Disclaimer

This paper reports the results of research and analysis undertaken by Census Bureau staff. It has undergone a Census Bureau review more limited in scope than that given to official Census Bureau publications. This report is released to inform interested parties of ongoing research and to encourage discussion of work in progress.

The National Research Council Panel on Poverty and Family Assistance (Citro and Michael, 1995) recommended poverty thresholds be calculated for a reference family (two related adults with two children) by specifying " . . . a percentage of median annual expenditures for such families on the sum of three basic goods and services-food, clothing, and shelter (including utilities)-and apply a specified ultiplier to the corresponding dollar level so as to add a small amount for other needs" (Citro and Michael 1995, p. 6).

Following this recommendation, the Panel’s treatment of housing (as opposed to shelter that includes utilities) is the same for owners and renters. Although the Panel only used out-of-pocket expenditures to define the thresholds, they also referred to consumption and needs in their discussion of the basic needs threshold and adjustments for different family types (e.g., see Citro and Michael, 1995, p. 102) If the purpose of the poverty threshold is to provide a level of expenditure that represents the consumption costs for food, clothing, housing, and utilities, then we suggest that the valuation or cost of housing consumption be re-examined before a final decision is made concerning the production of the threshold./1

It is likely there is general agreement that expenditures for food, clothing, and utilities are good approximations of the consumption costs associated with these commodities. However the same cannot be said for the expenditure and consumption cost of housing. It is unlikely that the out-of-pocket expenditures for homeowners with low or no mortgages represent their consumption of housing. The Panel’s approach treats the consumption of these owners in the same way as they treat the consumption of owners with mortgages and renters (see Citro and Michael, 1995, p. 148). While, homeowners with low or no mortgages have relatively low out-of-pocket housing expenses, their consumption costs are expected to be more like those of other homeowners and renters. For such low mortgage households, part of the costs of their housing consumption is being met through the implicit cost of the equity investment in their owned housing unit. If reference families are primarily composed of homeowners with low or no mortgages, the out-of-pocket housing expenditures used in the production of the thresholds would be relatively low compared to their expected consumption costs. Following the out-of-pocket approach would result in an underestimate of these owners’ housing consumption costs because it ignores the implicit cost of their equity ownership of the housing unit. If reference families were primarily composed of homeowners with newer mortgages, their out-of-pocket housing expenditures would be relatively high compared to the expenditures of other owners and renters. If this were the case, an overestimate of the cost of housing consumption could result. Using the out-of-pocket expenditures for owners with mortgages could also result in an overestimate of housing costs because owners with mortgages are allowed to take a tax deduction for mortgage interest paid, thus reducing their “true” costs for housing. Using out-of-pocket housing expenditures also ignores the implicit benefit of house price appreciation, which is one of the primary advantages of homeownership. Furthermore, thresholds based on owner out-of-pocket expenditures are likely to be more sensitive to fluctuations in interest rates and decisions to refinance.

With regard to poverty thresholds, basing owner housing costs on the actual outlays when the estimated housing costs are lower could mean, theoretically, that some owners could quite easily be considered poorer than renters only because these families own their homes and their out-of-pocket housing expenditures are higher. Such could be the case if different thresholds were produced for owners with higher mortgages, for owners with low or no housing costs, and for renters. Producing thresholds by housing status (e.g., own with mortgage, own without mortgage, renter) was an alternative mentioned by the Panel (Citro and Michael 1995, p. 245). We think it is counterintuitive that owners would be more likely to be poor than would renters, given the same amount of housing and other expenses. When out-of-pocket expenditures are higher for owners than for renters living in similar types of dwellings and in the same areas, and only one threshold is produced (using all reference families’ expenditures as is recommended by the Panel) rather than different ones based on housing status, renters would implicitly be “allocated” the higher expenditure amount for their housing consumption. This means that conceivably renters could spend more on other goods and services represented by the threshold.

The Panel acknowledged some of the problems associated with using actual out-of-pocket housing expenditures as reported in the U.S. Consumer Expenditure Interview Survey (CE), however they used these expenditures for processing convenience. They stated that “a preferable definition would include actual outlays for mortgage payments, taxes, insurance, and maintenance and repairs, together with an imputed amount for the estimated rental value of the home net of such outlays. Such a definition would treat homeowners with low or no mortgage payments in a comparable manner with other homeowners and renters” (Citro and Michael 1995, p. 148). The Panel noted that such an approach would account for the implicit costs of housing consumption of owners with low or no mortgages more appropriately. We contend that a better approach than this would be to estimate the housing consumption costs for owners regardless of their out-of-pocket expenditures for mortgage payments, taxes, insurance, and maintenance and repairs. These latter costs would be included in the owners’ reported rental equivalence or imputed housing costs. By following this approach, the housing costs of all owners and renters living in similar housing and the same areas would be treated conceptually the same.

We propose that a consumption approach for owner occupied housing be applied in the production of any new poverty threshold. Such an approach would be based on the costs of the consumption flow of housing services, rather than on out-of-pocket expenditures, for owner occupants. This approach is consistent with other major federal statistical programs including the U.S. Consumer Price Index and Personal Consumption Expenditures of the National Accounts.

In this paper we describe two approaches for estimating the costs of consumption flows of housing services which account for the occupancy of owner occupied housing. One is based on rental equivalence values reported by consumer units participating in the CE Interview. For the other we estimate a value for the flow of services from owner occupied housing using a hedonic approach and renter information. In this paper, we do not deal with the issue of accounting for the value of owner occupied housing in resources.

We compare reference family (families of two adults and two children) medians and thresholds based on out-of-pocket housing expenditure, homeowner reported rental equivalence, and imputed homeowner housing expenditures. These latter two approaches are briefly reviewed and supported in total or in part in the Panel’s report. Participants at the 1998 Brookings workshop on Housing and Geographic Issues in the Measurement of Poverty support our exploration of these approaches for poverty measurement. In addition, Conveners of the Working Group on Revising the Poverty Measure sent an open letter on revising the office measure of poverty (August 2, 2000) that supports additional research on poverty measurement. Signers of the letter include individuals from the 1998 Brookings meeting, a University of Wisconsin conference held in the spring of 1999, and other interested parties. In the letter “Determining how best to treat the flow of services form owner-occupied housing in measuring poverty” is identified as a priority area for additional research (Conveners, 2000, p. 4).