Record Home Prices That Aren't Really Records

The news

Why "real" prices matter

The gap between the two views is large over long periods. Nominally, national prices are 81.1% above the bubble peak. In real terms they are only about 8.3% above it, and the Composite 20 is only 0.5% above. A rough inference from those two figures, not one the article makes: consumer prices overall have risen by about two-thirds since the bubble peak, because 1.811 ÷ 1.083 ≈ 1.67.

The same logic explains how prices can be "flat" and "falling" at once. An illustration with assumed numbers: if a home's price is unchanged over a year while CPI rises 3%, its real price has fallen about 3%. That is roughly what McBride expects for 2026.

The pattern after a boom

McBride points to a pattern he has written about before under the title "House Prices: 7 Years in Purgatory". After a sharp run-up, real prices usually take several years to reach a new high. Nominal prices tend to stall rather than crash, and inflation slowly brings real prices down. He says there is "nothing magic" about seven years; it just made a good headline. The current stretch, at 51 months, fits the pattern so far.

The 2020–2022 run-up was steep, so on this reading a long flat spell is normal and not a sign of trouble.

The 2026 outlook

McBride's forecast at the start of the year was "mostly flat to a small decline in prices nationally in 2026." He now thinks nominal prices will end the year mostly flat, with real prices falling further. He gives two reasons:

• More inventory. More homes are for sale, which gives buyers more choice and less reason to bid up prices.
• Higher mortgage rates. Higher rates cut how much buyers can borrow on a given income, which reduces demand.

Still expensive by history

Even after four years of real declines, McBride stresses that real prices are historically high: about 8.3% above a bubble peak that was itself seen as extreme. He adds a qualifier. Real house prices have tended to rise slowly over the long run, so some gain over 20-plus years is expected. That makes the bubble peak an imperfect benchmark, but he still judges today's level as elevated.

What's left out and the caveats

A general caveat that applies to this kind of analysis: CPI includes shelter costs, which follow rents. Deflating house prices by headline CPI therefore partly measures housing against itself. Analysts sometimes use CPI excluding shelter for this reason, which would give somewhat different real figures. The article uses headline CPI. Case-Shiller is also a three-month moving average released with a lag, so the August figure reflects deals mostly struck in early summer.

For someone deciding whether to buy, the practical point is that a record in dollars is not a record in value. In real terms, buyers today pay a little less than in 2022, though still a lot by long-run standards. And borrowing costs, not the sticker price, often decide affordability.

Carlson's answer has two parts. On the reader's own terms, the numbers really are poor. But Carlson argues those are the wrong terms.

How affordability broke

That balance ended quickly. Prices rose more than 50%, and mortgage rates more than doubled. His updated chart shows average monthly payments now more than double their late-2010s level.

To see why the two combine so sharply, here is an illustrative calculation (ours, not Carlson's), counting principal and interest only. A $400,000 house with 20% down needs a $320,000 loan. At 3% over 30 years, that costs about $1,350 a month. Raise the rate to 7% and the same loan costs about $2,130. Raise the price by 50% as well, to $600,000 with a $480,000 loan, and the payment is about $3,190. That is roughly 2.4 times the original. Each change pushes the payment up, and together they multiply.

The rest of the bill has grown too. Carlson notes home insurance is up almost 50%. The 20% down payment is a fixed share of a bigger price, so it is a larger dollar amount. Property taxes are generally based on assessed value, so they rise with it. His verdict is that affordability is "about as bad as it's ever been."

Why returns probably cool from here

He lays out the forces on each side:

• Demographics, for now: The large millennial generation is in its prime home-buying years, which supports demand.
• Supply in the 2030s: Carlson expects "a decent amount of supply" to reach the market as baby boomers die, downsize or move to Florida.
• Underbuilding: The main bull case is that the U.S. simply isn't building enough homes.
• Lower rates: If mortgage rates fall to a more reasonable level, housing activity could surge. Carlson's footnote: unless rates are falling because of a financial crisis.

Location shapes all of this. Some cities are far less affordable than others, and some regions are more attractive. In some places, renting and investing the difference in stocks wins financially, and you avoid the burdens of ownership. In others, the better move is to lock in a 30-year fixed rate, refinance if yields fall, and build equity. A buyer also has to be comfortable owning a concentrated, single, undiversified, leveraged asset. No stock investor would build a portfolio that way.

The leverage point

Carlson's reframe: housing is consumption

Carlson grants that a home can be a financial asset. You build equity over time. You can borrow against it to renovate. And it gets favorable tax treatment; in the U.S., for example, a large part of the gain on selling a primary home is excluded from capital-gains tax. But he does not treat a primary residence as an investment "per se." You need somewhere to live either way, so housing is mainly consumption.

Carlson says the psychological side can't be left out, despite the reader's request. People buy to put down roots, to get a stable housing payment that hedges against inflation, to control their living situation, to own something outright in retirement so their biggest monthly expense goes away, and to leave something to their heirs.

What it leaves open

Carlson admits he doesn't know whether housing will deliver solid gains. His baseline guess is that prices keep up with inflation. He doesn't run an actual rent-versus-buy comparison for the reader. He also doesn't respond to the reader's comparison with high bond yields or option-income funds, which have their own risks that the reader doesn't mention. The practical advice is short: run the numbers for your own market, but don't "live and die" by the return you expect a house to earn.

Richard Quinn, a retired benefits consultant who writes regularly for HumbleDollar, has collected survey results showing how little Americans know about Social Security. His argument: this ignorance does more than embarrass. It leads people to claim benefits at the wrong time, plan retirement around bad assumptions, and resist reforms they don't understand. "It's hard to solve a problem if we refuse to accurately understand it," he writes.

What people get wrong about timing and size

How much it replaces. About one in five Americans think Social Security alone will cover their retirement. Quinn notes it replaces about 40% of an average worker's pre-retirement earnings. The share is higher for low earners and lower for high earners, because the benefit formula is progressive: it replaces a bigger fraction of the first dollars of lifetime earnings.

The benefit formula itself. Only 27% of respondents in a Social Security Administration study understood how benefits are calculated. Two things matter most: your earnings history and the age you claim.

Family benefits. People often miss benefits for spouses, for ex-spouses (if the marriage lasted 10 years or more) and for survivors. Quinn says this leads households to claim at the wrong times. One common example from general practice: a higher earner who claims early also shrinks the survivor benefit the widowed spouse will live on later.

Taxes: two opposite mistakes

How it actually works: under long-standing federal rules, once combined income passes about $25,000 for a single filer ($32,000 for a married couple), up to 50% of benefits count as taxable income. Above $34,000 ($44,000 for couples), up to 85% does. That income is then taxed at your ordinary rate. So for a retiree in the 12% bracket, the most they pay is about 10% of their benefit (85% × 12%), not 85%. These thresholds aren't indexed to inflation, which is why more retirees cross them every year.

Funding and "bankruptcy"

He also pushes back on two common claims:

He adds that Social Security is insurance, not just a pension. It also covers disability, spouses and survivors. Debates that treat it purely as a retirement investment leave that out.

Quinn's own numbers, and their limits

He also claims most people get back what they paid in SS taxes within six years of starting benefits. The article gives no source for this. It probably counts only the worker's share of the tax, and nominal dollars that aren't adjusted for inflation. His own example is a long-lived married couple, which is a favorable case. Your own result depends on your earnings, how long you live and whether you're married. Your contribution history on SocialSecurity.gov is the place to check.

What's left open

This is a short opinion post, not a study. It cites several surveys but doesn't link them or describe how they were done. It argues that the cap fix isn't enough but doesn't say what mix of tax increases and benefit changes Quinn would support. His core point does hold up: many popular beliefs about Social Security are wrong, and people pay for those mistakes in their own claiming decisions.

Economist Timothy Taylor, who writes the Conversable Economist blog, has a short post with a simple complaint about how Americans debate health care. The US system's problems are well documented: high costs, people without insurance, unequal access to care, and health outcomes worse than in many other high-income countries. But the alternative people offer is usually vague: "national health insurance, like all the other countries in the world," or "Medicare for all." Taylor calls these alternatives "nebulous" because the slogans seem to promise all the care anyone wants, with all or nearly all of the bill paid by someone else. He argues that neither Medicare nor foreign systems actually work that way.

What Medicare actually looks like

Medicare is the US federal health program for people 65 and older. Taylor points to three features that the "Medicare for all" slogan usually leaves out.

It has a long-term funding gap. Under current law, projected Medicare payments over the next 75 years exceed projected revenues by about $60 trillion. Taylor's dry suggestion is that it "might be prudent" to fix Medicare's long-term solvency before promoting it as the solution for everyone.

"National health insurance" names many different systems

The second half of the argument is that "what other countries do" is not one thing. National health insurance systems vary a great deal even among rich countries.

Canada vs. England: the same headline, different systems

Americans most often compare themselves with other English-speaking countries, especially Canada and England. The headline numbers look alike:

Taylor sees two sides to the spending gap. Spending 11% of GDP instead of 18% leaves more for consumption of other goods and services. But he suspects any concrete US proposal to cut total health spending by about a third, down to British or Canadian levels, "would go down in flaming defeat." A slogan can promise savings. A specific plan has to say whose income or care shrinks.

Below the headline numbers, the two systems are built differently:

So "do what Canada does" and "do what Britain does" point to different policies on insurance, governance and doctors' pay.

Taylor also notes that Germany, the Netherlands and Switzerland build universal coverage around individual health insurance policies, with regulated competition among insurers. People hold their own policies from insurers that compete with one another, under rules set by the government. That is a long way from both the British single national service and Canada's provincial public plans.

What the post does and doesn't do

Taylor's conclusion is that complaining about the US system isn't enough. Reformers can learn from other countries, but the systems differ, so a serious debate has to get into the details, and the Commonwealth Fund profiles are a reasonable place to start.

The post is short, and its limits are worth stating. It doesn't compare health outcomes, waiting times, prices or administrative costs across systems. It doesn't say which model fits the US best or why US spending is so much higher. It also doesn't break down the $60 trillion Medicare projection. What it does is make the vague debate more specific: before arguing for "national health insurance," say which country's version you mean, and how it handles private insurance, governance and paying doctors.

Why the question is hard

A second difficulty is selection. If people who use AI produce more, that could be because AI helps, or because productive people are the ones who adopt it.

Three sources of evidence

The authors combine three datasets:

Finding 1: after ChatGPT, resource gaps narrowed

ChatGPT's public release in late 2022 gives a natural before-and-after comparison. Afterwards, NBER authors without RAs published a larger share of their papers in top-tier journals. Authors from smaller teams increased their working-paper output and moved papers to publication faster. The natural reading is that AI gave researchers who lacked assistants a cheap version of the help better-resourced colleagues already had.

Finding 2: on average, little effect; underneath, a split

• More AI exposure narrows the output gap between researchers with and without RAs.
• More AI exposure widens the gap between senior and junior researchers.

Ruling out the obvious objection

The model: execution versus conception

To explain the pattern, the authors build a model with three assumptions:

• Diminishing returns to execution. The first bit of coding and data help matters a lot. Each additional bit matters less.
• Increasing returns to conceptual capacity. More judgment and expertise pays off disproportionately.
• Complementarity. The two inputs make each other more valuable.

Here is an illustrative analogy (not from the paper). Two cooks get a free kitchen assistant. A home cook who used to do all the chopping gains a lot, because the worst bottleneck is gone. A head chef already had prep staff, so the assistant adds less labour. But the chef can now try more dishes, and those dishes benefit from a lifetime of knowing what works. The assistant closes the gap between kitchens with and without staff. It widens the gap between novice and master, because the master turns freed-up time into more valuable output.

In the paper's terms, AI is a cheap supply of execution. Because execution has diminishing returns, AI helps most where execution was scarce, among researchers without RAs. Because execution complements conceptual capacity, which has increasing returns, AI also raises the value of seniors' accumulated expertise.

What may come next

The authors suggest the early convergence may not last. Once AI has removed execution constraints for nearly everyone, the levelling effect runs out. If AI also gets better at conceptual work, the inputs that seniors already have in abundance become more productive still. On that reading, the overall effect could move from narrowing gaps to widening them.

What to keep in mind

This is a job-market paper, which means it is new and has not been peer-reviewed. The abstract gives no effect sizes, sample sizes for the NBER analysis, or detail on how LLM-assisted writing is detected. A survey of 227 economists is modest and may over-represent people interested in AI. Publication in top journals depends on referees and editors, who may themselves be changing how they judge AI-assisted work. The broader point still applies outside academia. A tool that substitutes for routine help mainly benefits people who lacked that help. A tool that complements judgment benefits the people who already have the most judgment.