Throwing Money at the Poverty Problem Only Works To a Point
A couple of weeks ago the Council of Economic Advisers released a report recommending tougher work requirements in the nation’s public assistance programs. Midway through the report, readers encountered a startling chart that tracked poverty since the late 1950s. One line in the chart traced the official poverty rate, which fell steeply between 1959 and 1973. Since then we’ve seen fluctuations in poverty, but the long-term trend unmistakably pushed the poverty rate up from its low point in the‘70s. The surprising line in the chart showed the trend in poverty when it is measured based on family consumption rather than family income. In contrast to the official poverty rate, the consumption poverty rate has fallen steadily over the past several decades. It dropped almost 10 percentage points after 1980, falling to just 3.3 percent in 2016.
It is surprising to come across a Trump Administration report acknowledging that the nation’s anti-poverty programs have been a smashing success. The Council’s preferred estimate of the poverty rate has fallen about 27 percentage points—or nine-tenths—since 1960. A sizeable slice of consumption by the nation’s poorest citizens is financed with means-tested benefits, public health insurance, and Social Security, so it’s hard to claim government programs played only a minor role in fueling the long-term decline in poverty. Yet that is a common complaint of conservative critics of the social safety net, who often argue government anti-poverty programs produced effects that are the opposite of those intended. Despite trillions spent on the social safety net, the official poverty rate is higher today than it was in the early 1970s. Progressives who vigorously defend anti-poverty programs sometimes accept the official poverty statistics at face value. Either the nation does not spend enough on anti-poverty efforts or the nation’s harsh version of market capitalism has gradually worsened the plight low-skill breadwinners and their children.
The two lines in the Council’s chart tell starkly different stories about the long-term trend in poverty. Which is more likely to be true? While I’m skeptical of both stories, the official poverty statistics tell the more improbable tale. Both measures of poverty—the first based on family income and the second based on family consumption—establish an absolute threshold of minimal adequacy. Families with too little income or too little consumption to reach the threshold are counted as poor.
The problems of the official poverty statistics are widely known among specialists. For many years the poverty thresholds were annually adjusted in line with a price index that overstated consumer inflation. This error led to an overstatement in how fast incomes had to climb in order to reach the poverty line.
When totaling up a family’s income to determine whether it is above or below the poverty line, the official measure excludes noncash income sources, such as food stamps, housing assistance, and public health insurance. These noncash benefits obviously contribute to a family’s ability to meet its basic needs. By ignoring them the official poverty statistics overlook their sizeable impacts in reducing privation.
What is worse, the government’s provision of noncash assistance grew enormously over time. In contrast, cash assistance for the poor has languished. One exception is the refundable Earned Income Tax Credit, which now provides cash payments to millions of low-income breadwinners every year. Unfortunately, the tax credit payouts are also ignored in estimating the official poverty rate.
When the official poverty thresholds were first proposed in the 1960s, an overwhelming percentage of government aid to the poor and near-poor consisted of cash benefits, and these benefits were all counted in determining who was poor. By the early 1980s an overwhelming fraction of government aid to the poor took the form of noncash benefits. None of the new forms of aid were counted. If we ignore most government benefits when determining who is poor, we cannot reasonably complain that government benefits have had little effect on the poverty rate.
Another problem with the official poverty statistics is that they are derived from incomes reported in household surveys. This wouldn’t be an issue if survey respondents gave accurate information to Census interviewers. However, many respondents underreport the income they receive from government transfer programs. Between 2000 and 2012, only about 46 percent of workers’ compensation benefits, 50 percent of TANF cash benefits, and 67 percent of unemployment insurance benefits were reported on the Census survey used estimate the official poverty rate. Misreporting is not limited to government benefit programs. Census Bureau statisticians recently compared reports of interest and dividend income and pension income obtained on a Census survey with tax return information on the same income items reported to the IRS. For 2012 incomes, they found that survey respondents past age 65 failed to report about one-third of their interest and dividend income and over half their workplace pensions and withdrawals from retirement savings plans.
Some respondents report more income from a given source than they actually receive. However, under-reporting is much more common than over-reporting, and income under-reporting has become a worse problem over time. When a family’s reported income on a survey is cross-checked against tax return data and information from other government sources, the corrected data typically shows more income than originally reported on the household survey.
In sum, the official poverty statistics overstate the percentage of families with incomes below the poverty thresholds. The error occurs, first, because the statistics fail to count benefits from many government transfer programs and, second, because they fail to measure accurately the incomes actually received by many low-income families. Both problems have grown worse over time, suggesting the trend in poverty is less discouraging than implied by the official statistics.
Are the Council’s statistics on consumption poverty more credible? Probably, but they suffer shortcomings of their own. The distribution of family consumption may be even harder to measure than the distribution of income. Few of us can remember the details of our spending. Our income is easier to recall, especially if it is stable and derived from just one or two sources. Further, part of our consumption may be purchased by someone outside our immediate family, for example, through a health insurance plan. We may know what we spend out of pocket for medical care and insurance premiums, but how many of us know the amount our insurance carrier spends in our behalf? Not surprisingly, the consumption data obtained in government surveys is subject to even greater measurement error than income data. Like the income data, consumption data obtained in household surveys has been subject to worsening measurement problems over time.
Regardless of the appeal of a poverty definition that is based on family consumption, the nation will continue to rely on income statistics to determine who is poor. We can improve the definition and measurement of income, however. In determining a family’s resources, we should include the in-kind benefits provided by the government and employers, so long as the benefits help families achieve an adequate standard of living. At the same time, we should stop relying exclusively on household survey responses to measure familiy incomes. Survey responses should be verified using reliable income reports available in tax records and government benefit files.
Throwing money at the poverty problem has worked, or at least has worked better than the official statistics suggest. To gauge actual progress against poverty, however, will require a broader income definition than we now use and a better method for ascertaining and verifying family income.
