December 8, 2020

    Back to Policy Briefs

    December 2020

    By Lydia Wileden and Afton Branche-Wilson

    View the brief in full (Adobe PDF)


    The Detroit Metro Area Communities Study (DMACS) is a University of Michigan initiative designed to regularly survey Detroit residents about their perceptions of neighborhood conditions, economic opportunity, local priorities, and overall quality of life. Since 2016, DMACS has conducted citywide surveys on topics including crime and policing, community health, housing and blight, entrepreneurship, and the impacts of the COVID-19 pandemic.

    In fall 2019, DMACS deployed a survey focused, in part, on capturing baseline data to measure the impact of Detroit’s Strategic Neighborhood Fund (SNF) initiative. The SNF initiative is an effort by the City and its non-profit and private sector partners to stabilize neighborhoods and attract new residents through projects aimed at boosting economic opportunity and improving quality of life. The City and its partners selected ten Detroit neighborhoods—all outside of greater downtown (see Figure 1)—for the SNF program based on their likelihood to quickly benefit from localized investment as measured by their population densities, the presence of a central commercial corridor, strong local leadership, and proximity to public assets.1 Announced in 2014 and formally launched in 2016, the first round of SNF investments targeted improvements in the Southwest/Vernor, Livernois-McNichols, and Islandview/Villages neighborhoods (referred to here as SNF 1.0 neighborhoods). In 2018, the second round of SNF investments was announced for seven additional neighborhoods: Northwest/Grand River, Warrendale/Cody Rouge, Russell Woods/Nardin Park, Campau/Banglatown, Jefferson Chalmers, East Warren/Cadieux, and Gratiot/7 Mile (referred to here as SNF 2.0 neighborhoods). This report draws on fall 2019 DMACS data to offer a snapshot of neighborhood conditions and resident perceptions at the early stages of the SNF program. Deployed as initial investments were coming to fruition in SNF 1.0 neighborhoods and prior to much of the planned development in SNF 2.0 neighborhoods, the survey captures the perspectives of 1,855 Detroit residents, including an oversample of those living in SNF communities.2 The data have been weighted to reflect city demographics. The resulting data enable comparisons between SNF and non-SNF resident perspectives as well as between SNF 1.0, SNF 2.0, and other Detroit residents.

    By highlighting these differences in resident sentiments early in the SNF initiative, this report provides a baseline against which to measure the impact of forthcoming investments in parks, streetscapes, commercial corridors, and housing as well as neighborhood-level feedback on how the City can most effectively and equitably plan for continued economic development in these neighborhoods and across Detroit. Additionally, we highlight suggestive findings on how initial investments in SNF 1.0 neighborhoods may be impacting residents’ perceptions of their neighborhood and offer neighborhood level feedback on how the City can most effectively and equitably plan for continued economic development in these neighborhoods and across Detroit.

    Key Takeaways

    • Residents of the first three SNF neighborhoods (SNF 1.0)—where strategic investment started in 2016—are significantly more likely to report greater overall neighborhood satisfaction and improvements to neighborhood quality of life than other Detroiters. SNF 1.0 residents are also more likely to be satisfied with housing conditions and the availability of neighborhood parks and stores.
    • These same SNF 1.0 neighborhoods have seen more positive signs of neighborhood change— more people moving in, more businesses opening, increasing property values, and increasing neighborhood attractiveness—than other neighborhoods in the city.
    • Despite these reported improvements, residents of SNF 1.0 neighborhoods are no more likely than residents in the rest of the city to express fears over being displaced due to rising housing costs.
    • Perceptions of safety are stronger in SNF 1.0 neighborhoods than other Detroit neighborhoods. SNF 1.0 residents are also more likely to sense their neighborhoods are growing safer, though they remain dissatisfied with the amount of local crime.
    • In the second set of SNF neighborhoods (SNF 2.0)—where strategic investments are only in the planning stage— perceived changes in quality of life closely match the rest of the city. Comparing SNF 2.0 and non-SNF areas going forward can help demonstrate how SNF investments may impact resident perceptions over time.
    • On many measures of neighborhood quality, residents of the 10 SNF neighborhoods do not express greater satisfaction with neighborhood amenities than other Detroit residents, despite their neighborhoods being chosen for SNF in part based on preexisting local assets. These baseline similarities suggest that future comparisons in resident perceptions between SNF and non- SNF neighborhoods may offer evidence of the effectiveness of the SNF strategy.

    See the full results.

  2. Detroit’s Eviction Machine

    October 30, 2019

    Back to research publications


  3. Spending on Government Anti-poverty Efforts: Healthcare Expenditures Vastly Outstrip Income Transfers

    October 18, 2019

    Back to Policy Briefs

    October 7, 2019

    H. Luke Shaefer, Poverty Solutions at the University of Michigan
    Kate Naranjo, Poverty Solutions at the University of Michigan
    David Harris, Children’s Research and Education Institute and Center on Poverty and Social Policy at Columbia University

    Download printable version of brief (Adobe PDF)
    Response from Cato Institute Senior Fellow Michael Tanner (Adobe PDF)
    Accounting for spending on anti-poverty programs (Read-only Microsoft Excel)


    • A widely cited 2012 study by Michael Tanner of the Cato Institute concluded that government spending on anti-poverty measures in the United States totaled almost $1 trillion, with federal expenditures of $668 billion
    • Excluding expenditures on means-tested healthcare programs—which mirrors the construction of most poverty statistics—leads to an estimate of federal spending that is approximately half of what Tanner’s method yields
    • The federal government spends nearly three times as much on healthcare provision for low-income Americans as it does on means-tested cash transfers
    • Many programs that are assumed to target the poor also serve families well above the poverty line
    • Adjusting for these factors considerably reduces estimates of per-person federal spending on low-income Americans


    The government social safety net in the United States is complex, with more than one hundred programs spread across many agencies and levels of government. These dynamics make it surprisingly difficult to answer a seemingly simple question: how much does the federal government spend on anti-poverty programs?

    A 2012 study by Michael Tanner of the Cato Institute concluded that such spending totaled almost $1 trillion, with federal expenditures of $668 billion, and a roughly-estimated $284 billion by states.¹ This analysis led then-Senator Jeff Sessions to conclude that we spent more than $60,000 annually or $168 per day, per household below the poverty line.²

    Adapting Tanner’s approach and updating for 2018, we find that federal expenditures reached $857 billion, the increase since 2012 largely driven by increases in the cost of health care provision and expansion of Medicaid through the Affordable Care Act.³

    A careful examination of the programs included in this estimate raises important questions about what “counts” as spending on poverty. Does the federal work-study program that subsidizes the wages of lower-income college students count? Should spending on low-income working families above the poverty line through the Earned Income Tax Credit (EITC) be included? What about programs to support services on Native American reservations?

    The most consequential decision by far is whether to include spending on healthcare programs for low-income Americans in a list of anti-poverty programs. The Census Bureau’s official poverty metrics excludes resources from government healthcare spending, instead focusing on government programs more directly targeting the economic resources of low-income Americans. Adopting this stance in a tally of federal anti-poverty expenditures leads to an estimate of $393 billion in 2018, approximately half of what Tanner’s method yields for the same year. Moreover, a sizable portion of this spending goes to people who are living above the poverty line.

    The stark difference between these two estimates brings into sharp relief the divide between what the federal government spends on healthcare versus what it spends on income support. It also shows that the answers to thorny questions of what does and does not count as anti-poverty spending have major implications for one’s conclusions.

    Spending on Healthcare, the Elephant in the Room

    Healthcare in the U.S. is exceedingly expensive. Annual spending was estimated to be $3.5 trillion in 2017, more than $10,000 per person and nearly 18 percent of our national Gross Domestic Product. Much of this is subsidized by the federal government through preferential tax treatment for corporations and individuals.4 Because the U.S. lacks a universal healthcare system—despite having the most expensive healthcare system in the world—provision for low-income Americans consumes major federal resources.5

    The Federal government spends nearly three times as much on healthcare provision for low-income Americans as it does on means-tested cash transfers.6 Federal Medicaid expenditures in 2018 totaled $415 billion in direct vendor payments. Another $26.5 billion was eaten up by administrative costs and vaccine coverage, an amount that is roughly $10 billion more than the federal government spends on the nation’s primary cash assistance program for poor families with children, Temporary Assistance for Needy Families (TANF), and 30 percent more than we spend on school meal programs. Adding $17.3 billion in expenditures on the Child Health Insurance Program means that 54% of total federal expenditures on poverty goes to healthcare, when such expenses are included. If healthcare is considered poverty spending, then one way to seek cost efficiencies might examine the structure of our expensive healthcare system, rather than relatively efficient programs like Medicaid that have to operate within it.7

    Furthermore, a meaningful fraction of Medicaid spending supports families with vulnerabilities other than very low income. In 2014, the Kaiser Family Foundation found that $283 billion in state and federal Medicaid spending went to aged and disabled populations.8 As the Center on Budget and Policy Priorities writes, “much of Medicaid spending for long-term care is for people who were middle class for most of their lives, but whose long term care expenses now outstrip their savings.”9

    Healthcare spending certainly improves the circumstances of poor and low-income families. While the Oregon Medicaid Study showed inconclusive impacts on physical markers of health, it found that Medicaid coverage led to improved mental health, reduced financial hardship due to medical costs, “and almost completely eliminated catastrophic out-of-pocket medical expenditures.”10 Yet one might argue that the risk of catastrophic out-of-pocket medical expenditures is all but a uniquely American phenomenon because of the ways our system is constructed.

    Because of these complexities, spending on means-tested public health insurance for low-income Americans might be more easily considered spending on healthcare rather than poverty. Such spending has never been included in official poverty measures because of the analytic challenges that doing so presents. Even so, such programs clearly benefit poor and low-income families, and recognizing their costs and benefits is important. For these reasons, while our core estimate of federal expenditures on poverty excludes healthcare spending, we also continue to present estimates that include healthcare spending.


    Determining what counts as anti-poverty spending requires case-by-case decisions on hundreds of programs, decisions about which reasonable analysts will likely disagree.11 Tough questions must be asked: what fraction of the population can be served before a program is no longer for the poor? Should any program that serves lower-income Americans, among others, count?

    Many programs that are assumed to target the poor actually serve families well above the poverty line. For example, more than half of all infants in the United States receive the Supplemental Nutrition Program for Women and Children (WIC). While WIC targets low-income families, widespread utilization demonstrates its value to one out of every two infants in the U.S.12 The EITC, child tax credit, school food programs, and even the Supplemental Nutrition Assistance Program (SNAP) all serve substantial numbers of individuals above the poverty line. Should these be considered poverty spending? Should only the dollars going to families below the poverty line count?

    Tanner’s original estimates included Pell Grants, which enable millions of lower income students—many above the poverty line—to enroll in college, greatly expanding college access.13 Similarly, one might consider the Federal Work Study Program ($1.1 billion in 2018), which subsidizes students to work for their universities while in school. Both programs enable students above and below the poverty line to pursue higher education and improve their own economic mobility long-term. Is this anti-poverty spending?

    Tanner also includes numerous programs that serve Native Americans, which total $1.1 billion in expenditures in 2018. While this spending surely supports a population that is disproportionately low-income, not all of it does. In addition, one could argue it should be considered recompense for the many ways the U.S. has systematically disadvantaged this population, rather than poverty spending.

    Similarly, there are programs with even more universal benefits included in the original estimate. For example, a $8.7 billion 2018 expenditure on the Universal Service Fund—an FCC program that is focused on modernizing communications technology across the country, including a “geospatial information system solution.” This spending, while targeting rural communities that the market fails, is still improving infrastructure for everyone.

    While some might conclude that none of these programs should be included, others could argue that $857 billion is far too low. Agricultural subsidies clearly help lift some family farmers out of poverty. The poor are disproportionately likely to be incarcerated, thus perhaps corrections spending should be added. Social Security is the most successful anti-poverty program, keeping 15 million seniors out of poverty at a cost of roughly $1 trillion annually, so a strong argument could be made for adding some fraction of these expenditures.14,15 A similar argument could be made for unemployment insurance. Without any parameters, we run the risk of labeling nearly all public expenditures “poverty” spending.


    Below we offer a menu of options based on different decisions about what counts and what doesn’t.

    • Estimate A includes means-tested health care expenditures and adds in the refundable portion of the child tax credit to it, raising his federal poverty spending estimate to $857 billion
    • Estimate B concludes that healthcare spending is better accounted for in health rather than poverty, but counts everything else, for a total of $393 billion.
    • Estimate C includes only income support programs that would directly impact the official or supplemental poverty measure, for a total of $278 billion.

    Thus, three estimates of what the federal government spends on poverty differ by almost $580 billion.

    What does this equate to per person? The discussion above makes it clear that it is inappropriate to limit the denominator to the number of people in poverty in such a calculation, because so much of this spending goes to households above the poverty line. Thus, we compare results of per person spending for those in poverty to two, more accurate denominators: 1) the number of people below 150 percent of the federal poverty line and 2) the number of people below 200 percent of poverty. We use official estimates for 2017 (most recent numbers available). We use official poverty rates rather than supplemental poverty rates for simplicity (OPM rates do not include many of the income transfer expenditures included in the expenditures examined here).16

    We see that these assumptions matter a great deal for resulting estimates of federal anti-poverty spending per low-income person. Taking the full set of programs from the original Cato report and dividing it by the number of individuals below 150 percent of poverty yields a value of $12,704 spent per low-income person. This is well below per person estimates using only those below the poverty line.

    Simply taking out health expenditures immediately cuts this number by more than half (from $12,704 to $5,825) and reaching up to 200 percent of poverty cuts it by a factor of three (from $12,704 to $4,109). When we include only cash and near-cash transfers—in-kind resources or cash transfers that go directly to people—we see that spending equates to $2,908 per low-income person below 200 percent of poverty.


    The complexity of the federal social safety net makes it difficult for policymakers to evaluate the aggregate outcomes of anti-poverty spending, or even to agree on a dollar amount of that spending. It also makes it challenging for low-income individuals to navigate a confusing system. Simplifying the safety net is a policy goal that could lead to improvements over the current system. However, implementation of such a goal is fraught with challenges. Such changes could have major impacts on the well-being of poor and low-income families—for better or for worse—depending on what the changes are.

    At more than $450 billion, Medicaid and CHIP comprise considerably more than the federal government spends on means-tested income support programs. This leads to an important insight, which is that if healthcare spending is included—and policy makers want to curb costs—a focus on curbing the broader costs of our extremely expensive healthcare system is merited.

    What the federal government spends relatively little on is direct means-tested cash transfers, and much of this goes to low-income families above the poverty line and not to the poorest. In-kind and cash means-tested income transfers represent 32% of total federal anti-poverty expenditures overall.17 Cash transfers represent 18% of such federal expenditures. Given recent research on the efficacy of cash transfers, we argue that new aid may be more effective if it takes the form of cash income transfers.18

    While scholars will continue a vigorous debate about how much we, as a nation, spend on anti-poverty efforts, even asking the question can lead to a deeper understanding of the landscape, and uncover areas of agreement among scholars with very different perspectives.


    1 Tanner, M. (2012, April 11). The American Welfare State: How We Spend Nearly $1 Trillion a Year Fighting Poverty—And Fail. Retrieved from https://$1-trillion-year-fighting-poverty-fail

    2 Total Welfare Spending Equates to $168 Per Day For Every Household In Poverty (2012, December 12). Retrieved from https://www.budget.senate. gov/newsroom/budget-background/total-welfare-spending-equates-to- 168-per-day-for-every-household-in-poverty

    3 To begin our accounting, we used Tanner’s program list as the universe of poverty programs to provide as accurate a comparison to his original estimate as possible. Note that we add one program to his list: the refundable portion of the child tax credit, which is largely considered an anti-poverty program. Note also that we include all means-tested Medicaid expenditures, even long-term care, for consistency with the uniform data we use.

    We then identify each program using the Fiscal Year 2020 Budget Appendix (, produced by the Government Performance Office. This document “contains detailed information on the various appropriations and funds that constitute the budget” and provides the true expenditure from FY2018.

    As much as possible, we used the Budget Appendix as the sole source of budgetary numbers to avoid double counting any expenditures or using expenditure amounts from different years. When this was not possible, we used agency specific budgets to find expenditures. These are all noted in the linked spreadsheet. Many programs from the 2012 estimate have been phased out and/or changed names due to the change in presidential administration; this accounts for the zeroes in the expenditure column.

    Tanner’s original estimate includes Consolidated Health Centers at an amount of $2.2M; the correct value is $2.2B. We have included the 2018 expenditure of $4.2B.

    We do not attempt to measure state-level spending given the major uncertainty about such a number.

    A special thank you to Tiffany Loh, Natalie Peterson, and Kaitlyn Bimberg for their careful accounting work.

    4 Zernike, K. (2017, July 07). The Hidden Subsidy That Helps Pay for Health Insurance. Retrieved from health-insurance-tax-deduction.html

    5 It is worth noting that Medicaid, the primary health coverage program for low-income families, makes up only 17% of total healthcare spending. Source: Center for Medicare and Medicaid Studies. Retrieved from

    6 We include Supplemental Security Income (SSI), the EITC, the refundable portion of the CTC, and TANF in the category of means-tested cash transfers.


    8 Kaiser Family Foundation. (n.d.). Medicaid Spending by Enrollment Group. Retrieved from

    9 Parrott, S., Charite, J., Dutta-Gupta, I., & Sherman, A. (2013, February 25). Center on Budget and Policy Priorities. Retrieved from https://www.cbpp. org/sites/default/files/atoms/files/2-25-13pov.pdf

    10 Baicker, K., Taubman, S.L., Allen, H.L., Bernstein,M., Gruber, J.H., Newhouse, J.P., Schneider, E.C., Wright, B.J., Zaslavsky, A.M., and Finkelstein, A.N..(2013, May 02). The Oregon Experiment – Effects of Medicaid on Clinical Outcomes. Retrieved from NEJMsa1212321#t=article

    11 We used Tanner’s program list, adding the refundable portion of the CTC, as the starting point for this discussion.

    12 Martinez-Schifer, M. (2012, May 03). Half of all infants receive WIC. Retrieved from

    13 Protopsaltis, S. and Parrott, S. (2017, July 27). Pell Grants—a Key Tool for Expanding College Access and Economic Opportunity—Need Strengthening, Not Cuts. Retrieved from pell-grants-a-key-tool-for-expanding-college-access-and-economic-opportunity

    14 Romig, K. (2019, July 19). Social Security Lifts More Americans Above Poverty Than Any Other Program. Retrieved from research/social-security/social-security-lifts-more-americans-above-poverty-than-any-other-program

    15 Center on Budget and Policy Priorities. (2019, January 29). Policy Basics: Where Do Our Federal Tax Dollars Go?

    16 A shortcoming for all of these denominators is that they should be adjusted upward by subtracting the anti-poverty effects of programs included in the analysis. Poverty would be higher, if it were not for programs included in official poverty estimates such as TANF and supplemental security income. However, for simplicity, we use official numbers for this exercise. Given this, our preference is to use the <200 percent of poverty threshold.


    18 Shaefer, H.L., Collyer, S., Duncan, G., Edin, K., Garfinkel, I., Harris, D., Smeeding, T.M., Waldfogel, J., Wimer, C.,Yoshikawa, H.. A Universal Child Allowance: A Plan to Reduce Poverty and Income Instability Among Children in the United States. (2018, February). Retrieved from

  4. Hello world!

    August 10, 2018

    Welcome to Ford School of Public Policy Sites. This is your first post. Edit or delete it, then start blogging!