As a reader of the Pancake, it’s highly likely that you follow politics closely enough to agree that Medicare-for-All (MFA) is at or near the top of the agenda for most Progressive Democrats. It has been one of Senator Bernie Sanders’ mantras for years, asserting that “we must resolve that health care is a human right, not a privilege.”

As a Medicare recipient (i.e., at least 65 years old), I find it simple and effective, though I also need a private “supplemental” policy to cover what Medicare doesn’t. (For you young’ens, it pays for 80% of most doctor and hospital charges, but not drugs without a separate policy). Compared to private insurance with similar coverage, it’s a great value, costing a tad over $2400 per person this year, with small increases most years. It’s also funded by the 2.9% tax on wages, split equally by individuals and employers.

Medicare is a single-payer system. That is, a single public authority—in the U.S., the government — replaces multiple private health insurance companies to collect funds and pay medical bills. The system is primarily funded through public sources and taxes rather than private monthly premiums.

Typically in a single-payer system, the government pays the bills, but the healthcare is usually delivered by private doctors, clinics, and hospitals. The single payer sets standard fees as Medicare does, and negotiates drug prices for the entire population, which Medicare is just starting to do for a limited number of drugs. This is the model used in Canada. The United Kingdom is also single-payer, but the National Health Service is government-run, employs the doctors, and owns the hospitals, labs, and clinics. Patients never see a bill. It’s actually the model used by the Veterans Administration for its health system.

The U.S. spends about twice as much on health care as other OECD countries, with no appreciable advantage in outcomes. We know that the system is broken and that nobody knows what health care really costs. The broader discussion now should focus on the uncertainty about whether and how costs can be reduced.

Would Medicare work if simply expanded to include everyone?

Medicare today isn’t actually a self-contained, break-even system funded by senior premiums. So “just expand it as-is” runs into some real structural issues beyond the political debate it generates.

The reason is that while Medicare beneficiaries pay premiums, those premiums cover only a small fraction of program costs, roughly a quarter of Part B (doctors) spending, for example. The deficit is covered from general federal revenue. The payroll tax on current workers is earmarked for Part A (hospital insurance) expenses. Thus, retired seniors aren’t the ones “carrying” the system financially. They—we— are net beneficiaries of a program that working-age people are subsidizing

Therefore, if we simply applied senior premium levels to everyone, we’d capture only a small slice of the actual cost of care. The payroll tax base would need to expand dramatically, or other revenue sources would need to be found, to cover a population far larger than the traditional Medicare-eligible group.

Furthermore, Medicare pays providers less than private insurance does—often 40%- 60% less. This works today largely because providers can cost-shift, charging private insurers more to offset lower Medicare rates. If everyone were subject to Medicare rates, that cost-shifting cushion disappears. Some hospitals, especially rural ones, already operate at a loss on Medicare patients. Many providers could not remain financially viable if Medicare rates applied to their entire patient base.

Thus, expanding Medicare “as currently financed” to all wouldn’t be economically self-sustaining. It works now because it’s subsidized by taxpayers and cross-subsidized by private insurance rates. Any real Medicare for All proposal has to address financing and provider rates as separate, harder problems rather than assuming the current senior-focused funding formula would simply generalize.

How a Medicare for All system might be structured

Even Sanders recognizes that universal health care would require a structure apart from current Medicare. Here are some elements of his proposal, the Medicare for All Act.

The bill is a phased transition rather than an instant universal switch. Individuals age 18 or younger could enroll starting one year after enactment, while everyone else could buy into a transitional plan or an expanded Medicare program.

The benefits would be considerably richer than today’s Medicare. It’s designed to provide comprehensive benefits including primary care, vision, dental, prescription drugs, mental health, substance abuse treatment, long-term services and supports, and reproductive health care, with no cost-sharing for long-term care and a priority on home-based care over institutional care.

On the spending side, the bill leans heavily on government bargaining—or monopsony—power1. The Department of Health and Human Services (HHS) would be required to negotiate prescription drug prices, establish a formulary, and set rules for provider participation. HHS would administer payments and monitor costs, essentially standardizing rates the government pays across the whole population rather than the current patchwork of Medicare, Medicaid, and private rates.

Under the Sanders bill, with a genuine single-payer design and no opt-out option, we would no longer pay premiums, co-payments, or deductibles to private health insurance companies. That industry would, in effect, be phased out, which is one big reason adopting single-payer is so politically fraught.

So how is this version of MFA financed?

This is the meat of the issue. Alas, there is no definitive resolution because so many assumptions swing the economics from feasible to impossible.

The starting numbers

For scale, total U.S. health spending in 2024 was $5.3 trillion.2 Of that, private health insurance spending was $1.66 trillion (31% of the total), Medicare was $1.12 trillion (21%), Medicaid was $931.7 billion (18%), and out-of-pocket spending was $556.6 billion (11%). So the private-insurance-plus-out-of-pocket pool that could be replaced by a single-payer system is roughly $2.2 trillion, a large chunk of the system, but only part of it.

Sanders’ office has floated a menu of financing options rather than one locked-in formula. For example, one option could be a 7.5% employer payroll tax, estimated to raise about $3.9 trillion over ten years, while also saving employers the expense they now absorb for their share of private insurance for employee coverage.

Independent estimates of the overall cost of a Sanders-like single-payer system vary widely depending on assumptions about provider rates, drug prices, and utilization. The Committee for a Responsible Federal Budget (CRFB) found net spending increase estimates ranging from $17 trillion (the lowest up to $54 trillion over a decade. Landing on a “rough midpoint” figure, CRFB estimated financing roughly $30 trillion would require something equivalent to a 32% payroll tax, a 25% income tax surcharge, a 42% VAT, a $7,500 per-capita mandatory premium, doubling all income tax rates, cutting non-health spending 80%, or increasing debt by 105% of GDP. Clearly, taxes on high earners and corporations alone would not be sufficient.

That gap between the Sanders’ bill’s suggested financing menu and outside estimates of what full financing would actually require is really the crux of the economic debate. Critics argue the “menu” understates true costs, while supporters argue the outside estimates overstate them by not fully crediting administrative savings and drug price negotiation.

The Sanders bill doesn’t rely on current Medicare premiums expanded to everyone. It builds an entirely new financing architecture — payroll taxes, surcharges, and the elimination of tax breaks — precisely because the current premium-plus-payroll-tax model for seniors doesn’t generate nearly enough revenue to cover the entire population.

The case for a single-payer system

Obviously, the top rationale is to provide universal health care for all Americans. And a major component, say the advocates, is the administrative savings compared to today’s hodgepodge of private insurers.

Under any scenario, these savings are genuine. One detailed billing-cost simulation study found national billing and insurance-related costs would fall 33-53% under a Medicare-for-All design. The Congressional Budget Office’s (CBO) analysis estimated that single-payer would cut administrative costs by roughly $500 billion per year.

Moreover, there’s that $556 billion that Americans would save out of pocket on deductibles, co-pays, and other expenses that would be included under single-payer. In effect, that’s a pool that could be tapped through taxes to help fund the system.

Less obviously, under M4A, hospitals would collect payment for services provided to currently uninsured and underinsured people, which are now written off as uncompensated care or bad debt. Uncompensated care costs are already a live budget item which hospitals, in particular, have to absorb. Under universal coverage, that line item shrinks or disappears.

The economic hurdles to implementing single-payer in the U.S.

The political barriers to replacing our current system with a complete Medicare-for-all system are formidable. But they might be less so if the economics of a new system were clear-cut. They’re not. I’ll start deep in the data weeds and move to the overarching reason that even a fully formed single-payer system would not bring overall costs down, to say what Canada spends on healthcare.

I’ll start with payments to hospitals. Today’s Medicare, on average, pays about 82 cents for every dollar hospitals spend caring for Medicare patients. Hospitals recorded a 12.1% loss in 2024, with projected margins remaining around negative 10% in 2026. These losses are partially covered by cross-subsidies from payments from commercial-rate private insurers. This is a well-documented feature of U.S. hospital finance.

That said, a significant subset of hospitals, roughly 13% of all hospitals, categorized as efficient because they perform well on both cost and quality, has been shown to break even on Medicare payments. This suggests that at least part of the negative margin picture reflects operating inefficiency rather than an unfixable rate problem. Rural hospitals, however, face greater challenges for such performance. Overall, however, for-profit hospitals are profitable, with Medicare simply being the lowest-priced segment.

The Sanders bill and most serious M4A proposals anticipate negotiated rates somewhat above current Medicare levels precisely because architects are aware of the underpayment problem. But that also means the cost of providing the same services as today will increase for current Medicare reimbursements and for everyone added to coverage.

Overall, commercial insurers charge and pay hospitals about twice as much as Medicare does. Thus, despite the administrative savings, out-of-pocket costs, and the demise of employer-insurer payments, the rate gap remains substantial. That gap is too great to close purely through efficiency and volume effects. That a few efficient hospitals can break even on Medicare payments doesn’t prove all hospitals could adapt at the same pace, especially smaller or rural ones without the same economies of scale. So the underlying concern, that provider payment design is the load-bearing question for any transition, is often underweighted in single-payer advocacy.

The bottom line: Assume we took all the premiums for private insurance paid by employees and employers, plus out-of-pocket expenditures (i.e., deductibles and co-pays) of the privately insured, and factored in the real, if disputed, efficiency savings to hospitals and doctors from a single-payer system. Now, replace all that cost with a single-payer system. Wouldn’t there be a strong case that, overall, this would cost less than aggregate costs today, even assuming payments to hospitals and doctors are greater than current Medicare rates?

This is exactly the exercise CBO, RAND, and others have run. But when they run it rigorously, the answer isn’t a clean “yes, cheaper.” It’s “it depends enormously on a couple of specific assumptions,” and the range of outcomes is wide enough that confident claims in either direction usually reflect which assumptions the analyst picked, not settled math.

What the simple framing of Medicare-for-all tends to miss

1. Eliminating cost-sharing increases utilization — a lot, and this is the single biggest wildcard. This is the key to what most back-of-the-envelope versions skip. When people face $0 out-of-pocket cost at the point of care, they use more care. This is one of the most well-established findings in health economics, documented in the RAND Health Insurance Experiment. RAND’s own modeling of a Medicare-for-All-style plan found total health expenditures would be about 1.8% higher in 2019 relative to the status quo — not because administrative savings didn’t materialize, but because they were roughly offset by the surge in utilization from covering the uninsured and eliminating cost-sharing for everyone else. The Urban Institute similarly noted that due to the demand effect, a plan with such a large increase in coverage and essentially no cost-sharing will result in higher national expenditures, even when private-sector payment rates are cut to cost.

2. Covering the uninsured is new spending. The framing that takes premiums plus out-of-pocket costs and swaps in single-payer implicitly assumes just reallocating money that’s already circulating. But currently uninsured people, still tens of millions, generate relatively little formal spending today. Covering them for the first time adds real new utilization and costs.

3. The range across serious studies is wide, and it hinges on two variables: utilization growth and administrative/negotiation savings. The CBO modeled five illustrative single-payer designs and found that the effect on national health expenditures ranged from a $700 billion decrease to a $300 billion increase by 2030, all using the same conceptual system but applying different design choices (provider rates, cost-sharing levels, negotiating power). Thus, the honest answer is that single-payer is plausibly cheaper, yet equally plausibly a bit more expensive.

4. Provider payment rates are the pivot point. If you pay providers above Medicare rates, one preserves more of the hospital-margin problem but shrinks the savings side of the ledger. Every serious model shows an almost direct tradeoff: the higher you set provider payment rates to protect hospital viability, the smaller the net system-wide savings become. There’s no assumption set that gives you both large savings and generous provider rates. There’s no free lunch.

5. Supply-side capacity is the thing none of these dollar-figure models capture well. Even if the dollars net out to roughly break-even, a large utilization surge run through the same physician and hospital capacity we have today doesn’t necessarily translate into higher spending. Instead, it can show up as longer wait times, harder-to-get appointments, or rationing by queue rather than by price. Foregone or delayed care is a real economic cost that doesn’t appear in these calculations.

Wait. How can Canada and most of Western democracies make it work?

These systems built their cost structure over 60+ years with single-payer rate-setting embedded from the start (Canada's system dates to the 1960s-70s). The U.S. is trying to retrofit a similar destination onto an entrenched, high-margin system with powerful incumbent stakeholders (hospital systems, device makers, pharma, insurers, specialist physician groups) who built their business models around current price levels. That's a political-economy problem more than a pure economics-of-design problem. Therefore, it’s a big part of why U.S. models project higher costs than these countries actually spend today. They're not just modeling a single-payer system, they're modeling "converting a system already inflated to U.S. price levels."

When a Canadian or European single-payer system negotiates rates, it's negotiating from a baseline that never had U.S.-level prices to begin with. A U.S. transition, by contrast, has to claw down from an already-inflated price structure all built up over decades under a system with weak aggregate bargaining power. That's a much harder political and economic lift than simply "having a single payer," and it's exactly why serious U.S. proposals, constrained to pay providers something politically viable and above Medicare rates, can't fully capture the savings Canada or others enjoy, which stem from never having let prices rise that high in the first place.

No, it’s far from simple—or maybe even feasible

So, yes, Medicare has much lower administrative overhead than private insurance, and a single payer gets negotiating leverage to bring down drug and procedure prices, potentially offsetting some of the revenue gap. Meanwhile, the economic analysis is weak in weighing the moral issues underlying the recognition of health care as a human right.

However, from an economic perspective, the funding gap is too large to close through efficiency gains alone. Lower provider payments would cause access problems, such as fewer providers accepting patients, longer wait times, or providers even exiting practice. Universal health care is desirable, but would require an expansion of the number of health care providers. Any simple expansion of Medicare undersells how much new tax revenue would actually be required, while a complete do-over, à la Sanders, is as likely to raise overall medical spending above our already global-leading sum.

Leave a comment

Share

Read this post on Substack.

Medicare today isn’t actually a self-contained, break-even system funded by senior premiums. So “just expand it as-is” runs into some real structural issues beyond the political debate it generates.

Expanding Medicare “as currently financed” to all wouldn’t be economically self-sustaining. It works now because it’s subsidized by taxpayers and cross-subsidized by private insurance rates. Any real Medicare for All proposal has to address financing and provider rates as separate, harder problems rather than assuming the current senior-focused funding formula would simply generalize.

Leave a comment

The infatuation with democratic socialism that has become salient among many Democrats in the past year could have serious repercussions in both the short term and, if it persists, the long term. In the immediate future, the success of several socialist or socialist-adjacent candidates in this year’s primaries could squander the opportunity to elect congressional majorities for the Democrats to impose some limitations on Donald Trump. And I’m mystified by this turn toward a socialist agenda. My sense is that most voters don’t know what that really entails, and their Republican opponents will use their money advantage to frame all Democrats in competitive races with the socialist moniker.

Lunch at the marina

Over lunch with a longtime friend at a bistro overlooking the South Portland marina on a gray afternoon this summer, the conversation turned to the political fixation the Democrats seemed to have on the Democratic Socialists. My liberal friend seemed sympathetic to their cause, though it struck me that he didn’t really know what socialism entails.

I suspect that he, like many on the progressive end of modern liberalism, views socialism as a benign form of the welfare state, with Sweden as their mental model. If you’re thinking, “Yeah, that’s about right,” my objective, then, is to explain or refresh your understanding of this economic model and, along the way, point out, if you hadn’t been keeping track, that Sweden, the poster-society for the social welfare state, has largely abandoned socialism as a failed model for sustaining economic growth and creating the platform for broadbased oportunity.

What “Socialism” Traditionally Means

Before wading into the Democratic Socialists of America (DSA), it’s worth resetting the baseline, because the word “socialism” gets used so loosely in American politics that it has almost stopped meaning anything specific.

Classically, socialism is an economic system in which the means of production — factories, land, major industries, capital — are owned or controlled collectively, usually by the state. It is an alternative to capitalism, under which private individuals or shareholders own and control most property and production. The core ideology of socialism is that production should be prioritized above all to meet human needs rather than to generate profit for owners. Historically, there have been several flavors.

Marxist-Leninist socialism, most commonly popularized as communism, is the most extreme form of socialism. The state controls all aspects of the economy (often achieved through revolution). Though in theory a communist economy could exist within a democratic political sphere, in every application that has been practiced, there has been a single “party” that determines what gets produced, how it will be distributed, at what price, and how much all workers will be paid. This was the basic tenet in the Soviet Union, Maoist China, and Castro’s Cuba. Private property in productive assets is abolished or very tightly restricted.

In a social democracy, such as in the Scandinavian countries, in particular Sweden until the 1990s, private markets and private property remain largely intact, but the state heavily regulates markets and aggressively redistributes wealth through taxation to fund a generous welfare state, including universal healthcare, generous pensions, free education beyond high school, free child care, and full unemployment support. This Nordic model is what people usually have in mind, even though most Nordic countries don’t describe themselves as “socialist” today.

Democratic socialism is simply a commitment of socialists to work through the democratic process, differentiated from historical socialist states run under an authoritarian one-party system. It holds that social democracy isn’t sufficient to achieve its ends. Thus, workers should actually own and democratically control workplaces and major industries, rather than merely taxing their owners more heavily. In principle, this is meant to be achieved through the ballot box rather than revolution.

The distinction that matters most for the rest of this post is this: social democracy tames capitalism through taxation and regulation; democratic socialism aims to replace private ownership of the economy with collective or worker ownership. Sweden, as I’ll get to, is a social democracy that has moved even further toward markets in recent decades — not the democratic-socialist model the DSA is aiming for.

DSA’s electoral wins: From the fringe to the mainstream?

For decades, the democratic socialist position was voiced on a national scale by Bernie Sanders, who joined theYoung People’s Socialist League as a student at the University of Chicago and first came to prominence by winning an election (by about 10 votes) as mayor of Burlington, Vermont, in 1981. He’s been joined in Congress by a handful of actual DSA acolytes, notably Representatives Alexandria Ocasio-Cortez and the so-called Squad: Ilhan Omar, Ayanna Pressley, and Rashida Tlaib.

What makes 2025–2026 different from the DSA’s earlier history is the sheer number of candidates who’ve actually won, not just run:

Zohran Mamdani was elected Mayor of New York City in 2025. After gaining more votes in the Democratic primary than former Governor Andrew Cuomo. Melat Kiros is a 29-year-old first-time candidate and DSA-endorsed democratic socialist who is now the Democratic candidate for a House seat from Colorado, having unseated a 15-term incumbent Representative, Diana DeGette in a very blue district, even though DeGette herself has been a longtime progressive who supported Medicare for All and abolishing ICE.

Darializa Avila Chevalier and Claire Valdez won congressional primaries, also in heavily Democratic districts. Chevalier narrowly defeated five-term incumbent Adriano Espaillat, chair of the Congressional Hispanic Caucus, and Valdez won the open-seat race, both in New York City. Chris Rabb, a Pennsylvania state representative and democratic socialist, won the May Democratic primary for a safe Philadelphia U.S. House seat.

Abdul El-Sayed, who won a very close Democratic primary in Michigan earlier this month, has distanced himself from the democratic soicalist label. “I believe in capitalism; I just believe that capitalism has to be regulated," was his self-identification in a CNN interview in July. He told The Wall Street Journal, “I'm not technically or practically DSA," Nonetheless, he has a long history of attending and speaking at events hosted by the Democratic Socialists of America, and many of his positions ape those of the DSA

Many of El-Sayed’s key positions are congruent with the DSA, such as Medicare-for-all, free child care, labeling Israel’s actions in Gaza “genocide,” and dismantling the Immigration and Customs Enforcement Agency. He has also played footsie with Hasan Piker, the political commentator and online streamer, who self-identifies as a socialist and anti-imperialist, and whose commentary is infused with a Marxist economic critique,

Whatever one thinks of its platform (see below), this is no longer a marginal protest movement. DSA-aligned candidates have gone from city council long-shots to unseating multi-term incumbents in Congress and running America’s largest city.

What the DSA’s platform actually says

The DSA is the largest socialist organization in the U.S., with membership that has grown from roughly 6,000 in 2015 to over 100,000 in 2026. In June, its governing National Political Committee adopted a new platform called “Workers Deserve More,” which lays out the organization’s current agenda in unusually explicit terms.

Highlights of the platform include:

• Public/worker ownership in place of private corporate ownership in key sectors

• Abolishing the U.S. Senate and replacing the presidency and Supreme Court with an executive and judiciary chosen by, and subordinate to, Congress

• Drafting a new U.S. Constitution

• Allowing all permanent residents (not just citizens) to vote

• Abolishing ICE and moving toward decarceration (”abolishing the carceral forces of the capitalist state”)

• State-run healthcare, free public college, and expanded rent control

• Reparations for slavery and amnesty for undocumented immigrants

It’s an admirably transparent, if radical, document. This isn’t a wish list of higher taxes and expanded benefits, but a proposal to restructure the constitutional order itself alongside the economy. Notably, reporting has pointed out that the platform doesn’t necessarily match the individual policy agendas in the stump speeches of DSA-endorsed candidates; many downplay or simply never mention the more structural planks, such as abolishing the Senate or writing a new Constitution.

Let’s talk Sweden

Sweden is a country “still remembered fondly by progressives everywhere as the poster child for democratic socialism.” Sweden was a genuine pioneer of the cradle-to-grave welfare state. The Swedes even had a word for it, folkhemmet, “the people’s home.” But to deliver these services, it taxed its citizens heavily, sometimes to an absurd degree. In 1976, children’s author Astrid Lindgren, of Pippi Longstocking fame, faced a marginal tax rate of 102%, meaning she owed more than she earned.1

In the words of Swedish economist Johan Norberg, by 1970, Sweden had become one of the world's wealthiest countries and began experimenting with socialist ideas.

From 1970 through 1990, government was massively expanded, taxes were raised, and the economy was regulated. This was not the golden era of socialist nostalgia, but more like an Atlas Shrugged moment: Swedish companies like IKEA and Tetra Pak and lots of successful entrepreneurs left Sweden and not a single net job was created in the private sector. It was the one moment in Sweden’s modern history that we lagged behind other countries.

That era ended in crisis. In 1991, Carl Bildt, who became Sweden’s first conservative prime minister in six decades, oversaw a government that “slashed public spending, privatized state monopolies and carried out what was known as the tax reform of the century.”

The numbers since are stark: the top income tax rate has fallen from nearly 90% in the 1980s to 50% today; public social spending (education, healthcare, pensions) has fallen to roughly 24% of the economy — near U.S. levels and far below France or Italy; government debt sits at just 36% of GDP versus over 120% in the U.S.; and a country of under 11 million has produced Spotify, Skype, and Klarna. According to Lars Jonung, Bildt’s former chief economic adviser, “We have put a capitalist revolution into a former, very strongly socialist-oriented society.”

There were some consequences. Wealth inequality has widened under the new model, and Sweden’s partial privatization of education has produced a bitter fight over “free schools,” publicly funded, privately run — they sound like some charter schools in the U.S.— which critics accuse of hiring less-qualified teachers to pad profits.

Sweden still has a bigger welfare state than the United States. However, Kjell-Olof Feldt, who was the Social Democratic Minister of Finance from 1983 to 1990, reflected, “What we believed in as young socialists simply turned out to be impossible in practice.”

Why the DSA’s Vision Is Misguided

The case studies of the Nordic countries, led by Sweden, should be a red flag to those who are attracted to the DSA and its fellow travelers. Democratic socialists routinely hold up Scandinavia as proof of concept. But Sweden’s own trajectory runs in the opposite direction: away from state monopolies and 90%-plus marginal tax rates, toward markets, privatization, and fiscal discipline. Above all, it was reversed democratically. That is, the very citizens who supposedly benefited from the cradle-to-grave state voted it out. The cost/benefit equation was negative. If the most-cited real-world example spent the past 30 years dismantling the parts of its system Americans associate with “socialism,” that’s not a footnote; it’s the headline.

Although not all progressives buy into the entire DSA platform, they certainly carry the full DSA baggage if they accept the political support of DSA leaders. The platform confuses two different critiques of American life. Much of what draws people to DSA candidates — the cost of housing, healthcare, and childcare — is a legitimate affordability issue that plenty of non-socialists worry about too. But the “Workers Deserve More” platform bundles those genuine grievances with a set of structural constitutional changes (e.g., abolishing the Senate, drafting a new constitution) that have nothing to do with the cost of living and everything to do with the concentration of power. Voters backing Mamdani or Kiros for affordability reasons are not thereby endorsing the abolition of the Senate — but the platform doesn’t distinguish between the two, and neither, often, do the candidates on the trail.

It’s also ill-advised to conflate success at the ballot box with success in governance. Winning a primary against a less charismatic incumbent, or winning a mayoralty on a message of affordability, says a great deal about voter frustration with the status quo and relatively little about whether rent freezes, state-run healthcare2, or public ownership of industry actually deliver better outcomes once implemented. Sweden’s own history is instructive here too: its most heavily socialized period in the 1970s and 80s ended not in triumph but in a currency and banking crisis severe enough to force the reforms away from socialism.

DSA supporters might reasonably respond that Sweden’s welfare state itself (universal healthcare, subsidized childcare, strong labor protections) remains largely intact even after the market reforms, and that what Sweden abandoned was really “socialism” in only the loosest sense while keeping the redistributive core. They might also point out that there were real costs to Sweden’s pivot. And I haven’t heard of any DSA candidate proposing Soviet-style central planning, although tight state control of big business, such as airlines, could stifle innovation and price competition to a similar effect.

My big problem with the new crop of DSA, the DSA-aligned, and the DSA-curious candidates is that, as David Frum articulates in a recent Atlantic essay, they are more focused on promoting their ideological agenda than on corralling Donald Trump. To do that, the challenge for the Democats is to win elections in those moderately red and purple districts and states that Trump won in 2024. And it’s unlikely that many disenchanted MAGA and independent voters are going to switch to a candidate promoting open borders, abolishing ICE, transgender rights, and blaming Zionists for all our problems. The only issue that should matter in November is resistance to Trump.

Writes Frum:

The Democratic left is making a complicated—and quite ironic—internal gamble here. The faction’s efforts to hijack the larger party to their own unpopular causes will work only if they can convince Democratic voters that Trump is not the most important and urgent problem facing Americans. Yet the only reason Democratic centrists would coalesce around the party’s leftist candidates in November is if they believe in the absolute urgency of stopping Trump.

The tension, in the end, isn’t really “socialism vs. capitalism.” It’s between a movement pointing backward at policies most of the world’s actual social democracies have been quietly abandoning, and the voters it’s currently winning over for reasons that have very little to do with the platform’s more radical planks. The risk is that the far-far-left could win some primary battles but lose the war for control of Congress.

Leave a comment

Share

The infatuation with democratic socialism that has become salient among many Democrats in the past year could have serious repercussions in both the short term and, if it persists, the long term. In the immediate future, the success of several socialist or socialist-adjacent candidates in this year’s primaries could squander the opportunity to elect congressional majorities for the Democrats to impose some limitations on Donald Trump. And I’m mystified by this turn toward a socialist agenda. My sense is that most voters don’t know what that really entails, and their Republican opponents will use their money advantage to frame all Democrats in competitive races with the socialist moniker.

Many Progressives view socialism as a benign form of the welfare state, with Sweden as their mental model. If you’re thinking, “Yeah, that’s about right,” my objective, then, is to explain or refresh your understanding of this economic model and, along the way, point out, if you hadn’t been keeping track, that Sweden, the poster-society for the social welfare state, has largely abandoned socialism as a failed model for sustaining economic growth and creating the platform for broadbased oportunity.

• Share

  Leave a comment

Since moving to Portland, Maine, I’ve noticed so many people covered in tattoos. In the past, a man or woman may have had one or two tats, maybe a rose on the ankle, the name of a girlfriend, or “Mom” on the arm. The many sightings today may be in part because it’s summer, so more skin is exposed.

But it seemed to go beyond that, compared to what I saw in Cambridge, Boston, or Philadelphia. Am I just more aware, or is there a documented trend of more ink? If so, why, and what, if any, does this say about society? The Pancake wanted to investigate.

So that we’re all on the same page, the term "tattoo" refers to a permanent design or pattern created by injecting ink into the skin using a needle. Tattoos can be found on any part of the body, each with its symbolic meaning. For example, a forearm tattoo may represent strength, while a thigh tattoo can symbolize femininity. In some cultures, tattoo placement also carries spiritual meaning, such as among the Maori people of New Zealand, who use them to represent tribal heritage and social status.

And speaking of the South Pacific, the English word “tattoo” for body art comes from the Polynesian word tatau (meaning “to strike” or "to tap” introduced to Europe by Captain James Cook in 1769.1

A brief history of ink

Tattooing is ancient, and it has appeared independently on nearly every continent. Charles Darwin observed that there was no country in the world that didn’t practice tattooing or some other form of permanent body decoration. My deep dive discovered that the oldest confirmed evidence is Ötzi the Iceman, a Neolithic mummy found in the Alps and dated to roughly 3300 BCE, whose body carried dozens of simple tattoos that may have been a form of ancient acupuncture. Ancient Egyptian mummies show evidence of early tattooing, with a pattern suggesting it may have largely been a female practice, thought to relate to fertility and protection during childbirth.

The meaning of tattoos has swung wildly across cultures and eras. The ancient Greeks used tattoos to communicate among spies as early as the 5th century. Romans later marked criminals and slaves with tattoos. Fun fact: the word “stigma,” now meaning a mark of social disgrace, comes from the Latin word for a puncture mark made by tattooing. Paradoxically, tattoos are now part of prison culture.

Polynesian traditions developed tattoos as sophisticated visual languages encoding ancestry, rank, and spiritual status, and New Zealand’s Māori ta moko tradition used deep facial tattooing to record genealogy.

Thus, tattoos seem to have carried opposite meanings—punishment or ownership in one tradition, honor and status in another—long before modern Western culture adopted them as personal self-expression.

Why people use their bodies as a canvas

As you might expect, academic literature explores a range of hypotheses explaining the affinity for body art. Several motivations are at work.

• The one I find most compelling is the drive to seek identity and uniqueness. One well-cited study frames tattooing through the Theory of Uniqueness, which holds that people seek uniqueness on the attributes most central to their self-concept, and tattoos function as a visible expression of that need for identity and distinctiveness.2

• People often use tattoos to commemorate loss, survival, or major life passages. This pattern shows up repeatedly in trauma and identity-formation research, where tattoos are examined as tools people use to cope with trauma, resist stigma (using the modern meaning), and reconstruct identity. A Pew Research Center survey found that this was by far the most common reason people give for getting tattooed, cited by 69% of tattooed adults

• There’s a documented “collector” pattern where one tattoo leads to another. The planning and anticipation phase itself stimulates dopamine, and for some people the pull isn’t really about any single tattoo but the ongoing cycle of anticipation, completion, and renewed desire. Of course, the initial motivation may still have ben one or both of those identified above.

I’ll get to the pain factor shortly. However, alongside the pain involved, evolutionary anthropology research frames motivation around what’s called costly signaling theory.

The core idea holds that a signal is trustworthy precisely because tattoos and piercings injure the skin and carry real infection risk. Thus, they’re inherently costly and hard to fake, which is exactly what makes them useful as honest signals of fitness or affiliation. A costly signal separates the genuinely strong or fit from the merely posturing. Thus, it is not the tattoo itself, which could, after all, be a temporary transfer, but what it signals: the pain the person had to endure to get the ink. Researchers point to Polynesian tattooing traditions as the clearest historical case, with extreme pain, real risk of death by infection, and healing that sometimes took over a year, while those who couldn’t endure it lived in social shame. Ouch!

Overall, a review of the clinical literature found tattoos are associated with both positive effects—increased self-worth and a sense of empowerment—and negative ones, particularly regret when a design feels impulsive or no longer represents who the person is. (See Platner, Graham)

Who’s getting inked

Almost a third of U.S. adults have at least one tattoo. A 2023 Pew Research Center survey of more than 8,400 adults found that over a fifth have multiple tattoos.

For younger adults, the proportions are higher: 41% of those under 30 and 46% of those 30–49 have at least one tattoo. Not a surprise: only 13% of those 65 and older are inked. I’m in the other 87%.

And my informal observation that I see more body art work more often is borne out by the Pew data. A 2010 Pew survey found only 23% of adults had a tattoo. Other trackers estimate the “most rapid period of normalization” occurred between about 2012 and 2015, when U.S. prevalence reportedly jumped from roughly 21% to 29% in just three years. Attitudes have shifted even faster than the ink itself: eight in ten U.S. adults say society has become more accepting of tattoos over the past 20 years, versus just 7% who say it’s become less accepting.

Men vs. women

Historically, more men than women sported tattoos—you know, sailors, truck drivers. In a striking recent reversal, far more women (38%) now have at least one tattoo than men (27%). Race and ethnicity also show meaningful gaps: 39% of Black Americans have a tattoo, versus 35% of Hispanic Americans, 32% of white Americans, and 14% of Asian Americans.

Geography: within the U.S. and around the world

Within the U.S., Pew didn’t find much of a rural/suburban/urban split, nor a significant partisan gap. Internationally, the most commonly cited comparative study put Italy at the top with roughly 48% of the population tattooed, followed closely by Sweden and the United States.3 That same body of research found New Zealand has the highest share of tattooed men in the world at about 29%, while Australia leads for tattooed women at about 31%. On the low end, majority-Muslim countries show the lowest tattoo prevalence worldwide, largely reflecting religious prohibitions that classify permanent body modification as forbidden. Japan and Iran stand out as places where tattoos still carry significant social stigma despite deep historical tattoo traditions (Japan’s is arguably the most refined tattoo art tradition on earth, yet still associated there with organized crime).

Pain

Modern tattooing utilizes a motorized machine that drives a cluster of tiny needles up and down 50 to 3,000 times a minute. Surface tension and capillary action pull ink down the needle grouping, and each puncture deposits pigment into the second layer of skin, the dermis. Yeah, doesn’t sound like my idea of a fun afternoon. A small tattoo at a local shop might take 1.5 to 2 hours. Here’s a link to a YouTube video that explains how this all works.

Where the ink goes affects the pain involved. It’s generally agreed that areas with more fat, muscle, and fewer nerve endings hurt the least, and areas over bone or with thin skin hurt the most. Thus, the outer upper arm and shoulder — the classic first-tattoo spot — typically rate only 2–3 out of 10 on the pain scale. Ankles and shins consistently rank among the most painful spots because thin skin sits right over bone. Ribs, armpits, hands, feet, elbows, knees, and the spine round out the high end. There are some sex differences too — pain in the breast and stomach area tends to be more intense for women, while back, hip, and groin tattoos tend to rank higher for men. Complexity matters as well. So heavily filled pieces with thick, saturated lines hurt more than delicate line work, simply because they require more needlework in the same area.

What’s the cost?

Naturally, size, color, complexity, and an artist’s reputation all move the price. Rates for an entry-level artist run about $100–$150/hour, mid-level experienced artists $150–$250/hour, and senior/in-demand artists with a following can charge $250–$700+/hour. Color generally costs more than black and grey, since mixing and applying multiple ink shades properly takes more time and skill. Although larger areas often take more time, a small but extremely detailed tattoo can cost more than a larger, simpler one.

I priced a 3-inch color mermaid using the Tattoo Build website. That’s solidly in small tattoo territory, with shading and multiple colors adding time. It would typically take 90 minutes to two hours, thus $225–$500, though with a well-known artist that could jump to $400–$700 or more.

At the other extreme, you’ve likely spotted one or more full sleeves (I’ve picked up the jargon), where the entire arm is covered with ink. That can cost $2000 to $7000, though a top artist with complex art could hit $12,000. It takes five or six sessions of four to seven hours each, and three to six months for healing and recovery.

Permanence?

Removing tattoos is harder and less complete than most people expect, though modern lasers have improved. They break the tattoo pigment into tiny fragments, which the body then clears through the lymphatic system. It may take many sessions, spaced a month or two apart to let the skin heal, and the result may be more of a fade than a full erase. It’s also not cheap, generally $1500 and more, depending on complexity and size. Bottom line: think of the tattoo as forever before committing.

There’s a thriving subculture in ink world

To start, the best-known (though not to me) mainstream example of artistic competition is the TV show Ink Master television series, which has produced recognizable names (again, not to me) in the field, like Ryan Ashley Malarkey, who turned the title into studios, a jewelry line, and over a million Instagram followers. Beyond TV, there’s a whole convention circuit. Events like the Gods of Ink Tattoo Convention in Frankfurt draw thousands of attendees and more than 400 top artists for live tattooing and juried competitions. The 7th Annual New England Tattoo Expo will be at Mohegan Sun in Connecticut next April,4 headlining Ink Master winners and other award-recognized artists. Award categories at these shows get quite specific — best portrait, best color, best black-and-grey, best cover-up, and so on — and doing well can be a legitimate career lever, drawing new clients and raising an artist’s hourly rate.

Social benefits — is there real evidence?

There isn’t strong evidence that tattoo culture offers much in the way of Western societal benefits, writ large. However, some data and common sense suggest body art can benefit some individuals. The amateur psychologist in me sees validity in the Theory of Uniqueness, that some people seek uniqueness in the attributes most central to their self-concept.

One frequently cited study recruited subjects who planned to get their first tattoo. They were measured on, among other characteristics, their self-esteem and appearance anxiety. The same characteristics were then measured three weeks after the tattoo. Findings included significantly lower appearance anxiety and significantly higher self-esteem.

However, that boost isn’t universal or permanent: the self-esteem effect appears to fade faster in women than in men, who, in the same study, reported higher social anxiety about their appearance three weeks post-tattoo. Broader review work frames tattoos as serving real psychological functions beyond decoration — aesthetic enhancement, self-expression and identity, social affiliation, bodily autonomy, and psychological adaptation.

Social perception lags the subculture

Even as tattoos have gone mainstream, British researchers found most of us still rate visibly tattooed individuals as more rebellious and less reliable, though this varies substantially by tattoo placement. Face and neck tattoos, for example, read as far more extreme than, say, an ankle piece. Social perception hasn’t fully caught up with self-perception. Research shows tattooed people, and tattooed women in particular, are still perceived somewhat more negatively by others, which creates a gap between how tattoos make people feel and how they’re read by the world. That’s a real tension, not a contradiction.

Personally, I can see many other outlets for expressing uniqueness than inking: temporary tattoos that wash off, henna, a Mohawk haircut, dyeing one's hair green, unusual clothing ensembles (think Senator John Fetterman, Cardi B, Lil Nas X). All these are quite reversible if you change your mind.

Leave a comment

Share

Since moving to Portland, Maine, I’ve noticed so many people covered in tattoos. In the past, a man or woman may have had one or two tats, maybe a rose on the ankle, the name of a girlfriend, or “Mom” on the arm. The many sightings may be in part because it’s summer, so more skin is exposed.

But it seemed to go beyond that, compared to what I saw in Cambridge, Boston, or Philadelphia. Am I just more aware, or is there a documented trend of more ink? If so, why, and what, if any, does this say about society? The Pancake wanted to investigate.

Leave a comment

Share

Subscribe now

Have you ever finished your meal at a self-service eating establishment—it could be McDonald’s, Panera, even Hudson News at an airport—and, standing in front of two receptacles, one marked “Trash,” the other “Recycle,” faced momentary indecision? Your greasy napkin certainly hits the Trash bin; the empty tin can of Coke goes to recycle. What about the three extra napkins you pulled from the dispenser but didn’t use? Trash or perfectly recyclable paper?

My guess is that you wouldn’t dwell on it long. But it is a decision that has been made. I’ve hesitated at that point with the sticky remains of a strawberry yogurt parfait. The plastic container shouts “Recycle.” The messy innards whisper, “This is trash.”

There are two types of costs that must be weighed to justify recycling programs: environmental as well as financial. Policy needs to consider the cost and effort of putting more bins in our driveways, more receptacles on the streets, additional containers at the airport, and more trucks on the road to collect multiple streams of trash, against quantifiable environmental benefits. As is often the case at The Two-Sided Pancake, the answers are not as straightforward as the knee-jerk responses usually are.

Recycling isn’t as modern as you may think

Scrap metal drives and rag-collecting existed for centuries out of economic necessity. The widespread use of curbside and consumer recycling became a cultural phenomenon only in the late 1980s and early 1990s. When I was growing up in the 1950s and 1960s, there was just one category: trash.

No. Actually, there was more: garbage. It went into a smaller trash can, and once or twice a week farmers would come into my Philadelphia neighborhood, drive down the alleys between rows of homes, and take away our pails with apple peels and watermelon rinds to feed pigs on their farms.

And there was some recycling, though we didn’t refer to it that way. The milkman (always a man) would leave a few quarts and pick up the rinsed-out glass bottles he left at the previous delivery. At our house there was also the soda man, who left off a few bottles of Frank’s soda and took away the empty returns.

In any event, the actual catalyst that accelerated the recycling movement we know today came in 1987, when a garbage-laden barge named the Mobro 4000 spent months floating up and down the U.S. Atlantic coast, continuously turned away by cities fearing it contained toxic waste. It became a massive media spectacle, triggering widespread panic that America was running out of landfill space.

In reality, we had plenty of space for landfills, just not always in places convenient to where we were producing the most trash.1

My initial questioning of the value of Big Recycling dates back to the 1980s. Procter & Gamble, the makers of Huggies, commissioned Arthur D. Little, Inc., a top consulting firm that played an oversized role in developing the innards of the internet, to examine energy, water, wastewater, atmospheric, and solid-waste impacts to address the high-visibility environmental debate of the day: cloth diapers or disposable diapers.

The findings, in short, found that laundering a cloth diaper over its lifetime consumes up to six times the water used to manufacture a disposable. The study also found home laundering produced nearly ten times the water pollution of disposable manufacturing. In contrast, the manufacturing of the disposable diapers and their dumping in landfills created a different type of environmental impact. Thus, neither disposable nor cloth diapers are better or worse for the environment; they just affect it in different ways. A later 2005 UK Environment Agency study reached the same “no substantial difference” conclusion.

Today, we can explore similar dichotomies. Do municipal recycling programs for single-use plastics consume more energy and financial capital than they save? If so, modern sanitary landfills may be a more pragmatic environmental choice. What about recycling paper products? Do the deinking steps for newspapers, books, and magazines balance the savings on virgin pulp? How about aluminum?

The Objective: Why We Started Doing It

The overarching goal of recycling programs was to transition from an economy where we put into trash anything we used and no longer needed to a circular system that lowered the externalized costs of mass consumerism. Advocates had differing objectives that included:

• Conserving landfill space by slowing the rate at which cities bury their garbage.

• Replacing virgin materials by creating alternative production chains that reduced the need for logging, oil drilling, and mining from the rates that would have been required without recycled materials.

• As manufacturing goods from raw earth requires vastly more heat, electricity, and fossil fuels than melting down existing materials, we could reduce energy needs and therefore carbon emissions.

• And, thus mitigate extraction pollution from toxic runoff and heavy emissions generated when mining bauxite ore, refining petroleum, or manufacturing raw paper pulp.

Some recycling makes more sense than others

The true societal and economic value of consumer recycling depends heavily on the specific material being processed. When weighing the energy used to collect, transport, wash, and sort commodities against the cost of making new ones, recycling is highly beneficial for metals and paper, but often a net negative for plastics and glass. The Environmental Protection Agency estimates recycling aluminum saves about 95% of the energy of virgin production, paper saves about 60%, and glass saves about 30%. And then there’s plastics.

Aluminum—the big winner

Aluminum recycling is highly cost-effective because it uses such a small fraction of the energy needed to produce new metal from raw bauxite ore. While extraction requires intensive mining, chemical refining, and heavy electricity for electrolysis, recycling simply melts down existing scrap metal, allowing the material to be reused infinitely without losing quality. Specifically, primary production requires roughly 200–250 MJ/kg of energy, whereas recycling takes only 5–10 MJ/kg.2

Paper recycling and deinking

Here the evidence leans more clearly toward recycling still winning, even after accounting for deinking energy — though the margin is narrower than for aluminum.

Paper recycling saves up to 60% of the energy required to produce virgin paper, with a broader literature range of 28% to 70% depending on grade and process assumptions, translating overall to roughly a 25% cut in lifecycle greenhouse gas emissions per ton.

Deinking is real overhead. Most of the deinked sludge, a wet mixture of detached microscopic ink particles, short unusable cellulose fibers, mineral fillers (like clay and calcium carbonate), and processing chemicals, is sent to landfills or incinerated for energy recovery. Increasingly, mills are diverting it for reuse as agricultural soil amendments, compost mixtures, or raw material for brick and cement manufacturing. However, its cost and environmental impact don’t erase the gain because pulp-making from virgin fiber is more complex than deinking pulp and requires higher energy consumption. Fossil-based energy input is the main contributor to environmental impact in both systems (virgin paper making and recycled).

Note that there’s a structural ceiling on how much paper recycling can ever displace virgin pulp, independent of economics. Technical limits mean only 70% of fibrous inputs can ever be supplied by recovered paper so some fraction must always come from virgin wood.

On the one hand, recycled paper costs more upfront due to deinking and related processing. However, it uses roughly less energy and water than virgin paper, saving about 17 trees per ton. Virgin paper, made from fresh wood pulp, is cheaper to produce but has a higher environmental cost. Thus, the economic case for recycling is weaker than the energy/environmental case.

Glass? It cuts both ways

Glass recycling is often not cost-effective for municipalities due to its heavy weight, high transport costs, and tendency to break and contaminate other recyclables like paper. However, it provides massive energy and resource savings for manufacturers who use crushed glass (known as cullet) instead of raw sand.

Crushed glass melts at a lower temperature than raw sand and limestone, the ingredients for glass making, which saves factory energy. Unlike paper, glass can be recycled infinitely without losing any quality or strength. And in localities where there are dedicated drop-off programs (like color-separated bins) reduced contamination makes the process much cheaper.

However, economics weigh heavily. To encourage recycling, most cities have employed single-stream recycling, which means residents can toss out their glass in the same bin as paper, aluminum, and plastic. When glass breaks in a mixed recycling bin, it ruins other materials (like cardboard) and damages sorting machinery. As already noted, the high weight adds to the cost in fuel and labor to haul compared to plastic or aluminum. Therefore, processed recycled glass often sells for far less per ton than the actual cost to collect and sort it.

Plastics: the deal breaker

Mr. McGuire may have touted “plastics” to Benjamin Braddock,3 but for recycling, plastics are the real outlier. A peer-reviewed comparison found an energy cost saving from recycling is proven for many recyclable materials — excluding plastics — and that given current technology, real energy recovery from plastic essentially requires waste-to-energy conversion instead.

The financial picture for recycling plastic changed sharply after 2017. Between 2013 and 2018, China introduced a series of policies, most critically the National Sword, that substantially restricted importation of solid waste, based on contamination. Before these restrictions, the U.S. shipped most of its plastic scrap overseas, particularly to China.

Controlling for oil prices, producer price index, and amount of plastic scrap exported, researchers at the University of Buffalo found that following the changes in China, the amount of plastic recycled in the U.S. significantly declined while the amount landfilled increased by 23.2%. The same study found that with the need to improve their sorting after China’s new import restrictions, the operating costs of many materials recycling centers (MRFs) roughly doubled in some locations.

So, for plastics specifically, the “recycling saves energy and money” case has weakened substantially since 2018, largely for market/geopolitical reasons (loss of the China export outlet) as much as physics. For aluminum, paper, and — more debatably — glass, recycling’s energy case remains intact. It is not accurate to generalize “recycling doesn’t pay” across all recycling; it’s fairly accurate for plastics specifically in the current market.

Summary of the case for recycling, by type of material

The paradox of single stream recycling

Single-stream recycling has inadvertently killed the economics of the system it was meant to expand. In the first iteration of city-wide recycling (after bring-it-yourself to the recycle station failed to make a dent), many cities required some separation. I recall Cambridge mandated separate bins for paper/cardboard and for glass/plastics/aluminum. Wine bottles had to have the foil wrapper on the necks removed. It was a pain. Too many households had no room in their small city apartments or homes for multiple bins, or could not take the time to comply.

To encourage more recycling, in the 1990s, a few innovative cities introduced single stream recycling, shifting the work of sorting from households to new, large central sorting stations. By 2014, 80% of municipalities had adopted single stream. This made sense because it lowered their collection costs while increasing the volume of materials for which they received profitable payment.

But when China halted buying plastics, the economics shifted. Suddenly cities had all this waste that was now a net cost to them. For example, before 2017, Boston earned about $20+ per ton for paper and mixed recyclables, resulting in a small addition to the city treasury. Today, Boston pays $125 to $160 per ton. In Boston, and many other cities, the “tipping fees” for landfills is less, in this case $80 – $95 per ton.

A reversal is underway, with cities from San Francisco and Seattle to Boise and Hoboken starting to recognize that the small towns that had maintained separate bins at their “transfer stations” (i.e., dumps) were more cost efficient, and they are returning to multiple stream requirements.

Recycling now a one-sided benefit

When cities and towns first started picking up recyclables, it was a win-win proposition: the cities received more for the recyclable waste than it cost to collect it. And in removing tons of aluminum, glass, paper, and plastic from trash, the environment benefited because diverting any material saves raw natural resources, prevents landfill build-up, and reduces greenhouse gas emissions from mining virgin bauxite or cutting down forests.

The equation changed after 2017 for two reasons. First, the revenue stream shrank with China refusing most plastic waste. Some of that plastic now ends up in landfills; some is used as fuel in Waste-to-Energy (WTE) plants. Both have an environmental impact. Yes, plastics (including mixed or lower-grade recycled plastics) can be used in Waste-to-Energy as a high-heat fuel source. Nonetheless, it is generally not cost-effective or energy-efficient to use dedicated recycled plastic even this way because cleaning, sorting, and transporting it for combustion costs more than the value of the electricity produced.

The second factor may be termed the aluminum subsidy illusion: For decades, cities relied on aluminum to secretly subsidize the losses incurred by sorting glass and paper. But as plastic packaging largely displaced aluminum and glass in grocery aisles, the ratio shifted. Municipalities are now drowning in low-value plastic and glass, while lacking the high-value aluminum volume needed to balance the books.

(Speaking of landfills as the “pragmatic” fallback: modern, lined, monitored sanitary landfills are indeed a different animal than older dumps. The EPA has noted municipal solid waste landfills cause roughly one additional cancer case every 13 years, and today’s landfills are lined with clay/plastic, have drainage and gas-collection systems, and are monitored daily.)

The Bottom Line

The invention of the single-stream blue bin was marketed as a triumph for consumer adoption. In reality, it created a cross-contamination nightmare. Broken glass downgrades the value of paper, while grease and liquids destroy plastic batches. By making recycling “easier” for humans, we made the physical output economically worthless.

If we sorted only aluminum and paper, recycling would be an unambiguous home run both financially and environmentally. Today, curbside municipal recycling remains a heavily taxpayer-subsidized operation rather than a self-sustaining business. It is universally agreed that reducing consumption and reusing items provides a vastly higher net benefit to society than the actual act of recycling. But that remains a tough sale on a mass basis.

Leave a comment

Share The Two-Sided Pancake

There are two types of costs that must be weighed to justify recycling programs: environmental as well as financial. Policy needs to consider the cost and effort of putting more bins in our driveways, more receptacles on the streets, additional containers at the airport, and more trucks on the road to collect multiple streams of trash, against quantifiable environmental benefits. As is often the case at The Two-Sided Pancake, the answers are not as straightforward as the knee-jerk responses usually are.

I’ve mentioned here that I’ve recently relocated to Portland, Maine. Lots of stories about the move itself, but not for this venue. What I do want to share is the challenge of finding new health care providers, starting with primary care. It’s been time-consuming and frustrating. It shouldn’t be so hard.

An early warning came a few years ago, when my long-time primary care physician (PCP) in Cambridge retired and his practice shut down. I found a new PCP, but he moved on after just one year. The practice found me a replacement, who was great, but now I had to start over.

First, often getting through to anyone is a challenge. You’ve had the experience: press 1 for this, 2 for that. Often the press leads to more automated voices and finally, “Leave a message.” One large practice quickly cut to the quick: they did not accept Medicare. Others had filled waiting lists. I even tried a few “concierge” practices, with no success. I was open to any MD or DO available.

How did primary care become such a bottleneck? Contrary to some popular myths, as I’ve learned, the cost of medical school and student debt is not the major problem. What is being done and can be done to expand primary care?

GIGO (garbage in, garbage out) meant bad policy

The single biggest reason there are too few doctors relative to population today traces back to a BIG policy mistake made more than 45 years ago. In 1980, the Graduate Medical Education National Advisory Committee (GMENAC), a federal advisory panel, issued what, in retrospect, was a poorly designed study that infamously predicted the U.S. was headed for a physician surplus, and in response medical schools imposed a voluntary enrollment moratorium that lasted until 2005.

Over those 25 years, the U.S. population grew by 70 million people, while the number of MD graduates from U.S. medical schools stayed essentially flat at about 16,000 per year. This wasn't for lack of qualified applicants; the number of applicants for those 16,000 spots ranged from roughly 27,000 to 47,000 over the period, meaning thousands of qualified Americans were turned away every year while schools held enrollment steady. The resulting gap was quietly filled by international medical graduates, whose share of the U.S. physician workforce hit 25% by 2006.

By 2005, the Association of American Medical Colleges (AAMC) reversed course, admitting the surplus prediction was wrong and began pushing for enrollment growth. Total U.S. medical school enrollment has now reached nearly 100,000 students, an all-time high, and there are 155 accredited MD-granting and 37 DO-granting schools today, up substantially from a few decades ago.

However, that expansion only really got underway in the last 15–20 years, so the country is still working off a 25-year hole in the supply pipeline, compounded by the fact that residency slots (largely capped by federal Medicare funding since 1997) haven't grown proportionally. Therefore, more graduates doesn't automatically mean more practicing doctors.

More reasons why it’s so hard to find a PCP in most cities

It gets worse. Interest in primary care among medical graduates fell from 36% in 1982 to 22.5% by 1989 — a decline of nearly 38% in just seven years, and it has never really recovered. About 30% of medical graduates now pursue primary care specialties, a share that’s been declining for years across countries, not just the U.S.

But even that number overstates actual primary care output, because a lot of people who match into “primary care” residencies don’t end up practicing primary care. A widely cited study nicknamed “the Dean’s Lie” found that of 52,000+ graduates from 2013–2015, 50.5% entered a primary care residency, but five to seven years later only 27.2% were actually practicing primary care. While nearly 92% of family medicine graduates stayed in primary care, only about 55% of pediatrics graduates and under a third of internal medicine specialists did, with the rest subspecializing into cardiology, GI, pulmonology, etc. Internal medicine residency is the biggest “leak.” Historically, only about 25% of internal medicine graduates stay in general primary care practice, down from 54% a decade earlier.

One key reason shouldn’t be surprising: the pay gap with other specialties. In 2025, average compensation for primary care physicians was $298,000 (not too shabby), versus $417,000 average for specialists. General surgeons specifically earned more than $442,000, and cardiologists averaged $575,000. Orthopedics and orthopedic surgery topped the profession, at $611,000 annually.

In large measure, this is baked into the system by how Medicare, and hence private insurers, price physician labor. Since 1992, Medicare has paid physicians through a “resource-based relative value scale” that was supposed to give equal pay for equal time/effort/complexity. In practice, it has grossly undervalued primary care since its inception. The Centers for Medicare & Medicaid Services (the CMS (why isn’t it known as the CMMS?), which sets the relative values, is dominated by specialty societies, and a longstanding criticism is that the fee schedule rewards procedures and technology-heavy services far more generously than office visits and other “cognitive” evaluation-and-management services that primary care mostly bills.

CMS itself has now acknowledged this is a problem, with the 2026 Medicare fee schedule explicitly calling out “long-standing overvaluation of certain procedures and undervaluation of time-intensive services like primary care.” However, the correction is only just beginning.

How much did the ACA contribute to the problem?

The Affordable Care Act (ACA) was signed into law in 2010, and its major provisions took effect in 2014. Within five years, there were 20 million new previously uninsured adults, most of whom could now seek out a primary care physician, adding to demand while the supply of primary care doctors was just starting to nudge up from the old moratorium. ACA expansion led to a 29% increase in Medicaid visits and Medicaid patients among primary care physicians. One national study found expansion states saw significant increases in patients delaying care due to unavailable appointments or long wait times, likely reflecting added strain on providers.

But other, more granular studies found the system absorbed the shock reasonably well, in part by shifting a growing share of appointments to nurse practitioners and physician assistants. Overall, the ACA added real demand, but it looks more like it strained an already-thin supply rather than being the primary cause of the shortage

For seniors with Medicare, an added layer of difficulty. A detour into Medicare Advantage

I mentioned in passing at the start that one large practice of primary care physicians here in Portland does not accept seniors with traditional Medicare. They did accept Medicare Advantage (MA) subscribers. What’s going on with this?

Slightly over half of all Medicare beneficiaries participate in Medicare Advantage. This is the private-insurance alternative to traditional Medicare. Instead of the government paying doctors and hospitals directly for each service (fee-for-service), the government pays a private insurer — UnitedHealthcare, Humana, Aetna, etc. — a fixed monthly amount per enrollee, and that insurer takes on the job of covering all the Part A and Part B services, often bundling in Part D drug coverage, and more, as well.

These insurers do not receive the Medicare Part B premium. Rather, that premium paid by MA enrollees goes to the Medicare Trust Fund, same as always. What the insurer receives instead is a separate, negotiated per-member payment from CMS. I’m not going to take you down the rabbit hole of the funding and payment mechanism of MA. You can read about it here. As best I can tell, most medical practices dislike dealing with the MA insurers: hospital systems in Georgia, medical groups in San Diego, groups in Louisville, and nearly a third of Nebraska hospitals have all refused to accept certain MA plans, citing low payment rates, onerous prior-authorization requirements, and too many after-the-fact claim denials. MA plans pay physicians roughly 10–15% less than traditional Medicare, take about twice as long to reimburse, and layer on more prior authorization and denials on top.

HOWEVER…there are specific situations where a practice may take on MA patients while being effectively closed to new traditional Medicare patients.

• A practice can simply decide it’s at capacity and stop accepting new traditional Medicare patients, while continuing to honor an existing multi-year network contract with an MA insurer. That’s an administrative/capacity choice, not necessarily a statement about which pays better.

• Practices that are paid a flat monthly amount per MA patient regardless of visit volume that can manage a population efficiently find that the revenue model can beat fee-for-service traditional Medicare. These practices are often structurally built around MA and Medicaid managed care and don’t do traditional Medicare FFS billing at all.

• A growing number of MA plans are co-branded with the hospital system that employs the physicians. In that case, taking the system’s own MA plan keeps referrals and admissions inside the system — a strategic reason with nothing to do with the per-visit rate.

• Finally, some MA plans do succeed in negotiating rates for those services where traditional Medicare tends to overpay, like labs and imaging. A practice with significant ancillary revenue might do fine on an MA contract even if the physician-visit rate itself isn’t better.

Is med school debt a major factor pushing away from primary care?

In 2024 medical school graduates left school with an average of $212,341 in educational debt, with 56% of indebted graduates owing at least $200,000. However, that should be viewed in the context of the earnings I described above. Even for a PCP earning $300,000, a $200,000 debt load should be manageable to pay off in five or six years with modest sacrifice, and even less onerous against lifetime earnings that can run into the millions.

Counter-intuitively, a comprehensive review of the academic literature has found little to no connection between specialty choice and economic factors such as debt and income potential, and this “powerful but unfounded claim [that earnings push doctors into high-paid specialties] persists in academic medicine circles.” True, individual studies are genuinely mixed. One retrospective study did find high educational debt deters graduates of public medical schools from choosing primary care, but does not appear to influence private school graduates in the same way. And students from lower-income families are more strongly influenced by debt. But that’s a narrow, income-contingent effect, not a broad “debt pushes everyone toward specialization” story. Multiple other studies found essentially no relationship at all.

Is free med school tuition the answer?

In a word, no. With a $100 million donation in hand, New York University’s Grossman School of Medicine made tuition free for its entire main campus in 2018, no strings attached, hoping it would nudge students toward primary care. The result: the percentage of NYU medical students who went into primary care was about the same in 2017 and 2024. NYU’s Dean of Admissions downplayed this: “We got a lot of press for the hope that this will impact primary care, and that’s great, but it’s not our primary mission here... it allows students to pursue career choices based on their passion.”

However, NYU also runs a second, small, tuition-free medical school on Long Island that recruits and admits students for primary care rather than just subsidizing whoever gets in. The result: 67% of its graduating class entered primary care residencies, a sharp contrast with the main campus. What’s different is the program’s design, with targeted admissions coupled with a targeted mission, yielding a wildly different outcome. That’s a clear real-world confirmation that undifferentiated debt relief doesn’t move the needle, but debt relief bundled with a primary-care commitment and selection process does.

The National Health Service Corps funds tuition, fees, and a living stipend in exchange for a commitment to work one year for each year of funding received, with a minimum of two years at an approved site in a medically underserved community. It currently supports over 18,000 active clinicians and is widely regarded as one of the more effective tools for staffing underserved primary care. It helps confirm the proposition that philanthropic dollars modeled on debt avoidance or forgiveness contingent on years of primary-care service are more effective than blanket free tuition for expanding primary care physicians.

There is more needed than just medical school: primary care residencies

One more place philanthropy might do more good than funding tuition: residency slots. A less-discussed bottleneck is that Medicare, pays for the large majority of residency training. It has capped the number of funded residency positions per hospital since the Balanced Budget Act of 1997, fixing the geographical distribution of residency funding across the country even as the population has shifted dramatically since then.

Although the U.S. has been expanding medical school class sizes for two decades, residency positions have stagnated. Thus, more debt-free graduates doesn’t help if there’s nowhere for them to train. Philanthropic money underwriting new primary-care residency positions at teaching hospitals, which often train more residents than Medicare reimburses them for and therefore face real financial barriers to expansion, could plausibly move the physician pipeline more than another round of “make med school free.”

I haven’t addressed here the growing availability of nurse practitioners as well as physician assistants that are filling some of the void of M.D.s and D.O.s. That’s important, but not a full solution. A workable and effective path for improving primary care availability that has received substantial confirmation is policies that condition philanthropic endowments and scholarship aid on commitments to primary care by students and residency programs.

That said, I doubt any of this will help me find a PCP in Portland today.

Leave a comment

Share

How did primary care become such a bottleneck? Contrary to some popular myths, as I’ve learned, the cost of medical school and student debt is not the major problem. What is being done and can be done to expand primary care?

You’re likely aware of the high-profile issue around the placement and building of large data centers to house the equipment that comprises “compute,” shorthand for computational power, specifically, the processing capacity (mostly GPU/AI accelerator chips, networking, and memory) needed to train and run AI models. Data centers are the physical buildings; compute is the resource inside them. Think of it like the difference between a power plant (the building) and electricity (what it produces).

Construction of large data centers has become controversial: the large tracts of land needed, the energy consumed, the water required in exchange for tens of thousands of total construction jobs and thousands of permanent employees. Besides the very real impact these have on people and the environment, there are also real national security concerns. Unfortunately, the Trump administration cites national security as the rationale for all sorts of tangential—at best—executive orders it wants to enforce, from immigration to pausing offshore wind farms to building a White House ballroom. Nonetheless, the development of AI does have very real national security implications.

Will these data centers help make compute less expensive? Might data centers make higher-power LLMs more available to more people? What are the larger political and economic implications if a handful of large tech companies control the data centers?

I’m not going to address all this here. Rather, I want to hone in on the data center construction piece.

Massive and growing

Of the 10 largest data center operators, the five at the top are based in the U.S.: Amazon, Microsoft, Google, Meta and Apple. Three of the second five are in China: Alibaba, Huawei, and Tencent. IBM and Oracle complete the list. All were running data centers for cloud services before AI became a driving force for expansion.

There are already about 12,000 live data centers globally, many a legacy of the cloud computing trend over the past two decades. AI data centers are essentially cloud computing facilities on steroids. With 4,532 of these1, the US alone has almost eight times more data centers than any other country.

The scale of investment is beyond impressive. Developers are planning to spend nearly $700 billion on data center projects in 2026 alone. That amount is staggering, equal to almost the entire U.S. defense budget in 2020. By 2030, total global data center investment is projected to reach $6.7 trillion.2

There are around 500 data centers in Northern Virginia, more than any other state. Other areas that already have major data centers include Atlanta, which accounts for 16% of data center inventory nationwide, and the Dallas-Fort Worth area, which houses about 11% of the total US data center market. Chicago accounts for 9%.. What makes these localities so attractive? One reason is the location’s interconnectivity, particularly its superior fiber network that enables low latency. Virginia has many critical submarine cables that land on its coast. It’s also close to several major cities, like New York, Philadelphia, and Washington, D.C.

Most of these data centers were built or started construction while the issue was still under the radar. As AI in general has risen higher in the stack of public policy issues, and the scale of the cumulative impact of all the proposed data centers has become recognized, the pushback has become more intense.

Although existing data centers are mostly in urban areas, two-thirds of the 1500 or so centers now under construction or proposed are situated in rural counties, where presumably land is cheaper, though infrastructure (e.g., electrical lines) will have to be added.

The arguments for supporting new data centers

The case supporting the developers of just about any new data center revolves around the same key benefits:

• Tax revenue is often at the top of the list and is not without merit.. Nearly every proposal leads with projected tax revenue for the host jurisdiction. In California’s Imperial County, a $10 billion, 950,000 sq ft AI campus was projected to boost county revenue by $28 million annually, a sum equal to 3.7% of the current year’s projected revenue. In Jay, Maine, where a mill that closed in 2020 had accounted for 22% of the town’s tax base, the proposed data center would make a substantial property tax contribution.

• Jobs, both construction and permanent, are prized and real. Every case cites job numbers. Imperial County developers predicted 1,600 construction jobs; Jay, Maine expected 800 construction plus 100 permanent high-paying jobs. It’s the same narrative for Howell Township, Michigan, and Franklin Township, Indiana. In each case, there would be hundreds of temporary construction jobs, followed by permanent operational jobs that would be fewer but high-paying.

• There is a FOMO undercurrent. A data center may fill an economic gap that few other ventures would offer. And if our town or county turns down the developer, another region will capture the boom. This shows up in two flavors: distressed places betting on data centers as the only available large-scale investment (such as Jay, Maine’s shuttered mill and Imperial Valley’s poverty rate), and growing places betting on data centers to keep pace with a fast-moving industry before competing states or counties capture it.

• Developers consistently argue the law on land use is already on their side. In reality, existing zoning and legal compliance is often the hurdle. The developer in Imperial County claims a ministerial permit, meaning the land is already zoned and permitted for industrial development. In Lowndes County, Georgia, the property owner sought straightforward rezoning approval for residential land to light manufacturing, and in Wake County, North Carolina, the developer relied on existing agricultural-to-industrial pathways. In every case, they were either denied ot folded their tent when faced with delaying tactics and denials.

• In many cases, local elected officials as well as business leaders are encouraging of the project. Almost all can point to some local institutional backing — a chamber of commerce, county commissioners, a planning vote that initially passed — even when public sentiment at large is hostile. This creates the recurring split where officials and civic organizations support a project that a majority of residents attending hearings oppose.

The case against is gaining momentum

As the number of data centers increased and the resources they required started adding up, multiple constituencies arose—some of the usual suspects, others truly grassroots. “In the first three months of 2026, local opposition blocked or delayed at least 75 data center projects worth roughly $130 billion,” according to Data Center Watch.

• Among the latter are neighborhood groups and some municipal leaders in states like Texas, Maryland, Virginia, and New York who have successfully pushed for local zoning restrictions, lawsuits, and state-level construction moratoriums. They show up at county commissioner and zoning board meetings and make their displeasure known loudly.

• So-called grassroots coalitions include environmental advocacy groups. Food & Water Watch has organized the Stop Data Centers Coalition, which includes Our Revolution, a progressive national organization spearheading a coordinated pushback against Big Tech; among the 500 organizations that have signed on include Greenpeace, Friends of the Earth, Americans for Financial Reform, Popular Democracy, People’s Action Institute, Good Jobs First, Honor the Earth, . In the “strange bedfellows” department, the Trump-aligned Humans First3 (the footnote is worth reading) is also active in opposition. Who said bipartisanship is dead?

• Federal and state lawmakers have become involved. Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez introduced federal legislation, the “AI Data Center Moratorium Act,” to halt large new projects, while Representative Rashida Tlaib targeted projects on federal lands. Just this month, New York Governor Kathy Hochul signed a bill enforcing a one-year moratorium on data center construction in the state.

NIMBY arguments vs. those unique to data centers

The pushback incorporates classic NIMBY arguments that are cited for many industrial projects, such as Amazon warehouses:

• Rezoning fights over converting agricultural/residential land to industrial use

• Concerns about noise, truck traffic during construction, and visual impact of large windowless buildings

• Worry about property values

• Suspicion of secretive developers who won’t disclose basic project details (this shows up in the Valdosta, GA and Menomonie, Wisconsin cases alike, and is a pattern with any unpopular land use, not unique to data centers)

• The scale objection (”1,000 acres is too big for our township”) is the same objection any oversized single land use provokes, whether it’s a landfill, a prison, or a warehouse district.

However, the data centers debate has its own distinctive issues:

• For one, cooling large AI clusters, especially with liquid cooling, consumes water in ways that register differently than a typical factory, especially in drought-affected regions like Central Texas.

• Another is the strain on the electric grid and the impact that hits everyone’s utility bill. This shifts the constituency of opposition beyond the people who live nearby. Polls show 71% of the public opposes data centers being built in their area, a strikingly high number. This was specifically the case in San Marcos, Texas.

• Although there are thousands of centers operational and under construction, the speed and scale have gotten the attention of the grassroots and hence the politicians. The Data Center Watch group claims active opposition groups grew from 396 to 833 in the first three months of 2026, spanning 49 states. This is a much faster-moving, more geographically dispersed backlash than typical local land-use fights, because the same handful of hyperscalers are hitting dozens of communities simultaneously with similarly sized proposals.

• Because of the ferocity of opposition, some developers now conduct extensive due diligence, land assembly, and shell-company registration before disclosing the project publicly—this is a documented consent-management strategy that’s less common with, say, a new subdivision. Thus secrecy has become a structural feature, not an incidental one. The quieter initial process has thus become a flashpoint in many instances.

• A newer and more troubling development is a small but growing turn to violence in the opposition. Someone fired 13 rounds into an Indianapolis council member’s home after he supported a data center project, and a Michigan township treasurer resigned after receiving death threats over the project. That level of intensity is unusual even for contentious zoning fights.

But data center development pushes ahead

Despite the hurdles and pushback, development continues.

In Effingham County, Georgia, local county commissioners, school district officials, and the Industrial Development Authority jointly approved plans for a massive $20 billion OpenAI data center campus, spanning 1,400 acres. In Columbia County, Georgia, the local Board of Commissioners and Economic Development Authority officially announced that Google will be the sole operator of a newly approved data center facility.

The Columbus, Georgia city council recently approved general data center zoning rules to pave the way for project-specific reviews of a planned 600-megawatt facility, though further review is required. In Texas, some proposals have been delayed or paused this year, but in mid-May, the Red Oak City Council officially approved a massive 800-acre data center campus, despite local citizen protests regarding potential property value decreases, noise, and environmental impacts.

(NB: Hours after this column was published, The New York Times ran this story: “How Meta Got Everything It Wanted in a Secret Louisiana Data Center Deal.” It describes how Meta used private talks with local officials to start a project that covers nearly six square miles.)

Data center developers have also responded positively to the energy complaint. In their own self-interest, they are voracious promoters and buyers of clean energy and are thus instrumental in pushing for clean energy development. One high-profile deal was Microsoft’s agreement with the owner of the Three Mile Island nuclear power plant to restart a reactor that had been shut down for economic reasons. It will add 835 megawatts of power to the Pennsylvania grid.4

The International Energy Agency estimates that “over the next five years, renewables meet nearly half of the additional demand, followed by natural gas and coal, with nuclear starting to play an increasingly important role towards the end of this decade and beyond.”

A real values disagreement, not a question with a fact-checkable answer

I only touched on national security in this debate in my introduction, but I put down a marker in that realm. Although the overwhelming majority of the specific objections raised — water use, grid strain, secrecy, rezoning process, scale mismatch with the community — would apply almost identically to a semiconductor fabrication plant, a bitcoin mining operation, or a conventional warehouse campus of the same footprint, I detect a distinct current about AI specifically. It shows up in two forms.

First, as an ideological opposition to AI’s growth. This is a minority position but a real one. Some activists oppose data centers because they oppose AI's trajectory (job displacement, safety, concentration of power), and the buildings are simply the most tangible, locally fightable node of that larger objection. The violence — the shooting at an Indianapolis council member's home, the attack on Sam Altman's residence — suggests that for a small fringe, this has become a proxy fight about AI itself, not zoning.

The second form is anxiety about who benefits. Several cases of opposition to individual proposals show a specific worry that isn't really about the technology but about extraction: that a small number of already-dominant companies will capture the gains while communities absorb the water, power, and noise costs. That's less "I oppose AI" and more "I don't trust this industry's promises," which is a rational response to the secrecy and contract-reversal patterns.

I’m personally ambivalent. You?

If, like me, you think the AI race with China is close, high-stakes, and genuinely sensitive to a few years’ delay in the U.S. buildout, and that AI will be a net benefit to society, then the urgency argument carries real weight, and streamlining review via federal preemption, faster permitting, and standardized environmental review starts to look attractive even if it means less local control.

However, if you think the U.S.’s capital and hardware advantage is durable regardless of a handful of county fights, that the more important risk is a small number of companies extracting local resources without accountability, and that AI’s spread should be slowed down, then insistence on process looks less like sand in the gears and more like the only leverage communities and activists have.

I suppose we can also plead for some middle ground. Easier said than done.

Leave a comment

Share The Two-Sided Pancake

Construction of large data centers has become controversial: the large tracts of land needed, the energy consumed, the water required in exchange for tens of thousands of total construction jobs and thousands of permanent employees. Besides the very real impact these have on people and the environment, there are also real national security. Unfortunately, the Trump administration cites national security as the rationale for all sorts of tangential—at best—executive orders it wants to enforce, from immigration to pausing offshore wind farms to building a White House ballroom. Nonetheless, the development of AI does have very real national security implications.

“Our country is better than our government.”—William Kristol

Gallup’s most recent country-favorability poll found that Americans’ overall view of Israel has fallen to among the least favorable levels the organization has ever recorded. Conversely, Americans now sympathize more with Palestinians than with Israelis. This is the first time in the that poll’s history, a 10-point swing in a single year.

Pew Research Center’s tracking tells a similar story. Favorable views of the Israeli people among U.S. adults dropped from 67% in 2022 to 52% this year, while unfavorable views nearly doubled, from 25% to 42%. And this slide is not a Democratic phenomenon that Republicans are simply immune to. Pew found that the share of Republicans holding a negative view of Israelis climbed from 17% to 31% over the same period, driven mostly by voters under 50; among Republicans aged 18–49, a majority now view Israel unfavorably. A new AP-NORC poll released this month found that only 20% of U.S. adults hold a favorable view of Netanyahu personally, and that roughly a third of all adults — including about half of Democrats — now believe Israel has committed genocide in Gaza.

Of course, none of this happened in a vacuum. Support for Israel was already softening before October 7, 2023. Gallup’s two-state-solution numbers among Republicans, for instance, had been drifting down for a couple of years. But the war that followed Hamas’s pogrom, executing 1200 Israelis and abducting 250 more on October 7, accelerated the trend dramatically, particularly among younger Americans and political independents.

The war to eliminate Hamas in Gaza was accompanied by further crackdowns on Palestinians on the West Bank, bolder West Bank settlers often making their vigilante battles with seemingly little accountability from the Israeli government.

There is much to criticize of the Netanyahu government. However, there seems to be more underlying much of the anti-Israel and anti-Zionism rhetoric that seems to place them as the number one global pirriah state: worse than Russia, which has been lobbing missiles at civilians in Ukraine for four years, which, unlike Hamas, did not attack Russia and slaughter its civilians? Worse than China’s oppression of the Uyghurs? Worse than the years-long atrocities of Al-Shabaab, with hundreds of thousands killed in Somalia?

The convenient collapse of categories

Here is where I want to draw a distinction that too many people, especially in the American progressive circles I long ago traveled in before wising up, refuse to draw: the difference between despising a government’s policies and believing the country itself is illegitimate.

We do this instinctively with our own country. Plenty of Americans think Donald Trump’s immigration raids, his tariff whiplash, his attacks on universities and the press, his personal corruption—it’s a long list— are genuinely dangerous. Yet many of us, me included, still believe in America, still call ourselves patriots, still think the country is worth fighting for and improving rather than dissolving. No one is seriously suggesting we return the land to those who lived here before Europeans settlers arrived. Nobody accuses today’s sober critics of incoherence. We understand, almost without being taught, that a government and a nation are not the same thing, and that accepting the second doesn’t require excusing the first.

Somehow that basic distinction dissolves the moment the subject is Israel. Netanyahu’s coalition — arguably the most illiberal in Israel’s history — has become, for a large and vocal slice of the American left, simply “Israel,” full stop, and Zionism itself has become a stand-in for something closer to a crime than a movement. The “from the river to the sea” chant that became a fixture of campus protests and city council resolutions is not, whatever its defenders claim, a demand for equal rights within a shared state. It’s the same territorial claim Hamas’s own charter makes: a single Palestinian state where Israel currently exists, with no room for today’s Jewish citizens.1 People are, of course, free to chant it. But they should be honest that they are echoing the platform of an organization whose founding document calls for the eradication of a state its people rely on for their physical survival — not offering a peace plan.

I’m not interested in defending that instinct, and I don’t think we have to look away from Israel’s real problems to reject it.

What deserves real criticism

Let’s be plain about what’s wrong, because a defense of Israel that pretends nothing is wrong isn’t a defense — it’s propaganda, and it will convince no one worth convincing.

The settlement enterprise in the West Bank is a moral and strategic disaster. It has, for decades, entrenched a system in which Palestinians live under military rule with no vote in the government that controls their lives, while nearby settlers enjoy full Israeli citizenship — and settler violence against Palestinian farmers and villages has often gone unpunished, sometimes with the tacit blessing of ministers in the current government. The scale of destruction in Gaza, whatever one concludes about proportionality and intent, has been staggering, and the absence of any coherent postwar plan for who governs and rebuilds Gaza has been, by the admission of Israeli officials themselves, a genuine failure of strategy.2 Netanyahu’s years-long refusal to articulate a “day after,” his cynical history of quietly tolerating Hamas as a foil to Palestinian statehood ambitions, and his government’s assault on judicial independence are not fringe critiques — they are held by large numbers of Israelis themselves, many of whom spent the year before October 7 in the streets protesting exactly that judicial overhaul.

You don’t have to take a progressive American’s word for any of this. Rahm Emanuel, President Obama’s Chief ot Staff, former Mayor of Chicago, Ambassador to Japan, is hardly a stranger to Israel. His father fought in Israel’s war of independence, and his uncle is buried on the Mount of Olives. You may have seen or read his blunt speech at Tel Aviv University earlier this month, in which he told his audiencs, both those in the room as well as thosein the diaspora:

Unconditional support has produced a Prime Minister who has presumed that his strategic interests would incur no cost if he ignored America’s concerns about settlements and sparked a regional war. Unconditional support has allowed you to deny food and medical relief to innocent Palestinians suffering in Gaza, leaving the world to conclude that Israelis not only want to kill the Palestinians but that they are completely indifferent to their death, destruction, and suffering. Unconditional support has girded a political coalition in the Knesset that learned it can burn Palestinian farmland in the West Bank and terrorize Palestinian families without consequence. Given all of this, we need a fundamentally new and different approach to the alliance.

He proposed that Americans stop treating Israel’s defense budget as a permanent subsidy and instead require it to buy U.S. arms under the same commercial terms as any other ally. He even proposed a sanctions regime targeting Israelis who attack Palestinian civilians or property, and the banks and construction companies that finance settlements. This is not the language of an apologist.

But Emanuel drew the line I’m drawing here, too. In the same speech, he called both the “river to the sea” chant and the fantasy of a “Greater Israel” annexing the West Bank equally self-destructive delusions. Fanatics on both ends, he said, chasing outcomes neither will ever achieve. His critique of Netanyahu’s government was fierce. However, his commitment to Israel’s existence and its alliance with the United States was not in question for a second.

Ultimately, my take-away from the speech wasn’t the criticism but the alternative solution Emanuel sketched. Rather than the familiar two-state framework, he floated what he called a “23-state solution”: the Arab League’s 21 member states, alongside Israel and a future Palestinian authority, taking joint responsibility for building something durable. Arab governments would be pressed to stop rewarding those who kill Jews. In exchange Israel would end unilateral West Bank annexation and Arab states would extend full diplomatic recognition to Israel. He tied it to the India-Middle East-Europe Economic Corridor, arguing that regional economic integration would give every party a stake in stability that isolated, zero-sum nationalism never could.

Actually, it is a two-stat proposal, with the added layer of the Arab League as the catalyst and de facto guarantor—notably not the U.S. Whether or not that specific plan is realistic, it’s a useful model for what serious criticism of Israel looks like: it names the failures bluntly, it proposes concrete costs — sanctions on settlement financiers, an end to discounted arms sales — for continuing them, and it never once treats the country’s existence as up for negotiation. That’s a fundamentally different exercise from a slogan whose endpoint is the state’s dissolution.

Why “apartheid state” doesn’t hold up

This is where the “pariah state” and “settler-colonial apartheid regime” framing, increasingly standard on the progressive left, actually fails as description, not just as rhetoric.

Israel has roughly 21% of its citizens who are Arab — Muslim, Christian, and Druze — who vote, serve in the Knesset, sit on the Supreme Court, practice medicine and law, and organize their own political parties, some of which have sat in governing coalitions. That doesn’t erase any discrimination Arab citizens of Israel face in housing, budget allocation, and policing, and it doesn’t touch the fundamentally different legal status of Palestinians in the occupied West Bank, which is a separate and much more serious problem deserving of its own reckoning.

But a state where a fifth of the population is a legally protected, enfranchised minority with its own representatives in the national legislature is not what apartheid meant in South Africa, where the entire architecture of the state existed to disenfranchise the racial majority. Collapsing “deeply flawed democracy grappling with an unresolved occupation” into “apartheid state” isn’t moral clarity. It’s a category error that conveniently forecloses the possibility of reform, since if the whole state is the crime, nothing short of its dissolution counts as a remedy.

It’s worth being honest, too, that the occupation itself is genuinely contested terrain rather than a settled verdict. Serious human rights organizations have applied the apartheid label to describe conditions in the West Bank specifically, and that argument deserves engagement on its merits rather than reflexive dismissal. But there’s a difference between debating whether a particular legal regime governing millions of stateless residents of occupied territory meets a technical definition, and declaring that the sovereign, internationally recognized state next door — with its Arab Supreme Court justices, its Arab members of Knesset, its Arab-Jewish medical staffs treating patients side by side in Israeli hospitals — is itself an apartheid regime.

Much of the “settler-colonial” rhetoric popular on the American left elides that distinction entirely, folding pre-state Zionism, the 1948 war, the 1967 occupation, and today’s Netanyahu coalition into a single unbroken story of dispossession with one prescribed ending. That’s a coherent ideological narrative. It is not the same thing as an accurate description of a country where nearly a quarter of the population is a protected minority participating in its democracy.

The receptor-site problem

There’s a sharper way to put all this, and I owe it to Michael Feuer, former Dean of George Washington University’s Graduate School of Education, who recently responded to a Dana Milbank post in NOTUS in his Evidence Matters Substack. Millbank complains that Bibi Netanyahu and the right-wing zealots in his government have caused a lot of damage to the soul of Israel and are testing the love and loyalty of many American Jews. He’s on point about that. But he contends that Netanyahu’s government “has done more damage to the Jewish State than Hamas or Hezbollah could ever hope to do? “ Feurer is incredulous:

We’ve seen many examples of “moral inversion” since October 8, 2023, when crowds almost instantly blamed Israel for bringing on itself the worst pogrom since the Shoah and preemptively held it responsible for a retaliation that wouldn’t start for another three weeks. But this one really stretches credulity.

Feuer’s point is only in part the high pitch of Millbank’s specific claim, extraordinary as it was. It was also about a failure of persuasion. He borrowed a phrase from pharmacology: a message, however well-intentioned, has to find its “receptor sites” in the listener, or it does nothing — it just bounces off.

This notion is worth its own Pancake. Smug, sweeping condemnation might feel satisfying to write and gratifying to the readers who already agree with it, but it forecloses exactly the kind of listening that might actually move someone. Worse, Feuer argued, that kind of rhetoric ends up landing instead with people who don’t want reform of Israeli policy at all. Many want the whole project gone, because it hands them a mainstream-sounding quote to repurpose toward an end the original writer never intended.

Feuer’s essay draws precisely the distinction I’m making here. He’s unsparing about the damage Netanyahu’s government has done to Israel’s standing and to its own citizens’ faith in their institutions. But he refuses to let that shade into an equivalence between a democratically elected government making serious, criticizable mistakes and organizations explicitly committed, by charter, to Israel’s annihilation. Muddying that distinction, he suggests, isn’t just intellectually sloppy, it actively disarms the people best positioned to press Israel toward better policy, because it gives them nothing to work with except a demand to disown the country altogether. Feurer singled out Emanuel’s speech as an example of criticism that actually landed in Israel, precisely because it came from someone whose loyalty to the country was never in doubt.

Holding both truths

None of this requires pretending Israel is beyond reproach, and none of it requires signing onto everything Emanuel or Feuer proposes as policy — I don’t, entirely. But it does require American critics, especially on the progressive left, to do something they’ve largely stopped doing: separate their legitimate, often necessary fury at a specific government’s specific choices from a verdict on whether the country has a right to exist and defend itself against enemies who have said, repeatedly and in writing, that they intend to destroy it.

The polling tells us this distinction is eroding fast, on both sides of the aisle, especially among people who came of political age after October 7. That’s worth taking seriously rather than either denying or celebrating. But the answer isn’t to choose between reflexive defense of whatever Israel’s government does and treating Israel itself as a pariah. It’s the same answer we’d give if someone asked us to make that choice about our own country under a government we despised: you can think the leadership has badly failed its people, work to change it, and still believe the nation itself deserves to exist, to defend itself, and to be judged as a flawed democracy rather than condemned as something else entirely.

Leave a comment

Share

There is much to criticize of the Netanyahu government. However, there seems to be more underlying much of the anti-Israeland anti-Zionism rhetoric that seems to place them as the number one global pirriah state: worse than Russia, which has been lobbing missiles at civilians in Ukraine for four years, which, unlike Hamas, did not attack Russia and slaughter its civilians? Worse than China’s oppression of the Uyghurs? Worse than the years-long atrocities of Al-Shabaab, with hundreds of thousands killed in Somalia?

Leave a comment

Today’s baseline five-day, 40-hour work week evolved from the 10 (or more) hour, six-day norm that came with the Industrial Revolution in the early 19th century. To condense a long history into a sentence, the path to today’s 40-hour work week came about in different countries at different times over the next century. The main catalyst in the United States was the demand for labor during World War I. This was more broadly codified during the Great Depression. In 1938, the U.S. adopted the Fair Labor Standards Act, which required overtime pay for work beyond 40 hours by October 1940. This officially made the five-day, 40-hour schedule the legal standard.

And it has stuck ever since, save for an experiment here, an outlier there. However, there have been noises in recent years advocating that the time has come for a shorter work week, specifically 32 hours over four days. In 2024, Sen. Bernie Sanders introduced federal legislation to mandate just that.1

This isn't a fringe idea being tested in a handful of boutique companies anymore; it's edging into the realm of labor law, which means the version that gets written into policy — reduced hours versus compressed hours, protected for all workers — actually matters.

The pancakeable issue

Most debates over the four-day week get argued as a referendum on whether the idea "works," as if there's one answer. However, the four-day week isn't a single policy; it's a response determined differently depending on what kind of work one does, what technology is transforming any specific job right now, and whether any given employer is one of the well-resourced early adopters volunteering for a pilot or one of the millions of hourly employers who would only adopt it if forced to by law. A serious look has to examine both the genuinely strong evidence for it and the largely underexamined ways in which a mandated 32-hour week could exacerbate differences its advocates say they want to fix. That tension, not a verdict, is the point of this piece.

The case for a true four-day week

I set aside the compressed-schedule 40-hour/4 day gimmick and look only at genuine hour reductions. The evidence here is stronger than skeptics often admit.

The well-being data is close to unambiguous. The largest and most rigorous test to date is a six-month, six-country randomized controlled trial published in Nature in 2025. It found that workers were happier, less burned out, and reported better mental and physical health — and were just as productive as before. This wasn’t a self-selected startup anecdote; it was a controlled, cross-national study designed specifically to test whether the earlier, smaller pilots would hold up.

A trial in Iceland is the closest thing to a national-scale proof of concept. From 2015 to 2019, Iceland ran large-scale trials that put 2,500 public-sector workers on 35–36-hour weeks with no pay cuts. “Productivity stayed the same or improved, and, again, worker well-being improved dramatically across measures of stress, burnout, health, and work-life balance.” The follow-through is the more striking part: this wasn’t a pilot that quietly died. Today, roughly 86% of Iceland’s workforce has access to shorter hours or is actively working them. That’s not a press-release statistic; it’s a labor market that actually moved.

The largest test in the private sector, involving 61 companies over six months in the UK found real economic results, aside from happier employees: nearly 3,000 workers across these employers moved to a four-day week, and over 90% of participating companies kept the policy after their trials ended.

Microsoft’s experience in Japan remains the most-cited productivity experiment. After closing offices on Fridays and cutting meeting times in half, it recorded a 40% productivity gain — a model the company has continued offering employees since.

The academic literature backs this up beyond the individual pilots. A study in the Academy of Management Journal found the four-day workweek does not reduce labor productivity and may in fact increase it, and the American Psychological Association’s own review of the research base concluded that pilot studies have consistently found that switching to a four-day workweek improves worker well-being and job satisfaction while reducing certain organizational costs.

Finally, advocates make an explicitly historical argument, not just a wellness one. Groups like Inequality.org frame the push as a direct descendant of the labor movement that won the eight-hour day. The position of the advocacy groups, including many unions, isn’t just “wouldn’t it be nice.” They argue that productivity gains over the last several decades have flowed disproportionately to the top while wages stagnated2 and hours crept up. Thus, reducing hours without cutting pay is one direct way to claw some of that value back for workers rather than shareholders.

Taken together, this is a genuinely strong case — rigorous, replicated across countries and company sizes, and grounded in both well-being outcomes and a coherent political argument about who captures productivity gains. It’s exactly why the harder question isn’t “does this work,” but “who is this evidence actually about.”

The catch: Whose data is this?

I always want to know the source and assumptions behind the data: is it from an advocacy group that has a point of view? If sampling data, is it from a large scientific sample? Who was sampled: that is, what was the universe the sample was drawn from? Or was it a poll and not a scientific sample at all—just whoever responded.

The various trials cited above were vetted and reviewed. Nonetheless, understanding the data sheds light on possible biases or shortcomings. Start with who’s actually in these trials. Multiple reviewers of the UK pilot flagged “selection bias,” in that participating organizations were mostly small, disproportionately white-collar, and concentrated in professional services, nonprofits, and creative industries. These are sectors that lean on human capital rather than machinery or unskilled labor. The American Psychological Association’s own review of the research literature points to the same selection problem: companies that volunteer for a four-day-week trial usually already suspect it will work for their business, which is a very different population from a mandatory, economy-wide rule.

This does not undercut the findings, but does suggest a boundary on what they can tell us. That is, how generalizable are they?

Limits of the findings

Push the four-day week out of the office and into a hospital floor, a warehouse, a retail counter, or an assembly line, and the mechanics change entirely. Healthcare is the clearest case: patient care runs 24/7, and a shorter week for staff means hiring more people to cover the same hours. Though workers may support the finding on improved work-life balance, it is unlikely that overall productivity could increase today. Helping residents take showers or administering medications in an assisted living facility do not lend themselves to productivity gains. Researchers estimated that hospitals in the Netherlands would incur an additional $187 million in added staffing costs. That would have to be offset by reduced absenteeism and turnover. That’s a high-stakes wager, not a free lunch.

Manufacturing runs into a more literal problem: production lines and machinery run on fixed physical timelines, and cutting a day of operation risks production slowdowns and missed deadlines unless compensated for with sufficient automation, if it even exists for some tasks.

Then there’s the plainest problem of all: money. Hourly workers who depend on overtime pay, or who are already juggling unpredictable just-in-time schedules, don’t have the luxury of “same pay, fewer hours.” As one workforce-flexibility researcher put it, front-line and blue-collar workers lack the same scheduling flexibility as office workers, and industries built around hourly labor may resist adding compensating hours if it triggers overtime costs.

The two-tier question

This is where real tension may develop. A panel discussion at this year’s World Economic Forum pointed out that for knowledge workers, a four-day week is a straightforward win. But for many other workers, what they actually need is predictable hours so they can arrange childcare and guarantee enough hours to make ends meet. Ignoring that gap, said the panelists, raises inequality issues. They noted that roughly 60% of the workforce in developed countries doesn’t have the option of remote or flexible work at all. They were the ones showing up in person as “essential workers” throughout the pandemic while offices went remote.

Labor advocates on the left, oddly, converge with center-right skeptics on this exact worry, if from different angles: progressive commentators argue a four-day week without a legislated, universal right — one tied explicitly to protecting hourly wage floors — risks becoming another perk that flows to already-comfortable professionals Without that guardrail, the four-day week doesn’t level anything — it just adds another axis of stratification on top of the remote-work divide the pandemic already exposed.

The right-of-center critique arrives at a similar worry, but from the opposite direction: not that the market underdelivers a shorter week, but that a federal mandate forcing one is the wrong tool and would produce its own version of a two-tier outcome. The American Enterprise Institute’s case against the 32-hour federal proposal points to a labor market already running historically tight — more than 10 million job openings against roughly 9 million unemployed — meaning a mandate risks forcing employers to either expand headcount or pay overtime just to hold output steady, with unpredictable knock-on effects for wages and inflation. Its preferred alternative is explicitly the market-driven version of the same reform: let workers and employers negotiate their way toward shorter weeks as technology and productivity gains allow, rather than imposing it by federal law.

So the left worries a weak legal right entrenches the divide, and the right worries a strong legal mandate creates fresh distortions — but both routes plausibly land in the same place: knowledge workers get it first, hourly workers get it last or never, unless the policy design deliberately closes that gap rather than assuming it will close itself.

AI as force multiplier: knowledge work vs. manual labor

So far, I’ve bushwhacked through this analysis without saying much about AI, though itis strongly implied when looking at productivity in knowledge work. The knowledge-work case for AI-as-multiplier is already well documented and fairly strong. A review of task-level productivity studies covering writing, coding, customer support, and translation found consistent gains — typically 20%–60% in controlled studies and 15%–30% in real-world work settings. Google’s UK research found that AI saves administrative workers about 122 hours a year, and customer support studies found that teams using AI assistants became 15% more productive by handling more tickets in less time.

At the macro level, a Federal Reserve Bank of Atlanta survey of nearly 750 corporate executives found labor productivity gains from AI are positive, vary across sectors, and are expected to strengthen through 2026, with the largest effects concentrated in high-skill services and finance. PwC’s 2026 Global AI Jobs Barometer, drawing on more than a billion job postings, found something that complicates the “AI kills jobs” narrative: productivity growth is 40% higher at companies most exposed to AI, and those companies are raising both wages and headcount faster than less-exposed companies.

No surprise: the manual-labor case is close to the opposite. A widely cited occupational-impact study found the least-AI-impacted jobs centered on physical and manual labor in agriculture, transportation, food service, and construction, where tasks don’t involve the kind of abstract reasoning large language models are good at. Anthropic’s own usage data reinforces this. For Claude conversations, usage is highest for professions like software development and lowest for professions involving physical labor.

Enter robotics

However, there’s a further wrinkle to the knowledge vs. physical occupation dichotomy. One study tracking AI exposure and employment outcomes found occupations reliant on complex reasoning and cognitive tasks show the greatest adverse effects from AI, while some manual and physical-task occupation even showed employment gains. That’s a genuine inversion of the “AI helps knowledge workers, threatens manual workers” framing I expected. The mechanism may not be that manual work is safe from automation; it’s that language-model AI specifically doesn’t touch it, while a different technology — embodied AI, i.e., robotics — is the one actually aimed at that work.

The distinction that resolves the apparent contradiction: generative AI is a cognitive-task technology; robotics is a physical-task technology, and they’re advancing on different clocks. Just as digital AI may disrupt white-collar work, embodied AI is poised to disrupt blue-collar work, though on a more extended time scale.

Further out: What if manual work itself disappears?

This, then, leads to what lurks under this whole debate: the general assumption that retail clerks, plumbers, nurses, and machine operators are a fixed feature of the economy that a four-day week either accommodates or doesn’t. But two technologies now advancing in parallel—humanoid robots and 3D-printed construction — raise the possibility that some of these jobs shrink in number before the timing question even gets settled.

Humanoid robots have moved past demo-stage hype into actual factory and warehouse floors. BMW, Amazon, Mercedes-Benz, and GXO are running real, measured pilots today — tote-moving, parts loading, kitting, and inspection tasks that sit in the gaps between existing automation systems. Costs are falling fast: from over $150,000 for advanced models down to roughly $16,000 for entry-level platforms, a drop of at least 40% between 2022 and 2024. But nearly every serious industry analysis lands on the same caveat: dexterity is the bottleneck. Fine motor tasks — tight-tolerance assembly, cable routing, fastening, connector insertion — are still much harder than basic pick-and-place work. Foolproof manipulation for tasks like wiring and plumbing remains an open research frontier, not a solved problem.

For the immediate future, humanoids doing plumbing or finish carpentry are a goal rather than a near-term plan. Unit costs need to fall from the current $150,000–$500,000 range down to $20,000–$50,000 before broad deployment is economically competitive with local labor, on top of the safety certification and legal frameworks that don’t fully exist yet, but are not out of reach. Most forecasts converge on the same rough timeline: meaningful scale is a late-2020s-to-2030s story. Thus, not ready for prime time, but on the planning horizon.

Further along in one narrow sense, 3-D printing is already displacing labor on the structural shell of a house. Companies like ICON have completed a concrete shell in as little as 24 hours, and cost reductions of 30–50% versus conventional framing are now well documented across multiple projects. Nonetheless, the “greatly reduces the need for construction workers” framing overstates it in one important way: printing only replaces the framing/shell crew. Traditional crews still handle plumbing, electrical wiring, drywall, insulation, and finishing.

So: realistic, yes, as a gradual, partial, decades-long erosion of specific tasks within these trades, not a near-term replacement of the plumber or the baker. If anything, that timeline argues for treating the four-day-week debate and the automation debate as sequential rather than substitutable: the equity question about hourly workers being frozen out of the shorter week is a live 2020s problem, while robots doing their jobs for them is more plausibly a 2030s-and-beyond one. A two-tier economy could plausibly manifest itself before automation offers any physical workers a path to benefit.

The Pancake take: Implications for a 32-hour work week

If genuine AI force-multiplication is concentrated in knowledge work right now, then a true four-day week is more economically defensible for desk workers precisely because AI is quietly absorbing some of the output we’d otherwise lose by working one less day. Simply, the productivity gain offsets the hour reduction. Manual and hourly workers get no equivalent AI offset today, so a federally mandated hour reduction for them is a straight cost, not a productivity wash, until robotics — on its slower, still-uncertain timeline — eventually closes that gap. That’s a sharper version of the two-tier argument than “knowledge workers just have nicer jobs”: one group has a realistic technological subsidy today, making the shorter week nearly free to the economy, and the other doesn’t yet, but long trends suggest a positive outcome if we are politically patient. That’s not a sure thing.

Leave a comment

Share

There have been calls in recent years advocating for a shorter workweek, specifically 32 hours over four days. In 2024, Sen. Bernie Sanders introduced federal legislation to mandate just that.

This isn’t a fringe idea being tested in a handful of boutique companies anymore; it’s edging toward the realm of labor law, which means the version of it that gets written into policy — reduced hours versus compressed hours, protected for all workers — actually matters. Most debates over the four-day week get argued as a referendum on whether the idea “works,” as if there’s one answer. However, the four-day week isn’t a single policy; it’s a response determined differently depending on what kind of work one does

“Rebuild Local News,” a large coalition of newspaper nonprofits and community newspaper associations, warns that the loss of local news outlets is a “crisis” that cuts “communities off from essential civic information.” The “State of Local News Initiative” at Northwestern University’s journalism school also sees a crisis at work, one that is actually more urban than rural.

Is there a crisis? If so. what makes it so? And, if there is a crisis, what is to be done to uncrisis it? In the past I’ve written here that there are so many crises that the word may have become diluted. “Crisis” is the go-to word for anything that seems to demand immediate remedy.

Among the applications of the word can be “an emotionally significant event or radical change of status in a person’s life,” as in provoking a mid-life crisis. It may refer to “an unstable or crucial time or state of affairs in which a decisive change is impending,” as in a financial crisis.

Or crisis can denote “a situation that has reached a critical phase,” such as a local newspaper crisis. It is this last application that I want to address, though I propose that we consider this less a crisis than a potential problem. But I digress.

The narrative of local news decline as a democracy emergency is near-universal among media observers. There is a skeleton of a contrarian case worth making: local TV (still widely watched) covers many of the same beats; hyperlocal digital outlets, Substack newsletters, and community Facebook groups have partially filled gaps; and some of the “loss” involves consolidation of outlets that were duplicative or mediocre. The harder question is whether local election turnout, civic engagement, or corruption actually worsens when a paper closes. I attempt to tease out what the data actually show.

Subscribe now

The scale of the loss

The most-cited tracker is Northwestern’s Medill Local News Initiative, which has run the “State of Local News” report for a decade. Their 2025 report found that so-called news deserts are widening, newspaper closures continue unabated, and independent publishers are shutting down. On the plus side, local digital-only news sites are multiplying.

A news desert is a geographic area—such as a rural town, urban neighborhood, or entire county—that lacks access to credible, original local news and information. These information vacuums occur when local newspapers, broadcast stations, or digital newsrooms shut down or severely shrink their reporting operations.

Counties and towns with little or no local news source are not a new phenomenon. Twenty years ago, there were about 150 news desert counties covering roughly 37 million people. By 2025, the number of news desert counties rose to 213. Another 1,524 counties are down to a single remaining news source. Together, now roughly 50 million Americans have limited or no access to local news. The trend line is clearly worsening, not stabilizing.

Over the past two decades, the country has lost nearly 3,500 newspapers and more than 270,000 newspaper jobs, and almost 40% of all local U.S. newspapers have vanished since 2005. The closures aren’t slowing — 136 closed this past year, up from 130 the year before, a rate of more than two a week — and notably this year’s wave hit smaller, independently owned papers rather than chain-owned ones, which matters because those independent owners tended to actually live in the communities they covered and were the most trusted local sources.1

What the research says about consequences

There is some empirical evidence that there is at least a correlation, if not direct causation, between loss of local news and negative consequences for the residents of such deserts.

• Money: A widely cited academic study found that after a newspaper closes, municipal borrowing costs rise 5–11 basis points (finance-speak for .05%-.11%), costing the municipality about $650,000 per bond issue. In this instance, the authors found that the effect was causal, not merely correlated with the underlying economic decline. The mechanism is information: less local coverage means more friction for lenders trying to evaluate the quality of public projects and the officials running them. The demise of the local newspaper was also associated with higher government wages, higher deficits, and more costly refinancing deals. The conclusion: less accountability leads to more waste.

• Corporate/environmental accountability: Researchers found that a local newspaper’s closure is associated with a measurable 1.1% increase in local corporate violations and a 15.2% increase in penalties, suggesting that local journalism constrains both private misconduct and government behavior. An innovative variable detected that plant-level toxic emissions increase after a newspaper’s closure.

• Civic participation: A 2010 study found Los Angeles-area residents with access to a daily paper were more likely to vote regularly than those without, and a 2019 report found mayoral elections less competitive where newsroom cuts were steepest. Counties that lost a paper also showed fewer split-ticket votes in 2012, a proxy for rising straight-line partisanship once local accountability reporting disappears.

These findings are hard to wave away with “well, people get local news from Facebook now.” Facebook groups don’t do investigative accountability reporting, don’t attend the zoning board meeting, and don’t get quoted by municipal bond underwriters as a credibility signal. Individuals on Facebook may undertake any of these functions, but don't have the credibility usually associated with an established newspaper.

Do the alternatives actually substitute?

What about the perspective that local TV, hyperlocal digital outlets, Substack newsletters, and Facebook groups may fill gaps left by a shuttered local newspaper? For example, as I’m writing this, I noticed an email from Nextdoor with some news posted by residents of my Portland neighborhood. I do subscribe to the local daily, but I doubt it would have informed me about this meeting or even covered it.

What have we learned about how substitutes have evolved to replace, or sometimes improve on, the legacy newspaper?

A fresh Local News Initiative survey (Feb 2026) directly tested this. The answer is nuanced in a way that cuts both for and against either side.

• People in news deserts consume news at nearly the same overall rate as people in areas with a local paper, and they don’t perceive themselves as deprived. Indeed, they’re satisfied with social media and TV filling the gap. Social media news groups and local TV each reach about 4 in 10 news-desert residents daily, followed by search engines, friends/family, and influencers.

• On the other hand, people in news deserts trust the news media considerably less than people in served areas (46% vs. 59%), and the loss isn’t obvious to residents because “unvetted, un-fact-checked information” is bouncing around social media instead of professional reporting. The difference between social media sources and the local newspaper is who’s doing the informing and how accurate it is.

• A community-resilience study of local Facebook groups reached a similarly split verdict: audiences feel resilient and adapt to fill the local-news vacuum, but the study’s conclusion was that “social media is not a substitute for local news,” partly because content shared by familiar local ties carries a credibility bump that makes misinformation more, not less, dangerous there.

• In Colorado, the Colorado News Mapping Project found three counties as having no local news source other than a Facebook group. The Facebook groups share everything from local events to police updates to weather warnings. In these communities, many volunteer Facebook group moderators are serving as de facto part-time citizen journalists. It does beg the question about their roles in the news media, and the thin line between reposting and reporting.

• UK research found a similar pattern: every community studied had local Facebook pages and groups, but respondents said a wide range of local issues and institutions simply weren’t getting sufficient coverage, and the groups were also a steady source of rumor and misinformation.

• Are online editions of newspapers a workable option? I would have thought so. However, in rural areas like much of Colorado, low-connectivity realities actively work against “just go digital” as a fix. Rural newspaper owners in eastern Colorado say the internet isn’t strong in many of these places, and if their coverage were online-only, it wouldn’t be accessible to their less tech-savvy readership. I would think that access to news sites, “if we only know how,” would be a motivator to learn. Apparently not a motivator enough.

• Even in small towns, at the least municipal websites typically post agendas and hearing notices, and municipal announcements. This, too, may help offset some of the loss from the closure of newspapers, though it still faces connectivity and tech literacy challenges. Moreover, while a real improvement in raw access, it doesn’t replace the interpretive function a reporter might serve.

Some bright spots

The successes are real but narrow. Medill’s own “Bright Spots” and Nieman Lab’s rankings point to genuine replacements — but overwhelmingly in metro or college-town markets with money and educated audiences, not the places actually becoming news deserts.

• Lookout Santa Cruz launched in 2020, won a 2024 Pulitzer for its flood coverage, and had 8,500 paid subscribers by 2023. Its founder, Ken Doctor, is explicit that this enterprise “needs to be a replacement for a daily newspaper... The journalists need to be good enough, so it’s going to take more money.” This is a heavily capitalized, intentional rebuild of a newspaper, not a spontaneous Substack.

• Charlottesville, Virginia, was not exactly a news desert, with the three-times-weekly The Daily Progress in town. However, like so many newspapers, the legacy newspaper cut its journalism staff from 42 to 18 to stay afloat, and with it, coverage of the city. A local resident started Charlottesville Tomorrow as a blog to pick up some of the slack, initially focusing on planning and land use and attendant business and environmental issues. Over 20 years as a non-profit it has grown from a two-person operation into a full newsroom. Its reporting is always free, never paywalled.

Probably the mental image of a “news desert” is of remote rural towns. Yes, those exist. However, the vast majority of news deserts and thus local news startups are in metro areas. Four-fifths of the 300+ startups launched in the past five years are digital-only, concentrated in high-density areas, not the rural counties actually losing papers. A perfect example played out in Cambridge, Massachusetts- not exactly one’s image of a news desert. Cambridge is a city of 120,000 and a budget of almost $1 billion, across the Charles River from Boston, a distance of as little as 200 feet. Although The Boston Globe has run occasional stories focused on Cambridge, the bulk of its coverage has been of Boston and the state.

For decades, the weekly Cambridge Chronicle did a barely adequate job of covering the city’s government, social, business and entertainment areas. By the time it was sold to Gannett, it had no local staff or coverage. In 2024, a civic-minded group of Cambridge residents came together to buy a one-man site, Cambridge Day, and established it as a not-for-profit enterprise, enabling it to accept grants from foundations and tax-deductible contributions. Since then, it has hired several full-time journalists and, as importantly, a small business staff. It even publishes a tabloid weekly print edition, available for free at some local stores.

It does not take deep analysis to understand why localities like Cambridge, Charlottesville, or Santa Cruz are more likely to be the loci of local journalism. They share characteristics of a well-educated and relatively affluent market. It’s no news flash that economics plays a critical role.

Assuming the lack of a newspaper equals news desert is not wrong, but not correct.

The alternatives exist and, in many cases, are quite functional, if limited in scope. Volunteer Facebook and Nextdoor group moderators are not exactly the makings of a Front Page, but, truth be told, many of the dead legacy small-town dailies and weeklies did not always live up to our image of civic engagement and investigative exposés, either.

In many cases, the Facebook groups are limited to aggregating posts from others in the community. For example, Sara Hevner, in Cheyenne County, Colorado, started a Facebook page after the local paper folded in 2022. She's explicit that she isn't a journalist, but she says that before she reposts what others share, she checks whether posters are local. The majority of her feed's actual content is local school sports scores and game updates — which is exactly the "high school sports" beat that was often among the best-read sections of the former newspaper.

Nonetheless, need often leads to innovation. The Documenters Network was started 10 years ago in Chicago, and is now active in about 25 communities, some deserts, some not. It trains young journalists as well as civic leaders to specifically “attend local government meetings, distill what matters, and bring actionable information back to their neighbors.”

Local staff know their communities best. They determine which meetings to cover, manage editorial workflows, and build the relationships and programming that make Documenters valuable beyond the assignment. What local sites develop and learn gets shared across the network—and what the network learns sharpens what sites do at home.

Documenters are paid on a freelance basis for their reports, and the organization, initially funded through a Kickstarter campaign, seems to be supported by grants and donations.

So, are news deserts a crisis?

These examples suggest that the absence of a newspaper is not a prima facie requirement for designating a news desert, though they do point to the need to examine an area more thoroughly. The variety of substitutes is not evenly distributed, and even where they exist, they don’t functionally replace what a single paper did.

First, compared to a newspaper, aggregation of local TV or radio plus Facebook groups, plus Nextdoor, plus a local Substack or blog, plus a service such as Documentor has a real and measurable cost, not necessarily financial but in time. Aggregation requires proactivity a single paper never did. The trust gap found by the State of Local News Initiative (46% vs 59%) and that “people don’t realize they’re uninformed” are evidence that these options aren’t fully substituting.

Second, the accountability function has no real substitute at scale. Research has shown that municipal bond costs, the number of corporate violations, and local government wages are hard economic metrics indicating that something breaks structurally when the newspaper watchdog leaves, regardless of whether residents feel informed. They don't know what they don’t know.

Finally, where it does work, it’s expensive and rare. Lookout Santa Cruz needed millions in seed philanthropy and a Pulitzer-chasing strategy; Charlottesville Tomorrow needed 20 years. The Documenters has had to raise $1 million to pay its citizen-journalists. If “success” requires that, most communities losing their paper will never get a real substitute other than a Facebook group. After all, the local newspaper shut down because the community could not financially sustain it.

And another “finally”

Let’s not lose sight of the reality that today’s information environment is an improvement over the pre-internet era, even in the worst-served places.

This is true regardless of whether a newspaper exists or not — it’s not really about newspaper decline, it’s about the baseline improving underneath the whole system. A few structural facts:

• Court records, property records, permit filings, and budget documents that once required a physical trip to a courthouse or clerk’s office are now frequently searchable online, even for an ordinary resident or the sole surviving reporter, or a citizen-journalist Facebook moderator. They can now do in ten minutes what once required in-person digging that most citizens simply never bothered to do. This is the exact mechanism that made a local paper indispensable in 1985. Today the access part of that bottleneck is gone even where the time and motivation part still exists.

• Meetings that were once genuinely invisible unless a reporter physically sat in the room are, per the open-meetings-law requirements, now routinely recorded and archived — so a scandal or a controversial vote is discoverable after the fact by anyone who thinks to look, not lost the moment the room emptied.

• Statewide and national outlets, wire services, and aggregators can now surface a small-town story instantly if it resonates. Pre-internet, a genuine outrage in a 3,000-person town might never travel beyond the one weekly paper’s circulation area no matter how serious it was.

Once again, however, none of this replaces the proactive watchdog function — someone whose job is to notice the story before it becomes a scandal, not just make the paper trail available after the fact to someone who already suspects something’s wrong. Raw information access is necessary but not sufficient for accountability; the bond-yield and corruption research from earlier suggests it’s specifically the proactive monitoring, not the availability of records, that moves those numbers. The internet has made the worst-served communities better off on the axis of “can a motivated person find out what happened,” while leaving them no better off on the axis of “does anyone with the time, skill, and incentive actually go looking.”

Crisis or problem?

The “State of Local News Initiative” survey data does find that people in news deserts don’t feel worse off and consume news at similar volumes . So there’s a genuine empirical case that subjective welfare hasn’t collapsed, even if objective accountability metrics have. That’s worth naming rather than ignoring. But it’s not a solution.

Given my characterization of “crisis,” I couldn’t make a case that news deserts are a societal crisis. That’s too glib. But they are a problem.

Leave a comment

Share

The narrative of local news decline as a democracy emergency is near-universal among media observers. There is a skeleton of a contrarian case worth making: local TV (still widely watched) covers many of the same beats; hyperlocal digital outlets, Substack newsletters, and community Facebook groups have partially filled gaps; and some of the “loss” involves consolidation of outlets that were duplicative or mediocre. The harder question is whether local election turnout, civic engagement, or corruption actually worsens when a paper closes. I attempt to tease out what the data actually show.

If you first got online with 300 or 1200-baud dial-up modems (beep-beeeep got you into The Source, CompuServe, or AOL), you’ll remember when broadband replaced dial-up. Connections got faster, more reliable, and cheaper. In 1995, it cost roughly $.10 to transmit 1 mb of data. As the infrastructure providers upgraded their networks, they became more efficient, and the cost per mb for the network providers dropped exponentially. Today, it is so low that it is close to free. In spite of the lower cost of transmission, we are using far more energy to run and use those networks because, of course, with lower cost came far more use and thus terabytes more data transmitted.

I don’t recall anyone responding by advocating that we now needed to use the internet more efficiently. Instead, we invented YouTube, then Netflix, then Zoom. The faster and cheaper each megabyte became, the more megabytes got consumed — by orders of magnitude. By the time 5G and gigabit wired connections arrived, whatever the “savings” from efficiency had been buried under an avalanche of demand that faster and cheaper connections had themselves created.

This is not a paradox. It is economics. The Victorian economist William Stanley Jevons observed it first in 1865, watching Britain’s coal consumption soar as steam engines became more fuel-efficient.1 Cheaper energy per unit didn’t reduce total consumption. It expands the universe of things energy gets used for. The savings get eaten by growth.

What I will discuss today is AI data centers, energy consumption, and costs. Keep Jevons and coal consumption in mind as I work through the AI energy debate, because almost nobody covering it seems to have heard of him.

Subscribe now

What the debate looks like from the outside

By now you’ve probably at least seen the news reports about data centers being built and proposed— the vast warehouse complexes of servers that power everything from your Google search to ChatGPT. They are consuming electricity at a staggering and accelerating rate. U.S. data centers consumed 183 terawatt-hours in 2024, more than 4% of the country’s total electricity use, roughly equivalent to the annual consumption of Pakistan. By 2030, that figure is projected to grow by 133%. Globally, the International Energy Agency (IEA) projects data center electricity demand will nearly double, growing at 15% per year — more than four times the rate of every other sector combined.

The environmental advocates say this is a climate catastrophe in the making. The tech companies say they’re on it: Google reports that the energy per median Gemini prompt fell 33-fold in a single year. Microsoft, Amazon, and Meta are signing nuclear power deals. The future, they assure us, is carbon-free.

Both sides are telling us something true. Both sides are obscuring something important.

The Jevons problem hiding in plain sight

Here is the number that doesn’t appear in the tech companies’ press releases: in the same year that Google reduced its per-prompt energy consumption by 33 times, its total data center electricity consumption grew by 27%. Cheaper AI means more potential users will sign up for a service. More AI means more total energy, even as each individual query costs less. The paradox is not a paradox at all — it is exactly what economics predicts.2

The IEA’s projection that data center energy for use will nearly double by 2030 already incorporates expected efficiency gains. In other words, the analysts building that forecast know about the 33-fold efficiency improvement and have factored it in — and they still arrive at near-doubling of total consumption. Efficiency is real. It is just being lapped by demand.

This is not an argument that AI is bad or that data centers shouldn’t exist. It is an argument that the industry’s preferred framing: “we’re getting more efficient, so the energy problem will solve itself,” is precisely backward. Historically, the greater efficiency tech enterprises deliver leads to greater consumption. This has been true of automobiles, refrigerators, air travel, and home heating. There is no particular reason to think AI is the exception.

The variable nobody is calculating

The conventional debate , which has been almost entirely about energy supply, misses what may be its most consequential dimension: what does all that computing actually gets used for?

There is a legitimate and serious case that AI, deployed toward the right problems, could be a net positive for global energy and climate. Applying AI to optimizing electrical grids in real time could reduce transmission losses. Accelerating materials science could produce better batteries. Could AI-assisted drug discovery reduce the substantial energy waste embedded in clinical trials that don’t pan out? Some researchers have suggested, and the IEA has acknowledged in exploratory modeling, that AI applied to energy optimization alone could offset the carbon cost of running the data centers that power it.

This is not wishful thinking. It is a genuine open question.

The honest answer, which almost no one in this debate is offering, is that the net impact of AI on energy and climate is deeply uncertain, because it depends almost entirely on what AI is used for, at scale, over time. If the marginal AI query is generating a cat video—or a Substack post— the ledger looks bad. If it is modeling nuclear fusion containment, the ledger looks very different. We are, at the moment, generating a lot of cat videos. (Hard to tell about Substack.)

Who decides — and who pays?

There is a third dimension to this story that doesn’t fit neatly into either the environmental or the economic frame. Decisions of enormous public consequence, such as which power sources get priority, which communities host these facilities, what gets built on the national grid, are being made almost entirely by five or six private companies, with meaningful public deliberation late to the party.

For example, in Virginia, which already has become the world’s most concentrated data center market, those facilities now consume more than a quarter of the state’s total electricity. Dominion Energy projects that the average residential electric bill will roughly double by 2039, driven primarily by data center demand. That projection did not emerge from a public debate about energy priorities. It emerged from private power-purchase agreements negotiated between utilities and corporations. Ordinary ratepayers are picking up a substantial portion of the infrastructure costs. They were not asked.

The national security argument, that America must build AI infrastructure at any pace to compete with China, is real and not to be lightly dismissed. China is building data centers and deploying renewable energy simultaneously at a scale that should make American planners uncomfortable. It doesn’t have to deal with the niceties of policy debates, at least not in public. But the national security framing is also being used to preempt questions that deserve answers: built where, powered by what, at whose expense, and toward what ends?

The two sides don’t balance equally — but both exist

The case that the current data center buildout is creating genuine problems — for electricity costs, for local communities, for carbon trajectories — is substantially stronger than the industry’s reassurances. The efficiency gains are real but insufficient. The emissions pledges have been largely abandoned in practice. The governance structure is essentially private.

But the doomsday framing has its own blind spots. Data centers currently account for about 1.5% of global electricity consumption and roughly 0.5% of global CO₂ emissions — serious numbers, but not, on their face, civilization-ending ones. The nuclear deals being struck, whatever their motivation, are reviving an energy source that deserves revival. And the possibility that AI accelerates solutions to the very problems it creates is not zero.

The debate we are having is about megawatts and pledges and coal plants. The debate we should be having needs to incorporate issues of governance, the demand side of the equation, and the basic premise in economics: that making something cheaper reliably produces more of it, not less.

Jevons knew that in 1865. The data centers are proving it now.

As always, the Pancake welcomes the other side. What am I missing?

Leave a comment

Share The Two-Sided Pancake

Subscribe now

By now you’ve probably at least seen the news reports about data centers being built and proposed— the vast warehouse complexes of servers that power everything from your Google search to ChatGPT. They are consuming electricity at a staggering and accelerating rate. U.S. data centers consumed 183 terawatt-hours in 2024, more than 4% of the country’s total electricity use, roughly equivalent to the annual consumption of Pakistan. By 2030, that figure is projected to grow by 133%. Globally, the International Energy Agency (IEA) projects data center electricity demand will nearly double, growing at 15% per year — more than four times the rate of every other sector combined.

What I will discuss today is AI data centers, energy consumption, and costs.

Leave a comment