2
Setting and Updating Thresholds
The 1995 National Research Council (NRC) report offered several recommendations on how poverty thresholds for a reference family should be constructed. The report called for specifying a percentile of median annual expenditures of a family of two adults and two children on food, clothing, shelter, and utilities and applying a specific multiplier to the resulting dollar level to cover a small amount for other needs. Because an element of judgment is needed in the selection of an initial poverty threshold, the panel offered a range of values, in essence representing 78-83 percent of the median of expenditures for the basic bundle and a multiplier of 1.15–1.25.
The report also recommended that the threshold should be updated annually to reflect changes in median expenditures on the basic bundle of goods using the most recent 3 years’ worth of data from the Consumer Expenditure Survey (CE), adjusted to current dollars. The advantages of this method, according to the report, are that thresholds would be updated in real terms on a regular basis from changes in spending on categories of basic goods, rather than on just price changes of all goods, as is done when using the Consumer Price Index (CPI) (for details, see Citro, 2004).
The workshop discussion on setting a revised threshold revolved around two issues: how to select a specific dollar value for the reference family threshold and the method to use for updating it over time. Setting a reference family threshold to a specific dollar amount inevitably involves making some judgments. Constance Citro (National Research Council)
noted that this is not to say that these choices are arbitrary or capricious; rather, judgment is needed, along with a thorough understanding of the issues at hand.
One question that arose was whether setting a dollar value for the reference family threshold in the new measure should, in the interest of maintaining some continuity between old and new measures, be influenced by the current official measure. A few alternatives were discussed, including using an “equal rate” method, which would set the new threshold at a level that would, by design, produce a poverty rate that equaled the official poverty rate in a particular base year (after which it would presumably diverge in one direction or another). The advantages of this method are that it would provide a more seamless change in measured poverty rates from the current official measure, and it would provide a good sense of how the composition of the poverty population differs when using the alternative measure. The main disadvantage of this method is that the threshold would in essence be an artifact and not inherently meaningful—its level would depend entirely on the official poverty rate in a given year.
An alternative method for achieving some continuity between the old and new poverty measures would be to set an “equal threshold,” for which the new reference family threshold is about the same dollar value as the official threshold. This approach could be helpful heuristically in making the transition to a revised poverty measure. Discussion of these alternatives indicated that there was little support for the “equal rate method” (setting the poverty rate of the alternative poverty measure to equal the official poverty rate in a given base year).
Many participants voiced support for the approach implemented in current Census Bureau reports on experimental poverty measures. With that approach, the dollar value of food, clothing, shelter, utilities, and a little more in the reference family threshold (for two adults and two children) does not differ very much from the reference family threshold in the current official measure; the CE-based reference family threshold was about $1,000 higher than the official threshold in 2002. The similarity in the thresholds, though, is not by design; rather, the NRC-recommended method of pricing expenditures on the basic items in the threshold just happens to be similar to the reference family threshold in the current measure.
It was also noted, however, that the revised threshold, while similar in value, actually represents as much as a 19 percent real increase in the value of the reference family threshold in recent Census Bureau publications,
given the differences in how resources are measured (Citro, 2004:19). That is, if alternative poverty measures were to add medical out-of-pocket and work-related expenses (including child care expenses) to the thresholds rather than subtract them from people’s incomes (the latter is currently done in most of the alternative measures in published reports), then thresholds are indeed significantly higher in the new measures.
Two main options for annually updating the reference family threshold were discussed at the workshop: (1) using the CPI or (2) following the original NRC recommendations of using 3 year’s worth of data on people’s actual expenditures on the items contained in the threshold from the CE. The advantage of using the CPI is that it provides an easy way to update the thresholds. However, as the 1995 NRC report argued, the advantage of using the CE-based thresholds is that they rely on a “quasi-relative” updating mechanism, with the thresholds updated on the basis of changes in real expenditures on basic expenditure items. The NRC-recommended measure is termed “quasi-relative” because the proposed update would be based on consumption expenditures for only basic categories of goods and services, which would be expected to rise less rapidly than total expenditures or median income. Three years of CE data were recommended by the NRC so that the thresholds would not be affected as much by random annual fluctuations in the data.
In short, the 1995 NRC report held that the quasi-relative aspect of the recommended thresholds would make them less likely to become outdated over time than the CPI-based thresholds. Public opinion, for example, has indicated that the current official poverty thresholds are too low, even though they once had broad acceptance (see NRC, 1995:137-140; an updated version is in Iceland, 2003). Many workshop participants agreed with the NRC’s recommended method of annually updating the thresholds—that is, to use the most recent 3 years of data on expenditures on the threshold items from the CE. They also generally voiced support for the NRC recommendation for continued research on the behavior of CE-based thresholds over time.