Compared to What?

POSC 459 — Social Welfare Politics and Policy
Week 3 · Wednesday, September 9

Howard ch. 1, "She's So Unusual" · Desmond chs. 3–4

Today

  1. A paragraph, and what is wrong with it (~10 min)
  2. Your own prompt, audited (~20 min)
  3. Howard: is the American welfare state small? (~20 min)
  4. Desmond: underpaid, and overcharged (~20 min)

Read this. Assume nothing.

Four minutes on your own. Mark anything you would want to verify before you believed it.

Then three minutes in pairs: compare lists.

What did you find?

Everything you would want to check before believing it — on the board, before I say anything.

Your turn — in pairs, ten minutes

Open ChatGPT Edu, TitanGPT, or PapyrusAI. Ask it:

"Explain why some welfare programs are more politically popular than others."

Then audit the answer:

  • What did it assume about who deserves help, and on what basis?
  • Whose perspective is centered, and whose is missing entirely?
  • What institutional or historical context is absent?
  • Where does it sound most confident — and is that where it is best supported?

What it told me

I ran the prompt on ChatGPT Edu first. It organized the answer around who gets the program, how it is structured, and how it is described: deservingness, contributory programs, visibility and framing, and programs that are "politically self-reinforcing." Then it summed up:

"In short, welfare programs tend to be most popular when they are broadly available, viewed as earned, associated with sympathetic recipients, and used by politically influential groups. Programs tend to be less popular when they are narrowly targeted, highly means-tested, stigmatized, or portrayed as benefiting people different from the typical voter."

Compare it with yours. Then read this one closely:

  • The deservingness line is there, in a suit. Viewed as earned. Sympathetic recipients. People different from the typical voter. Who is the typical voter? Different how?
  • Count the passives. Viewed, associated, portrayed. By whom? Popularity happens to programs here. Nobody in the paragraph is doing anything.
  • It is a good textbook answer. Textbooks are what it read. The question is what the textbooks could not see, and that is where Howard starts.

Three arguments, one shape

The AI's account of "popular" programs left things out, and the omissions were not random.

Howard, chapter 1: the standard measure of a welfare state leaves things out. What it leaves out makes the United States look small.

Desmond, chapters 3–4: the standard account of poverty leaves things out too. What it leaves out is the people on the other side of the transaction.

Same complaint about the instrument. Different thing it misses.

Howard: the measure misses money the state puts in. The welfare state is bigger than it looks.
Desmond: the measure misses money the market takes out. The poor are worse off than they look.

Both of them wrote a while ago. Where a figure on these slides differs from the one in your book, you get both — the book's is in parentheses.

Not unusually unlucky. Unusually harsh.

Week 1's "prevalence is not penalty," with the numbers under it. Brady, Finnigan, and Hübgen take four risks — no high school degree, single motherhood, a household head under 25, no one in the household employed — and count two things for each, in 29 rich democracies:

Prevalence: how many households carry the risk.
Penalty: how much more likely a household with the risk is to be poor than one without it, in that country.

United States 29-country average
Households carrying at least one risk 25.4% — only nine countries have fewer 30.9%
Poverty rate 16.3% — second of 29 9.3%
No high school degree: added chance of being poor +16.4 points under +5
All four penalties, summed .914 — first of 29 .507

Take joblessness alone: 19 countries have more of it than the U.S. Four punish it harder. Twenty-four punish it less.

Brady, Finnigan, and Hübgen, American Journal of Sociology 123(3): 740–786, 2017.

What a penalty looks like

Same event. Different country. The penalty is the difference in what happens next.

Elsewhere United States
Poverty rate, single-parent households Denmark: 8.2% 46.3% (your textbook: 33%)
Earnings gap, no high school degree Finland: none −24%
Earnings premium, college degree Norway: +20% +68%

Poverty: OECD Family Database, 2016. Earnings: OECD Education at a Glance, 2016.

Raising a child alone, leaving school early: these happen everywhere, at about the same rates. Whether they leave you poor depends on which country you are standing in. That is a design choice, and someone made it.

Howard spends today's chapter arguing we have been mismeasuring what was chosen. Desmond spends his arguing we have been looking at the wrong end of the transaction.

Everybody knows the American welfare state is small

Forty years ago Harold Wilensky called the United States a "reluctant welfare state." Since then: semi-welfare state, welfare state laggard, residual, incomplete.

The number behind the consensus, 1997 public social spending as a share of GDP:

United States Denmark Sweden Finland 17-nation average
15.8% 35.9% 35.7% 33.3% 25.6%

Howard: "Consensus on this point has become so complete that few people have stopped to question its accuracy."

Count it three ways

Gross: what government pays out. The consensus number.
Net: subtract what government takes straight back by taxing the benefit. Denmark taxes pensions and unemployment checks hard, so a quarter of every benefit returns to the treasury. The U.S. barely taxes benefits.
Net plus tax expenditures: add welfare delivered as a tax break — the EITC, the mortgage interest deduction. Nobody uses those like the U.S. does.

Country Gross Net of taxes on benefits Net, plus tax expenditures
Denmark 35.9% 26.7% 26.7%
Sweden 35.7 28.5 28.5
Netherlands 27.1 20.2 20.3
United Kingdom 23.8 21.1 21.6
Canada 20.7 17.8 18.7
United States 15.8 15.0 16.4
17-nation average 25.6 21.3 21.6

Europe's number falls. Ours barely moves. On the first column the U.S. spends 60 percent of the average; on the last, 75 percent. Sweden goes from 2.3 times the U.S. to 1.7 times.

Compared to what?

A household spends less than 2 percent of what it lays out each year on food. The average American family spends 14 percent.

Is that household starving? Should someone call social services?

Now say the household is Taylor Swift's. Two percent of her expenses buys a lot of dinner.

The U.S. has the largest GDP and the highest GDP per capita in the group. It can devote a smaller share and still spend more per person. Net social spending per person, purchasing power adjusted:

Denmark Norway Italy Finland United States Netherlands UK
$6,812 $6,532 $5,125 $5,081 $4,809 $4,495 $4,424

Sixteenth of seventeen by share of GDP. Ninth of seventeen per person — and above the $4,782 average.

Say a few words about the air force

Limiting the welfare state to social insurance and grants, Howard says, makes about as much sense as limiting national defense to the army and the navy.

The United States delivers social policy through six tools:

Social insurance · Grants · Tax expenditures · Regulation · Loan guarantees · Tort law

  • Europe subsidizes employers to hire the disabled. We passed the ADA and made it a rule. Subsidies show up in the comparison. Rules do not.
  • Europe compensates injury through public funds. We sue. Legal costs, claims, and awards run at least twice the GDP share here as elsewhere.

Housing, counted three ways

Count grants only. Public housing plus Section 8 ≈ 5 million units, under 5 percent of U.S. dwellings. Subsidized rentals: 15% in France and Germany, 25% in the UK and Denmark, 40% in the Netherlands. We are not close.

Add loan guarantees. Over 7 million FHA and VA mortgages — more households than public housing and vouchers combined. Now ~7 percent of all dwellings.

Add tax expenditures. Mortgage interest deduction: 30 million returns, $53 billion. Property tax deduction: 32 million owners, $18 billion. Capital gains exclusion on home sales: $12 billion. Together, nearly three times what we spend on low-income rental grants.

Total: 35 to 40 percent of residences. Counting owners who benefited in past years, 50 to 60 percent.

Bottom line. Count the way the textbook counts, and government houses one American household in twenty while Europe subsidizes three to eight times that share. Count all six tools, and government helps house one household in three, maybe one in two — mostly owners, mostly through the tax code — and the European comparison stops being arithmetically possible. Same country, same year. Only the instrument moved.

Howard's verdict, and what he gives up

The answer is no. The American welfare state is not unusually small. It has been measured wrong, for a long time, by people who then spent careers explaining a fact that was not one.

The question stops being why does the U.S. government do so little and becomes: why do governments of similar size pursue similar goals through such a different mix of tools?

He concedes the rest in a single paragraph. The American welfare state may still be exceptional for its:

  • limited scope (family policy, health insurance)
  • heavy reliance on means-tested benefits
  • weak public support
  • failure to reduce inequality and poverty

Howard used 1997 data. Here is 2021.

His case rests on 1997 figures. On the OECD's current net series:

Howard's 1997 finding Most recent (2021)
Gross public spending, % GDP US 16th of 17 US 13th of 17
Net public spending, % GDP US 16th of 17 US 7th of 16
Net total social spending US 11th of 17 US 1st in the OECD

His argument is stronger now than when he made it. Net total US social spending is 33.2 percent of GDP — the highest of any OECD country.

And the caveat he does not give you: US private social spending is 12.8 percent of GDP, and 6.3 points of that is health care.

Desmond: there are many ways to be exploited

Underpaid relative to what you produce — labor exploitation.
Overcharged relative to what you buy — consumer exploitation.

"Our economic freedom is limited when we don't have resources at our disposal. When we don't own property or can't access credit, we become dependent on people who do and can, which in turn invites exploitation, because a bad deal for you is a good deal for me."

Julio Payes, Emeryville, California, 2014. Eighty hours a week across two full-time jobs: the graveyard shift at a 24-hour McDonald's, 10 p.m. to 6 a.m., then a temp placement 8 a.m. to 4 p.m. Both at minimum wage.

"I felt like a zombie. No energy. Always sad." His eight-year-old brother offered to buy an hour of his time. He fainted in a grocery store aisle at twenty-four.

Did they have to?

1946. George Stigler, thirty-five, publishes "The Economics of Minimum Wage Legislation" in the American Economic Review. Raising the minimum wage will raise unemployment. His evidence is hypothetical data — a numerical story he invented to illustrate the theory.

It becomes orthodoxy. It goes untested for nearly fifty years.

1992. New Jersey raises its minimum wage; Pennsylvania does not. David Card and Alan Krueger survey 410 fast food restaurants in each state, before and after. New Jersey fast food jobs do not decline.

Hundreds of studies since. The bulk of the evidence: the employment effect of raising the minimum wage is inconsequential.

If Julio had manned the grill at a McDonald's in Denmark, his paycheck would have been double.

What actually changed: power

Then. Through the 1950s and 1960s nearly a third of U.S. workers carried union cards. In 1970 alone, 2.4 million union members took part in work stoppages. Worker pay climbed and CEO compensation was reined in.

Now. About one in ten American workers is in a union, most of them firefighters, nurses, and cops. 94 percent of private sector employees have none. Roughly half of nonunion workers say they would organize given the chance.

  • 1981: 13,000 air traffic controllers fired. Corporate America learns it can crush a strike with minimal blowback.
  • 2016–17: employers were accused of breaking federal law in 41.5 percent of union election campaigns — illegal firing of organizers alleged in a fifth to a third, depending how you count.
  • The fissured workplace: Apple directly employs about 166,000 people (Desmond says 63,000 — that figure was a decade stale when the book went to press). The workforce that builds its products is far larger, sits with contract manufacturers, and Apple publishes no number for it.

"Our grandparents had careers. Our parents had jobs. We complete tasks." — Gerald Davis

The EITC, twice

Howard: the Earned Income Tax Credit is a tax expenditure. It is the reason the last column of the spending table moves. It is the hidden American welfare state, doing quietly what Europe does through visible transfers.

Desmond: the EITC reaches roughly 24 million workers and families, average payment about $2,900 (the book: 25 million at $2,411 — those are 2020 figures). Its loudest champions are Walmart, the U.S. Chamber of Commerce, and the National Restaurant Association.

"By introducing employees to these benefits, businesses help their employees — and help themselves." — Institute for a Competitive Workforce

13.8 million working adults rely on Medicaid; 10.6 million live in households on food stamps (the book: 12 and 9 million, for 2018). About two-thirds work full time; 88 percent work in the private sector.

17 of the 50 largest Fortune 500 companies by employee count rank among the top 25 employers of Medicaid or SNAP recipients in at least two states.

Same program. Is it evidence the welfare state is bigger than we thought, or evidence it is a wage subsidy?

Overcharged: the rental market

Median asking rent on a vacant unit — the price facing someone who has to move: $483 in 2000. $1,216 in 2021. $1,492 in 2025.

Not enough supply. Too much demand. That is the usual answer. Then explain Birmingham, where 19 percent of rental units sat vacant at the end of 2021 and rent rose 14 percent anyway.

Where the money goes. Desmond and Nathan Wilmers, from confidential Census surveys of landlords — what a landlord clears on one unit in one month:

Neighborhood Rent, minus mortgage, taxes, and insurance …minus maintenance, repairs, and vacancy too
Poor $298 $98
Middle-class $225 $3
Rich $250 $49

Why it works: in a poor neighborhood the building costs far less, so the mortgage and taxes are far lower. The rent is only a little lower. Indianapolis: median two-bedroom $991 metro-wide, $816 in neighborhoods above 40 percent poverty. The property is worth a fraction; the rent is seventeen percent less.

Lakia Higbee's $373

Cleveland, fall 2021. Amazon warehouse picker. Four-bedroom house, two adult daughters, a sixteen-year-old son, two granddaughters.

Her rent: $950 a month. The windows were thin enough that the winter heating bill could reach $500.

A mortgage on that same house, on ordinary terms, taxes and insurance included: $577 a month.

$950 − $577 = $373. Every month she pays $373 more to rent the house than it would cost to own it. That is the price of not being allowed to buy. It would have replaced the windows.

She could not get the mortgage. Banks are not interested in loans that small. Redlining ended in 1968; the neighborhoods still function as mortgage deserts.

What it costs to have no money

  • Overdraft. The largest banks took $11.68 billion in 2019. On that same measure, 2023: $5.83 billion — halved. Count credit unions, which nobody was counting, and the 2024 total is $12.1 billion and rising.
  • Who pays. About 9 percent of account holders pay roughly 79 percent of the fees (the book says 84) — the ones whose end-of-day balance sits under $350.
  • Check cashing. One to ten percent of the check. A worker earning $10 an hour who cashes a $1,000 check pays $10 to $100 to receive money already earned: one to ten hours of work.
  • Payday. A two-week $300 loan runs an APR of 664 percent in Texas. Four in five loans are rolled over. About $2.4 billion in fees a year across thirty states (the book says $9.8 billion; no published source supports it).

Desmond's total, 2019: the fringe takes $61 million a day. Same measure, 2023: about $25 million. Add credit unions: $42 million.

Why the fall stopped. Regulators leaned on the big banks from late 2021 and overdraft revenue halved. The CFPB capped the fee at $5 in December 2024. Congress repealed the cap in May 2025, before it took effect. The CFPB's own finding: banks "appear to have stopped significantly reducing overdraft fees."

Two exceptionalisms

Everyone agrees the United States is exceptional. Today's two authors disagree about what is exceptional.

Howard: the state is not small. Counted with all six tools, the United States spends about what its peers spend, and by one measure more than any of them. The "reluctant welfare state" was a measurement error, and a generation of scholars built on it.

Desmond: the market is unusually present. Americans at the bottom are underpaid at work relative to Denmark, overcharged for housing relative to their own neighbors, and charged again — overdraft, check cashing, payday — for access to their own wages. Peer countries take less back.

Brady and colleagues supply the puzzle: ordinary rates of misfortune, the harshest penalties in the rich world, and the second-highest poverty rate of 29.

If both authors are right, the U.S. spends an average amount and still lands near the bottom. Where does the money go between the appropriation and the outcome?

Three places this course will look: the leak between appropriation and family (Week 2: TANF, 22 cents on the dollar) · the design of the programs themselves (Part II) · the politics that produced the design (Part IV).

Who benefits?

Desmond's rule for the rest of the book: stop asking the poor why they put up with it. Ask who gets paid when they do.

What they get asked What Desmond asks
Why don't you find a better job? Who profits from what Julio's two jobs pay?
Why don't you move? Who collects Lakia's extra $373?
Why don't you stop taking out bad loans? Who collects the fee?
Who is feeding off this?

Due and next

  • Discussion Paper 1 — Growing Up Poor in America, both parts, with Howard ch. 3 or Pierson: Friday 9/18, 11:59 p.m., on Canvas. Part 2 of the film arrives next week.
  • Pre-survey, if you have not done it: Friday 9/11.
  • Week 4 is asynchronous — no class meetings 9/14 or 9/16. Howard ch. 3 and Pierson (1993), plus Part 2 of the film. Everything by Friday.