expenses to families in the CPS. The technique of assigning medical out-of-pocket expenses to families in the CPS was based on a regression model designed to replicate the full distribution of actual expenses from the 1987 National Medical Expenditure Survey and then inflating the aggregated level of out-of-pocket expenses to equal benchmarks from the National Health Accounts administrative data. Characteristics used in the regression model include age of householder, medical insurance status, family size, poverty status, and race.

Current experimental poverty reports by the Census Bureau (e.g., Short, 2001a) offer several ways of accounting for medical out-of-pocket expenses: one subtracts estimates of actual medical out-of-pocket expenses from family income; another adds expected expenses to the thresholds; and a combined method does both. The experimental poverty reports also incorporated two main changes from the original report recommendations that affect all of these alternatives: (1) not to inflate medical out-of-pocket expenses to meet administrative benchmarks, as such benchmarking is not currently done with any other element in the experimental poverty measures; and (2) to use data from both the Consumer Expenditure Survey (CE) and the Medical Expenditure Panel Survey (MEPS).

The workshop discussion on medical out-of-pocket expenses centered on whether the poverty measure should include “actual” or “expected” expenditures. That is, should the measure conceptually attempt to take into account people’s actual reported medical expenses, as the 1995 NRC report recommended, or their expected out-of-pocket medical spending needs, based on their demographic and health characteristics (see Banthin, 2004, for more details).

A related question, whose answer in large part depends on the answer to the one above, is whether medical expenses should be accounted for by subtracting “actual” (or imputed) out-of-pocket costs from resources (the NRC-recommended method) or by adding expected need to the threshold, or some combination of the two methods. Another related question is whether out-of-pocket expenses should be adjusted for the underconsumption of medical care by the uninsured. In particular, uninsured people often appear to spend little on medical expenses, though some argue that this does not necessarily reflect less need. Adopting the view that the poverty measure should incorporate expected out-of-pocket expenses (treating medical care as a basic need) would tend to lead one to accept adjusting expenses for underconsumption by the uninsured, while those preferring to