Affordability Crisis

⬅️ Folder: 60.3.7 Economics · Hub: Politics MOC · Cross: Capitalism · Health Care · Regulations · Social Benefits · Taxes paid by class · 47 - Trump Presidency

Research workspace on the 2025–26 “affordability crisis”: what is actually rising, what only feels like it is rising, which claims by politicians survive contact with the data, where life has quietly gotten cheaper, and how the whole thing relates to 50 years of inequality. Every section leads with the question it is trying to answer. Figures are dated; anything I could only get from secondary reporting is flagged in Verify before citing.

The five numbers that frame everything

• Prices are ~3.5% higher than a year ago (CPI, June 2026) but roughly 25–30% above the 2019 level — the pain is the level, not the current rate. Latest primary release verified in this file: June 2026 (published July 14). July was scheduled for Aug 12, but the official archive had not posted it during the Aug 17 research pass; refresh before citing.
• Median existing home: $440,600 (NAR, June 2026), +54% since 2020, at a 6.69% 30-year mortgage — price-to-income ~5× vs ~3× in the 1990s.
• 49% of renters (22.7M households) are cost-burdened; 12.1M severely (JCHS, 2024 data).
• ACA marketplace premiums +26% for 2026, and with the enhanced tax credits expired, average net premiums rose ~58% ($113 → $178/mo).
• Electricity +6.9% in 2025, more than double headline inflation, with data centers ≈ 40% of demand growth.

READ FIRST — deflate every number in this note before using it

Every “+30%”, “+40%”, “+54%” below is nominal. General prices rose ~25% and nominal wages rose ~28% over the same window. Most of the scary headline numbers are just… inflation. A figure only means something if it beat the two benchmarks.

The arithmetic (it is division, not subtraction)

Prices +30% against wages +20% is not a 10% squeeze. It is:

1.30 ÷ 1.20 = 1.083 → a real increase of 8.3%

Subtracting is a decent approximation at small numbers and drifts badly at these magnitudes. Always divide.

The two benchmarks, 2019/2020 → 2025/26

Benchmark	Change
General prices (CPI)	~25% above 2019 (~23.5% on KFF’s 2020–2025 window)
Nominal wages	~28% (2019→2025)

Wages roughly matched or slightly beat general inflation. Aggregate real wages are modestly positive since 2019 — not the collapse the headline numbers imply.

What that does to this note’s own figures

Deflated by ~25% prices and ~28% wages. Approximate and directional — the windows differ slightly by sector.

Line item	Nominal	Real vs prices	Real vs wages
Medical-care prices (CPI)	+13%	−9%	−12%
Employer family premium	+26.5%	+1%	−1%
Groceries	~+30%	+4%	+2%
Child care	~+30%	+4%	+2%
Electricity	~+40%	+12%	+9%
Natural gas	~+54%	+23%	+20%
Home price	+54%	+23%	+20%
New buyer’s monthly payment	~+130%	+84%	+80%

Read the table twice. Medical-care prices fell in real terms. Employer premiums were flat. Groceries and child care rose about 2–4% real across six years — real, but not a crisis. The crisis is concentrated in four rows, and one of them dwarfs everything else.

Do not over-correct — four things this reframe does not fix

1. The entry problem survives deflation completely. A new buyer’s payment is still up ~80% in real terms. No wage story rescues that. This is why the note’s thesis is about entry, not general inflation.

2. “Average wages” is not “your wages.” Retirees on fixed incomes, anyone who did not switch jobs, and workers in the 45 states where real wage growth ran negative over a recent 12-month window got no such raise. The average conceals who actually received it.

3. But the distribution helps rather than hurts here. Cleveland Fed finds that from 2020:Q1 to 2025:Q3, 10th-percentile real wages grew 9.7% against 4.5% at the 90th — low-wage workers did more than twice as well. That is the wage compression in Autor, Dube & McGrew, already in this file’s references. It weakens the “workers fell behind” framing specifically at the bottom.

4. Sentiment tracks salience, not arithmetic. People observe grocery prices weekly and their own salary once a year. A correct real-terms calculation does not make anyone feel better, and dismissing the feeling with it is how this argument gets lost.

The defensible bottom line: most of the headline increases are nominal illusion. The genuine, post-inflation, post-wage story is housing entry costs, household energy, and whoever did not get the average raise — not a broad collapse in living standards.

---

The framing question: level, rate, or entry?

Three different claims get compressed into the phrase “affordability crisis,” and they have different answers:

1. Rate claim — “prices are still rising fast.” Weak. 3.5% is elevated but not a crisis rate.
2. Level claim — “prices never came back down.” True and unfixable by design; the 2021–23 step change is permanent, and household sentiment tracks levels, not first derivatives. This is why sentiment stayed bad through a good labor market.
3. Entry claim — “you can no longer buy in.” Strongest. The squeeze is concentrated on people entering durable-asset markets: first home, first child, first unsubsidized insurance policy. People who already hold the assets (a 3% mortgage, an employer plan, grown kids) are largely insulated — and they vote.

Working thesis to test: this is not a general inflation story. It is a story about a handful of supply-constrained, third-party-paid, or land-anchored sectors (housing, health care, child care, insurance, now electricity) inflating against a background of cheap tradable goods, with the cost falling almost entirely on new entrants and renters.

Before saying “the 1950s were better”

The historical comparison strengthens this note’s thesis rather than defeating it. America is vastly richer, safer, healthier, and more comfortable than in 1950: inflation-adjusted disposable personal income per person is about 4.7× as high, while measured real consumption per person is about 4.8× as high; food takes roughly half the income share; life expectancy is about 11 years longer; infant mortality is down about 81%; and amenities such as complete plumbing, air conditioning, safe cars, instant communication, and modern medicine are ordinary rather than rare or nonexistent. These are aggregate-per-person measures, not claims that every family became 4.7× richer; the companion audit defines the measure, checks consumption, and separates averages from the median.

But the nostalgic case contains a real signal: starter housing was much cheaper relative to income, public college cost less, unionized industrial labor had more bargaining power, and family formation followed an easier economic sequence for many households admitted to the postwar compact. Recorded mid-1950s homicide was also low enough that crime should be called mixed, not claimed as an easy modern victory.

Full audit: Were the 1950s Better Than Today - A Standard of Living Audit

Definitions and the poverty-measurement layer: What Poverty Means - Absolute, Relative, and Why Cross-Country Numbers Lie — why cheap manufactured goods and expensive necessities coexist without contradiction, and why a falling price of televisions proves nothing about rent.
The companion note tests economics, the one-income-family claim, lifestyle, crime, basic comforts, medicine, civil rights, and the things the 1950s genuinely did better. Its conclusion: the 1950s made several gateways into adulthood easier for the favored middle class; today makes the contents of life richer, safer, longer, and open to far more people. The policy goal is to recover postwar entry affordability and social infrastructure without recovering exclusion, inferior medicine, or lower material living standards.

The missing household-balance-sheet layer

The note’s entry thesis survives, but it needs a second axis: stock vs. payment vs. distress. “Household debt is at a record” is a nominal stock statement, not a diagnosis. The correct sequence is:

debt stock → denominator → required payment → interest rate → delinquency → distribution

The 2026 Q1 New York Fed total was $18.794T, of which mortgages were $13.191T (70.2%). Auto loans were 9.0%, student loans 8.8%, credit cards 6.7%, HELOCs 2.4%, and other debt 3.0%. Credit cards are therefore not where most debt is; they matter because 20%+ revolving rates make them a sharp measure of failed cash flow. New York Fed.

The counterintuitive result is that aggregate required payments are not at a 1980s record. On the Fed’s archived same-method series, the household debt-service ratio was 10.89% of disposable income in 1984 Q4 vs. 9.78% in 2024 Q1. The broader financial-obligations ratio was 15.95% in 1984 Q4 vs. 14.20% in 2023 Q3, its final observation. A larger debt stock can coexist with a similar or lower payment share because of fixed rates, longer maturities, refinancing, higher incomes, and debt composition. Federal Reserve archived ratios.

That aggregate calm conceals the real crisis: renters, first-time buyers, recent mortgages, weak-credit auto borrowers, revolving cardholders, student borrowers after the reporting restart, and households without liquid savings. The Fed’s 2025 household survey found 73% doing okay or living comfortably, but 28% worse off than one year earlier, 58% saying prices harmed their finances, and 19% spending more than income in the prior month. The universal-collapse story is false; the concentrated cash-flow story is strong. Fed SHED.

The 1984–2024 household-budget comparison also rejects the easy “luxuries did it” answer. Entertainment’s average expenditure share was essentially flat (4.8% → 4.6%), while housing (30.4% → 33.4%), health care (4.8% → 7.9%), and personal insurance/pensions (8.6% → 12.5%) gained share; food, apparel, and transportation lost share. BLS 1984 · BLS 2024.

Full household-finance audit: Affordability, Household Debt, and the Modern Household Budget

The companion separates the consumption economy, commitment economy, and asset-and-entry economy; maps where debt sits; distinguishes aggregate solvency from marginal distress; compares the 1984 and 2024 budgets; and explains why modern luxuries, higher real income, and genuine entry hardship can all be true at once.

---

Q1 — What is actually rising, and by how much?

Ranked by how much of a household budget it moves, not by how often it trends.

Housing (the whole story’s center of gravity)

• Median existing-home price $440,600 (NAR, June 2026); +54% since 2020 (JCHS); both new and existing medians above $400k.
• Price-to-income near record: median existing single-family price ≈ 5× median household income in 2025, vs ~3× as the 1990s norm (JCHS).
• Existing-home sales at three-decade lows — the market cleared by freezing, not by repricing. Mortgage lock-in is the mechanism.

🔍 Mortgage lock-in, explained — the mechanism behind almost everything else in this section

What it is

A US 30-year fixed mortgage is a rate you keep for as long as you keep the house — but it is not portable. Move, and you surrender it and borrow at today’s rate.

So when market rates roughly doubled, every owner holding a 3% loan acquired something valuable and non-transferable. The relevant variable in the literature is Δr — the gap between the rate you locked at origination and the current market rate. The larger that gap, the more moving costs you, entirely apart from the price of the new house.

This is why the note treats housing as bifurcated: lock-in is the machinery that splits owners from entrants.

How strong the effect is

FHFA Working Paper 24-03 (Batzer & Coste) quantifies it:

Finding	Magnitude
Each 1 pp that the market rate exceeds your origination rate	−18.1% probability of sale
Reduction in sales of fixed-rate-mortgaged homes, Q4 2023	−57%
Sales prevented	~1.33 million

The counterintuitive part — lock-in made houses more expensive

The instinct is that doubling mortgage rates should crush prices. It didn’t, and lock-in is why. FHFA decomposes the two opposing forces:

• Direct effect of high rates (demand suppression): −3.3% on prices
• Indirect effect of lock-in (supply strangulation): +5.7% on prices

Net: lock-in raised prices roughly 6%, outweighing the price relief that high rates should have delivered. Sellers are also buyers; freezing them removes a listing and a bid, but the listing is removed first and more completely. The market cleared by freezing, not repricing — which is the note’s line at three-decade-low sales volumes.

This is the single best answer to “why didn’t prices fall when rates doubled?”

It is not only a housing problem — it degrades labor reallocation

Fonseca & Liu (October 2023) use individual credit-record data and variation in origination timing:

• A 1 pp decline in Δr reduces moving rates by 0.68 pp — about 9% of the sample mean.
• The relationship is non-linear: once Δr is large enough, refinancing-without-moving becomes the attractive option and further increases in Δr stop mattering for mobility.
• Critically, lock-in blunts households’ responsiveness to job opportunities requiring a move — measured as wage growth in counties 50–150 miles away, instrumented with a shift-share design.

That last finding is the one economists care most about: people stop chasing better jobs. It converts a monetary-policy side effect into a productivity cost, since labor stops flowing to where it is most valuable. The authors frame it as an unintended consequence of monetary tightening in an economy built on long-term fixed-rate mortgages.

Where it stands now, and why it unwinds slowly

Lock-in decays as low-rate loans are paid off, refinanced, or ended by life events. That decay is happening — but slowly, and unevenly:

• ~80% of outstanding mortgages still carry rates under 6%; roughly 69% at or below 5%; about half at or below 4%.
• The sub-3% share fell below half of all outstanding mortgages in Q1 2026 for the first time since Q3 2020 — down from a peak above 65% in Q1 2022. Mortgages above 6% now outnumber those below 3%, a genuine turning point.
• But the unwinding stalled in 2026 as holders of below-3% loans stopped paying them off, and about 35% of owners with sub-6% rates say they would not give the rate up under any circumstances.

FHFA’s own caution: even if rates decline, lock-in will not vanish quickly. The stock of cheap loans takes years to work through, and every month it persists is another month of suppressed listings.

Three things to get right when arguing this

1. Lock-in is a transfer, not a cost saving. Existing owners hold a below-market asset; would-be buyers pay ~6% more for houses because of the resulting supply drought. The gain and the loss are different people. Nobody legislated this — it is an emergent property of a non-portable 30-year fixed rate meeting a rate spike.

2. It cuts against the “high rates will fix prices” argument. Anyone predicting that expensive money would restore affordability has to explain the FHFA decomposition, where the supply effect dominated the demand effect.

3. Falling rates are not an unambiguous fix. Lower rates release locked-in sellers (more supply, disinflationary for prices) and revive buyers (more demand, inflationary). Which dominates is genuinely uncertain — do not assert either direction confidently.

• First-time buyers: median age 40, share of all buyers 21% (all-time low) per NAR. Redfin puts the age at 35 — a real methodological fight worth resolving before citing either.
• Renters: 22.7M cost-burdened (49%), 12.1M severely (>50% of income).
• Construction cost +43% since early 2020; tariffs add ~$10k per home; ~30% of the construction workforce is immigrant and 33% of firms report labor disruption from immigration enforcement.

”How much more expensive is housing since 2020?” — the answer ranges from cheaper to more than double

Housing is the most bifurcated cost in the economy. Unlike health care, where the spread between measures is a measurement artifact, here the spread is real and it tracks a single variable: when you locked in. This is why the note’s framing question is *level, rate, or entry*.

Measure	Change since 2020
Median existing-home price	+54% — to $440,600 (NAR, June 2026)
30-year mortgage rate	~3.11% → 6.69% — roughly doubled (Freddie Mac, week ending Aug 6, 2026)
Monthly P&I for a new buyer	≈ +130% — my calculation, assumptions below
CPI shelter	Rising but lagging; YoY peaked ~8.2% (Mar 2023), now 3.3% (June 2026). Cumulative figure not established — see Verify before citing
Market asking rents (ZORI)	Spiked far earlier and harder than CPI — ~15–20% annualized in early 2022 while CPI rent read 5.5%
Existing owner with a 2020–21 mortgage	Nominal P&I unchanged → roughly −20–25% in real terms

The illustrative payment calculation (label it as such if you use it): 20% down, 30-year fixed. 2020 — ~$290k at 3.11% → ≈$992/mo. 2026 — $440,600 at 6.69% → ≈$2,272/mo. That is +129%. JCHS publishes a higher figure, $2,570/mo, because it assumes first-time-buyer terms (3.5% down). Both are defensible; they answer slightly different questions.

The two headline answers:

• “How much more does it cost to buy in today?” → the monthly payment has roughly doubled. Price explains about half of that; the rate explains the rest.
• “How much more are households paying?” → far less, because most owners never re-entered the market. Mortgage lock-in is why sales hit three-decade lows: the market cleared by freezing, not by repricing.

Three traps specific to housing numbers

Price ≠ payment. A 54% price rise with a doubled rate produces a ~130% payment rise. Quoting the price increase alone understates the entry cost by roughly half.

CPI rent ≠ market rent. Over 80% of the CPI rental sample are continuing tenants, so CPI lags market turning points by 12–18 months. In 2022 that gap was enormous — ZORI ~15–20% against CPI rent at 5.5%. Whichever you cite, name it; they are not interchangeable and the choice can swing the story by triple digits in relative terms.

“Existing owners are fine” is only true for P&I. Their principal and interest is frozen, but homeowners insurance is +40–46% since 2020–21 and property taxes have followed assessments upward. The lock-in protects one component, not the bill.

The distributional fact that follows: an owner who locked at ~3% in 2020–21 has seen their housing payment fall by roughly a fifth in real terms while a would-be buyer faces double. That single split explains most of the divergence in housing sentiment — the same national market is genuinely cheaper for one group and genuinely unaffordable for the other, and both are describing it accurately.

Health care (the fastest-moving 2026 line item)

• Marketplace gross premiums +26% for 2026 (30% on HealthCare.gov states, 17% on state exchanges) — before subsidy changes.
• Enhanced premium tax credits expired end of 2025: average net premium +58% ($113 → $178/mo); deductibles ~$1,000/person higher; households above 400% FPL ($62,600 single) lost subsidies entirely and eat the full increase. Healthier enrollees exited → worse risk pool → next year’s filings are again double-digit for 2027.
• Employer coverage: total benefit cost per employee +6.5% in 2026, the largest increase since 2010 (Mercer). Family premium ≈ $27,000, worker share ≈ $6,850.
• Note the measurement trap: when employers hold premiums down by raising deductibles, CPI sees less health inflation while households feel more.

”How much more expensive is health care since 2020?” — depends entirely which measure you mean (13% to 29%)

Three different things get called “health care costs,” and they have moved at very different rates. This is the single easiest place to be caught overstating.

Measure	2020	Latest	Change
Total national health spending	$4.1T	$5.3T (2024)	+29%
Employer family premium	$21,342	$26,993 (2025)	+26.5%
Employer single premium	$7,470	$9,325 (2025)	+24.8%
Worker contribution, family coverage	$5,588	$6,850 (2025)	+22.6%
Average single-plan deductible	$1,644	$1,886 (2025)	+14.7%
Medical-care prices (BLS medical-care CPI)	—	—	≈ +13.3% (end-2020 → end-2025)

Medical-care CPI, Dec-to-Dec: 2.2% (2021) · 4.0% (2022) · 0.5% (2023) · 2.8% (2024) · 3.2% (2025). The 3.2% is confirmed against the BLS December 2025 release; the ~13.3% five-year figure is compounded from the annual changes, not a directly published number.

The two headline answers:

• “How much more expensive has health care become for households?” → roughly 20–25%, depending on the component.
• “How much has America increased total health spending?” → ~29%, 2020 to 2024.

The caveat that keeps this honest

The 29% is not a price increase. Total spending also rises with population growth, aging, higher utilization, more expensive treatments and drugs, and coverage expansion. CMS attributes the 2023–24 surge primarily to non-price factors — demand and service intensity — with personal health care spending growing 8.9% annually across those two years, the fastest since 1991–92.

2020 is also a distorted baseline. COVID drove a surge in federal health spending while suppressing ordinary utilization. CMS notes that without federal public-health and COVID programs, 2020 national health-spending growth would have been just 1.9%. Any “since 2020” figure inherits that distortion.

Watch the windows. The spending figure is 2020→2024 (four years); the premium figures are 2020→2025 (five). Not directly comparable.

Context that cuts against the crisis framing: KFF reports family premiums rose 26% from 2020–2025, against 23.5% general inflation and 28.6% wage growth over the same period. So employer premiums outpaced consumer prices but slightly trailed wages across that specific five-year window — a genuinely awkward fact for the strongest version of the health-cost argument, and one worth conceding before an opponent uses it.

Child care

• Center-based national average ≈ $15,000/child/yr (infant ~$17,000).
• 35% of median single-parent income vs 10% for married couples — the single sharpest distributional fact in the whole file.
• Federal affordability benchmark is 7% of income; infant/toddler center care clears it in zero states.

”How much more expensive is child care since 2020?” — +29% or +23%, and the baseline you pick flips the verdict

Measure	2020	Latest	Change
National average, all ages (CCAoA)	~$10,174 (~$200/wk)	$13,184 (2025)	≈ +30%
Center-based infant (CCAoA)	>$12,300	—	—
CCAoA’s own framing	—	—	+29% (2020→2024) vs +22% general inflation
Alternative window	—	—	+23% (2021→2025) vs +24% overall prices

Same data, opposite conclusions. Measured from 2020, child care outpaced inflation by ~7 points. Measured from 2021, it trailed inflation slightly. Both are true statements about real series; the difference is entirely the starting year.

Why the baseline matters so much here — and it is not arbitrary

ARPA’s $24 billion in child care stabilization grants ran from March 2021 to September 30, 2023, supporting ~220,000 programs and holding tuition down for the duration. A 2021 baseline therefore starts inside a large subsidy window; a 2020 baseline starts before it. One month after the funds expired, 29% of families reported a tuition increase in the prior month.

This is the same structural problem as the health-care baseline above: 2020–21 was distorted by emergency federal spending in both sectors, and “since 2020” figures inherit that distortion in opposite directions — COVID inflated the health-spending baseline while ARPA suppressed the child-care one.

Source discrepancy in the figures above this callout

The $15,000 / infant ~$17,000 cited in the bullets is Care.com data — a survey of ~3,000 parents reporting what they actually pay (their infant figure is ~$332/week ≈ $17,300/yr). CCAoA reports $13,184, drawn from state Child Care Resource & Referral agencies reporting published provider rates.

Neither is wrong; they measure different things. Parent self-report captures true out-of-pocket including premium and metro-weighted care; agency rate surveys capture listed prices across all licensed providers. The gap is ~25%, so always name which one you are using — this is the same methodology fight already flagged for NAR vs. Redfin on first-time-buyer age.

The distributional figure is the one that survives all of this: 35% of median single-parent income versus 10% for married couples (CCAoA, 2024). That ratio is robust across both methodologies, because it is a share, not a level — which is why it remains the sharpest fact in this file.

Insurance (the sleeper)

• Auto insurance +64% since 2020; homeowners +40–46% since 2020–21; average homeowners premium projected $3,057 in 2026.
• Matters more than its CPI weight suggests because it is non-negotiable, escrowed into housing costs, and rising fastest exactly where housing already broke (FL, TX, CA, LA).

Electricity (the newest)

• +6.9% in 2025, ~+6% more through 2027 (Goldman); data centers ≈ 40% of demand growth; PJM attributes $6.3B of consumer cost increases over three years mostly to data-center load.
• State extremes, May 2025 → May 2026: IL +28.4%, HI +26.7%, DC +24.3%.

”How much more expensive is electricity since 2020?” — ~40%, and natural gas is worse

Measure	2020	Latest	Change
EIA average residential price	~13.2¢/kWh	17.29¢ (2025) · 18.44¢ (Aug 2026)	+31% to 2025 · ~+40% to Aug 2026
CPI electricity	—	—	~+41% since Jan 2020
CPI piped utility gas	—	—	~+54% since 2020
Current pace	—	Electricity +4.0% YoY, gas +3.0% YoY (June 2026)	—

The EIA/CPI gap is an endpoint artifact, not a real conflict

EIA’s +31% and CPI’s +41% look contradictory. They aren’t — they stop at different places. EIA’s figure runs 2020→2025; the CPI figure runs Jan 2020→2026. Carry EIA forward to its own August 2026 reading and it gives 18.44 ÷ 13.2 ≈ +40%, which lands on the CPI number.

(That reconciliation is my arithmetic — see Verify before citing.) The lesson generalises across this whole file: before treating two figures as contradictory, check whether they simply end in different years. Electricity has been rising fast enough that a single year of endpoint difference moves the answer by ~9 points.

Natural gas has outrun electricity — and that matters more than it looks. Gas heats most US homes, so the ~54% rise hits the winter bill directly. It is also far more volatile than electricity: gas spiked hard in 2022 on the Ukraine shock and partly retraced, so a “since 2020” gas figure is unusually sensitive to which month you stop in. Electricity’s climb has been steadier and is therefore the more reliable series to cite.

Rate ≠ bill, and geography dominates everything

The bill is price × consumption. A 40% rate rise does not mean a 40% bill rise: usage moves with weather, appliance efficiency, and household size. Anyone quoting a rate increase as though it were a bill increase is overstating. Reference point: at 2020 prices, a household using ~855 kWh/month paid roughly $1,400/year.

Dispersion is the widest of any cost in this file. State moves over a single recent year ranged from IL +28.4% to national averages near 4%. National figures are close to meaningless for any specific household — always pair them with the state number if the argument is about a real person.

What makes this line item different from the others in Q1: health care, housing, child care and food all trace their post-2020 shocks to the pandemic or the policy response. Electricity’s driver is new and forward-looking — data-centre load, ≈40% of demand growth, with PJM attributing $6.3B of consumer cost increases over three years mostly to it. That means this is the one category where the causal story points at continued increases rather than at a fading shock, which is why the note flags it as “the newest.”

Food

• Food-at-home +2.7% YoY (June 2026) — but ~32% above Jan 2020. Beef is the salient outlier (wholesale +12.7% YoY). Restaurant prices have outrun groceries.

”How much more expensive is food since 2020?” — groceries ~25–32%, restaurants more, and the level–rate gap is widest here

Measure	Since 2020	Now (YoY)
Food at home (groceries)	~25–32% — sources conflict, see below	+2.7% (June 2026)
Food away from home (restaurants)	+39.3% — but measured Jan 2019→Jan 2026, a 7-year window	+3.4%; full-service +3.7%
Share of disposable income on food	9.7% in 2025 — down from 10.4% in 1997	—

This is where the level-versus-rate gap does the most damage to political argument. Grocery inflation is running near 2–3%, i.e. roughly normal. The level is ~30% above 2020. Anyone saying “inflation is back to target” and anyone saying “groceries are unaffordable” is describing the same series — one the derivative, one the integral. Neither is lying; they are answering different questions.

Item dispersion is enormous, and it wrecks anecdotal reasoning

Over a single year (Jan 2025 → Jan 2026): ground beef +21.8% ($5.55 → $6.75) while eggs −48.0%, tomatoes −12.6%, potatoes −10.5%, chicken −1.0%, bananas +5.2%, cheese +5.0%.

Eggs are the cautionary case in both directions: they spiked spectacularly during the avian-flu shock and then collapsed. Whichever item you pick, someone can pick another that says the opposite. Beef is currently the salient outlier driving perception; a year ago it was eggs. Cite the index, not the item — and expect the other side to cite the item.

2020 is a distorted baseline here too — and in a third distinct way

USDA ERS reports that total food spending fell 8.2% in 2020, the sharpest annual drop since 1967, because restaurants closed. So the food-away-from-home baseline is artificially depressed, and any “since 2020” restaurant figure overstates the underlying trend.

That completes a pattern worth naming across this whole file: COVID inflated the health-spending baseline, ARPA suppressed the child-care baseline, and lockdowns suppressed the restaurant baseline. “Since 2020” is doing different work in every sector, and it is never neutral.

[!check] The share-of-income measure cuts hard against the crisis framing — and it is the strongest steelman in this block
USDA ERS’s Food Expenditure Series (data as of June 2026) shows the share of disposable personal income spent on food has fallen over the long run: 10.4% in 1997 → 9.7% in 2025.

	1997	2025
Total food	10.4%	9.7%
Food at home	6.1%	4.8%
Food away from home	4.3%	4.9%

The decline is driven entirely by groceries falling from 6.1% to 4.8%, more than offsetting the rise in restaurant spending. Restaurants overtook groceries as a share of income somewhere in the mid-2010s and the two are now roughly level.

How to hold both facts at once. Grocery prices are ~30% above 2020, and groceries are simultaneously a smaller share of income than in 1997. Both true: prices rose, but nominal incomes rose more over the longer horizon. The crisis is real on a 5-year view and absent on a 28-year view — so the window you choose determines the answer, and anyone citing only one is choosing a conclusion.

Two caveats on the series itself: shares are nominal, not inflation-adjusted, and ERS notes the data are subject to revision (see below — they already were).

One unresolved conflict in the food figures

Grocery cumulative since 2020. This note’s ~32% is not reconciled with a secondary source reporting 25.2%. Compounding the annual CPI food-at-home changes lands near 30%, which favours the higher figure — but that is my arithmetic, not a published series. Pull CUUR0000SAF11 from BLS and compute it directly.

Resolved — the ERS share-of-income conflict was a series revision

Earlier drafts of this note flagged an apparent contradiction between 9.7% (2025) and ERS’s own 11.3% (2022) / 11.2% (2023). The current ERS chart (Food Expenditure Series, data as of June 2026) settles it: on the current vintage, total food never approaches 11%, running near 10% in 2022 and 9.7% in 2025.

The 11.2–11.3% figures are a superseded vintage, not a different concept. ERS revised the series downward. This is a useful general warning: the same ERS chart ID can return materially different numbers depending on when it was pulled, and older secondary write-ups will keep circulating the pre-revision figures indefinitely. Always cite the vintage date.

---

Q2 — What is historically the least affordable? (Top 3)

Judged by share of a median budget × trajectory over 40 years × share of households with no substitute:

Runner-up, and the one to watch: electricity, because it has just acquired a national demand shock (AI/data centers) hitting captive residential ratepayers through regulated rate cases — structurally similar to housing’s supply-constrained setup.

Deliberately excluded — and this is the interesting part: college. Sticker price rose enormously, but inflation-adjusted net tuition and fees at public four-years peaked at $4,450(2025$) in 2012-13 and fell to ~$2,300 in 2025-26; public two-year net is negative (~$1,190 grant surplus); private nonprofit net fell from $19,810 (2006-07) to $16,910. Over the last decade real published tuition fell 7% (public 4-yr) and 10% (public 2-yr), >half of public-university BA graduates finish with zero debt (+13pp), and average debt among borrowers is down ~20%. College belongs in Q4, not here — which is itself a finding, since college is the canonical example people reach for.

---

Q3 — What are politicians pointing at? Are they right?

Pattern worth naming: politicians on both sides gravitate to the most salient bills (eggs, gas, groceries — ~10% of disposable income) over the largest ones (housing, health, care). Salience buys votes; the big three require fights with homeowners, hospitals, and licensing boards.

---

Q4 — What do people call unaffordable that is historically cheap?

The honest version: the goods people buy weekly got cheaper; the services and assets people buy once a decade got dramatically more expensive. Weekly purchases are what people notice and complain about; once-a-decade purchases are what actually break a budget.

---

Q5 — Where has life gotten much better without recognition?

• Rents are falling right now. 35 straight months of YoY asking-rent declines nationally; Austin −16.2% from peak; Minneapolis −11.1%. Almost no one describes 2026 as a good year to be a renter. Caveat: Austin will complete ~9,000 units in 2026, −72% from the 2024 peak — the supply that produced the relief is switching off, so this improvement is on a timer.
• The bottom of the labor market had its best run in 40 years. 10th-percentile real wages +~15% (2019–2024) — more than the cumulative prior four decades — vs median +6%, p90 +7%, unwinding roughly a quarter of the four-decade rise in 90-10 wage inequality (Autor–Dube–McGrew). Caveat: low-wage real wages fell −0.3% in 2025; the compression may be over.

🔍 “A quarter of the four-decade rise in 90-10 inequality” — what the claim actually means

What “90-10 wage inequality” is

The wage gap between a worker at the 90th percentile (near the top; earns more than 90% of workers) and one at the 10th percentile (near the bottom). It is the standard summary measure of overall wage dispersion, because it ignores the extreme tails — no CEOs, no outliers — and just asks how far apart a well-paid worker and a low-paid worker are.

“Log points” is how economists express the gap so changes are additive. For intuition: the 33-log-point rise means the 90/10 ratio grew by roughly 39% over the four decades. Small log-point numbers are close to percentages; large ones drift.

The arithmetic behind “a quarter”

Quantity	Value
Rise in the 90-10 log wage gap, 1979 → 2019	33 log points
Fall in that gap after 2020	9 log points
Share reversed	9 ÷ 33 ≈ 27% — “approximately one-quarter”

The 33-point figure decomposes as a rise in the log 90/50 gap from 0.69 to 0.91, plus the log 50/10 gap from 0.54 to 0.65 (EPI’s CPS analysis, cited by the authors). The late 1970s are the conventional start of what the paper calls the Great Divergence — four decades of almost uninterrupted rising inequality. This is the first material reversal in that entire period.

It compressed from both ends — which is the part people miss

Between January 2020 and September 2022:

• 10th percentile real wages: +6.4%
• Median real wage: −5.6% below its May 2020 level
• 90th percentile: −6.7%

So the gap narrowed partly because the bottom genuinely gained, and partly because inflation ate the real wages of the middle and top. Anyone citing this as an unambiguous good-news story is telling half of it. It is simultaneously the strongest evidence for “the bottom did well” and evidence for “most workers lost ground to inflation.”

The mechanism: this is a monopsony story, not a minimum-wage story

The authors’ argument runs through employer market power. Normally firms face labor supply that is far from perfectly elastic — workers do not easily leave — which lets employers “mark down” wages below what competition would produce.

Post-pandemic tightness broke that temporarily:

1. Quits and job-to-job switching surged, especially among young non-college workers (high school or less, under 40).
2. The wage-separation elasticity rose — workers became more willing to leave for better pay. That elasticity is the empirical measure of labor-market competition.
3. Higher elasticity → smaller wage markdowns → employers had to pay more to retain and attract.
4. Gains concentrated among workers who actually changed employers and industries, moving up a “job ladder” from lower- to higher-paying firms.

In the decomposition for young high-school workers, a rising industry-mover premium contributed ~1.3 log points to wage growth and a rising move rate a further ~0.4.

They tested the obvious alternative explanation and it does not hold

Objection: the 10th percentile was already rising before 2020 — maybe this is just state minimum-wage increases continuing.

Their test: split the states. Before the pandemic, the 10th-percentile rise appeared only in the 31 states with minimums above the federal floor — consistent with the minimum-wage story. But after 2020, the 10th percentile jumped sharply in both groups, including the 20 states that never raised their minimum. Median and 90th-percentile trends were comparable across both groups.

That is a genuinely well-designed check, and it is why the paper attributes the post-2020 compression to labor-market competition rather than to policy.

Cite the vintage — the number changed between drafts

The March 2023 working paper says “approximately one-quarter” (and 27% in the text). The revised NBER abstract says “around one-third.”

The estimate was revised upward as later data came in. This note’s “a quarter” is correct for the original draft; if you cite the current NBER page you must say one-third. Same paper, same ID, different number — exactly the vintage trap already flagged for the USDA food series. Give the draft date whenever you use the figure.

Limits worth stating before you deploy this

• It may be over. The note’s own caveat: low-wage real wages fell −0.3% in 2025. Compression required a tight labor market; as tightness faded, so did the mechanism. This was a cyclical windfall, not a structural repair.
• A quarter reversed still leaves three-quarters standing. Four decades of divergence remain overwhelmingly intact.
• Wages are not income or wealth. This measures hourly pay among workers. It says nothing about capital income, non-workers, or the wealth distribution — where concentration continued rising.
• Relative gains, absolute pain. The bottom gaining relative to the top is compatible with everyone feeling worse, which is what the top-of-note deflation table shows.

Why it matters to this file: it is the single strongest piece of evidence against a simple “workers fell behind” narrative — and its strength is precisely bounded. Best used as: the bottom did unusually well for about three years, largely because it briefly became easy to quit for a better job, and that window has probably closed.

• College net price fell for a decade and majority-zero-debt is now the norm at public universities (Q2).
• Crime collapsed. Homicides −21% in 2025, −44% from the 2021 peak; H1 2026 down another 18% → likely a record low. Public perception has not moved with it.
• Out-of-pocket drug costs for seniors. Part D $2,000 annual cap (2025); $35 insulin cap — share of Medicare patients paying ≤$35 per 30-day insulin went 48% (2019) → 75% (2023), mean OOP $50.87 → $21.98.
• Uninsured rate near record low: 8.3% (2025), vs the 7.7% record in early 2023 — though this is the one item on the list that is about to reverse.
• Energy technology. Battery storage LCOE −27% in 2025 to $78/MWh, a record low — the mechanism that could eventually cap the electricity problem.
• Quality that never shows up in the complaint. Cars that last 200k+ miles, houses with central air and more square footage per person, near-free access to the world’s information — the Cass/Winship fight (below) is fundamentally about whether to count this.

---

Q6 — How does this compare to inequality? Has it gotten much worse in 50 years?

Wealth: unambiguously more concentrated.

• Top 1% net worth share ≈ 31.6% (Jan 2026), vs ~23% in the late 1970s; peaked above 32% in 2021 (highest since the series began in 1989).
• Bottom 50% hold ≈ 2.5% of net worth (Oct 2025).
• Generational: boomers ≈ 51.7% of household wealth vs millennials 9.8% (2024). Gen Z homeownership 4.5% vs boomers 73.1% — the widest recorded gap, though a chunk of that is simply age, and cross-age comparisons like this are a common cheat.

Income: genuinely contested, and the disagreement is methodological, not ideological.

• Piketty–Saez–Zucman: top 1% after-tax-and-transfer share 9% (1960) → 15% (2019).
• Auten–Splinter, same tax data: 8% → 9%, i.e. essentially flat after taxes and transfers.
• The gap comes from allocating unreported income, treating transfers and retirement income, and handling the post-1986 shift in how high earners take compensation (when top rates were 70%+, income was retained inside businesses rather than paid out).
• Do not cite one without the other. The honest statement is: pre-tax market income inequality clearly rose; how much survives taxes and transfers is unresolved among specialists.

Wage inequality specifically: rose for four decades, then partly reversed 2019–2024 (Q5), which no popular narrative on either side has absorbed.

The connection that matters more than the correlation: in supply-constrained markets, inequality is not a parallel story to affordability — it is the transmission mechanism. When housing supply is fixed, prices are set by the bidding capacity of the top of the local income and wealth distribution, so rising inequality converts directly into unaffordability for everyone below. That predicts the affordability crisis should be worst where high inequality meets inelastic supply (coastal metros) and mildest where supply is elastic (Austin, Sun Belt) regardless of inequality. That is a testable claim and it should be the next thing checked in this file.

---

My own questions (to sharpen the picture)

On measurement

1. If real median earnings are up ~1.5–2.4% since 2019 and goods are cheaper, why does the median household report being worse off? Candidate answers to separate: price level shock, salience of recurring bills, cost-shifting into deductibles that CPI treats as quality-neutral, housing lock-in, partisan media (the sentiment partisan gap widened from <20 index points in 2008 to 55 points in 2026), social comparison.
2. Does CPI adequately capture insurance deductibles, escrowed premiums and property taxes, HOA fees, and child care? Where it doesn’t, official inflation understates the cash-flow squeeze without being “fake.”
3. Is the correct denominator income, or income after non-discretionary commitments? Build a “committed expenditure share” series (housing + health + care + insurance + debt service) and compare 1985 / 2000 / 2019 / 2026. The companion defines the series and the household types it must separate, but a consistent historical dataset still needs to be built.

On mechanism
4. Why do the inflating sectors share three features — third-party payment, licensing/credentialing, and land or grid constraints — while every sector exposed to import competition deflates? Test against cash-pay medical markets (LASIK, cosmetic, direct primary care), which should behave like goods if the theory holds.
5. Is it Baumol cost disease (labor-intensive services must track wages) or subsidized demand into restricted supply? These predict different fixes: the first says accept it, the second says deregulate supply.
6. How much of “unaffordable” is that the goods themselves got more expensive vs. that the one-earner option disappeared? Child care becomes a $17k line item precisely because both parents work; the second income and the child-care bill are the same phenomenon.
7. Debt is doing hidden work: delinquencies, negative auto equity, BNPL. First answer: aggregate leverage is not at a payment-burden record, but high-rate marginal debt and delinquencies reveal concentrated cash-flow stress. See where the debt is, credit cards, and BNPL.

On politics
8. Who benefits from the current price level, and does that explain the policy stasis? Homeowners with sub-4% mortgages, 65+ asset holders, incumbent renters under rent stabilization. Affordability politics is usually framed as rich vs poor; it may be better framed as incumbents vs entrants.
9. Which policies have ever measurably lowered a real price in the US? Airline and trucking deregulation, generic drug entry, Part D caps, Austin/Minneapolis upzoning. What do the winners have in common, and does anything currently proposed resemble them?
10. Both parties now run on affordability. Are they converging on the same diagnosis (supply) or just the same slogan? Track the Build America Caucus and CEQA-style reforms as the tell.
11. Is electricity the next housing? Same structure: national demand shock, captive ratepayers, multi-year supply lag, regulated pricing. If so, the political fight arrives ~2027-28.

On the counterfactual
12. Would a household prefer 1985 prices with 1985 goods, incomes, medicine, and job options? Nobody actually chooses this, and yet “it was affordable then” is the core rhetorical move. Force the bundle comparison explicitly.
13. Which of these problems are American, and which are shared? The affordability revolt is global (incumbent parties losing across democracies), which is evidence against purely US-specific explanations like local zoning or one administration’s policies — and for shared ones like the post-2021 price level and global housing supply constraints.

---

Steelman: the two strongest opposing readings

The decline case (Oren Cass / American Compass, Cost-of-Thriving Index)

A single male earner needed ~40 weeks of work in 1985 to buy the middle-class bundle (housing, health care, transport, education, and a savings margin) and ~62 weeks by 2022 — a bundle that now takes more than a year to earn. The argument’s force is that it prices the bundle a family actually needs to enter the middle class, not a consumption basket weighted by what people currently buy. Cheap TVs do not offset a house you cannot buy.

The achievement case (Scott Winship & Jeremy Horpedahl, AEI)

Correcting the same index: the increase is ~10 weeks, not 22, and ~4 weeks after adjusting for quality improvements in the tracked goods (better medicine, safer cars). Account for lower federal taxes, or use all full-time workers rather than men only, and the after-tax cost of thriving falls by ~7.5 weeks. The decline story depends on ignoring taxes, ignoring quality, and holding fixed a single-male-earner household that stopped being the norm.

Both are right about something: Cass is measuring entry costs; Winship is measuring consumption capacity. They are not answering the same question, and the affordability debate is mostly people talking past that distinction.

---

Three traps to avoid when arguing this

1. Averages vs medians vs the marginal buyer. NAR’s affordability index says a median-income family spends ~25% of income on a median home (Aug 2026, 20% down) while NAHB says 32% for a median new home and JCHS says the median home now demands a ~$120k income. All three can be arithmetically correct; they assume different down payments, incomes, and homes. Always ask whose purchase.
2. Nominal vs real vs time price. Nominal price changes are nearly meaningless over decades; real prices are the minimum bar; time prices (hours of work per unit) are the most honest and least used.
3. Stock vs flow. Most affordability statistics describe transactions (new mortgages, new leases, new policies) while most people are in existing positions. Both the crisis and the calm are real, for different populations.

---

Verify before citing

Figures below came through secondary reporting during this research pass and should be checked against the primary series before use in writing or argument:

• June 2026 CPI category detail beyond headline (3.5% YoY) and shelter (3.3%) — the June release is the latest primary release verified in this file. July was scheduled for Aug 12, 2026, but was not present in the official archive during the Aug 17 research pass; retrieve and incorporate it before treating the top callout as current.
• Employer family premium $27k / $6,850 worker share and the projected “$8,900 paycheck deduction in 2026” (KFF vs consultant projections — different measures).
• First-time buyer median age: NAR 40 vs Redfin 35 — resolve the methodology before using either.
• Millennial-vs-boomer net worth at the same age (sources mix averages with medians; the comparison is fragile).
• Bottom-50% wealth share 2.5% (Oct 2025) — confirm against the Q1 2026 DFA release.
• Auto insurance +64% since 2020 and homeowners +40–46% — get to the underlying BLS/industry series.
• Autor–Dube–McGrew: state which draft. The March 2023 working paper says “approximately one-quarter” (27% in the text); the revised NBER abstract says “around one-third.” The estimate was revised upward. Q6 uses the original figure — if you cite the live NBER page instead, the number is one-third.
• The deflation table in the top callout is my arithmetic, built on a ~25% price and ~28% wage benchmark. Both benchmarks came from secondary sources (~28% nominal wage growth 2019→2025 against ~26% CPI). The sector windows differ slightly — some run 2019→2026, others 2020→2025 — so the real-terms column is directional, not precise. Pull CPI-U and average hourly earnings (or median usual weekly earnings) for matched windows before publishing any specific real-terms figure.
• Electricity: the EIA/CPI reconciliation is my arithmetic. EIA’s +31% (2020→2025) and CPI’s ~+41% (Jan 2020→2026) reconcile once EIA is carried to its Aug 2026 reading (18.44 ÷ ~13.2 ≈ +40%) — but the ~13.2¢ 2020 baseline is itself back-derived from EIA’s own +31.38% figure rather than read off the source table. Pull EIA’s residential average price series directly and confirm the 2020 value before publishing either figure.
• CPI electricity ~+41% and piped gas ~+54% since 2020 came from a secondary compilation, not from BLS directly. Confirm against the BLS series (CUSR0000SEHF01 electricity, CUSR0000SEHF02 utility gas).
• Food-at-home cumulative since 2020 — unresolved. This note says ~32%; a secondary source says 25.2%. Compounding annual CPI changes lands near 30%, favouring the higher figure, but that is derived arithmetic. Pull the BLS food-at-home series (CUUR0000SAF11) and compute 2020→2026 directly.
• USDA food share of disposable income — conflicting ERS figures. Resolved. The 11.2–11.3% figures are a superseded vintage; the current series (as of June 2026) gives 9.7% for 2025 and ~10% for 2022. Cite the vintage date whenever using ERS expenditure shares — older secondary write-ups still circulate the pre-revision numbers.
• Food-away-from-home “+39.3%” is measured Jan 2019 → Jan 2026 — a seven-year window, not comparable to the six-year “since 2020” figures used elsewhere in Q1.
• Child care: resolve CCAoA vs. Care.com before citing a level. The $15,000 / infant $17,000 in Q1 is Care.com parent self-report; CCAoA reports $13,184 from state agency rate surveys. A ~25% gap. Pick one, name it, and use it consistently — the share-of-income figures (35% single parent / 10% married) are stable across both and are the safer citation.
• Child care 2020 baseline figures (~$10,174 average, >$12,300 center-based infant) came via secondary reporting of CCAoA’s 2020 data, not from the 2020 report itself. The +29% vs. 22% inflation framing is CCAoA’s own; the competing +23% (2021→2025) figure came from a secondary source and its underlying series was not confirmed.
• Cumulative CPI shelter since 2020 — not established. Only the year-over-year path was confirmed (little change in 2020, peak ~8.2% in March 2023, 3.3% in June 2026). Pull the shelter, rent-of-primary-residence, and OER index levels from the BLS series and compute the 2020→2026 change directly before citing any total.
• The ~130% monthly-payment figure is my calculation, not a published statistic: 20% down, 30-year fixed, ~$290k at 3.11% vs. $440,600 at 6.69%. The $290k 2020 median is itself derived from the JCHS “+54%” figure rather than taken from NAR’s 2020 series — confirm the actual 2020 median before publishing the number. JCHS’s own $2,570/mo uses 3.5%-down first-time-buyer terms.
• Sources not opened directly (they block automated access; figures came via search results): JCHS State of the Nation’s Housing 2026 takeaways, Zillow ZORI methodology, MBA payment index.
• Medical-care CPI, ~13.3% end-2020 → end-2025. The Dec-2025 figure (+3.2% over 12 months) is corroborated, but the five-year number is compounded from annual Dec-to-Dec changes, not a published series. BLS blocks automated access, so the release pages were not opened directly — pull the medical-care series from the BLS CPI database and confirm the annual figures (2.2 / 4.0 / 0.5 / 2.8 / 3.2) before citing the compounded total.

Primary sources to pull directly: BLS CPI detailed tables · FRED (DFA wealth shares WFRBST01134, WFRBSB50215) · Harvard JCHS State of the Nation’s Housing 2026 and America’s Rental Housing 2026 · KFF Marketplace and Employer Health Benefits surveys · USDA ERS food expenditure series · College Board Trends in College Pricing 2025 · Yale Budget Lab tariff updates · Council on Criminal Justice crime trends · Autor/Dube/McGrew NBER w31010 · Auten–Splinter and Piketty–Saez–Zucman.

---

Handwritten Notebook Pages

Source

Open Affordability crisis notebook pages — start at notebook p. 142 (file not included in this release).

References

Housing — JCHS State of the Nation’s Housing 2026 · price-to-income map · America’s Rental Housing 2026 · NAR first-time buyer share 21%, median age 40 · Redfin: typical first-time buyer is 35 · NAHB Q1 2026 affordability

Wages vs. prices (the deflators for the top callout) — USAFacts: are wages keeping up with inflation? · Hamilton Project: has pay kept up with inflation? · Brookings version · BLS Real Earnings summary · Cleveland Fed: real hourly wage growth across the lower half of the wage distribution (10th percentile +9.7% vs 90th +4.5%, 2020:Q1–2025:Q3) · CNBC: cumulative inflation since 2020 by category

Electricity & household energy, cost growth since 2020 — EIA Electricity Monthly Update: end-use prices (average residential cents/kWh) · EIA Electric Power Monthly · BLS: measuring price change for household energy · FRED: average price, utility (piped) gas per therm · Wolf Street: household energy inflation breakdown (the ~41% electricity / ~54% gas cumulative figures — secondary, confirm against BLS) · American Action Forum: how much are electricity prices rising and why · Choose Energy: electricity rates by state (state-level dispersion) · CNBC: cumulative inflation since 2020, by category

Food, cost growth since 2020 — BLS Consumer Price Index summary (food-at-home and food-away-from-home YoY) · USDA ERS Food Price Outlook · ERS: share of disposable income spent on food, 1997–2025 — the current vintage: 9.7% total in 2025 (4.8% at home, 4.9% away), down from 10.4% in 1997 · ERS chart reporting 11.2% for 2023 and 11.3% for 2022 — superseded vintage, do not cite · ERS: food-away-from-home price growth outpaced groceries over the past decade · USAFacts grocery price tracker (item-level year-over-year detail) · NBC News grocery price tracker · National Restaurant Association menu price index

Child care, cost growth since 2020 — CCAoA: Child Care in America — 2024 Price & Supply (the +29% 2020–2024 vs. 22% inflation figure; 35%/10% income shares) · CCAoA methodology (three-method national average from state CCR&R agency rate surveys) · CCAoA 2024 affordability analysis PDF · FFYF on CCAoA’s 2020 prices (the >$12,300 infant baseline) · Care.com Cost of Care report (the higher parent-self-report series) · TCF: child care funding cliff at one year · NWLC: lessons from pandemic relief funding (the $24B ARPA stabilization window)

Housing, cost growth since 2020 — Freddie Mac Primary Mortgage Market Survey (the 30-year rate series) · JCHS: Ten Takeaways from the 2026 State of the Nation’s Housing ($2,570 monthly payment at first-time-buyer terms) · JCHS press release, 2026 report · FHFA House Price Index · FRED: Case-Shiller U.S. National Home Price Index · MBA mortgage application payment index · Census: renters in 20% of counties paid more in 2020–2024 than the prior five years

Housing, mortgage lock-in — FHFA Working Paper 24-03, The Lock-In Effect of Rising Mortgage Rates (Batzer & Coste) · full PDF — the −18.1% per point, 1.33M prevented sales, and the +5.7% / −3.3% price decomposition · Fonseca & Liu, Mortgage Lock-In, Mobility, and Labor Reallocation (Oct 2023) — the 0.68 pp / 9% mobility effect and the labor-reallocation finding · Liu, Unlocking Mortgage Lock-In: Evidence From a Spatial Housing Ladder Model (SF Fed) · NBER working paper version · Wolf Street: the unwinding stalled in 2026 · Mortgages above 6% now exceed the share below 3%

Housing, the CPI-vs-market-rent measurement problem — NBER: understanding the lag between CPI shelter inflation and market rents · Brookings: how the CPI accounts for the cost of housing · Cleveland Fed: disentangling rent index differences · Zillow ZORI methodology · BLS: rent and rental equivalence

Supply experiment — Pew: Austin’s construction surge drove down rents · Multi-Housing News on Austin · Realtor.com June 2026 rent report · Abundance (Klein & Thompson)

Health care, cost growth since 2020 — CMS National Health Expenditure historical data (2024: $5.3T, $15,474/person, 7.2% growth, 18.0% of GDP) · CMS NHE highlights PDF · CMS press release on 2020 spending (the $4.1T baseline and the 1.9%-without-COVID-programs point) · KFF 2020 Employer Health Benefits Survey (2020 premium baselines) · KFF 2025 Employer Health Benefits Survey · KFF 2025 summary of findings PDF · BLS CPI news release, December 2025 (medical-care index +3.2% over 12 months) · BLS: measuring price change in the CPI — medical care

Health care — KFF: 2026 marketplace enrollment, premiums, deductibles · KFF: 2027 preliminary rate filings · Mercer: highest cost increase in 15 years · SHRM: family premiums near $27k · CDC: uninsured rate 2025 · JHU: Medicare insulin caps

Energy & goods — CNBC/Goldman: electricity prices and AI demand · Newsweek: state-level electricity costs · Bloomberg: battery storage costs −27% · USDA ERS: 9.7% of disposable income on food (2025) · AEI: Chart of the Century · Human Progress: air travel · Gas prices adjusted for inflation

Politics & claims — New Dem Affordability Agenda · Yale Budget Lab: State of US Tariffs · Trump EO exempting food from tariffs · EO 14393, Promoting Access to Mortgage Credit · 50-year mortgage paused · Texas Tribune: Talarico economic plan · Forbes on Mamdani’s city groceries · CNN: greedflation and the SF Fed research · Time: the global affordability crisis

Wages & inequality — Autor, Dube & McGrew, The Unexpected Compression (NBER w31010) · Pew: have wages kept up with inflation? · FRED: top 1% net worth share · Auten & Splinter, top 1% income shares · Tax Policy Center: primer on the inequality measurement debate · AEI: The Cost of Thriving Has Fallen (Winship & Horpedahl)

Sentiment — Chase: are Americans too pessimistic? (2026) · University of Michigan Surveys of Consumers · Council on Criminal Justice: crime trends mid-2026