November 5, 2019 | The Groundwork Collaborative, Center on Poverty & Social Policy at Columbia University
The Costs of Being Poor: Inflation Inequality Leads to Three Million More People in Poverty
It is widely recognized that income inequality has skyrocketed in recent decades. Incomes at the top of the distribution have grown rapidly, far outpacing income growth at the bottom. Recent research also shows that prices have risen more quickly for people at the bottom of the income distribution than for those at the top — a phenomenon dubbed “inflation inequality.” An implication of this new finding is that we may be underestimating income inequality and poverty rates in the United States — two national statistics that rely heavily on the annual inflation rate as part of their calculation. In this brief, we utilize an adjusted inflation index that accounts for inflation inequality across the income distribution and re-estimate recent trends in poverty and income inequality from 2004 to 2018. Our adjusted inflation index indicates that 3.2 million more people are classified as living in poverty in 2018, and that real household income for the bottom 20 percent of the income distribution actually declined by nearly 7 percent since 2004. These results show that inflation inequality significantly accentuates both the incidence of poverty and income inequality.
How many people in the U.S. are poor? How much more income do the richest have than the poorest among us? Is income inequality rising or falling? To most, these would seem fairly basic economic facts, easily estimated from government-collected data. But as with most economic indicators, the devil lies in the details. One of these key details involves how to measure inflation, or the changes in prices faced by consumers over time. If a household, for example, had $25,000 in income in 2018 and $27,000 in 2019, their income obviously increased. But because the goods and services these individuals purchase with their income could cost more (or less) in 2019 than in 2018, the difference in so-called “real” income, or nominal income adjusted for inflation, may not actually be $2,000. Accounting for inflation is thus key to determining trends in the economic well-being of the workforce and the population.
Government statistics, from poverty rates to household income to wages, rely on measures of inflation to compare the economic well-being of people today to that of people in the recent and more distant past. The most commonly used measure of inflation is some version of the Consumer Price Index (CPI), calculated by the Bureau of Labor Statistics. Some economists prefer alternative measures, such as the Personal Consumption Expenditure (PCE) index, which accounts for price changes differently than the CPI. But what all of these measures have in common is the assumption of a single rate of change in the average price of goods and services faced by the population as a whole.
But what if that assumption does not hold? Recent evidence suggests that it may not. Just as aggregate measures of GDP may mask variation in economic growth at different points in the economic distribution, aggregate measures of inflation may mask the fact that the prices and price changes faced by the poor may be fundamentally different from the prices and price changes faced by the middle class, which in turn may be fundamentally different from the prices and price changes faced by the rich. In a 2019 paper in the Quarterly Journal of Economics, the London School of Economics professor Xavier Jaravel (and coauthor of this brief) finds just that. To examine differences in annual inflation rates faced by lower- and higher-income Americans, Jaravel uses several price and expenditures datasets, including scanner data collected in retail stores from 2004 to 2015. He finds that annual inflation rates for those at the bottom of the income distribution are substantially higher than for those at the top of the income distribution, effectively increasing income inequality.
Jaravel points to increasing income inequality as the root cause of this new “inflation inequality.” Soaring income inequality in recent decades has been driven almost entirely by increasing incomes at the top of the income distribution. He shows that 2 because companies are increasingly interested in competing for the dollars of these wealthy individuals, prices for goods that wealthy people buy are actually decreasing relative to the prices of goods that lower-income families purchase. In other words, as income inequality has increased, companies have increasingly catered to families with high incomes, driving down prices for the goods they buy, and further increasing real income inequality. In the meantime, poor families face prices and price changes that are “business as usual.”
This brief poses a relatively straightforward question: How would recent trends in poverty and inequality differ if we accounted for the differential inflation trends between the rich and the poor that Jaravel found? To answer this question, we utilize an adjusted inflation index produced using data from Jaravel’s paper and apply this index to official poverty thresholds and household income measures from 2004 to 2018. The results are striking. When you properly account for variation in prices by income, poverty is noticeably higher and income inequality widens.
We calculate revised measures of poverty and inequality using inflation estimates derived from the data and methodology developed in Jaravel (2019), starting from 2004. With the baseline approach applied between 2004 and 2015, Jaravel’s paper finds that the annual inflation rate is 0.44 percentage points higher for the bottom income quintile compared with the top income quintile, on average. We apply this 0.44 percentage point correction to the aggregate measure of inflation conventionally used to calculate official poverty line, the threshold below which individuals are considered to be in poverty and become eligible for certain government support programs. Furthermore, we 1. The bottom fifth of the income distribution is a reasonable proxy for the poor population since 11.8 percent of Americans were found to be poor in 2018. assess trends in “real” household income using this 0.44 percentage point correction.¹ The Appendix provides a complete description and discussion of the methodology.
1 The bottom fifth of the income distribution is a reasonable proxy for the poor population since 11.8 percent of Americans were found to be poor in 2018.
We begin by examining how recent trends in poverty would differ if more accurate measures of inflation were used. The poverty line, or threshold, was developed in the 1960s and was based, at the time, on the cost of food as a proportion of family budgets. Since then, the poverty line has been updated each year for inflation using the CPI. In 2018, the poverty line for a family of four is just over $25,000. A single individual is not considered poor if they make just a bit over $12,000. By most domestic standards, these are very low levels of need. So, when eligibility for important government benefits is tied to these poverty lines (for example, the Supplemental Nutrition Assistant Program and Medicaid), small changes to how those lines are calculated can make a big difference. To measure the implications of utilizing an inflation measure based on the price increases faced by the lowest income individuals, we use Jaravel’s data and apply to the poverty threshold an inflation index that adjusts for inequality across the income distribution.
Figure 1 shows the number of individuals in poverty under the official measure and according to our adjusted measure. With the adjusted CPI, the number of people in poverty in 2018 is about 8 percent larger than under the official measure, which corresponds to an increase in the poverty rate of approximately 1.0 percentage point. This is a large difference. Using the adjusted CPI translates into over 3.2 million more people classified as living in poverty, or about the population of the entire state of Iowa.² This means that millions of people who might reasonably qualify for benefits like food and housing assistance from antipoverty programs that could help them and their families, do not qualify. It also means that we have a too-optimistic portrait of the number of families who are struggling to make ends meet. Keep in mind as well that this divergence in poverty rates is apparent just when using a relatively narrow thirteen-year window. Over a longer time period, this divergence is likely to compound further.
We can also break out poverty changes by various demographic subgroups. Tables 5-8 in the Appendix provide the number of individuals in each subgroup newly classified as poor if we considered Jaravel’s income specific inflation rates as the relevant measure. Child poverty rates would be 1.5 percentage points higher in 2018, corresponding to 1.1 million additional children in poverty in that year. We see similar trends for women-headed households with children, by race and ethnicity, and by gender respectively. 1.2 million more individuals in women-headed households would be newly counted as poor. This is especially significant given that mothers are now the breadwinners in 40 percent of all households.³
Furthermore, 1.2 million more white non-Hispanic individuals would be newly counted as poor when using the updated inflation measures, increasing the white poverty rate by 0.6 percentage points to 8.7 percent in 2018. In addition, 650,000 more Black non-Hispanic individuals would fall under the poverty line increasing the Black poverty by rate by 1.6 percentage points to 22.7 percent in 2018. The Latinx/Hispanic poverty rate would increase by 1.8 percentage points to 19.4 percent in 2018, meaning 1.1 million more Latinx/Hispanic individuals in poverty. Additionally, 1.7 million women, and 1.5 million men would be newly counted as poor.
We repeat the analysis using the deep poverty rate, with deep poverty defined as living below half the poverty line. Since the poverty line for a family of four is only about $25,000, this is a marker of fairly severe deprivation. Again, we see greater deep poverty over time when using a more accurate measure of the prices that low-income families pay. By 2018, this means over 800,000 more people would be classified as living in deep poverty relative to the standard, official measure of poverty. (see table 3 in Appendix for more details.)
It is clear that relying on a population-wide measure of inflation when calculating the poverty rate understates the deprivation faced by the most vulnerable adults, children, and families. To develop policies that improve the economic well-being of low-income individuals, we must first accurately measure their well-being. That should start with an accurate understanding of the prices they pay for the goods and services they consume.
We can also use updated inflation measures to reassess recent trends in the divergence of household incomes. Every year, the Census Bureau publishes statistics on household income changes, again adjusted using a version of the CPI. It is now widely recognized that income inequality has skyrocketed in recent decades due to extreme growth in income at the top of the distribution. Incomes at the bottom have stagnated, unless you count the resources families receive from government income-support programs that try to compensate for low market incomes. There is a large and growing gap between the incomes of the top quintile of households and the bottom quintile between 2004 and 2018, the period of our study. When we apply inflation corrections to household incomes during this period, we see that this gap is growing even more over time than one finds using the CPI alone.
Figure 2 reports changes in income across quintiles of the income distribution, using the conventional CPI and our adjusted inflation rates. All figures are adjusted for household size before dividing households into quintiles. Using conventional estimates based on the CPI, we see that income in the top fifth of the distribution grew by 16.6 percent between 2004 and 2018. Concurrently, income at the bottom fell by about one percentage point. Thus, income inequality between the top and bottom quintiles widened by more than 17 percentage points over the period (2004-2018) using conventional measures of inflation. If, instead, we use our adjusted inflation measure, we find that incomes in the bottom quintile actually declined by 6.7 percent.4 Put another way, according to our adjusted measure purchasing power fell significantly at the bottom of the distribution, which the conventional CPI largely misses. Using the adjusted inflation measure, income inequality between the top and bottom quintiles widened by about 23 percentage points, or about 33 percent more than with the conventional measure of inflation (17.6 vs. 23.4 percentage points).
2 If we instead use a less conservative adjustment of 0.66 percentage points per year (see the methodology appendix), 4.6 million more Americans would be classified as poor, as opposed to the 3.2 million found here with the more conservative approach.
3 Glynn, Sarah Jane, “Breadwinning Mothers Continue To Be the U.S. Norm” (Center for American Progress, 2019), available at https://www.americanprogress. org/issues/women/reports/2019/05/10/469739/breadwinning-mothers-continue-u-s-norm/.
4 As previously, we assign the CPI-U to the top income quintile, and we correct inflation rates for the bottom income quintile using Jaravel’s data.
In sum, our results show that if we take seriously the idea that inflation varies across different points in the income distribution, a different picture of the economic health of those with low incomes emerges. Jaravel’s research shows that, at least in recent years, inflation is steeper at the bottom of the income distribution. If we apply this steeper inflation to the poverty threshold, we see that millions more people would be classified as living in poverty. The divergence in household incomes that has been underway for decades is also notably larger than we thought.
Despite this evidence, which shows that our current practice of using aggregate inflation measures significantly understates income inequality and poverty, recent proposals by the Trump administration seek to use smaller inflation rates to adjust the poverty threshold. Doing so would result in a lower poverty line and lower official poverty rates over time. This would, in turn, mean that fewer and fewer low-income Americans would find themselves eligible for federal benefits, which have been shown to reduce the poverty level substantially.
Taking seriously the actual inflation rates faced by those at the bottom of the income distribution indicates that we currently have a too-rosy view of the actual levels of deprivation faced by those at the bottom of the income distribution. We should be doing more, not less, to help them make ends meet.
For our baseline estimates, we follow the conservative approach described in Jaravel’s paper, which is based on data from the CPI combined with the Consumer Expenditure Survey. This approach has two advantages. First, it covers the full consumption basket of American households, and second, it closely follows the official methodology of the CPI to compute inflation. This approach may understate the required adjustment, however, because the correction for inflation inequality is found to be larger when using more granular data available for products with barcodes.
With the baseline approach applied between 2004 and 2015, Jaravel’s paper finds that the annual inflation rate is 0.44 percentage points higher for the bottom income quintile compared with the top income quintile. With barcode-level data (which may not apply to all goods purchased by Americans), the annual inflation rate difference increases to 0.66 percentage points. As discussed in Jaravel’s paper, with barcode-level data the correction is larger because inflation inequality turns out to exist even within detailed consumption categories (e.g., between organic spinach and regular spinach), while the baseline approach can only capture the part of inflation inequality that arises between product categories (e.g., between spinach and beef). In order to be conservative, we focus on the baseline inflation inequality estimates derived from the CPI-CEX data. We also briefly report the (larger) correction derived from the alternative approach using more granular data for products with barcodes.
With both approaches, we take the CPI-U as our reference point. The expenditure weights used in CPI-U are the aggregate expenditure shares for the whole economy, which effectively track changes in the cost of living for a fairly affluent household close to the top quintile of the income distribution (see Deaton 1998, Hamilton 2001, and Almas, Beatty and Crossley 2019). Therefore, we assign the CPI-U to the top income quintile, and we correct inflation rates for other income quintiles using Jaravel’s data.
With both the baseline approach and the barcode-level data, the correction for inflation inequality is very similar in magnitude for all years of the sample, therefore we use a constant correction factor for the whole period. Results remain unchanged when using year-specific inflation-inequality estimates.
See Sections 2 and 3 of Jaravel (2019) for a description of the data sources and of the methodology to compute the adjustment for inflation inequality (in particular, Figures 2 and 3, and Tables 2 and 3).
Table 1. Inflation Rates
Table 2. Poverty Rates and Counts
Table 3. Deep Poverty Rates and Counts
Table 4. Poverty Thresholds
Table 5. Child Poverty Rates and Counts of Children in Poverty
Table 6. Poverty Rates and Counts of Individuals in Female Headed Households
Table 7. Poverty Rates and Counts by Race/Ethnicity
Table 8. Poverty Rates and Counts by Gender
Table 9. Average Income in 2004 Adjusted Using the CPI-U and the Income Group Specific Inflation Rates for Higher and Lower Income Households
Table 10. Percent Reduction in Average Income Between 2004 and 2018 CPI and Income Specific Inflation
Table 11. Poverty Rates and Counts – Two Version of Inflation Inequality
Christopher Wimer is co-Director of the Center on Poverty and Social Policy (CPSP) at Columbia University. He conducts research on the measurement of poverty and disadvantage in both local and national contexts, as well as historical trends in poverty and the impacts of social policies on the poverty rate. His work pays particular attention to the role of government policies and programs and their potential impacts on the wellbeing of low-income families and children.
Sophie Collyer is a Research Analyst at the Center on Poverty and Social Policy (CPSP) at Columbia University. Her work evaluates the impacts of national and local anti-poverty policies, with a particular focus on reforms to the tax code. Collyer began working with the CPSP as a graduate student at Columbia where she completed a dual-degree (MPA/MSW) at the School of International and Public Affairs and the School of Social Work.
Xavier Jaravel is an Assistant Professor of Economics at the London School of Economics. His work examines the factors that can help create and sustain inclusive growth, with a particular focus on innovation, inflation, and trade. Prior to joining the LSE he was a postdoctoral fellow at Stanford University and obtained his PhD from Harvard University.