Measuring Poverty Using Household Outlays

Featured in print Digest


Measuring poverty is a long-standing challenge. The official US poverty rate is based on households’ before-tax-and-transfer income. Alternative measures, derived from household expenditures, have been proposed as more-informative indicators of well-being among disadvantaged households.

In The Supplemental Poverty Measure: A New Method for Measuring Poverty (NBER Working Paper 30056), John Fitzgerald and Robert A. Moffitt develop a new metric for identifying poor households. They use information on all household outlays — which they alternatively term expenditures — reported in the Consumer Expenditure Survey to construct the Supplemental Expenditure Poverty Measure (SEPM). This measure considers not just spending on goods and services, but also outlays such as contributions to retirement accounts, loan payments, and savings — resources that could have been used to buy the minimum necessities, such as food, clothing and housing, that are needed to avoid being classified as poor. Some households classified as poor under standard definitions could, by incurring more credit card debt or dropping their contributions to retirement plans, raise their outlays by enough to rise above the poverty line.

A new measure of poverty derived using all household outlays tracks existing aggregate poverty measures well while also providing new information on poverty rates among population subgroups.

The researchers compare their new measure to the Census Bureau’s Supplemental Poverty Measure (SPM), an after-tax poverty measurement based on income data reported in the Current Population Survey. They calculate what they term a “net” SEPM poverty series that mimics the SPM series, which adjusts for in-kind transfers, various costs of working, and significant out-of-pocket medical costs, but uses net outlays instead of net income. The two track each other closely over time. The average SEPM and SPM rates for the period 2017–19 are 13.3 and 13.0 percent, respectively. Both typically exceed official poverty rates published by the Census Bureau, which are defined quite differently and do not deduct any costs from income.

However, the SEPM and SPM diverge during this period when measuring the poorest of the poor. Households in deep poverty and near poverty are those with annual income and spending below 50 and 150 percent of the poverty line, or about $13,000 and $39,000. The average SPM deep-poverty rate was 4.4 percent, 3.3 percentage points greater than the expenditure-based SEPM deep-poverty rate. There are many more families with very low incomes than very low expenditures, possibly because incomes are underreported. In contrast, the SEPM near-poverty rate was about 5 percentage points greater than the corresponding income-based SPM poverty rate because there are many more families with only modest levels of expenditure than modest levels of income. About a third of the population is poor or near poor according to the spending-based metric.

The two alternative poverty measures are within 1 percentage point for most demographic groups. However, SEPM child poverty rates are greater than SPM child poverty rates after 2010, with the differential reaching 2 percentage points between 2010 and 2013. Removing government transfers would raise both alternative poverty rates significantly. After 2010, the removal of in-kind transfers, including Supplemental Nutrition Assistance Program benefits, would increase both measures by about 3 percentage points. Overall, tax credits and transfers reduce the net SEPM poverty rate by between 4 and 5 percentage points.

Using an expanded definition of the SEPM, the researchers construct an upper bound for the potential resources a household could spend by including potential drawdowns from liquid bank accounts and unused credit card borrowing capacity. Bank balances are low for households at the bottom of the spending distribution — in the bottom quartile, for instance, households with heads younger than 65 have a median balance of zero. Balance drawdowns only have a modest effect on poverty levels. However, adding unused and potential credit card borrowing to potential resources lowers poverty rates by between 3 and 4 percentage points from a base of just over 13 percent. The researchers find that almost 10 percent of households, including over 31 million individuals in 2019, could not buy the minimum bundle of goods, even after entirely depleting their balances and maximizing their credit card borrowing.

— Aaron Metheny