may indeed occur, but Richard Bavier (Office of Management and Budget) pointed out that erroneous poverty classifications resulting from this method were rather modest and the same error also applies to accounting for the cost of housing in the thresholds.
Using expected medical out-of-pocket expenses rather than actual expenses tends to produce slightly higher overall poverty rates—about 0.4 to 0.6 percentage points (Proctor and Dalaker, 2003). Elderly poverty rates are relatively lower and child poverty rates are relatively higher if expected costs are added to thresholds rather than subtracting actual costs. Jessica Banthin (Agency for Healthcare Research and Quality) asserted that this difference between the two methods increases the confidence of many researchers and policy makers in using expected costs.
A third approach to dealing with medical out-of-pocket expenses that was raised during the workshop and which has appeared in Census Bureau reports involves implementing a combination of the two methods above. First, expected medical out-of-pocket expenses are added to thresholds. Next, the difference between an estimate of actual out-of-pocket expenses and expected expenses is calculated. This net out-of-pocket amount is then subtracted from family income. This method has the advantage of replicating the distribution of actual expenses, though Gary Burtless (The Brookings Institution) argued that this is not necessarily a preferable feature, given that extreme expenses by the elderly in particular are often financed by assets or public funds rather than by income.
Many workshop participants indicated support for accounting for medical out-of-pocket expenses in a new poverty measure. While many participants expressed support for adding expected medical out-of-pocket expenses in the poverty thresholds, there was a lack of consensus on how exactly to do so. Many participants also voiced support for adjusting expected expenses for underconsumption among the uninsured, and for not having the calculation of expenses affected by extreme values sometimes observed in the data.
Rebecca Blank (University of Michigan) summarized her impression of the wide-ranging discussion: “We should account for … medical out-of-pocket expenses; we should do some adjustment for the uninsured; and we should top code the calculation, to get rid of those who really hit catastrophes, with the idea that that’s picked up in some other ways….” However, she noted, there really was substantial disagreement about whether medical out-of-pocket expenses should go into the threshold or should be imputed into people’s income.