Kentucky, Mississippi, Tennessee
7. W South Central Arkansas, Louisiana, Oklahoma, Texas
8. Mountain Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming
9. Pacific Alaska, California, Hawaii, Oregon, Washington