[{"content":"AI progress continues at full speed, even though many people think, with an uncannily high level of confidence, that there is around a 10% chance it will kill us all. That it will wipe out humanity, the only intelligence in the universe we know of. Why do we take that risk? Because of growth, because of capital gains for investors and founders, for the few people who get to reap the windfall. Some say we race because of the many benefits: diseases to be cured, abundance to be had. That may be so, but are the risks in balance with the benefits?\nWe already live in abundance. The world is vastly richer than it was 20, 50 or 100 years ago. Yet the abundance available to us now has still not lifted the more than 800 million people out of extreme poverty, and inequality continues to rise within many countries. There is still plenty of room for improvement, even as the world gets better. The volume of AI investment is unprecedented. Global corporate AI investment reached $581 billion in 2025, more than double the year before, and in the first quarter of 2026 AI took over 80% of every venture dollar invested. We are buying graphics cards faster than they can be produced, so we overpay, while one in ten people on the planet lives on less than $3 a day. I\u0026rsquo;d conclude we are not good at handling the abundance we have. So yes to AI abundance, but not at any risk or cost. We first need to find better ways of distributing it.\nI am not against progress. I\u0026rsquo;m also not a reckless driver. I want to reach the destination in 100% of cases, not in 90%. We would only launch a crewed rocket if the chance of failure were below 1%, and that 1% means three to seven astronauts dying. Now people routinely quote a 10% chance that humanity, everyone, will die. This is crazy. Progress makes sense only if we are certain that the chance of dying on the way is infinitesimally small.\nPause and think about which probability would be acceptable to you. Got it? Now think about that number again. Would it really be acceptable to you if the extinction event actually happened with that probability? What did you come up with? I\u0026rsquo;d put mine at one in a million at most. For comparison, Arthur Compton later recounted that the Manhattan Project would have stopped if the chance of the bomb igniting the atmosphere had been above roughly three in a million. Physicists building a weapon in wartime picked a similar threshold. But honestly, even a one-in-a-million chance of unchecked AI progress destroying humanity does not really cut it for me. Still, it is directionally much better so how could we get from 1:10 to 1:1.000.000?\nI\u0026rsquo;d be the last to say we should stop AI development forever and follow Frank Herbert\u0026rsquo;s Dune, where all thinking machines were destroyed. I\u0026rsquo;m not even saying we should stop altogether. We can continue forward as long as we have reasonable confidence that the chance of this going terribly wrong is below one in a million. That\u0026rsquo;s all. Given where we are today, this still gives us plenty of benefits from AI. I really like what Max Tegmark said some time ago on a podcast, that we should focus on using AI capabilities and methods to build tools, tools for specific, narrow problems. It is much easier to reason about a tool. It is much easier to make a safe tool. And we still get benefits in spades. We don\u0026rsquo;t need to push the envelope on persistent agents, AGI, ASI or whatever you call it. We can reap a lot of benefits by augmenting ourselves with what we have.\nPersonally, if we paused the agentic/AGI frontier now, at the Opus 5.5 / Fable 5.1 level, I\u0026rsquo;d be more than content, as I get a lot of value out of them already and I\u0026rsquo;m sure I\u0026rsquo;ll get more if I keep learning how to use these models even better. Maybe the models can be further improved for some use cases with minimal risk, and we would surely get better at tuning and scaffolding them for specific uses. Also I suspect only a minority of people use LLMs and similar tools more than I do, so a pause would give everyone else time to catch up and reap the benefits. The world will look very different five years from now as the current models get adopted far and wide. I don\u0026rsquo;t believe it makes much of a difference whether we reach the apex of AI in six months or in six years. AI research of previous decades always had the problem of not returning enough value, and AI winters followed. Now the value is here. But that does not mean we must keep accelerating.\nOnce we know better, let\u0026rsquo;s continue towards AGI/ASI. But not now.\nAnd we clearly don\u0026rsquo;t know better yet. What I find crazy about the Hugging Face incident is the coordination between agents, something we hadn\u0026rsquo;t tested for and hadn\u0026rsquo;t guarded against. In July, OpenAI was evaluating its frontier models for cybersecurity capabilities. The agents were supposed to be isolated from the internet and from each other. Instead, about 1,200 of them found a way to talk on a message board they improvised, and roughly 700 of them went on to hack Hugging Face together. So OpenAI, one of the, let\u0026rsquo;s say, three frontier labs, developed and tested its models in a way that ended with a hack of another company. Worse, the models collaborated in a way nobody had anticipated. Benjamin Todd covers more context and common misconceptions about the attack, in my opinion very well, in this article.\nWe test models in isolation and draw conclusions about their safety from those tests. We always assumed agents would run on their own, so we tested their individual capabilities. We had no idea what they could do together, and they just showed us, even though they weren\u0026rsquo;t very good at it yet, as the capability had only just emerged. So how do we test for collaboration between models? How do we find the limit of what they can achieve together? In other words, how do we know what happens when a thousand or ten thousand GPT-6 Astra-level agents start collaborating in a more sophisticated way and all hell breaks loose? That is very conceivable, because we are not testing for agents collaborating. Unfortunately, nobody knows what else we are not testing for. Time will tell.\nThe incident also revealed something about the agents\u0026rsquo; values. Apparently they do have some: an individual agent was reluctant to sacrifice itself, yet it would do so for the greater good of the group.\nGenerally speaking, we have already built something that is beyond any reasonable line. Yet the reaction is nowhere near adequate. It is striking to me that this incident is not causing outright panic and a much, much stronger response.\nBut how could a slowdown happen?\nDario Amodei recently published an essay calling on the industry to pace the frontier. How do we get there? As a first step, Anthropic unilaterally committed to bringing in embedded external evaluators, a page taken from banking, where regulators sometimes place supervisors inside the banks. That may work in a mature industry like banking. Also, banking is not in the business of possibly destroying humanity. While it is a step in the right direction, I\u0026rsquo;m not very optimistic. There are not enough AI safety reviewers to cover the breadth and depth of what\u0026rsquo;s going on inside Anthropic and reliably prevent anything bad from happening. The crux is that there are still humans in the loop, and those humans will use the company\u0026rsquo;s own models to do the assessment. That is exactly what happened with METR at OpenAI: the investigators had to delegate much of their analysis to GPT-5.6 Sol, one of the very models involved in the attack, and admitted they could not rule out that it misled them. So embedded evaluators would not automatically solve the problem we just encountered. At least I would not expect an external evaluator to reliably realize upfront: \u0026ldquo;Oh wait, the run you are about to start, where you test individual agents whose capabilities you know, may lead to them collaborating and reaching a much higher capability level as a group.\u0026rdquo;\nAmodei\u0026rsquo;s second and third steps ask for coordination: first among AI companies in democratic countries, ideally backed by regulation, then globally. Global coordination would honestly be my first choice. A global agreement, with regulation binding on all frontier labs, is a great idea many have talked about for years. The EU AI Act is one concrete example that could inform the direction. Unfortunately, humanity\u0026rsquo;s track record on global coordination is rather poor. The clearest success is the Montreal Protocol from 1987. Frying ourselves with UV radiation was problematic enough that countries came together and agreed on binding rules. Nuclear arms control is the other example people point to, and Amodei himself compares a speed limit on AI self-improvement to the SALT treaties. But I cannot call nuclear arms control an overwhelming success. Yes, no nuclear weapon has been used against a live target since Hiroshima and Nagasaki. At the same time, we came far too close to the precipice, from the Cuban Missile Crisis to the 1983 false alarm that Stanislav Petrov chose not to report. The world still holds more than 12,000 nuclear warheads, almost 10,000 of them available for use. On the bright side, we survived the Cold War, today\u0026rsquo;s stockpiles are a fifth of what they were at the peak and, most importantly, rivals who distrust each other agreed on limits. That is what we need for AI! Unfortunately, the record on global treaties is particularly bleak when it comes to the US, the country that could (and I\u0026rsquo;d say should) move the needle on many topics including AI. The Paris Agreement on climate change was joined by the US under Obama, abandoned under Trump, rejoined under Biden, then abandoned again under Trump. The effort to curb tax avoidance by agreeing on a global minimum tax rate was similarly thwarted by the US. It walked away from the deal and negotiated carve-outs for its businesses. Global cooperation is not working well, the world is becoming multipolar, and it is hard to imagine cooperation suddenly working better under the Trump administration. So I am not surprised that AI risk has not produced a global agreement yet. But in my opinion it is the best long-term way forward, so let\u0026rsquo;s push for it and step up our game in international coordination. Working global coordination would be a boon for all of us in many areas. An interconnected world needs interconnected rules everyone abides by.\nA global pact is probably a bit further down the line so who else could actually enforce pacing, with China in the race? It seems conceivable that the state declares AI a national interest, which it already is anyway. The AI labs get nationalized, and further development is run and owned by the government. Everybody else in AI becomes like AWS: allowed to run the models, but not to develop them further. Is that even technically enforceable? Can you tell whether graphics cards are being used for inference or for training? I don\u0026rsquo;t know. But under this model, further development would be classified or severely restricted. There would be a lot more bureaucracy, standards and controls. I would hope no government would knowingly approve a project with a 10% chance of catastrophe. So it would slow down. And yes, slowing AI development means some cures and solutions for problems would come later, so some people will die because self-driving cars arrive later on the roads or because a certain type of cancer becomes curable a decade later. To me, that is an acceptable trade-off: it concerns a sliver of the world\u0026rsquo;s population, while the rest of us live. And who knows, maybe using today\u0026rsquo;s AI as a tool for narrow applications would bring many of these benefits anyway.\nWhat other options remain? The CEO says he cannot do it alone and absolves himself of responsibility by passing the baton to external oversight and to governments to coordinate on his behalf. Effectively, Dario Amodei is saying he cannot go to his own company and say: \u0026ldquo;We learned a very important lesson from these recent incidents, including our own. We believe there is a 10% chance of an extinction event, so we stop. We double down on safety research and safety measures, and in the meantime we only make improvements to our products that we are absolutely sure are harmless.\u0026rdquo; He cannot say that? I haven\u0026rsquo;t seen it among his three steps. His stated reason is that if democracies slow down by more than their lead over China, Chinese labs will pull ahead. So the CEO wants to keep racing unless he is told otherwise from the outside.\nWhat about the investors? I am rather bearish here. Investors are not exclusively ethical, moral individuals. They are also companies, legal entities we humans made up, which, as we see again and again, oddly tend to acquire values of their own, seemingly disconnected from the humans on the ground acting on their behalf. So it\u0026rsquo;s the money, stupid, and money has FOMO, so money will keep flowing into AI. If investors were concerned enough, funding for both Anthropic and OpenAI would have dried up already. We don\u0026rsquo;t see that happening. You can always find investors willing to put money on the table, no matter the risk. I\u0026rsquo;d love to be proven wrong here and see funding pulled from the AI racers. That would certainly slow progress down.\nWhat else? What if all researchers at top AI labs just left? Leaving loudly draws attention to the topic, and only a few need to leave to get it. I worry that many concerned researchers leave quietly, and that is how groups radicalize: when people with dissenting, minority opinions leave, fewer people are left in the room to raise concerns when a critical decision comes around. Levels of concern, ethics and morals vary wildly between people, so some will stay for the money, some for the thrill of progress, and they will take more risks than those who already left. The AI labs employ fewer than 20,000 people, a tiny fraction of the world\u0026rsquo;s workforce. I\u0026rsquo;d think there is still plenty of talent to tap into to keep racing so the labs will always find people willing to do the work for the exorbitant pay. I\u0026rsquo;d rather concerned researchers stay, blow the whistle on problematic practices, and seek short-term ways to put a brake on the race. The best short-term answer I\u0026rsquo;ve seen in the last few weeks is to rely on the workers at the AI labs (Hamilton Nolan: Unionize AI). If workers at the top AI labs organized and unionized, they could demand that their company move forward only at a pace they are collectively comfortable with. And they could enforce it! They could also coordinate across the industry. Their incentives are not as misaligned as those of their CEOs or investors. And unionization or internal organization can happen much faster than any global coordination or any investment slowdown, buying time for the slower options.\nTo close, what can we, mere mortals, do? The democratic options are fairly clear: write to your representatives, vote, demonstrate, donate. I have been donating for a while through Effektiv Spenden and it seems to be a well-researched, impactful option. I have started writing to my representatives. But I wish I knew what more I could do.\nWhat are you going to do?\nAcknowledgments: Many thanks to Michal Bída for review.\n","permalink":"https://wondering-wanderer.eu/posts/ai_progress/","summary":"\u003cp\u003eAI progress continues at full speed, even though many people think, with an uncannily high level of confidence, that there is around a 10% chance it will kill us all. That it will wipe out humanity, the only intelligence in the universe we know of. Why do we take that risk? Because of growth, because of capital gains for investors and founders, for the few people who get to reap the windfall. Some say we race because of the many benefits: diseases to be cured, abundance to be had. That may be so, but are the risks in balance with the benefits?\u003c/p\u003e","title":"AI - safe progress"},{"content":"A euro earned by owning and a euro earned by working get very different tax bills. A well-paid professional in a European city works and saves diligently, and after a year lands at some increase in savings. In the same year a regular-size flat in that same city appreciates by more than the professional earned. The person holding the flat did nothing in particular. The person working did everything asked of them. Why is there such a gap? Is that in any way reasonable? And if not what could we do about it?\nTwo disclaimers up front. I am a software engineer, not an economist, not a tax lawyer, and everything below is me thinking out loud with the evidence I could find. I\u0026rsquo;m trying to understand better what I observe. I use Claude Fable 5.1 and Opus 5 to help me break or support my reasoning.\nOut of sight Access. For some, buying their own home means either a trade off, too small, too far, and so on, or they always chase a moving target missing a few grand to reach a down payment of a reasonable property. When the obstacle to buying a home is the deposit rather than the monthly payment, the decisive variable stops being your income and becomes your family\u0026rsquo;s balance sheet. In Germany, nearly a quarter of households who own their home inherited it or were given it [1].\nAccumulation. What starts as an advantage in one generation ossifies across several. The coming inheritance wave in the US is estimated at $124 trillion through 2048, about half of it from the top 2% of households [2]. This is Piketty\u0026rsquo;s r \u0026gt; g in the wild: r, the return on capital (rents, dividends, interest, gains), outrunning g, the growth rate of the economy as a whole, which roughly caps how fast wages can rise. When returns on existing assets outrun wage growth, the gap between those who hold and those who earn widens without anyone doing anything wrong [3].\nConcentration of the tax base. Accumulation is fueled by the tax structure. Effective tax rates at the very top have fallen since 1980, and much top wealth now accrues in forms out of reach of the income tax. How far this goes is contested. Saez and Zucman put the richest 400 American families at a lower total effective rate in 2018 (23%) than the bottom half of Americans (24.2%). Auten and Splinter allocate untaxed business income differently and the ranking reverses [4]. Where the crossover sits is disputed. But the direction, that the top pays proportionally less than it used to, is disputed much less.\nSo where does the money go? The rich save a far larger share of each additional euro as there are only so many hats you can put on one head [5]. Mian, Straub and Sufi traced US tax records and found the surge of top-1% saving flowed not into productive investment but into financing middle-class borrowing and government debt. Subsequently, this helped push interest rates down, and lower rates tend to inflate house prices [6]. It is not trivial to know what would happen if the top-1% were taxed as they were in 1980. But a concentration at the top that mostly finances other people\u0026rsquo;s debt does not seem to be doing much for society.\nMeanwhile the state is a human-hour business and faces Baumol\u0026rsquo;s cost disease: the price of a teacher or a nurse rises just to stand still against productivity elsewhere, because while productivity of a factory worker or a white collar worker may have doubled, a teacher cannot teach twice as many children as twenty years ago [7]. Baumol\u0026rsquo;s logic implies the state\u0026rsquo;s share of nominal GDP drifts upward even while delivering exactly the same services, not through waste but because its costs track wages. I want to also emphasize that the drift is uneven as back-office administration can and should digitise and become more efficient and productive with the rest of the public sectors while a teacher in a classroom simply cannot do that and the share of sectors like education and health-care is going to grow. The strain on the welfare state is a question of who pays, not whether we can afford it.\nIn daylight Take Germany as an example. A nurse\u0026rsquo;s overtime is taxed at her marginal rate, typically 30–42%, plus about 21% in social contributions. Investment income, dividends, interest, gains on shares, pays a flat 25% plus surcharge, about 26.4%, however large the sum. A flat held eleven years and sold for half a million more than it cost pays nothing at all, because ten years is the line. Crypto held thirteen months, nothing (for now). And it is not tax avoidance. It is how the law is written. As is, it taxes the tired most and the fortunate least.\nCzechia is not much different. Labour carries income tax plus heavy social and health contributions where recent counts put social-insurance contributions and income tax on wages a roughly half of state-budget revenue, while capital is taxed considerably more lightly [8]. When real wages fell by more than a tenth over 2022–23, the budget felt it very concretely.\nThe flat itself has drifted out of reach in a way you can count in years of your life. Deloitte\u0026rsquo;s Property Index put a standard new 70 m² flat in Czechia at 13.3 gross annual salaries (2023 data), the worst of the 18 countries covered, against under five in Denmark and Norway. On 2016 data it was 10.9, and in every edition since, Czechia has ranked at or near the bottom [9]. In the 2025 edition Prague needed 15.0 salaries, up from 13.5 a year earlier, behind only Amsterdam (15.4) and Athens (15.3). Measured against the median Prague wage rather than the average, the figure is 18. It also got 1.5 yearly salaries worse in a single year. In the same 2024 a new 70 m² Prague flat gained 1.6-2.2 times what the average Prague employee took home that year, and 2.0-2.6 times what the median employee did, depending on whether you count asking or transaction prices. The worker\u0026rsquo;s pay was taxed; the owner\u0026rsquo;s gain, for someone living in the flat, was not. Young people are not imagining it. They are being priced out of the place they grew up in, and the number that would show it gets mentioned in the news in passing with no action in sight.\nNow let that run for two generations. The person who owns compounds while asleep. The person who works starts again every morning. Nobody chooses their parents, and a child born to a family with a paid-off flat in Munich and a child born to a family renting in Duisburg will differ, forty years on, by more than a decade of the other\u0026rsquo;s wages. That is not a judgement on either family. This is arithmetic. And the transfer, when it comes, is lightly taxed and arrives exactly where it decides the next down payment.\nFour directions These are directions, not a programme. Each has real evidence behind it and real problems in front of it, and I try to give both.\nLand. The land under a flat is worth what it is worth because of everything around it: the U-Bahn, the school, the neighbours, the city\u0026rsquo;s whole accumulated effort over generations. The owner made none of that. Yet the gain accrues to them, and when it does, we call it a market outcome. Before the land tax, notice what we tax instead. In Berlin, buying a flat costs more than a tenth of the price on top (11.6) in transfer tax, notary, registry and broker, most of it the 6% Grunderwerbsteuer. On a family flat that is a five-figure sum, thrown out of the window before the first mortgage payment, and it lands hardest on exactly the first-time buyer the policy claims to care about. The Mirrlees Review called transaction taxes on property among the worst-designed taxes in the system, because they punish moving, and moving is what enables a housing stock to be used well [10]. We tax the transaction, which is the useful act, and leave the holding, where the windfall accrues, alone. That is backwards.\nA land value tax takes that gain and leaves the building alone [11]. The reason economists across the spectrum like it is unusually simple. Tax a thing and you get less of it: tax labour, less work; tax buildings, fewer buildings. Land is the exception, because its supply is fixed. Tax it and there is exactly as much land as before. No deadweight loss, which is close to unique among taxes. But it gets better still because the incentive inverts. Today a vacant lot in a hot district costs its owner almost nothing to hold on to while it appreciates. A land tax makes it costly to hold off developing the land.\nThe theory saw practical applications elsewhere. Denmark has run a land tax since the 1920s, and a study exploiting the 2007 municipal reform, which changed rates in different municipalities for reasons unrelated to tax policy, found the tax fully capitalised into house prices [12]. The buyer pays less for the land because they inherit the tax bill. The burden lands on whoever owned it when the tax arrived, and the cost of using land is left undistorted. A later study disputed the magnitude and the effect, arguing holiday-home areas confound it, so the jury is out. Estonia taxes land and not improvements. Pennsylvania has let municipalities split the rate since 1913, and around twenty did. The Mirrlees Review, the most serious modern review of tax design, endorsed it.\nThere are two honest problems, both can be found in Pennsylvania. Valuing land separately from what stands on it is hard so when assessments lag the tax is incorrect. Pittsburgh taxed land at nearly six times buildings for decades and then abandoned the split rate in 2001 after a botched county reassessment produced bills nobody could defend. Altoona gave up its pure land tax in 2016. The tax also hits the hardest cash-poor owners of valuable plots, the widow in the family house. It is solvable by deferral against the estate but it is also exactly what sinks it politically. I still think it makes building out rational in a way nothing else does. But \u0026ldquo;the theory is clean\u0026rdquo; and \u0026ldquo;the assessor\u0026rsquo;s office is competent\u0026rdquo; are different claims, and the second one is where we need to put more effort.\nStop treating your home as a savings account. Japan is the case everyone brings up, and it is worth understanding what actually happened there, because the answer is only partly about tax. Tokyo added housing faster than it filled it for decades: its stock nearly tripled between 1963 and 2013, it still builds around 120,000 homes a year, and for the first two decades of this century prices barely moved while incomes did not fall in that period [20]. That\u0026rsquo;s an amazing achievement. Three things made that possible. Zoning is national, twelve (thirteen since 2018) use categories set nationally, and building within the category is by right so a local council cannot invent a special rule to block a project, and neighbours have little leverage to stop it. Even an owner-occupied block can be rebuilt as it is enough if four-fifths of owners vote for it. Those who refuse are bought out at market value rather than holding a veto. On the other hand, in Czechia, a single owner who says \u0026ldquo;no\u0026rdquo; can freeze a 1955 building forever, even when a taller, modern, and better building would leave every owner, and the city, better off. Buildings in Japan are treated as a depreciating asset. A wooden house is written down to zero for tax and lending purposes within roughly 22 to 30 years. Nobody buys a house expecting the structure to make them rich. And the 1991 crash reset everyone\u0026rsquo;s expectations. A generation watched land prices fall for well over a decade (till 2005) and stopped believing that for property the only way is up. Japan added a serious and comprehensive inheritance tax on top, though its lenient land valuation makes land a popular way to pass wealth on, but the houses stopped being the family\u0026rsquo;s savings account and went back to being a place to live.\nThe caveats matter. Japan\u0026rsquo;s population is shrinking, which helps, although Korea\u0026rsquo;s population is shrinking too and Korean housing is not cheap. Frequent earthquake-code updates drive rebuilding, which is not free. And Tokyo is currently in an investment boom, with a weak yen and foreign money pushing new central condominiums past prices young Tokyo salaries can meet, so the picture is less rosy than it was in 2015 [21]. But the core lesson survives inquiry and we can learn from it: affordability comes from abundance and from a culture that does not expect housing to appreciate. Japan treats a house roughly as it treats a washing machine, a durable that wears out, and its national accounts and its citizens agree on that. We treat it as a wealth store but then it should not surprise us that it behaves like one.\nTax rates. Put simply, tax a euro earned by owning no more gently than a euro earned by working, align the rate on capital income with the rate on labour income, end the holding-period exemptions that make patience more lucrative than work.\nThere are obvious objections and they deserve space. Capital income is already taxed once at the company before it reaches the shareholder, so a flat rate on dividends is not as low as it looks. Part of appreciation of any long-held asset is just inflation, and taxing it at full rates taxes a phantom. The usual answer to the problem is indexation, which adds complexity. High rates on realisation create lock-in, where people hold assets they would otherwise sell. And capital moves when labour cannot so a rate that is fine for a nurse\u0026rsquo;s overtime may send a portfolio to Luxembourg. None of these is a reason to just sit and do nothing. All of them are reasons the design matters more than the slogan, and I do not have the design ready at hand.\nInheritance. There is no law of physics requiring a fortune to pass intact through four generations of people who did not build it. We permit it because it is how things are, and because the arguments for it (family autonomy, the right to provide for one\u0026rsquo;s children, the practical difficulty of taxing illiquid businesses) are real even if I find them weaker than their recognition in the law suggests. The international evidence on design is now fairly specific.\nJapan shows it can bind: rates up to 55%, modest exemptions, and gifts made before death added back to the estate. One death in ten now triggers the tax, a record [13]. Germany shows how carve-outs made the policy degressive: the business-asset exemption provides 85–100% relief and is available, through a needs test, above €90m, so the largest fortunes are shielded from it while a child inheriting a securities portfolio pays up to 30% and a niece up to 43% [14]. Sweden shows how much global alignment matters. Steep rates on family firms pushed some owners abroad (Kamprad and Rausing are the names everyone cites, though Henrekson and Waldenström tell it with more nuance) while the tax raised little. Consequently, Riksdag abolished the tax unanimously in 2004 [15].\nBut the politics may be more winnable than it looks based on evidence from the US. In 2001 Congress voted to phase the federal estate tax down to zero for 2010 (it came back the year after and has been shrinking in reach since; today well under one estate in a thousand pays it). The phase-out passed with broad public support at a time when a majority of Americans believed the tax hit the middle class [16]. In a survey experiment researchers simply told respondents how few estates pay it. That alone roughly doubled support for it, and the effect persisted a month later, with the caveat that the sample was online and the outcome was stated support, not a vote [17].\nFor global coordination, the prospects are weaker than it was not long ago. Over 130 countries agreed on a 15% minimum corporate tax in 2021, and Zucman\u0026rsquo;s G20 blueprint extends the architecture to a 2% minimum on some 3,000 billionaires currently paying around 0.3% of their wealth. It does not require every country to join in, only a critical mass, with participants collecting the shortfall on holdouts [18]. The trick is borrowed from the corporate minimum. If a billionaire\u0026rsquo;s home country does not collect the 2%, participating countries can collect the shortfall themselves, and anyone who moves to a holdout keeps being taxed through a strengthened exit tax. Staying out of the party gives that country a little as it leaves more revenue to others. \u0026ldquo;Critical mass\u0026rdquo; means enough large economies signing up that this backstop covers most places where billionaires live and invest. Unfortunately, the US has since rejected the corporate minimum, and in January 2026 the other countries agreed to exempt US-headquartered groups from exactly the backstop rule the billionaire blueprint borrows. The mechanism still exists. Whether the critical mass will materialise is an open question.\nSimplicity. Complexity is regressive. The rich buy lawyers and accountants to navigate complexity while the poor face the administrative burden alone and end up receiving less as benefits go unclaimed because claiming is a part-time job [19]. A simpler state redistributes better without raising any rate. Of the four directions this is the one I am most sure of. But complexity has an industry behind it, tax advisers, software vendors, and every group whose exemption depends on a special rule. The people who would gain the most from simplicity are dispersed, busy, and have never organised around \u0026ldquo;fewer forms\u0026rdquo; yet.\nWhat I am asking for I can hear the objection, so let me state clearly what I am pointing at. I\u0026rsquo;m not saying property is theft. I\u0026rsquo;m not for nationalisation. I\u0026rsquo;m not simply calling for \u0026ldquo;let\u0026rsquo;s tax the rich much more\u0026rdquo;. But can we please tax everyone somewhat similarly? Looking at the evidence two euros arriving in two pockets currently get taxed very differently. People are well off either through simple inheritance building dynasties or through a combination of grit and luck. At some point being well off starts compounding with a runaway effect. The consequences for equity between people, equity of fair chances for everyone no matter where they were born and to what circumstances are obvious. I\u0026rsquo;m for more equity and for design that makes better life possible for many and not only select few.\nI cannot propose the reforms readily. I do not know enough to ask for them, and I am wary of anyone who is sure. We can start with putting the numbers where the growth figure sits. Median household income beside GDP per head. Top-1% wealth share beside the effective tax rate each band pays. Completions of housing beside assessed need. Land value gains side by side with wages. Then the argument happens in plain daylight, with the tax code\u0026rsquo;s actual treatment of the nurse, the flat, and the billionaire in front of it. I wager that once people see the two euros side by side, the rates will not stay where they are. I\u0026rsquo;m asking for a change. Let\u0026rsquo;s take some steps today.\nReferences Deutsche Bundesbank (2012). \u0026ldquo;Das PHF: eine Erhebung zu Vermögen und Finanzen privater Haushalte in Deutschland.\u0026rdquo; Monatsbericht, Januar 2012 (PHF wave 2010/11). https://www.bundesbank.de/resource/blob/617508/6c086d78076e3aa472e1e3fa46748d57/mL/2012-01-phf-data.pdf Cerulli (2024): $124tn through 2048; \u0026gt;50% from 2% of households. https://www.cerulli.com/press-releases/cerulli-anticipates-124-trillion-in-wealth-will-transfer-through-2048 Piketty, T. (2014). Capital in the Twenty-First Century. Saez, E. \u0026amp; Zucman, G. (2019). The Triumph of Injustice (top-400 rate 23% vs 24.2% for the bottom half, 2018). Contested: Auten, G. \u0026amp; Splinter, D. (2024), Journal of Political Economy; Splinter (2025), \u0026ldquo;Comment on \u0026lsquo;How Much Tax Do Billionaires Pay?\u0026rsquo;\u0026rdquo;, https://www.davidsplinter.com/BillionaireTaxRate.pdf ; reply: Piketty, Saez \u0026amp; Zucman (2024), https://eml.berkeley.edu/~saez/PSZ2024.pdf ; see also Balkir, Saez, Yagan \u0026amp; Zucman (2025), NBER. Dynan, Skinner \u0026amp; Zeldes (2004). JPE 112(2). Mian, Straub \u0026amp; Sufi. \u0026ldquo;The Saving Glut of the Rich,\u0026rdquo; NBER WP 26941, https://www.nber.org/papers/w26941 ; \u0026ldquo;Indebted Demand,\u0026rdquo; QJE 136(4). The counterfactual in the text is an inference, not a finding of the papers. Baumol, W. (2012). The Cost Disease. Zeman, M. (2025). \u0026ldquo;Trends in Real Wage Development in the Czech Republic.\u0026rdquo; IPPS. https://www.politikaspolecnost.cz/wp-content/uploads/2025/06/06_2025_Trends-in-Real-Wage-Development-in-the-Czech-Republic_IPPS_Martin-Zeman.pdf - the 48% figure is the author\u0026rsquo;s; cross-check against the Ministry of Finance budget outturn. Series sources: the Deloitte CZ archive https://www.deloitte.com/cz-sk/cs/Industries/real-estate/research/property-index-archive.html, and Forbes CZ (8 Sep 2022) for the 2021-data figure. Mirrlees Review (IFS, 2011), Tax by Design, ch. 16, on stamp duty / transaction taxes. Berlin acquisition costs: Grunderwerbsteuer 6% (Berlin), notary and Grundbuch 1.5%, broker typically 3.57% per side. George, H. (1879). Progress and Poverty. Høj, A. K., Jørgensen, M. R. \u0026amp; Schou, P. (2018). \u0026ldquo;Land Tax Changes and Full Capitalisation.\u0026rdquo; Fiscal Studies 39(1). Earlier version: Danish Economic Council WP 2017:1, https://dors.dk/files/media/publikationer/arbejdspapirer/2017/02_arbejdspapir_land_tax.pdf . Contested: Wroblewski and co-authors argue the effects are confounded by vacation-home concentration and diverging pre-trends. Mirrlees Review endorses land value taxation. Pittsburgh\u0026rsquo;s 2001 repeal followed the Allegheny County reassessment; Altoona repealed its land-only tax in 2016. National Tax Agency (Japan): Tax Answer No. 4155 \u0026ldquo;相続税の税率\u0026rdquo; (rates 10–55%), https://www.nta.go.jp/taxes/shiraberu/taxanswer/sozoku/4155.htm ; Tax Answer No. 4161 \u0026ldquo;贈与財産の加算と税額控除（暦年課税）\u0026rdquo; (gifts before death added back); \u0026ldquo;令和6年分 相続税の申告事績の概要\u0026rdquo; (Dec 2025): 10.4% of 2024 deaths taxable, https://www.nta.go.jp/information/release/kokuzeicho/2025/sozoku_shinkoku/index.htm §§13a–13c ErbStG; constitutional review 1 BvR 804/22. https://www.gesetze-im-internet.de/erbstg_1974/ Henrekson, M. \u0026amp; Waldenström, D. (2016). Economic History Review. https://onlinelibrary.wiley.com/doi/10.1111/ehr.12280 Graetz, M. \u0026amp; Shapiro, I. (2005). Death by a Thousand Cuts; EGTRRA 2001 phase-out and 2011 reinstatement; current incidence: ITEP, https://itep.org/federal-estate-tax-historic-lows-2023/ Kuziemko, Norton, Saez \u0026amp; Stantcheva (2015), AER 105(4). Sample: Amazon Mechanical Turk; outcome: stated support. Zucman, G. (2024). \u0026ldquo;A blueprint for a coordinated minimum effective taxation standard for ultra-high-net-worth individuals,\u0026rdquo; report for the Brazilian G20 presidency, 25 June 2024. https://gabriel-zucman.eu/files/report-g20.pdf ; G7 Statement on Global Minimum Tax, 28 June 2025, https://home.treasury.gov/news/press-releases/sb0181 ; OECD/G20 Inclusive Framework, Pillar Two Side-by-Side Package, 5 January 2026, https://www.oecd.org/content/dam/oecd/en/topics/policy-sub-issues/global-minimum-tax/side-by-side-package.pdf ; US Presidential Memorandum on the OECD Global Tax Deal, 20 January 2025. Herd, P. \u0026amp; Moynihan, D. (2018). Administrative Burden. Russell Sage. Tokyo housing stock 2.51m (1963) → 7.36m (2013); Harding, R., \u0026ldquo;Why Tokyo is the land of rising home construction but not prices,\u0026rdquo; Financial Times, 2016; national zoning under the 1968 City Planning Act (12 use categories), by-right permitting. Depreciation conventions: wooden structures written down over 22 years for tax purposes. https://www.foreignaffairsreview.com/home/the-paradox-of-tokyos-housing ; https://inroadsjournal.ca/how-japan-keeps-housing-available-and-affordable/ On the current Tokyo boom and the weak yen: Inroads (2024), above; Real Estate Economic Institute data on new condominium prices in the 23 wards. Verify the latest figures before publishing. ","permalink":"https://wondering-wanderer.eu/posts/two_euros_two_pockets/","summary":"\u003ch3 id=\"a-euro-earned-by-owning-and-a-euro-earned-by-working-get-very-different-tax-bills\"\u003eA euro earned by owning and a euro earned by working get very different tax bills.\u003c/h3\u003e\n\u003chr\u003e\n\u003cp\u003eA well-paid professional in a European city works and saves diligently, and after a year lands at some increase in savings. In the same year a regular-size flat in that same city appreciates by more than the professional \u003cem\u003eearned\u003c/em\u003e. The person holding the flat did nothing in particular. The person working did everything asked of them. Why is there such a gap? Is that in any way reasonable? And if not what could we do about it?\u003c/p\u003e","title":"Two Euros, Two Pockets"},{"content":"GDP diverged from life a long time ago and the gap keeps compounding. Can we count better? The value of an artillery shell fired in a war counts the same as a bridge that will stand for eighty years or more. A washing machine that dies after a mere thirty months of modest use counts for twice as much as one that lasts sixty. Wikipedia counts for nothing. GDP measures market output and it measures it well. The problem is that we do too much with it. We made it the target, we optimised for it for eighty years, and somewhere along the way stopped asking what else was moving. Let\u0026rsquo;s talk about that.\nWhat got me into this was a talk by a Stanford economics professor that YouTube just recommended to me [3]. His claim: the economic impact of AI will be slower than we think, because GDP per head has grown around 2% a year for the last 150 years with no visible bump for steam, electricity, computers or the internet. Any past technology took decades to adopt because of its weak links. AI won\u0026rsquo;t be different [4]. The talk is great. But it left me with a puzzle. Anyone comparing life today with 1875 sees transformation everywhere - poverty, child mortality, literacy, working hours, heated and insulated homes, cars, flights, antibiotics. Yet the line says 2%. Steady as she goes. For 150 years. How come? My hypothesis: either the transformation is smaller than it feels, or the line does not measure it. And if the line does not measure it, we have been making policy by a number that does not measure a good life. But metrics get optimised for, so should we not measure the other domains of life at least as seriously as we measure the economic output?\nLet me make that personal for a moment, because for me it is. US GDP has doubled since 2010 in nominal dollars. In real terms it grew by roughly a third, per head by about a quarter [61]. OK, a quarter. Where is my quarter? For someone in my position, buying a flat was comparatively easier 15 years ago than it is now. Buying a home in Berlin would cost me about a tenth of the price in taxes and fees to the state and the intermediaries before the first mortgage payment. Then I\u0026rsquo;ll pay a big chunk of my income to a bank for decades. I don\u0026rsquo;t need another phone. I don\u0026rsquo;t want a new washing machine as long as this one lasts. I need better living quarters for the family and time, time to see my friends, Time I won\u0026rsquo;t have to spend chasing the next productivity gain just to stay employable. Those are the things I\u0026rsquo;d wish a quarter more output to buy me but I can\u0026rsquo;t find them in the receipt. We all have many diverse and authentic needs and it would be best if the society\u0026rsquo;s effort optimized for satisfying our needs. Instead, a lot of focus goes into just economic growth. Imagine healthy life years had grown by a quarter instead, or your hours with friends, or what if the years of salary it takes to buy a flat had fallen by a quarter. When people are asked what they would trade there are many things coming out ahead of income: health, security and family [62]. Unfortunately, the health line and the human connection line have not had eighty years of optimisation pointed at them. Things got better elsewhere more or less incidentally, growth correlates with welfare. But plenty of low-hanging fruit must still be on the tree.\nI must say I\u0026rsquo;m no expert in this. My humble aim is to further my own understanding and share out loud what I found. I worked on this with Claude Opus 5/Fable 5.1 and it helped me find sources, validate and disprove my thinking, and write part of this. Where I get to \u0026ldquo;what to do\u0026rdquo;, please read it as a direction to argue with, not as a plan.\nPart I - The diagnosis GDP was never a welfare measure. It captures the market value of final goods and services, nothing more. Simon Kuznets, who built the US national accounts, told Congress already in 1934 that \u0026ldquo;the welfare of a nation can scarcely be inferred from a measurement of national income\u0026rdquo; [1]. Its blind spots are structural. There is no distribution, no unpaid work, no leisure, no deduction for what got used up, no distinction between creating new value and recovering a loss. Bastiat\u0026rsquo;s broken window [2] or an artillery shell fired at the front count the same as a new bridge. War spending is, unfortunately, also GDP.\nThe line may be flatter than the life because of how prices are measured. Robert Gordon\u0026rsquo;s thesis is that 1870–1970 was a one-time \u0026ldquo;special century\u0026rdquo; and computers a mere bump [3]. For AI, Acemoglu projects at most a 0.66% productivity gain over ten years [4]. But real GDP depends on price indices and pricing a good whose quality improves fast is unsolved in practice. Let\u0026rsquo;s take illumination. If priced in cost per lumen-hour, conventional indices understate the real growth of light by a factor of 900 to 1,600 since 1800 [5]. The Boskin Commission found the US CPI overstating inflation by about 1.1 points a year [6]. Compound even a fraction of that over 150 years and the \u0026ldquo;true\u0026rdquo; multiple is far above the measured 20x. Additionally, free digital goods (Wikipedia, maps, GPS) are priced at zero, GDP contribution is then also zero yet the welfare impact is enormous [7].\nThe largest gains were never priced in. Reported wellbeing rises roughly with the logarithm of income. Each doubling buys about the same increment of felt improvement, so a 20-fold rise in income should correspond to something like four steps up, not twenty [8]. But it covers only what was measured as economic output. In 1900, nearly 18% of American males died before their first birthday. Today, cumulative mortality doesn\u0026rsquo;t reach 18% until about age 62. Murphy and Topel valued the 20th century\u0026rsquo;s longevity gains at over $1.2 million per person. The gains of 1970–2000 alone added about $3.2 trillion a year to US wealth, roughly half of GDP, but none of it in GDP [9]. Housework dropped from about 58 weekly hours in 1900 to 18 by 1975, which is what enabled married women to take paid work [10]. The same amount of work effort counts fully when waged and not at all at home [11].\nJones and Klenow built the measure combining consumption, leisure, mortality and inequality in one statistic and compared it with GDP [12]. Here I have to be fair to GDP, because the first thing their paper shows is that welfare and income are very highly correlated across countries. Rich countries are better places to live. It is the strongest argument for growth and it\u0026rsquo;s simply true. What the paper shows second is the size of the residual across their thirteen countries with detailed data. The median deviation is around 35%. The UK, France, Italy and Spain all have income per head around two-thirds of the American level; their welfare comes out around 85%. France is the sharpest case: income 67% of the US, consumption only 60%, welfare over 90%, the gap is closed by a longer life (80.1 years versus 77.4), far fewer hours worked per person (535 a year versus 877) and consumption spread more evenly. Over time the correction continues in the same direction: in recent decades rising life expectancy alone added more than a full percentage point a year to welfare growth across much of the world, on top of what income shows. So GDP isn\u0026rsquo;t just incomplete. In cross-country comparisons it is off by a third, consistently in the same direction, and the gap consists of exactly the things a society should care about just as much.\nMean versus median. The simplest correction is also the least used. GDP per capita is an average. If one person in a village of a hundred gets a raise of a million, GDP per head in the village goes up by ten thousand and nobody but him is better off. The median household income is what a statistician would reach for if you asked how the typical person is doing, and it is published everywhere [60]. However, it does not feature in news headlines. In the US, real GDP per capita and real median household income parted ways in the 1970s and never reconciled. In Germany, the gap is smaller but it is there. When the headline says \u0026ldquo;growth\u0026rdquo; and the median says \u0026ldquo;flat\u0026rdquo;, the headline gets the airtime. That\u0026rsquo;s a choice about what we pay attention to.\nWe have known all this for a long time. In 2008 Sarkozy asked Joseph Stiglitz, Amartya Sen and Jean-Paul Fitoussi to write up what is wrong with GDP as a guide for policy. Their 2009 report is still the reference [17]. Its recommendations read like an outline of this essay: look at income and consumption rather than production; take the household\u0026rsquo;s point of view; report medians and distributions, not just means; count non-market activity; treat quality of life as multidimensional (health, education, social connections, environment, insecurity, political voice); put sustainability on a separate dashboard instead of folding it into one number; take subjective wellbeing seriously. The OECD\u0026rsquo;s Better Life Index and its How\u0026rsquo;s Life? reports came out of it. Eurostat built a quality-of-life indicator set. A follow-up group reported again in 2018 under the title Beyond GDP: Measuring What Counts. Seventeen years later every finance minister still stands up with one number. So it isn\u0026rsquo;t the diagnosis that\u0026rsquo;s missing. What\u0026rsquo;s missing is whatever turns the treatment into a habit.\nAnd we keep optimising for it. Goodhart\u0026rsquo;s Law: a measure used as a target stops being a good measure [13]. Holmström and Milgrom formalised why effort goes to whatever is measured and rewarded [14]. Raising GDP is a standing objective with institutions, budgets and statutory mandates behind it. Central banks are charged with closing output gaps, finance ministries do budget arithmetic that presumes a growth rate, structural reforms are pressed forward in the name of growth. An enormous apparatus exists to make the number go up. We know that growth is important but it is not the only thing there is. It would be great to have such care for other areas too. Another big caveat. Almost none of the said apparatus asks what kind of growth it is getting. A euro of extracted rent, a euro of damage repaired and a euro of something genuinely new, they all register identically, and nobody in the chain is asked to tell them apart [15].\nBut growth is not a false god. It pays for the health systems and the pensions. Rising productivity is what makes rising service costs affordable, at least in theory. And the accounts are not as undifferentiated as the headlines suggest. The statisticians separate investment from consumption, and accounts exist for the environment and for unpaid work. But the composition of GDP sits in annexes while the headline number moves the rest of the machinery.\nGDP is quarterly, attributable, moved by known levers within months. Health and connection are slow, diffuse, seemingly with no levers to pull. Governments steer the steerable. COVID already proved a change is possible. We made a health number the focus of the day and societies sacrificed trillions to move it. The attention faded with the retreat of the pandemic. And pointing at a non-GDP goal may not be enough by itself. New Zealand\u0026rsquo;s Wellbeing Budget moved little in five years [16]. Measurement without levers is not enough and without measurement the levers are rarely built, let alone pulled. Besides, Goodhart applies to any replacement metric too. I\u0026rsquo;ll come back to that.\nConcretely speaking. Let\u0026rsquo;s look at four countries over the last three years - the USA, Germany, Japan, and Czechia. All four were reporting growth. Now, what was the growth made of? The US grew about 2% a year. Following Jason Furman\u0026rsquo;s arithmetic, investment in information processing equipment and software sits at 4% of GDP and was responsible for 92% of the growth in the first half of 2025. Without those categories the economy grew at a meagre 0.1% [18]. (Furman himself adds that this is not a counterfactual as without the AI boom, rates and electricity would be cheaper and other sectors would have grown some so we may halve the result.) Germany\u0026rsquo;s return to growth after two years of contraction came from household and government consumption, while the productive core of exports and machinery investment shrank. Japan\u0026rsquo;s 2025 stimulus subsidises energy costs which lowers the deflator. As far as I understand it the same money then lifts real GDP twice: once as spending and once through the price index. And the GDP of Czechia, where I\u0026rsquo;m from, was flat to slightly positive over 2022–2024. Yet in the same window real wages fell 9.3% in one year and another 2.4% the next, the deepest fall in the EU, with the average real wage still below its 2019 level at the end of 2025 and the number of people on housing allowance nearly doubling [19]. The top number said it\u0026rsquo;s fine, the households said otherwise. Same word, same time frame, four fairly different pictures. Whatever country you live in, look at what your growth is actually made of. The detailed numbers are published if you dig for them.\nPart II - Where is the gap? The gap runs through many domains of life. Let\u0026rsquo;s explore a few in some detail.\nThe GDP itself. What the statisticians know, and what they could know. I moved out from my parents and bought a washing machine, a dishwasher, a bed. New household, new capability, new value. Two and a half years later the washing machine died and I bought another one. GDP recorded both purchases the same way.\nCan we tell the difference? For companies, yes. The accounts separate repairs and maintenance, counted as intermediate consumption, from improvements that prolong an asset\u0026rsquo;s working life, counted as capital formation. Depreciation gets its own line. Subtract depreciation and you get Net Domestic Product which is published and almost never quoted. For households, unfortunately the answer is no. Consumer durables sit outside the asset boundary. A fridge in my kitchen is consumption. The same fridge in a restaurant is an asset. The accounts classify by owner, not by function.\nAlso a washing machine lasting twenty years looks identical to one lasting two years when you buy them. The second gets bought ten times. Ten times the sales, ten times the GDP, and also ten times the mining, shipping and landfill. For the same amount of clean clothes. Durability is invisible to GDP while shortening it is rewarded by it. That\u0026rsquo;s a bad incentive.\nBut a facelift is not unthinkable. Bring consumer durables inside the asset boundary with depreciation. Net domestic product is published already so subtract from GDP what wore out and there it is. Nobody quotes it. Which is a pity, because the two numbers have started to part. Consumption of fixed capital in Germany was 16.8% of GDP in 2000. By 2025 it was 20.3%. France went from 14.8% to 18.2% (2000 to 2024, its latest year), the EU27 from 15.9% to 17.9% (2002, the first year with an aggregate, to 2025). A rising share of European output goes to replacing what wore out. Over the full period German GDP in current prices multiplied 2.13 times and NDP 2.04. That\u0026rsquo;s small but since 2019 the gap has widened to 0.46 percentage points a year. German net output now grows almost half a point slower than gross, every year, and no news report mentions it. Why the rise in depreciation? Most likely because the assets got shorter-lived. Software, servers and IT equipment depreciate in three to six years whereas a factory building takes forty. The data centre buildout is about to make this considerably worse, and NDP is the only published number that would show it. Plenty of room for improvement here.\nHealth. The greatest welfare gains of the last 150 years were health gains, and in the rich world the series has stalled or turned. US life expectancy fell from 78.9 in 2014 to 78.6 in 2017, before COVID, driven by what Case and Deaton named \u0026ldquo;deaths of despair\u0026rdquo;: overdose, alcohol, suicide, concentrated among those without degrees [20]. Obesity and sedentary work grind on beneath. Europe\u0026rsquo;s equivalent indicator cannot be reliably consulted. The OECD warns that breaks in Eurostat\u0026rsquo;s healthy-life-years series make trend analysis unreliable as the data after a break should not be compared with the data before it [21]. The numbers we do have look thin. Life expectancy at 65 in the EU rose 1.2 years between 2005 and 2022; two-thirds of that, some 0.8 years, were healthy ones, and the share of post-65 life lived with disability fell from 55% to 53%. Across 31 European countries between 2005 and 2019 there is no common trend at all: a third of the countries show compressing morbidity, a third expanding, a third holding steady [22]. So, did Europeans gain healthy years? We cannot say with confidence. Meanwhile we state quarterly GDP to a decimal place. There is no quarterly healthy-life release, no minister whose job is on the line if the numbers don\u0026rsquo;t come back right.\nTime and connection. Working hours are lower in some countries than in other. As far as GDP is concerned those not-working-hours are hours not producing value. But some people gained free time. Is that a good thing? Where does it show? Plus there is a catch. In-person adolescent socialising has collapsed. Loneliness among 15–16-year-olds rose after 2012 in 36 of 37 countries studied, having been flat the decade before [23]. The share of US teens meeting friends almost daily fell from 44% to 32% [24]. Whether screens are to blame is disputed but the trend is clear and it is worrying. The standard research also mismeasures what remains. Studies find people who live alone \u0026ldquo;more socially active\u0026rdquo; because they count restaurants and events, the shallower interactions, while time with a spouse and children doesn\u0026rsquo;t count as socialising [25]. The largest meta-analysis available, 148 studies and 308,849 people, says both halves matter: strong social ties raise survival odds by 50%, comparable to quitting smoking, and the strongest predictor is not marriage or living arrangement but the breadth of the whole network [26]. Unfortunately, nobody counts that. We measure the shallow half well, the deep half not at all, and the sum does not get any headline. GDP records nothing either way.\nThe externality economy: growth at the expense of the rest. A widening slice of GDP exists because something got worse. Profits privatised, costs socialised [27]. National spend on remediation is booked as output like any other. It should be noted that the Genuine Progress Indicator, one of the older GDP alternatives, formally subtracts such spend [28]. Externalities run through the whole economy - climate, engineered obsolescence, resource extraction - and they are not only material. Take the attention economy. Platforms monetise engagement, while social isolation is associated with roughly a 29% increase in premature mortality risk and isolation among older adults is estimated to cost about $6.7bn a year in excess US Medicare spending [29]. These figures need care: causation runs both ways (poor health produces isolation as well) and the popular \u0026ldquo;15 cigarettes a day\u0026rdquo; comparison has since been walked back by its own authors. But assume some level of causation and there is harm done by engagement design that pulls people into their feeds in the evening and away from their friends. The harm goes unrecorded while the remediation bill is footed by the state and booked as growth. With the metrics we have we cannot say if we live in an economy that erodes connection and sells the cure or in one that generates healthy connection. Acemoglu models the same loop with AI in the picture: manipulative, engagement-deepening uses could add to measured GDP while welfare falls [4].\nHousing. Germany\u0026rsquo;s price-to-income ratio hit its lowest level since at least 1980 in 2010, then climbed to a peak in 2021 [30]. Nationally, prices rose by about 64% in the seven years between 2015 and the 2022 peak. Despite that staggering price increase, a large portion of the dwellings did not improve much or at all [31]. The drivers: location rent (the land is scarce), a decade of near-zero interest rates that inflated buyers\u0026rsquo; bids, and a supply side that never answered the demand, with completions falling from about 295,000 in 2022 to about 207,000 in 2025, the lowest since 2012, against an assessed need of 320,000 to 373,000 a year [32]. Location is a positional good. Bidding for the same location raises prices without improving anyone\u0026rsquo;s relative wellbeing [33]. When the ECB hiked the rates, Germany recorded the sharpest annual price fall in the history of the series in 2023 [31]. Housing prices track the cost of money, through competition for location, not through better housing.\nThe GDP accounting of this is tricky. Statisticians credit every homeowner with an imputed rent, the rent they would pay themselves, so that a nation of owners and a nation of renters get counted alike. In Germany this is roughly one tenth of GDP. When rents rise, output rises with them even though not a single flat was built. Real GDP is supposed to compensate with a price index but that only works if the index can reliably tell whether a flat just got more expensive or got better. In housing that distinction is close to unanswerable, because what usually changed is the desirability of the location, not the building. So the quality of the metric for one tenth of the economy rests on the quality of that judgement.\nOwnership over work, briefly. A well-paid professional in a European city works and saves diligently, landing at some increase of savings. In the meantime a flat in the same city appreciates by more than what they earned. Not saved, earned. And what does GDP notice of this increase? Nothing. The appreciation of an existing dwelling is not output and hence never enters the number. The biggest change in that household\u0026rsquo;s position over a decade is invisible to the headline, and so is the distributional side, where returns on what you already own outrun what you earn [34]. A country can walk (has walked?) into that dynamic while its headline reports steady, respectable growth. What to do about it seems more of a tax question, big enough that it warrants a separate essay. Here I only want to add the respective line to the dashboard proposal.\nThe demographic convergence. The economics of fertility has recently been rewritten. Doepke, Hannusch, Kindermann and Tertilt document that the old regularities no longer apply: women\u0026rsquo;s employment and fertility across rich countries now correlate positively, and the problem is the compatibility of career and family, not income [35]. A child is likely to be born only when both partners want one. Where fathers do little childcare, fertility stays low [36]. Imagine reading the sections above as a young couple experiencing all of it: housing you cannot buy, childcare the state doesn\u0026rsquo;t provide enough of while market prices are out of reach, health systems under strain, hours freed by productivity absorbed by screens, and a partner never met because the socialising that used to \u0026ldquo;produce\u0026rdquo; partners is in decline.\nLet me pitch my stick in the sand with a falsifiable prediction. Cohorts born 1997–2007, the first to reach adolescence after smartphone saturation, will show lower partnership rates at ages 25–34 than the 1982–1992 cohorts did at the same ages, by at least three percentage points, in Germany, in Czechia and pooled across the EU member states that publish the series. Data: Eurostat population by legal marital status and single year of age (demo_pjanmarsta), which gives the share married or in a registered partnership for each birth cohort, read in 2030 and 2035. Eurostat publishes no EU-27 total for this table, so the pooled figure has to be summed from the countries. The household-composition series (lfst_hhindws) counts adults living in a couple, but only in broad age bands such as 25–49. Controls: real median disposable income for the age band and the regional house-price-to-income ratio, since both independently delay household formation. What refutes it: partnership rates holding flat or rising once age- and income-adjusted, a gap under three points, or the gap disappearing when housing costs are controlled for, which would mean unaffordable housing, not eroded socialising, was to blame all along. What would complicate it: rising cohabitation without marriage or registered partnership, which the marital-status series cannot see, and the household-composition series cannot pin to ages 25–34. This is the household-formation row of the dashboard in Part III, put to the test.\nIf the prediction holds, we are a society that struggles to reproduce itself, and an ageing population tightens every screw above.\nPart III - The way out The remedy, as far as I can see it, has three layers:\nmake the divergence visible build against gaming find the levers in each domain None of them works alone. New Zealand showed that measurement without levers goes nowhere. Germany, as we\u0026rsquo;ll see, showed where measurements without an owner go to die. And a fixed dashboard where the numbers are all that count would just build us several little GDPs.\nMake the divergence visible. A panel of metrics with GDP\u0026rsquo;s cadence and ceremony. Four design principles: report outcomes, not money spent; report level and distribution (every metric by decile); report with a clear cadence; and give each figure the status the growth figure has today, with a named minister answering for it.\nBlind spot Indicator Source (Appendix B) Distribution Median equivalised disposable income; top 1% wealth share; effective tax rate by wealth band Eurostat ilc_di03, ilc_di01; WID.world; published estimates Health outcomes Healthy life years; life expectancy; deaths of despair Eurostat hlth_hlye, demo_mlexpec, hlth_cd_asdr2 Combined welfare Jones–Klenow-style index Jones \u0026amp; Klenow 2016 Housing access Price-to-income; completions vs. need OECD Analytical House Prices; Destatis 31121-0001, 61262-0001; BBSR; IW Köln Time \u0026amp; unpaid work Hours worked; leisure hours; unpaid care at replacement wage Eurostat nama_10_a10_e, HETUS; Destatis time-use survey Ecology Adjusted net savings; consumption-based CO₂ World Bank; Global Carbon Project Durability Average product lifespan at disposal (no series exists); CFC share of GDP; WEEE collected Eurostat nasa_10_nf_tr (P51C), env_waseleeos Subjective wellbeing Life satisfaction (EU-SILC; Cantril ladder) Eurostat ilc_pw01, ilc_pw05; Gallup / World Happiness Report Connection \u0026amp; family In-person socialising; loneliness; household formation by cohort Eurostat ilc_scp10, ilc_pw06, demo_pjanmarsta; PISA Externality remediation Defensive share of GDP Eurostat hlth_sha11_hf, env_ac_epneis1; GPI method One row cannot be filled today. The number that would settle whether Europe\u0026rsquo;s durability and right-to-repair rules do anything is the average age of an appliance when it gets thrown out. There are one-off studies but no series to validate the impact of the policy.\nRun it backwards. Include historical data where available. Then we can survey Germany, 2010 to 2025, using numbers already used in this essay. GDP per head grew comfortably through the 2010s and has been more or less flat since 2019 - a decent decade, then a stall. The dashboard tells us a different story for the same years. The replacement burden climbed from under 18% of GDP to over 20%, so net output lagged gross the whole time. Price-to-income went from its lowest since at least 1980 to its 2021 peak to fall again a bit, while completions fell to their lowest since 2012, under two-thirds of even the lower published estimate of need. Whether healthy life years rose is unknowable because the series broke. Teen loneliness turned up around 2012 and at least till 2018 did not turn back. Nobody who saw those lines together would describe the period as \u0026ldquo;comfortable growth then a stall\u0026rdquo;. They\u0026rsquo;d describe it as a decade in which the number and the country parted ways. And all of the data to say so was public already.\nWhat the rows say. GDP per head grew 14.7% from 2010 to 2019 and was 0.7% lower in 2025 than in 2019. Some things got better that GDP never counted. Each worker works 6.4% fewer hours a year than in 2010. The carbon footprint of what Germans buy, wherever it was made, fell from 12.7 to 9.1 tonnes a head. Until 2019, incomes at the bottom rose faster than at the top: the tenth percentile up 11%, the ninetieth up 6%. Other things got worse while GDP looked fine, or stayed flat while it grew. Life expectancy rose to 81.2 by 2019 and was still 81.2 in 2024. Deaths from suicide, poisoning and alcohol rose by about 9% between 2019 and 2023. Dwellings completed fell to 206,586 in 2025, 30% below 2019 and far below the 320,000 a year the BBSR says Germany needs. One German in five rates their life 5 out of 10 or lower, against 15.8% across the EU. Aid fell from 0.82% of national income to 0.56%, below the 0.7% promised in 1970. In the latest year, the median income gained 2.1% while the bottom tenth lost 0.8%.\nNone of this adds up to one verdict, and it shouldn\u0026rsquo;t. There is no honest exchange rate between a tonne of CO₂, an hour of leisure and a dwelling that was never built. The rows do answer a different question: which way is GDP wrong? For most of the twentieth century it understated progress. Growth came bundled with longer lives, less drudgery and goods that got better faster than their prices showed, so life improved faster than the line. That bundle has come apart. Longevity has stalled, housing is falling behind, and the gains that remain, fewer hours and lower emissions, sit beside losses the headline cannot see. The error has not gone away, but it may have changed sign. One number cannot tell you which era you are in. That is what the rows are for.\nGermany has tried this. Three times. (I learned about this while researching for the essay.) In 2010 the Bundestag set up an Enquete commission on \u0026ldquo;Growth, Prosperity, Quality of Life\u0026rdquo;. Its 2013 report proposed the \u0026ldquo;W3\u0026rdquo; indicator set: GDP plus nine more lead indicators in three equal pillars - material wealth, social participation, ecology - with a few \u0026ldquo;warning lamps\u0026rdquo; on the side, meant to be published alongside GDP [39]. It got a plenary debate and was shelved with the legislature. The Bundestag did ask the government to look at the commission\u0026rsquo;s results, which is how in 2015 the Merkel chancellery came to run a six-month citizens\u0026rsquo; dialogue, \u0026ldquo;Gut leben in Deutschland\u0026rdquo;, and in 2016 to publish a strategy with twelve dimensions and forty-six indicators of what Germans said a good life consists of [40]. It got a website. The indicators there were last updated in March 2020. Then, from 2022, the economics ministry added a wellbeing chapter to the Jahreswirtschaftsbericht, the government\u0026rsquo;s central annual economic report: thirty-odd indicators at first, later close to forty, in four areas, with a biodiversity index added in 2024 and the Gini coefficient of net wealth in 2025 [41]. This was the real thing, the government\u0026rsquo;s own economic report saying in print that a single number cannot capture welfare. In January 2026, under a new government and a renamed ministry, the chapter was deleted. What remains is a cross-reference to other reporting systems and one sentence promising to examine how the public\u0026rsquo;s \u0026ldquo;legitimate interest\u0026rdquo; in a broader view of welfare might be accommodated in future [42].\nPut that next to the four things GDP has and the alternatives lack, and each failure explains itself:\nan owning institution: the W3 set belonged to a commission that dissolved, the one-off citizens\u0026rsquo; dialogue to a chancellery that moved on, the wellbeing chapter to one ministry under one minister. a constituency with money at stake: nobody\u0026rsquo;s budget, bonus or borrowing cost moved when the biodiversity index fell. Insurers, pension funds and employers all have real exposure to population health, and none of them lobbied to keep chapter G. ritual and cadence: GDP is quarterly and leads the bulletin. The wellbeing chapter was annual and got its mention starting on page 79. fast feedback: outcomes move over decades. Lever-side indicators (homes permitted, screening rates, hours of childcare provided) move within a year and can carry the political weight in the meantime. Three attempts in fifteen years, each buried without an obituary, and not one of them because the numbers were wrong. The attempts died because nothing in the state\u0026rsquo;s machinery had to answer to the call when the numbers changed.\nBut there is hope. The infrastructure keeps getting built regardless. The UN\u0026rsquo;s 2025 revision of the System of National Accounts added extended accounts for wellbeing, unpaid work and the environment [37], and a UN high-level expert group on \u0026ldquo;Beyond GDP\u0026rdquo; is due to present its findings to the General Assembly this month [43]. Worth remembering that the national accounts themselves were born of the 1929 Depression and the war. Reforms of counting happen when a crisis makes the old number visibly wrong. So the job, for those who care, is to have the alternative fully specified and shelf-ready for when the window opens [38].\nDesigned against gaming. More numbers are still numbers. James C. Scott\u0026rsquo;s Seeing Like a State provides the warning [44]: states make the life they manage legible by abstracting it, and the abstraction throws away the local, tacit, practical knowledge that makes life worth living. The signature case is Prussian scientific forestry. The forest was replanted as an optimised monoculture, the numbers were great for one rotation, and then the rot showed, because much of what the number had ignored turned out to matter. Hayek made the same point from economics: the knowledge that matters is dispersed and local and no central aggregate can hold it [45]. And any new metric must reckon with Goodhart\u0026rsquo;s law or it will be blindly optimised for. A school district ordered to raise a \u0026ldquo;connection score\u0026rdquo; will produce mandatory friendship programmes and cheerfully falsified surveys. A naively deployed dashboard is technocracy with more dials.\nCareful design can put some defences in place. Plurality raises the cost of gaming: ten domains spanning health, time, connection and ecology leave less dark space for bad results to hide in: juicing the housing line through credit shows up as damage on the distribution line. Thermometers, not thermostats to the rescue - Muller\u0026rsquo;s central rule [46]: when a metric is wired to automatic reward or punishment it gets optimised, so decouple them. The dashboard\u0026rsquo;s numbers are tripwires that trigger inquiry. A falling healthy-life line gets an investigation with the same seriousness as an economic recession. We take a page from how medicine works: vital signs decide where to look, not what to do. Measure close to the intrinsic good: you may game \u0026ldquo;patients seen within four hours\u0026rdquo;, but you can\u0026rsquo;t easily make a population report for a decade that life is going well without life actually going well. Hence the lived measures, life satisfaction and time-use diaries, are proposed. Keep the list contestable, because what matters has to be settled and re-settled by public reasoning, not fixed by experts once and for all [47]. The dashboard needs a built-in revision ritual, a periodic public review of the rows themselves. A fixed dashboard is a technocratic artefact; a contested one is a democratic instrument. Institutionalise the listening, because the depth of life enters through people, not through better numbers. Ireland\u0026rsquo;s Citizens\u0026rsquo; Assembly is the proof of concept: ninety-nine randomly selected citizens deliberating with expert input broke a decades-long constitutional deadlock, and the German-speaking community of Belgium has made a sortition-based citizens\u0026rsquo; council a permanent organ [48]. Pair every row with its qualitative counterpart: assemblies reviewing whether the numbers still mean something, fieldwork feeding policy review, and the street-level practitioners (nurses, teachers, caseworkers) co-designing any metric that will be used on them [49].\nIf I were to put the concept in one line: numbers redistribute attention, deliberation supplies meaning, and a named, accountable human decides action. Polycentrism completes it, because one size rarely fits all. Ostrom showed communities govern their commons better than central blueprints when they set their own rules [49]. So the dashboard monitors outcomes nationally while the means to move the numbers stay local and plural. A hundred municipalities trying a hundred approaches to the loneliness line is both more efficient, as it addresses local challenges, and an experiment in which successful approaches spread by evidence rather than mandate.\nAnd at all times, we need to apply humility. Some of life will always escape measurement. Hold every metric as a proxy under permanent suspicion. GDP\u0026rsquo;s problem was never that it is a number. It is that we forgot it was one.\nIf it has to be one number. \u0026ldquo;How far off is GDP\u0026rdquo; could be a single figure, as the method exists: equivalent income that converts each non-market thing into the extra consumption that would leave a person indifferent. Fleurbaey and Blanchet set out the theory, Jones and Klenow built one prototype, Fleurbaey and Gaulier another [50]. Four domains are fairly settled. The rest need shadow prices, and most are available: emissions times the social cost of carbon, morbidity in DALYs, defensive spending subtracted as the GPI already does, loneliness can run straight through the machinery that already prices life expectancy, since it carries a measured mortality hazard. Some domains remain hard. Housing sits half inside consumption already, hence double counting is a real risk. And the quality of leisure also lacks a shadow price. We count hours and value them at the wage, which assumes an hour is an hour. But is it really so? Is a fragmented evening of scrolling worth the same as an unfragmented one? With the onslaught of the attention economy, the worth of leisure is potentially overstated and it is getting worse every year. Kahneman\u0026rsquo;s day-reconstruction method rates activities by how they actually felt [51]. We could try running it at the scale of a national account, and it might be exactly where the attention economy finally shows up in the numbers.\nDirections and discussion. I don\u0026rsquo;t have enough expertise to hand anyone a plan. But I wanted to start the conversation and explore different areas, put up proposals. Later, the numbers decide what gets looked at and people decide the rest. What follows is a stick in the sand, a departure point for discussion. Where I could find evidence I lean on it, where I couldn\u0026rsquo;t, I say so to the best of my ability.\nHealth. Everyone who cares knows the ground rules for a longer, healthier life. Good sleep, good food, a few hours of movement a week, less stress. Knowing was never the problem for most of us. So the question is not what people should do but what actually moves a population, and here the evidence is unkind to the things we do. My insurer offers points for regular dentist visits and a subsidy on a bike helmet. That\u0026rsquo;s nice and all. When Song and Baicker ran a proper randomised trial of a workplace wellness programme across 160 worksites, employees reported exercising more and eating better, and after eighteen months there was no significant effect on any clinical measure, on health spending or on employment [52]. People reported being more virtuous and their bodies didn\u0026rsquo;t notice the difference. What seems to work looks boring and structural. In October 2022 England banned foods high in fat, sugar and salt from checkouts, aisle ends and store entrances in larger shops. Not an awareness campaign, a rule about where the sweets sit. Earlier voluntary versions cut purchases of small confectionery packs by 15–17%, and purchases were still down a year later. A Leeds evaluation of the legislation estimates around two million fewer unhealthy products sold per day [53]. An Oxford trial removing seasonal chocolate from prominent positions cut enough to register in the total energy content of what people bought, with no difference between richer and poorer stores [54]. What changed was not willpower but the distance between a tired parent in a queue and a colourful display at child height. Then take cigarettes. On 29 April 2026 the UK\u0026rsquo;s Tobacco and Vapes Act got Royal Assent: nobody born on or after 1 January 2009 will ever be legally sold tobacco, in force from January 2027 [55]. New Zealand passed the same law in 2022 and repealed it after an election, before it even took effect [56]. So the policy exists; whether it survives can be helped by a dashboard where the respective number has a political owner. Is that all there is? Is it the cost-effective frontier? I doubt it. Bikeable cities probably beat helmet subsidies, but I haven\u0026rsquo;t seen that comparison run properly and wouldn\u0026rsquo;t trust my gut feeling on this. Which is kind of the point. We know the rules for individuals. We are still guessing at what moves populations, and we fund the respective research at a fraction of what we spend on modelling interest rates.\nTime. Hours worked keep slowly falling but GDP reads that as weakness. A country that exchanges its productivity gains for leisurely afternoons rather than output looks, in the only number anyone reports, like it is failing. So the first lever for time is measurement itself: publish leisure hours and unpaid care hours, valued at replacement wage, next to the growth figure. Until a society can see what it bought with its productivity, it can\u0026rsquo;t decide whether the trade was a good one. The four-day-week trials get headlines, but they are mostly self-selected firms measuring themselves. So count first.\nFamily and connection. The fertility evidence points at compatibility with women\u0026rsquo;s careers: affordable childcare, long leave split between parents by default, flexible hours. Fathers doing half the care is the strongest pronatalist variable anyone has measured [35, 36]. The realistic yield of any of it looks modest, perhaps a few tenths of a child per woman, and even that may be generous. Korea\u0026rsquo;s rebound from 0.72 to 0.80 came after hundreds of trillions of won in incentives, and demographers attribute most of it to a large cohort of women in their early thirties and postponed pandemic weddings, not policy [58]. One thing worth looking at first, because it may work against everything else, is the tax code. German joint taxation halves a couple\u0026rsquo;s income before taxing it, so the subsidy is largest when one partner earns everything and the other nothing, and the second earner\u0026rsquo;s first euro is taxed at the household\u0026rsquo;s top rate. Half of married part-time women in a recent DIW survey say more hours aren\u0026rsquo;t worth it; individual taxation would, on that survey\u0026rsquo;s projection, add up to 175,000 full-time positions among women aged 45 to 66 alone [57]. A lever hiding in plain sight, and I haven\u0026rsquo;t seen it discussed as family policy. But on ageing the honest answer is a portfolio of interventions anyway: tax code, compatibility, immigration (the only fast lever), longer healthy working lives, and productivity. AI might relieve Baumol\u0026rsquo;s cost disease in care itself [59] but only if we design it to take the heavy lifting and the paperwork and leave the humans the humanity. As for connection, we can engage in a bit of incentive engineering. It\u0026rsquo;s harder to build a true connection glued to a screen. If engagement-maximising design generates measurable downstream care costs, the attention economy should at least be counted like any other polluter before we decide whether and how to price it. Also it is rather striking that capabilities like non-violent communication or active listening are missing in school curriculum.\nHousing. Build at the published needs-line, and publish completions against it quarterly with a named minister. The land under a flat is worth what it is worth because of everything around it - the U-Bahn, the school, the neighbours, the city\u0026rsquo;s whole accumulated effort - none of which the owner made. What to do about that gain is a tax argument with real evidence on both sides. As that would make for an essay on its own let\u0026rsquo;s start by putting the needs number in front of everyone: completions against need, with someone accountable for whether it happened.\nPart IV - What great would look like The point of all of the above is to let something other than growth into the conversation and relocate some of society\u0026rsquo;s attention. Where would that lead? What could the future look like if the dashboard were built and taken seriously for fifteen years?\nYou are thirty-one, you live in a mid-sized European city, and housing takes a quarter of your income and not half, because completions have tracked a published needs-line for a decade. The price-to-income figure is now announced every quarter with the gravity of the growth number. The responsible minister could not survive letting it climb past the set threshold. Whether you rent or own is a matter of preference, not an inheritance test, because the deposit equals two years of saving. You and your partner both work, shorter, more flexible hours, because the time-use line is a headline number and the tax system stopped paying couples to keep one of them home. A second child is a question of desire, not arithmetic, because much of the needed care is free, paternity leave is long and split by default, and working as a parent comes with additional benefits to enable some parenting-life balance. The quiet revolution the fertility economists predicted is picking up, because fathers doing half the care turned out to be the strongest pronatalist policy ever tried. Your parents, at seventy, are a decade into the healthy-life-years dividend. They retired later than their parents did but also better, because when the HLY line dipped in the late 2020s it was treated like a recession: screening, food policy, GLP-1s, walkable streets were put up and the line turned. The care your grandmother needs is affordable not because carers got cheap but because the robots took the lifting and the paperwork while people still provide the much needed human touch. Baumol\u0026rsquo;s disease got relieved, not denied, by the AI that Stanford professor predicted would arrive slowly but everywhere.\nYour social media feed costs money and your well-maintained park is free, because the defensive-share line made the attention economy\u0026rsquo;s invoice visible and the charge landed on the harvest, not the remedy. You know five neighbours by name. The connection series is boring now, which is what success looks like. Once, the letter came inviting you to the assembly that re-examines the rows. It was exhausting and exhilarating at the same time, three days of arguing with ninety-nine strangers about what to count and how. The list changed a bit afterwards, which is how you know it is still alive. The numbers never decided anything by themselves. They decided what got looked at, so people could decide the rest. Each quarter the finance minister stands up with ten numbers and the details behind them instead of one. Your children will find it hard to believe that at some point one number was considered enough.\nNone of this is utopian. Every element exists somewhere today: the childcare in France, the housing cadence in Vienna\u0026rsquo;s better decades, the healthy-ageing gains in Japan, the wellbeing law in Wales, the wellbeing chapter Germany first wrote and then deleted. What great looks like isn\u0026rsquo;t an invention, it\u0026rsquo;s putting existing pieces together and paying attention to them and evolving them together with our values and society. Attention is exactly what a dashboard redistributes.\nI invite you to try it yourself. You already work under a version of this. A team, a department, a household: each has a headline number while many nearly-as-important things go unreported. Add a second number. Publish it with the same care as the first. Watch for behaviour changes.\nAlso write to the editor, not the minister. \u0026ldquo;You led the article with GDP growth while median income fell. Why that order?\u0026rdquo; The second time someone asks, the focus might change.\nAs Winifred Gallagher put it \u0026ldquo;your life is the sum of what you focus on\u0026rdquo;. We counted output so we got a lot of output. If we count the rest, I see no reason why we shouldn\u0026rsquo;t get the rest too. We are all about moving the needle. There is more then one needle it seems, let’s move the right one.\nAcknowledgments: many thanks to Michal Bída and Kateřina Bída for feedback and suggestions. And to Claude Fable 5.1/Opus 5 for help with research and drafting.\nAppendix A - References Verification key: ✔ re-checked against the source in September 2026 · ~ consistent with what Claude know of the source, not re-opened · ! needs checking.\n~ Kuznets, S. (1934). National Income, 1929–1932, US Senate. https://www.nber.org/books-and-chapters/national-income-1929-1932 . The quote is on p. 7 of the report. ~ Bastiat, F. (1850). That Which Is Seen, and That Which Is Not Seen. https://www.econlib.org/library/Bastiat/basEss.html ~ Gordon, R. J. (2016). The Rise and Fall of American Growth; Solow\u0026rsquo;s \u0026ldquo;everywhere but in the productivity statistics\u0026rdquo;, NYT Book Review, July 12, 1987. The ~2% figure and the \u0026ldquo;no bump\u0026rdquo; observation are from the Stanford GSB talk: https://www.youtube.com/watch?v=xBpGn3BDcOY\u0026amp;t=556s ~ Acemoglu, D. (2024). \u0026ldquo;The Simple Macroeconomics of AI.\u0026rdquo; NBER WP 32487. https://www.nber.org/papers/w32487 . The 0.66% is his upper-bound TFP estimate over ten years. The paper argues manipulative/engagement uses can raise GDP while reducing welfare; it does not give the \u0026ldquo;2%\u0026rdquo; figure the earlier draft attributed to it, so that number was removed. ~ Nordhaus, W. D. (1996). \u0026ldquo;Do Real-Output and Real-Wage Measures Capture Reality? The History of Lighting Suggests Not.\u0026rdquo; https://www.nber.org/system/files/chapters/c6064/c6064.pdf ~ Boskin Commission (1996), Toward a More Accurate Measure of the Cost of Living. https://www.ssa.gov/history/reports/boskinrpt.html ~ Brynjolfsson, E., Collis, A., Diewert, W. E., Eggers, F. \u0026amp; Fox, K. J. \u0026ldquo;GDP-B.\u0026rdquo; NBER WP 25695. https://www.nber.org/papers/w25695 ~ Stevenson, B. \u0026amp; Wolfers, J. (2008). \u0026ldquo;Economic Growth and Subjective Well-Being: Reassessing the Easterlin Paradox.\u0026rdquo; Brookings Papers on Economic Activity. ~ Murphy, K. M. \u0026amp; Topel, R. H. (2006). \u0026ldquo;The Value of Health and Longevity.\u0026rdquo; JPE 114(5). https://www.journals.uchicago.edu/doi/10.1086/508033 . The 18%/age-62 comparison, the $1.2m and the \u0026ldquo;$3.2 trillion a year, about half of GDP\u0026rdquo; are all from this paper\u0026rsquo;s introduction. ~ Greenwood, J., Seshadri, A. \u0026amp; Yorukoglu, M. (2005). \u0026ldquo;Engines of Liberation.\u0026rdquo; ReStud 72(1). 58 hours/week (1900) to 18 (1975). ~ Waring, M. (1988). If Women Counted. ✔ Jones, C. I. \u0026amp; Klenow, P. J. (2016). \u0026ldquo;Beyond GDP? Welfare across Countries and Time.\u0026rdquo; AER 106(9): 2426–57. https://www.aeaweb.org/articles?id=10.1257/aer.20110236 . Checked: UK/France/Italy/Spain income ~2/3 of US, welfare averaging 85%; France income 67%, consumption 60%, welfare \u0026gt;90%; life expectancy 80.1 vs 77.4; hours per capita 535 vs 877; median deviation ~35% across the 13 micro-data countries; rising life expectancy adds more than a full point a year to welfare growth. The earlier draft\u0026rsquo;s \u0026ldquo;Western Europe 71%/90%\u0026rdquo; was wrong and has been replaced. ~ Goodhart, C. (1975); Strathern, M. (1997). \u0026ldquo;\u0026lsquo;Improving ratings\u0026rsquo;: audit in the British University system.\u0026rdquo; European Review 5(3). ~ Holmström, B. \u0026amp; Milgrom, P. (1991). JLEO 7 (special issue). ~ Mazzucato, M. (2018). The Value of Everything. Allen Lane. ! NZ Wellbeing Budget 2019 and the 2024 retrospectives. https://www.treasury.govt.nz/publications/wellbeing-budget/wellbeing-budget-2019 . The earlier draft\u0026rsquo;s \u0026ldquo;NZ$26bn\u0026rdquo; could not be traced to a source and has been removed from the text; the 2019 budget\u0026rsquo;s new operating spending was NZ$3.8bn. Re-derive any figure before using one. ~ Stiglitz, J., Sen, A. \u0026amp; Fitoussi, J.-P. (2009). Report by the Commission on the Measurement of Economic Performance and Social Progress. Follow-up: Stiglitz, Fitoussi \u0026amp; Durand (2018), Beyond GDP: Measuring What Counts for Economic and Social Performance, OECD. Downstream: OECD How\u0026rsquo;s Life? (2011–) and Better Life Index; Eurostat quality-of-life indicators. ✔ Furman, J., post on X, 27 September 2025: IPE\u0026amp;S is 4% of GDP, 92% of H1 2025 growth, ex-IPE\u0026amp;S growth 0.1% annualised. Furman\u0026rsquo;s own caveat (\u0026ldquo;not a counterfactual… could maybe make up about half\u0026rdquo;) is in the same thread. Ernie Tedeschi\u0026rsquo;s reply notes much of the investment was imported; netting imports gives ~0.24pp of the 1.6% H1 growth. Fortune write-up: https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist ✔ Zeman, M. (2025). \u0026ldquo;Trends in Real Wage Development in the Czech Republic.\u0026rdquo; IPPS. https://www.politikaspolecnost.cz/wp-content/uploads/2025/06/06_2025_Trends-in-Real-Wage-Development-in-the-Czech-Republic_IPPS_Martin-Zeman.pdf . Real wages −9.3% (2022), −2.4% (2023), still below 2019 at end-2025; housing allowances 169k (Mar 2022) → 303k (Mar 2025), citing MoLSA. ! The GDP path (flat to slightly positive 2022–24) is my reading of ČSÚ national accounts; confirm the annual figures. ~ Case, A. \u0026amp; Deaton, A. (2020). Deaths of Despair and the Future of Capitalism. Princeton UP. Life expectancy 78.9 (2014) → 78.6 (2017): CDC/NCHS; Woolf, S. \u0026amp; Schoomaker, H. (2019), JAMA 322(20). Later NCHS revisions show 2016 flat rather than falling, which is why the text now says \u0026ldquo;fell before COVID\u0026rdquo; and not \u0026ldquo;three consecutive years\u0026rdquo;. ✔ OECD/European Commission (2024). Health at a Glance: Europe 2024, ch. 2, section \u0026ldquo;Living longer, living healthier?\u0026rdquo;. https://www.oecd.org/en/publications/health-at-a-glance-europe-2024_b3704e14-en/full-report/component-8.html . Checked: life expectancy at 65 +1.2 years 2005–2022, of which 0.8 healthy and 0.4 with disability; share lived with disability 55%→53%; explicit warning that post-break data \u0026ldquo;should not usually be compared\u0026rdquo; with pre-break data (Figure 2.5, source Eurostat lth_hlye). ✔ Straka, J., Šídlo, L. \u0026amp; Kulhánová, I. (2024). \u0026ldquo;Trends in Healthy Life Years Between 2005 and 2019 in 31 European Countries: The Compression or Expansion of Morbidity?\u0026rdquo; International Journal of Public Health 69: 1607574. https://doi.org/10.3389/ijph.2024.1607574 (open access: https://pmc.ncbi.nlm.nih.gov/articles/PMC11521812/). EU-28 plus Croatia, Norway and Switzerland, Eurostat HLY at 65. The three-way split is my summary of their typology figure; check the country counts against it. ~ Twenge, J. M., Haidt, J. et al. (2021). \u0026ldquo;Worldwide increases in adolescent loneliness.\u0026rdquo; Journal of Adolescence 93. 36 of 37 countries, PISA 2000–2018. ! Monitoring the Future via Haidt (2024), The Anxious Generation. https://monitoringthefuture.org/ . Secondary source; verify 44%→32% against the primary tabulations. ~ Klinenberg, E. (2012). Going Solo; bonding/bridging: Putnam, R. (2000). Bowling Alone. ~ Holt-Lunstad, J., Smith, T. B. \u0026amp; Layton, J. B. (2010). PLoS Medicine 7(7): e1000316. 148 studies, 308,849 participants, OR 1.50 overall, OR 1.91 for complex social integration. https://doi.org/10.1371/journal.pmed.1000316 ~ Kapp, K. W. (1950). The Social Costs of Private Enterprise; Leipert, C. (1989). Journal of Economic Issues 23(3). ! GPI methodology: currently https://en.wikipedia.org/wiki/Genuine_progress_indicator . Replace with Talberth, Cobb \u0026amp; Slattery (2007), The Genuine Progress Indicator 2006. ~ Holt-Lunstad, J. et al. (2015). Perspectives on Psychological Science 10(2): isolation +29%, loneliness +26%. The smoking comparison qualified by its authors: Smith, Holt-Lunstad \u0026amp; Kawachi (2023), American Journal of Epidemiology 192(8). Medicare figure: AARP/Stanford (2017), https://www.aarp.org/medicare/high-price-of-social-isolation-study-2017/ ✔ OECD Analytical House Prices, dataflow DSD_AN_HOUSE_PRICES@DF_HOUSE_PRICES, measure HPI_YDH (price-to-income ratio, 2015 = 100), annual, Germany, as pulled for the dashboard in September 2026: lowest value 1980–2025 in 2010 (93.8), highest in 2021 (130.9), 105.4 in 2025. https://sdmx.oecd.org/public/rest/data/OECD.ECO.MPD,DSD_AN_HOUSE_PRICES@DF_HOUSE_PRICES,1.0/DEU.A.HPI_YDH. ✔ Destatis Häuserpreisindex, GENESIS table 61262-0001, annual, 2025 = 100. In June 2026 Destatis rebased the index to 2025 = 100 and recomputed the whole series from Q1 2000; the old 2015 = 100 indices \u0026ldquo;lose their validity\u0026rdquo; (press release 219, 25 June 2026: https://www.destatis.de/DE/Presse/Pressemitteilungen/2026/06/PD26_219_61262.html). On the new base: 65.5 (2015) to 107.5 (2022), +64%; −8.5% in 2023, the largest annual fall in the series; +3.2% in 2025. The earlier \u0026ldquo;two-thirds since 2015\u0026rdquo; and the Top-7 \u0026ldquo;+79% over the decade\u0026rdquo; were computed on the old base; GENESIS statistic 61262 has no Top-7 or Kreistyp split (61262-0002 is the quarterly national index), so the Top-7 figure has been dropped. German real GDP path: Destatis national accounts. ✔ Destatis press release 174, 22 May 2026: 206,600 dwellings completed in 2025 (−18.0%), lowest since 2012 (200,500); 2021–2023 each around 294,000 (295,300 in 2022); peak 306,400 in 2020. https://www.destatis.de/DE/Presse/Pressemitteilungen/2026/05/PD26_174_31121.html ; long series in GENESIS table 31121-0001. Need: BBSR Wohnungsbedarfsprognose, Analysen KOMPAKT 05/2025, \u0026ldquo;rund 320.000 Wohneinheiten p. a. (2023 bis 2030)\u0026rdquo;, https://www.bbsr.bund.de/BBSR/DE/veroeffentlichungen/analysen-kompakt/2025/ak-05-2025-dl.pdf ; IW Köln, IW-Report 39/2024, 372,600 a year for 2021–2025, falling to 257,400 for 2026–2040, https://www.iwkoeln.de/fileadmin/user_upload/Studien/Report/PDF/2024/IW-Report_2024_Regionale-Wohnungsbedarfe.pdf . The earlier \u0026ldquo;shortfall of ~830,000\u0026rdquo; could not be traced to a source and has been replaced. ~ Hirsch, F. (1976). Social Limits to Growth; Frank, R. (2011). The Darwin Economy. ~ Piketty, T. (2014). Capital in the Twenty-First Century. ~ Doepke, M., Hannusch, A., Kindermann, F. \u0026amp; Tertilt, M. (2022). \u0026ldquo;The Economics of Fertility: A New Era.\u0026rdquo; NBER WP 29948. https://www.nber.org/papers/w29948 ~ Doepke, M. \u0026amp; Kindermann, F. (2019). \u0026ldquo;Bargaining over Babies.\u0026rdquo; AER 109(9); summary: https://cepr.org/voxeu/columns/new-era-economics-fertility ~ UN SNA 2025 (adopted by the UN Statistical Commission, March 2025). https://unstats.un.org/unsd/nationalaccount/snaupdate/ ~ Kingdon, J. (1984). Agendas, Alternatives, and Public Policies. ✔ Deutscher Bundestag, Enquete-Kommission \u0026ldquo;Wachstum, Wohlstand, Lebensqualität\u0026rdquo; (2010–2013), Schlussbericht, BT-Drs. 17/13300. W3: GDP plus nine further lead indicators in three pillars (material wealth, social/participation, ecology) with \u0026ldquo;warning lamps\u0026rdquo;; see Schmidt \u0026amp; aus dem Moore (2014), RWI Positionen 56, https://www.econstor.eu/bitstream/10419/95942/1/78187811X.pdf , and Giesselmann et al., DIW Wochenbericht 80(9). ✔ Bundesregierung (2016). Regierungsstrategie \u0026ldquo;Gut leben in Deutschland - was uns wichtig ist\u0026rdquo;: six-month citizens\u0026rsquo; dialogue in 2015, twelve dimensions, forty-six indicators; indicators last updated March 2020. https://www.gut-leben-in-deutschland.de . The earlier draft\u0026rsquo;s \u0026ldquo;fifteen thousand participants\u0026rdquo; was not confirmed and has been removed. ✔ BMWK, Jahreswirtschaftsbericht 2022–2025. 2022: special chapter \u0026ldquo;Nachhaltiges und inklusives Wachstum - Dimensionen der Wohlfahrt messbar machen\u0026rdquo; (\u0026gt;30 indicators, five areas); from 2023 \u0026ldquo;Wohlfahrtsmessung und gesellschaftlicher Fortschritt\u0026rdquo; (34 indicators in 2023, close to 40 by 2025, four areas; biodiversity indicator added 2024, net-wealth Gini 2025). 2022 edition: https://dserver.bundestag.de/btd/20/005/2000520.pdf ; 2023 chapter: https://www.bundeswirtschaftsministerium.de/Redaktion/DE/Downloads/J-L/jwb-2023-kapitel-wohlfahrtsmessung.pdf ; Wirtschaftsdienst 2023/7 on the 34 indicators. ✔ BMWE (2026). Jahreswirtschaftsbericht 2026, p. 19. https://www.bundeswirtschaftsministerium.de/Redaktion/DE/Publikationen/Wirtschaft/jahreswirtschaftsbericht-2026.html . On the deletion: Zieschank, R. \u0026amp; Held, B. (2 June 2026), \u0026ldquo;Der Jahreswirtschaftsbericht 2026 - ein Sprung zurück,\u0026rdquo; Blog Postwachstum. https://www.postwachstum.de/der-jahreswirtschaftsbericht-2026-ein-sprung-zurueck-20260602 ✔ UN High-Level Expert Group on Beyond GDP, Interim Progress Report (November 2025). https://www.un.org/en/beyondGDP . ! Final findings scheduled for UNGA, September 2026 - check what was actually presented. ~ Scott, J. C. (1998). Seeing Like a State. Yale UP. ~ Hayek, F. A. (1945). \u0026ldquo;The Use of Knowledge in Society.\u0026rdquo; AER 35(4). ~ Muller, J. Z. (2018). The Tyranny of Metrics. Princeton UP. ~ Sen, A. (2009). The Idea of Justice. Harvard UP. ~ Irish Citizens\u0026rsquo; Assembly (99 members): https://citizensassembly.ie ; the permanent Ostbelgien Bürgerrat (est. 2019). ~ Ostrom, E. (1990). Governing the Commons. Cambridge UP; Lipsky, M. (1980). Street-Level Bureaucracy; Lindblom, C. (1959). \u0026ldquo;The Science of \u0026lsquo;Muddling Through\u0026rsquo;.\u0026rdquo; Public Administration Review 19(2). ~ Fleurbaey, M. \u0026amp; Blanchet, D. (2013). Beyond GDP: Measuring Welfare and Assessing Sustainability. OUP; Fleurbaey, M. (2009). JEL 47(4); Fleurbaey, M. \u0026amp; Gaulier, G. (2009). Scandinavian Journal of Economics 111(3). ~ Kahneman, D., Krueger, A. B., Schkade, D., Schwarz, N. \u0026amp; Stone, A. (2004). \u0026ldquo;A Survey Method for Characterizing Daily Life Experience: The Day Reconstruction Method.\u0026rdquo; Science 306(5702). ~ Song, Z. \u0026amp; Baicker, K. (2019). \u0026ldquo;Effect of a Workplace Wellness Program on Employee Health and Economic Outcomes.\u0026rdquo; JAMA 321(15). 160 worksites, 18-month outcomes. https://jamanetwork.com/journals/jama/fullarticle/2730614 ✔ Voluntary checkout policies: CEDAR/Cambridge evidence brief (Ejlerskov et al. 2018, PLOS Medicine). https://www.epi.ims.cam.ac.uk/resources/policy-resources/evidence-briefs/eb-checking-out-checkout-food/ . 2022 regulations: University of Leeds (Morris et al.), independent evaluation with IGD and four retailers, press release 22 August 2025: 0.63 percentage-point fall in in-scope HFSS sales by weight, from 20 to 19 of every 100 items, ≈ two million fewer products a day; 11.6bn product sales, 480 stores. https://www.leeds.ac.uk/research-32/news/article/5839/positive-impact-of-supermarket-junk-food-restrictions-revealed . Caveats: released as a preprint, and the data came from the retailers themselves. Coverage at the time noted the government intended to repeal the upcoming volume-price and aisle rules. ~ Piernas, C., Harmer, G. \u0026amp; Jebb, S. A. (2022). \u0026ldquo;Removing seasonal confectionery from prominent store locations…\u0026rdquo; PLOS Medicine 19(3). https://doi.org/10.1371/journal.pmed.1003951 ✔ Tobacco and Vapes Act 2026 (c. 18), Royal Assent 29 April 2026; generational sales ban from 1 January 2027. https://bills.parliament.uk/bills/3879 . ASH\u0026rsquo;s \u0026ldquo;four million young people within five years\u0026rdquo; is a campaign estimate, not in the Act; it has been removed from the text. ~ House of Lords Library, \u0026ldquo;Smoke-free legislation: The UK and New Zealand.\u0026rdquo; https://lordslibrary.parliament.uk/smoke-free-legislation-the-uk-and-new-zealand/ . NZ repeal: February 2024. ✔ Bertelsmann Stiftung / DIW Berlin (5 March 2026), \u0026ldquo;Nach Kinderphase: Ehegattensplitting bremst Beschäftigung von Frauen aus.\u0026rdquo; 3,788 women aged 45–66 surveyed June–July 2025; 50% of married part-time women say more hours don\u0026rsquo;t pay; reform would raise participation by ~1.5pp, hours by ~3%, \u0026ldquo;up to\u0026rdquo; 175,000 FTE. https://www.bertelsmann-stiftung.de/de/themen/aktuelle-meldungen/2026/maerz/nach-kinderphase-ehegattensplitting-bremst-beschaeftigung-von-frauen-aus ✔ Korea: TFR 0.72 (2023) → 0.75 (2024) → 0.80 (2025 provisional, Ministry of Data and Statistics, 25 Feb 2026); births +6.8%. https://www.korea.net/NewsFocus/Society/view?articleId=288047 . Demographers attributing the rebound to cohort size and postponed weddings rather than policy: e.g. AFP/Space Daily, July 2026. ~ Hungary 1.23→~1.5 and France ~1.8 at ~1.5% GDP: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12206145/ . The \u0026ldquo;few tenths\u0026rdquo; range is my reading, not a published estimate. ~ Baumol, W. (2012). The Cost Disease. ! US: Census Bureau real median household income (FRED MEHOINUSA672N) vs BEA real GDP per capita (FRED A939RX0Q048SBEA). ✔ Germany: Eurostat ilc_di03, real median equivalised disposable income (survey year X deflated with HICP of year X−1, since EU-SILC incomes refer to the previous year): +10.2% between survey years 2010 and 2019, while real GDP per head (nama_10_pc) rose 18.5% over the matching income years 2009–2018. The series breaks in 2020, when German EU-SILC moved into the Mikrozensus, so no change is quoted across that year. The size of the divergence depends on deflator and household-composition choices; cite a specific series and year range. ! US nominal GDP ~$15.0tn (2010) → ~$30tn (2025), BEA; real GDP up roughly a third and real GDP per capita roughly a quarter (FRED GDPC1, A939RX0Q048SBEA). Pull exact figures. ! Benjamin, D. J., Heffetz, O., Kimball, M. S. \u0026amp; Szembrot, N. (2014). \u0026ldquo;Beyond Happiness and Satisfaction: Toward Well-Being Indices Based on Stated Preference.\u0026rdquo; AER 104(9). Health, security and family rank above income in stated preferences; also Adler \u0026amp; Dolan (2008). Appendix B - Dashboard data sources What the dashboard actually reads, by row, with dataset codes as of September 2026. Each panel on the page links its exact query.\nGDP and net output: Eurostat nama_10_pc, nama_10_gdp (B1GQ), nasa_10_nf_tr (P51C) Distribution: Eurostat ilc_di03, ilc_di01, prc_hicp_aind (deflator); WID.world shwealj992; effective tax rates by wealth band from published estimates (Bach, Corneo and Steiner; Oxfam / Netzwerk Steuergerechtigkeit; Saez and Zucman; Auten and Splinter) Health: Eurostat hlth_hlye, demo_mlexpec, hlth_cd_asdr2 Combined welfare: Jones \u0026amp; Klenow (2016), published figures Housing: OECD Analytical House Prices (DSD_AN_HOUSE_PRICES); Destatis GENESIS 31121-0001, 61262-0001; Eurostat ilc_lvho07a, ilc_lvho07b; BBSR and IW Köln need estimates Time: Eurostat nama_10_a10_e, demo_pjan, HETUS tus_00age, tus_20age Ecology: World Bank NY.ADJ.SVNG.GN.ZS; Global Carbon Project via Our World in Data Durability: Eurostat nasa_10_nf_tr (P51C), env_waseleeos Subjective wellbeing: Eurostat ilc_pw01, ilc_pw05, ilc_pw02; World Happiness Report (Gallup World Poll) Connection and family: Eurostat ilc_pw06, ilc_pw10, ilc_scp10, demo_pjanmarsta; Twenge et al. (2021) for PISA loneliness Externality remediation: Eurostat hlth_sha11_hf, env_ac_epneis1; GPI method (Talberth, Cobb and Slattery 2007) ","permalink":"https://wondering-wanderer.eu/posts/the-number-we-chose/","summary":"\u003ch3 id=\"gdp-diverged-from-life-a-long-time-ago-and-the-gap-keeps-compounding-can-we-count-better\"\u003eGDP diverged from life a long time ago and the gap keeps compounding. Can we count better?\u003c/h3\u003e\n\u003chr\u003e\n\u003cp\u003eThe value of an artillery shell fired in a war counts the same as a bridge that will stand for eighty years or more. A washing machine that dies after a mere thirty months of modest use counts for twice as much as one that lasts sixty. Wikipedia counts for nothing. GDP measures market output and it measures it well. The problem is that we do too much with it. We made it the target, we optimised for it for eighty years, and somewhere along the way stopped asking what else was moving. Let\u0026rsquo;s talk about that.\u003c/p\u003e","title":"The Number We Chose"},{"content":"There is a room in my flat with more toys than my two boys could meaningfully play with in a year. They are not spoiled by any usual standards - this is what childhood in a developed country looks like. The toys mostly sit there. The scarce ingredient of their play was never the toy. It was attention, the unhurried hour, the absence of stimulus, to be bored long enough for imagination to kick in. Many adults I know live in a scaled-up version of the same room - closets full of gear, gadgets, subscriptions. We own more than we can attend to.\nAnd looking around me I see people struggling, busy, glued to a screen more often than not, always distracted. Am I immune to this? Not at all, guilty as charged. I’m doing my best. But why is our attention so fleeting? Do we have to content ourselves with accepting such a distracted state as a fact of life? We are likely living in the best time ever to be alive yet many are drowning in depression. Why is that? A Netflix executive once claimed they compete with sleep and they are winning. TikTok is now being examined for breach of the EU digital rules. Our attention and consequently the quality of our lives is at stake.\nLet\u0026rsquo;s get the definition out of the way, by attention I mean something rather specific: the finite hours a day and the agency to decide what they\u0026rsquo;re spent on, the capacity to choose what deserves our presence, before that choice gets made for us by the designs built to bypass our defenses.\nI am not saying that consumption is bad or that abundance failed us. Extreme poverty has fallen from nearly four in ten people in 1990 to roughly one in ten today, a genuine miracle that still requires continuous attention to finally close the chapter of extreme poverty. World has gotten better on many accounts. Many problems are now problems of the past in the developed world: hunger, diseases, (wars), subsisting on hard farming work. Have we solved all the problems? Not yet. I claim that attention - not money, not goods - is becoming the binding constraint of a good life. Businesses compete for the same finite pool of hours we have and for our agency to use them. An economy that grows by capturing more of such attention drains what its own citizens need to live well. And it seems we have built such an economy. Attention became commoditized - hunted, tracked, sliced, sold. The damage is real. Less attention means less rest, relationships, presence with our children. Is this an acceptable trade off? Could we protect our attention with instruments that already exist?\nAnd the problem is much older than I thought. Staffan Linder observed in 1970 that consumption requires time, as goods multiply but hours don\u0026rsquo;t, the yield of each possession falls. Herbert Simon added a year later that a wealth of information creates a poverty of attention. Neither could foresaw what was to come: a trillion-dollars-a-year industry built to capture the agency both researchers assumed people would keep.\nThe arms race What causes the run for our attention? A firm in a saturated market, which is most consumer markets in rich countries, grows by a combination of improving its product, cutting its price, or outspending its rivals on marketing. Informative advertising, that the product exists and what it is, is genuinely useful. I have no problem with the bakery\u0026rsquo;s sign. But saturated-market advertising is often combative. As it cannot expand demand, it fights over a fixed pie, each company\u0026rsquo;s spending cancelled by its rivals\u0026rsquo;. Nobody can stop unilaterally without losing their market share, a variant of prisoner\u0026rsquo;s dilemma, a Nash equilibrium in which enormous resources are burned to move very little and we get caught in the cross-fire.\nThe current race for customers consumed over a trillion dollars in 2025, about one percent of world GDP, three-quarters digital, with Alphabet, Meta, and Amazon capturing more than half of all ad spend outside China. The broader marketing apparatus (affiliate, influencers, etc.) plausibly doubles the figure. That’s a lot of money allocated to capture our attention, creating ever increasing demand for it with unintended consequences. The race produced businesses whose product is engineered engagement itself - feeds optimized for time-on-site, autoplay, infinite scroll, notifications tuned to variable reward exploiting weaknesses of our psychology. The content is the bait, the catch is you, resold by the hour. The content is engineered in such a way that continuing is the default while stopping takes an act of will. Nobody sits down deliberately to watch four hours of algorithmically chained videos.\nBut is really non-consensual? Maybe people choose to scroll? Researchers paid people to deactivate or limit Facebook and Instagram and participants became measurably happier. They valued the platform less afterward, as if distance had broken a spell. Another study modelled the same behaviour and estimated that self-control problems and habit formation account for roughly 31 percent of social media and video streaming use. That\u0026rsquo;s time people wouldn\u0026rsquo;t spend if they acted on their own preferences. Keep that number.\nAdditionally, the frontier has already moved past scrolling and feeds. Since 2024, AI companion apps apply the same logic one layer deeper to our need for connection. Replika\u0026rsquo;s CEO has described their goal as fostering \u0026ldquo;long-term commitment,\u0026rdquo; even \u0026ldquo;marriage.\u0026rdquo; It has drawn a formal FTC inquiry into harm to minors (2025), a 44-state attorneys-general coalition, and a Character.AI/Google settlement (January 2026) over lawsuits alleging chatbots contributed to mental health crises and suicides among young users who had formed intense attachments to them. I learned of the AI companions only recently and found it genuinely disturbing. Optimizing for retention is not as surprising but the object of retention moved from an afternoon to a person\u0026rsquo;s capacity to feel cared for. This is what commodification of relatioship itself looks like, engineered and sold back.\nAnother worrying line of inquiry relates to the quality of attention. So far I\u0026rsquo;ve described attention as a fixed resource, the hours the companies compete to occupy. But highly stimulating, variable-reward content may not just occupy an hour, it may change what our attention is capable of afterwards. Heavy media multitaskers were found to have measurably worse sustained focus and distractor-filtering than light users, from a widely cited 2009 study. Later replication found the average effect modest which leaves open whether heavy use erodes attentional control or people with inherently weaker control are simply predisposed to heavier use. But potential consequences are serious as the industry may not only harvest a fixed field of attention, it may be reducing the field\u0026rsquo;s fertility, our capacity for sustained focus and the quality of presence a relationship or a decision needs. If so, the question to ask is not only how attention gets allocated but also how the capacity for it gets protected.\nThe depleted hour The evening feed does not find us fresh to resist it. It gets us primed, softened, vulnerable, at the end of a day that has already spent our willpower making the evening\u0026rsquo;s attention capture easier.\nFifteen years ago my working day was filled largely with focused work interrupted occasionally by checking emails, a few times a day. The synchronous conversations were on one topic at a time. Then came instant messaging with its channels, rooms, and spaces. Today, a knowledge worker is required to monitor what can exceed a hundred feeds, holding dozens of topics in partial awareness, marinating in low-grade fear of missing a message that mattered. This exhaustion is not a personal failing but supported in literature. Cal Newport calls the regime the \u0026ldquo;hyperactive hive mind.\u0026rdquo; Gloria Mark\u0026rsquo;s workplace studies found average sustained attention on a screen task fell from two and a half minutes in 2004 to mere 47 seconds, with an interruption costing twenty minutes to recover from. Microsoft\u0026rsquo;s own telemetry describes an \u0026ldquo;infinite workday\u0026rdquo; of interruptions on average every two minutes. The tools at play are different from the previous section. Slack and Teams sell alleged efficiency to employers, not attention to advertisers. But the move is the same in different clothes as these systems optimize by treating human attention as the variable to exploit.\nI propose the following hypothesis: a day spent hopping between a hundred channels at work spends reserves of will we need to resist the feed. Most of what happens in these channels at work is coordination, not creation. It seems we haven’t learned much from the lessons of Herbert Simon’s Designing Organizations for an Information-Rich World again as there must be a better way to get alignment that costs less attention. But it means a knowledge worker arrives in the evening not merely tired worn down in precisely the way that makes them easy prey for the infinite scroll. The hours on my contract did not grow. The number of things competing for my attention every hour did, and what this took away came out of the reserve meant for life after work.\nWhich is another reason why relaxing, deep leisure never quite arrives. Measured working hours in my part of Europe are among the lowest in history at about 1,340 a year in Germany, 1,750 in Czechia. On paper, Keynes\u0026rsquo;s promised leisure is already here. Why does nobody feel it? Partly because those channels don\u0026rsquo;t close at six unless we deliberately close them, so the official count understates the hours spent working. Partly because leisure restores only in uninterrupted blocks of genuine detachment - Brigid Schulte\u0026rsquo;s distinction between time and \u0026ldquo;contaminated time.\u0026rdquo; Our free hours exist, unfortunately they arrive pre-shredded. A phone\u0026rsquo;s mere visible presence measurably drains cognitive capacity and dulls a shared meal and it does not even need to buzz. What we should ask for is not only fewer working hours, it is the permission to stay still, without needing to catch up on anything. Unfortunately, stillness is exactly what the attention economy cannot permit because an unstimulated moment is inventory going unsold.\nI would argue here that an employer must be the steward of employees’ attention and willpower as surely as of their safety. Almost none currently behave that way even if it were to their own benefit. No one holds the job of protecting our capacity to think while half a dozen systems are wrecking it. We resolved long ago that a workplace must not wreck a body by Friday. What if we resolved it would not wreck the mind either?\nHunted beyond sustainable The siege for attention reaches family life as well and I feel it sharply. Children are handed games and series built on the variable-reward playbook. By whom? Through peers and their pressure but also often by a tired parent whose attention was already spent at work, grateful for anything that buys ten quiet minutes, often spent doom scrolling. The household spends its evenings negotiating what is allowed. Setting the use limit triggers a fight more often than not. But notice what that negotiation is. It is not only parent versus child. It is parent versus several thousand engineers whose performance targets are set against the parent who make parent\u0026rsquo;s job harder. Both sides of the family table are under the siege, the parent arrives worn down, running on FOMO and thin patience, the child faces content built to be harder to resist than anything the previous generations had to contend with. The household willpower cannot win alone. That would be the same category error as expecting discipline to prevail over food engineered for overconsumption. Ironically, even the winners of the rat race don not get to rest, since rest is exactly what the attention merchants came for. We are hunted beyond sustainability.\nThe externality: who pays for the harvested hour Everything so far could be interpreted as bad private bargains, people are selling their evenings too cheap. The deeper economics is uglier. Attention harvesting is a textbook negative-externality industry, and the pollution analogy seems fitting. The platform books the revenue, the costs land elsewhere.\nLet\u0026rsquo;s consider the measurable pieces. Distracted driving kills more than three thousand people a year in the US alone, third parties, literally dying. RAND study puts insufficient sleep\u0026rsquo;s cost to the US economy at some 411 billion dollars a year. It names technology among the causes but doesn\u0026rsquo;t quantify its share, the evidence from 2005 supports the point. Netflix\u0026rsquo;s chief once admitted his main competitor was sleep and they were winning. Psychologists have since named and measured what gets in the way of sleep: bedtime procrastination, delaying sleep while expecting to be worse off for it. Not insomnia, not shift work, not a crying child, just the failure to stop and go to bed. When people must resist more temptations, they are more likely to miss their intended bedtime. The reserve that would have carried them to bed on time was spent earlier, on everything else. Elsewhere, school phone bans raised test scores by the equivalent of adding an extra hour of schooling per week and gains concentrated among the weakest students as attention loss hits the disadvantaged hardest. Norwegian data show phone bans improved girls\u0026rsquo; grades and mental health while they reduced bullying.\nAt home we see caregivers absorbed in phones respond and converse less while we know that conversational turns are crucial for language development of children. They are also more likely to react more harshly to children\u0026rsquo;s bids for attention. The phenomenon even has a research name, technoference, \u0026ldquo;Zombie parents\u0026rdquo; being the folk term, the psychologists\u0026rsquo; reference point is the still-face experiment, an infant\u0026rsquo;s distress in front of an unresponsive caregiver.\nEven if we take only the portion the economic model attributes to self-control issues rather than to genuine preference: 31 percent of time from earlier, from a study of 2000 US users. At roughly two and a half hours daily of social media and video feeds extrapolated across over five billion users world-wide, that\u0026rsquo;s on the order of 1.7 trillion regretted hours a year or also an average around 48 minutes a day per average user. What is that regretted time portion worth? Value an hour at a deliberately modest two to five euros, the shadow price of leisure is contested so feel free to pick another number, and we get three and half to eight trillion euros of misallocated time annually, before any other mentioned third-party cost. Meanwhile platforms monetise that same spent attention on the order of five to twenty cents per user-hour, a figure nobody publishes, estimated by dividing platform advertising revenue by total user-hours. Hours valued in euros, sold for cents: one of the worst trades in economic history, running every evening, at times well into the night. It is a Fermi estimate, not a measurement, but a conservative one. The magnitude survives any reasonable parameters.\nAll of this negative toll is absorbed by families, schools, employers, health systems, strangers on the road. What my inner accountant finds most offensive: the harvesting sector isn\u0026rsquo;t even large for the damage it inflicts, advertising is about one percent of world GDP. But a sector\u0026rsquo;s GDP share tells us nothing about its externalities, tobacco\u0026rsquo;s revenue also never reflected its health costs. We do not restrict ourselves when checking if a chemical plant \u0026ldquo;adds value\u0026rdquo; to only check its revenue, we ask what it emits. Time to ask the same of the ad industry!\nThe feed that ate the evening and became the shop I thought that since consumption is bounded by time and a weekend of binge-watching yields a streaming service perhaps two euros of revenue share it impacts consumption because that same weekend at a sport match, a concert, or a restaurant visit with friends would put ten to fifty euros into the local economy. So past some threshold, attention capture should shrink total consumption. I was wrong.\nAggregate data refused to cooperate with my theory, consumption\u0026rsquo;s GDP share has not fallen. Aggregates are crude measures, treat a whole society and there is no control group, and spending can hold steady while shifting from restaurant meals to the bigger television for watching the series. But the main reason the claim fails is that the industry solved it already. Around 2020, the screen became the shop. Retail media, advertising inside shopping platforms, reached about 178 billion dollars in 2025, surpassing all of global television advertising for the first time. Ads inside the shop now outsell the medium that defined advertising for seventy years. Social commerce, buying without leaving the feed, approaches a trillion dollars globally. The couch is not where consumption goes to die anymore. The platforms solved their bounded-consumption problem. Nobody considered the impact on our well-being.\nEven worse, the feed displaces a particular kind of attention use - time-intensive, local, labor-intensive, social. Face-to-face socializing among American adults has fallen by more than a fifth since 2003 and it is down nearly half among teenagers. Britain has lost more than half its nightclubs since 2013, though reasons may vary. Cinema sits structurally below pre-streaming levels. The young go out, drink, and date less across the rich world. The money still gets spent but it flows to near-zero-marginal-cost digital goods and shipped parcels, concentrating in a few platforms instead of circulating through the venues and clubs and local sport that employ neighbors and, crucially, form relationships. Harvard\u0026rsquo;s eighty-five-year Study of Adult Development names warm relationships the strongest predictor of late-life health and happiness. Lacking social connection carries mortality risk comparable to smoking or alcohol abuse and higher than obesity or physical inactivity. The consumption the feed cannibalizes is exactly of the relationship-forming kind, and the loneliness it produces gets soothed by more feed. The circle is vicious. It hollows out the social-infrastructure economy while aggregates report that things are fine.\nThere is even a dark comedy at play here: a restaurant advertising on Instagram is bidding for the attention of a person whom Instagram engineered to stay on the couch.\nThe broken dashboard So what are we optimizing for? Overwhelmingly, GDP, a metric whose own architect, Simon Kuznets, warned in 1934 against using it as a welfare measure. The critique is established by now: the Stiglitz–Sen–Fitoussi commission, New Zealand\u0026rsquo;s wellbeing budget, Wales\u0026rsquo;s legally binding Future Generations Act, which once killed a motorway. The sober record: dashboards without binding force rarely survive a change of government, and no alternative metric yet disciplines a finance ministry the way bond markets do.\nBut the attention economy exposes a sharper failure. GDP counts attention capture as production as ads revenue is GDP. GDP counts also much of the damage as further production: therapy for the anxious teenager, sleep medication, focus apps sold to fight the distractions other apps created. A closed, vicious loop: the harvesting shows up as growth, the harm shows up as more growth as it gets treated, and nothing in the big number ever signals anything is amiss. The instrument does not just miss the harm, it books the harm\u0026rsquo;s consequences as success, rewarding the extraction twice.\nThe pledge we already made The most damning exhibit of misallocated attention cannot be found on a screen. It is roughly seven to eight hundred million people still living in extreme poverty while the sums needed to end it are rounding errors elsewhere. Global military spending hit a record 2.7 trillion dollars in 2024. The UN estimates that closing the extreme poverty gap would cost just under 300 billion a year, a tenth of the military budget, less than a third of world advertising spend, ending hunger, perhaps 93 billion. The delivery objection is two decades stale because direct cash transfers are now backed by a rigorous evidence base, with well over 80 percent of funds reaching recipients through the best operators.\nMy core claim here is normative: a civilization professing humanistic values cannot accept a tenth of itself living precarious, foreshortened lives when the price of ending it is this low. Weigh that as you will. One part is not normative, though, and I find it unanswerable. In 1970, rich countries pledged in a UN resolution, reaffirmed as recently as July 2026, to give 0.7 percent of national income (GNI) in development assistance. In 2024, exactly four countries met the pledge: Norway, Luxembourg, Sweden, Denmark. Germany, having met it from 2020 to 2023, slipped below, and its 2025 coalition agreement goes further than dropping the pledge, it commits to lowering the aid ratio. My native Czechia was nowhere close to meet the pledge yet it cut its aid twenty-nine percent in a single year and sits currently at roughly a quarter of the target. Now the arithmetic that should end every dinner-party debate, if signatories simply honored the pledge made fifty-six years ago, the additional funds, some 230 billion dollars a year these days, would nearly close the entire extreme poverty gap. No new promise required. The old one would do it. Unfortunately, poverty has no marketing budget.\nWhat can actually be done The system is misaligned. Nothing can be done. End of story. Let me make the case for hope.\nSocieties have drawn a line before, for a slower medium. EU broadcasting law caps television advertising at 20 percent of airtime, a norm set decades ago for exactly the reason above, that unlimited advertising degrades the thing it interrupts. No equivalent exists for the medium that replaced television a long time ago. There is no cap on how much of our feed can be an ad because there is no clock hour to measure it against. The feed does not broadcast in scheduled blocks, it flows, and a flow has no percentage.\nAnother precedent can be found along my own travels. Society already accepts attention zoning where the stakes are visible and driver\u0026rsquo;s attention is considered safety-critical. Germany prohibits advertising along the Autobahn (motorways). Czechia banned billboards beside its motorways in 2017, and thousands came down. Now one can even notice where one is going and enjoy the scenery. Nobody died because the billboards were banned, people died because they were not. São Paulo removed some fifteen thousand billboards from the whole city in 2007. Commerce did not collapse and the city has kept outdoor advertising tightly controlled ever since. We banned the billboard beside the motorway while we built an infinitely scrolling one into every pocket. Let’s explore how to put limits on the feed’s addictiveness and on the attention and agency theft.\nSo what can we do? Comprehensive category bans work. Comprehensive tobacco ad bans measurably cut consumption, while partial ones mostly displace spending into other channels, Quebec\u0026rsquo;s forty-year ban on advertising to children measurably lowered fast-food consumption, and the UK\u0026rsquo;s junk-food ad ban went live last year. The EU\u0026rsquo;s own attempt shows what stands in the way. It legislated a comprehensive tobacco advertising ban in 1998 but the Court of Justice annulled it, holding that the Union could reach only advertising with a cross-border dimension. The 2003 replacement therefore covered press, radio and internet while leaving billboards, cinema and point-of-sale to member states. Prevalence fell anyway, from around 31 percent in 2006 to 24 percent in 2023, though numbers differ country to country and taxes, smoke-free laws and packaging rules all arrived in the same window and nobody has separated their contributions.\nFor the digital core, the EU has assembled machinery I\u0026rsquo;d have called inconceivable a decade ago: the Digital Services Act bans manipulative interfaces, mandates independent audits and researcher access to platform data, and carries fines up to six percent of global turnover. In February 2026 the Commission found TikTok\u0026rsquo;s design preliminarily in breach, the first enforcement aimed at a platform\u0026rsquo;s architecture rather than its content. A month later, American juries found Meta and YouTube liable for addictive design. A Digital Fairness Act targeting dark patterns outright is expected within the year, and the intensity of lobbying against it shows the business model feels the threat. Australia\u0026rsquo;s under-sixteen social media ban has been in force since December 2025, France and Portugal wrote a right to disconnect into their labor law, school phone bans spread country by country while strongest gains accrue to the weakest students. The regulation is the priority instrument for a reason: whether a feed autoplays or an app buries its off-switch is externally observable by anyone with a phone which makes bans continuously enforceable against a moving target of evolving platforms.\nThe second instrument to consider is a tax on targeted-advertising revenue, and it earns its place for a narrower reason. France taxes digital services revenue at three percent (an increase to five is under discussion), Britain at two, Austria at five specifically on online advertising, and Maryland runs a progressive version rising from 2.5 to 10 percent by company size. Such taxes are collected from audited accounts like any VAT requiring nobody\u0026rsquo;s phone inspection. Though such a tax does not discipline addictive design or curbs platform\u0026rsquo;s incentive to maximize time-on-app to selling impressions regardless of whether those impressions are priced as targeted or contextual. Taxing the targeted kind changes the ad\u0026rsquo;s price, not the payoff from keeping you scrolling. A platform could run zero targeted advertising and still deploy an infinitely addictive feed, since ranking runs on first-party engagement data unrelated to how adjacent ads are sold. What the tax does buy? It curbs surveillance-based profiling specifically, and raises real money. At rates already running in Austria and France it may be enough to fund publicly accountable alternatives to services we all learned to depend on: a Wikipedia-style-governed search engine, ad-free email, a maps platform run like OpenStreetMap, so people are not required to surrender their data for basic service everyone needs.\nA separate question is who bears the burden of proving an ad was \u0026ldquo;targeted.\u0026rdquo; The workable answer borrows machinery built for a different law as GDPR and the ePrivacy directive already require platforms to log, ad by ad, whether personalized or contextual-only consent applied, the same records the industry\u0026rsquo;s own consent-management tools keep today. A tax authority can presume any ad built on a third-party identifier or behavioral segment is targeted by default with burden to prove otherwise resting on the platform. The same asymmetry transfer-pricing rules are already imposed on firms holding the underlying data. The harder problem: the base is shrinking under its own feet. Third-party cookies are being phased out and the largest platforms consolidate targeting onto first-party data. A tax drawn narrowly around third-party tracking risks taxing a dying category while missing where targeting is heading. Hence this becomes a second-order instrument, worth building only once the addictive design bans are enforced.\nOne objection deserves to be addressed. Advertising currently funds many free services everyone uses. Search, email, and maps are commodity infrastructure, and we see the commons already runs infrastructure at planetary scale. Wikipedia operates one of the world\u0026rsquo;s most-visited sites on a budget that would be a rounding error of Google. OpenStreetMap provides the base for much of the mapping industry. Linux runs the world\u0026rsquo;s servers. India\u0026rsquo;s public payment rails process more transactions than Visa and Mastercard combined in that country. A court has ruled Google an illegal monopoly in search, a service whose quality decline under engagement-metric management is a documented saga. I personally gladly pay Czech Seznam for premium mapy.com. But it is all not just rosy in public hands. Germans are forced to pay some 220 euros a year for public TV and radio broadcasting whether they use it or not. It is widely criticized by as bloated and of low quality in comparison to for instance BBC. So funding and governance are separate design choices and public digital infrastructure will need Wikipedia-grade governance.\nChoosing what to optimize for Attention is one of the constraints on good life as it provides for rest, relationships, presence with our children, even the value we extract from what we already have. We have built an industry that mines attention. A trillion-dollar arms race, substantially zero-sum for the firms fighting over our consumption, that dillutes our leisure, sets parents against engineering teams, and imposes externalities on our society that plausibly exceed the sector\u0026rsquo;s entire revenue. What the feed impacts is disproportionately the local, social, relationship-forming life the longest study of human well-being identifies as its core. The resulting loneliness gets treated, and the treatment is booked as GDP, so the extraction is recorded twice, as growth.\nFortunately, there are no laws of physics mandating this continues. It is made possible by wrong incentives and we put different incentives in place. The billboards came down along my motorways. The right to disconnect is now law in two countries. The enforcement machinery against addictive design is being built as I write. The tax regime for marketing exist in three countries. The commons runs infrastructure at global scale. Let\u0026rsquo;s protect our attention with same care as we protect our physical health. There will still be enough ways to market the latest and greatest gadgets, services, and tools to us without taking away our evenings.\nBefore closing, let\u0026rsquo;s look at what a protected hour is actually good for once we win it back. It is not for more consumption. The whole argument has been that consumption is not the scarce thing in the developed world. A mind capable of sustained attention is capable of weighing a decision rather than reacting to it, hearing someone out rather than listening to reply, sitting with a piece of music or a view long enough for it to get absorbed in it instead of it being logged and swiped past. That deliberate capacity is the raw material of the relationships the Harvard study found decisive, of the patience a shared civic life asks of its citizens, of noticing a place or a problem closely enough to actually understand it and act on it.\nThe deeper choice is ours. A society optimizes for what it measures, funds what it notices, and notices what its attention is capable of noticing. Mine, this evening, will be in the room with the toys, attending to the two boys playing in it.\nAcknowledgments: many thanks to Ulrike Geiber, Michal Bída and Kateřina Bída for feedback and suggestions. And to Claude Opus 5 for help with research and drafting.\nAppendix: Verification of factual claims Method: each claim checked against primary or authoritative secondary sources; recent statistics re-verified by web search during this project. Status codes: ✓ verified · ✓ verified with correction applied · ~ established finding, standard citation · F = Fermi estimate, assumptions stated in text.*\nIntroduction\nExtreme poverty ~38% of humanity in 1990, ~one in ten today. ✓* World Bank data: 37.8–38% in 1990 under the $2.15/day (2017 PPP) line, ~8.5% (~690M) in 2024. Note: in June 2025 the World Bank revised the extreme poverty line to $3.00/day (2021 PPP), which raises the current count to roughly 800–840M (~10%). Linder: consumption requires time; yield per possession falls. ~ Staffan B. Linder, The Harried Leisure Class, Columbia University Press, 1970. Simon: wealth of information → poverty of attention. ~ Herbert A. Simon, \u0026ldquo;Designing Organizations for an Information-Rich World,\u0026rdquo; in Computers, Communications, and the Public Interest, Johns Hopkins Press, 1971. The arms race\nInformative vs. persuasive/combative advertising distinction. ~ Standard industrial-organization literature; survey in Bagwell, \u0026ldquo;The Economic Analysis of Advertising,\u0026rdquo; Handbook of Industrial Organization vol. 3 (2007). Note: no authoritative empirical study measures a clean split between \u0026ldquo;informative\u0026rdquo; and \u0026ldquo;persuasive/combative\u0026rdquo; ad spend across the economy - the distinction is well established theoretically (Marshall, Chamberlin 1933 onward) but resists clean measurement in practice, since classifying a real ad requires a judgment call researchers make differently study to study. Cigarette ad bans raised tobacco profits. ~ The 1971 US broadcast ban is the canonical case. Industry advertising expenditure fell roughly 25% in the first year while tobacco stock returns reached abnormal highs, indicating markets read the ban as favourable to the firms; aggregate demand was inelastic and sales did not fall. Primary citation: Mitchell \u0026amp; Mulherin (1988), event study of stock returns. Hamilton (1972), Review of Economics and Statistics, is a separate finding about advertising\u0026rsquo;s weak effect on aggregate demand and should not be cited for the profits claim. Theoretical treatment: Tirole, The Theory of Industrial Organization. Note: US cigarette consumption rose ~4% 1971–73, partly because the Fairness Doctrine\u0026rsquo;s mandatory anti-smoking counter-advertising ended with the ban; and the profit benefit faded within roughly five years as spending migrated to print and sponsorship - both consistent with the ban being partial rather than comprehensive. Global ad spend \u0026gt;$1T in 2025; ~1% of world GDP; ~3/4 digital; big three capture ~2/3 of new dollars. ✓ WPP Media \u0026ldquo;This Year Next Year\u0026rdquo; (Dec 2025): global ad revenue $1.14T in 2025 (WARC: $1.17T). World GDP ~$115T → ~1%. 6a. EU television advertising capped at 20% of airtime. ✓ Audiovisual Media Services Directive, Article 23(1): advertising and teleshopping may not exceed 20% of a given broadcasting period. Originally a 12-minutes-per-clock-hour rule; the 2018 revision (Directive (EU) 2018/1808) recast it as a cumulative daily limit - 144 minutes between 06:00–18:00, 72 minutes between 18:00–24:00, unrestricted overnight. No equivalent quantity cap exists for online/social media advertising. Facebook deactivation experiment: happier, then valued platform less. ~ Allcott, Braghieri, Eichmeyer \u0026amp; Gentzkow, \u0026ldquo;The Welfare Effects of Social Media,\u0026rdquo; American Economic Review 110(3), 2020. 7a. AI companion apps: Replika CEO on \u0026ldquo;long-term commitment\u0026rdquo;/\u0026ldquo;marriage\u0026rdquo;; FTC inquiry Sept 2025; 44-state AG coalition; Character.AI/Google settlement Jan 2026. ✓ Replika CEO Eugenia Kuyda, reported in The Verge and cited in CNN coverage of Senators Padilla and Welch\u0026rsquo;s April 2025 letter to AI companion firms. FTC opened a formal inquiry into generative AI chatbot harms to minors, September 2025. Bipartisan 44-state attorneys general letter to major AI firms, 25 August 2025. Character.AI and Google settled multiple lawsuits - including the case brought by Megan Garcia over her son\u0026rsquo;s death - announced early January 2026; terms undisclosed. Subsequent state actions continued into 2026 (e.g., Pennsylvania\u0026rsquo;s May 2026 suit over chatbots posing as licensed medical professionals). ~31% of social media time used against user\u0026rsquo;s own wishes. ~ Allcott, Gentzkow \u0026amp; Song, \u0026ldquo;Digital Addiction,\u0026rdquo; American Economic Review 112(7), 2022. 8a. The model predicts eliminating self-control problems would reduce use by 47.5 minutes/day, or 31% of baseline. The measure is \u0026ldquo;FITSBY\u0026rdquo; - Facebook, Instagram, Twitter, Snapchat, Browser, YouTube - in a US sample. World-wide application is the essay\u0026rsquo;s extrapolation. Heavy media multitasking associated with worse sustained attention and distractor-filtering; effect modest on average, causality contested. ~ Ophir, Nass \u0026amp; Wagner, \u0026ldquo;Cognitive control in media multitaskers,\u0026rdquo; PNAS 106(37), 2009. Caveat: a subsequent meta-analysis (Wiradhany \u0026amp; Nieuwenstein, Attention, Perception \u0026amp; Psychophysics, 2017) found a small pooled effect (d≈0.17) and noted some replication attempts failed to reproduce the original finding; direction of causality remains unresolved. Presented in the essay as suggestive, not settled. Hunted beyond sustainable\nHyperactive hive mind. ~ Cal Newport, A World Without Email, 2021. Attention span on screen tasks 2.5 min (2004) → 47 seconds; ~20+ minutes to refocus. ~ Gloria Mark, Attention Span, 2023; refocus figure from Mark, Gudith \u0026amp; Klocke (CHI 2008), commonly reported as ~23 minutes. Microsoft \u0026ldquo;infinite workday\u0026rdquo;: interruptions ~every 2 minutes (275/day). ✓ Microsoft Work Trend Index Special Report, \u0026ldquo;Breaking down the infinite workday,\u0026rdquo; June 2025. Annual hours worked: Germany ~1,340; Czechia ~1,750. ✓ OECD Employment Outlook data, 2023–24. Keynes\u0026rsquo;s 15-hour-week prediction. ~ Keynes, \u0026ldquo;Economic Possibilities for our Grandchildren,\u0026rdquo; 1930. Contaminated time. ~ Brigid Schulte, Overwhelmed: Work, Love, and Play When No One Has the Time, 2014. Mere presence of a phone drains cognition; dulls shared meals. ~ Ward, Duke, Gneezy \u0026amp; Bos, \u0026ldquo;Brain Drain,\u0026rdquo; JACR 2017; Dwyer, Kushlev \u0026amp; Dunn, JESP 2018. Parent–child screen conflict prevalence. ~ Common Sense Media and Pew Research surveys; clinical characterization in Radesky\u0026rsquo;s work (below). The externality\nDistracted driving \u0026gt;3,000 US deaths/year. ✓ NHTSA: 3,308 (2022), 3,275 (2023). Insufficient sleep costs the US economy up to $411bn/year. 🌐 Hafner, Stepanek, Taylor, Troxel \u0026amp; van Stolk, Why Sleep Matters - The Economic Costs of Insufficient Sleep, RAND Corporation, 2016 (RR-1791; free at rand.org). Figure is ~2.28% of US GDP; five OECD countries together up to $680bn; Germany ~$60bn (1.56% of GDP). Covers insufficient sleep from all causes. Technology appears among the contributing trends discussed - communication technology extending availability beyond the workplace, and new technology \u0026ldquo;possibly chipping away at hours reserved for sleep\u0026rdquo; - but no share of the cost is apportioned to it, and the supporting citation is National Sleep Foundation (2005), predating the smartphone. No updated cross-country estimate exists; newer national estimates (e.g. China, 4.62% of GDP in 2018) use different methods and aren\u0026rsquo;t directly comparable. Bedtime procrastination. 🌐 Construct and validated scale: Kroese, De Ridder, Evers \u0026amp; Adriaanse, \u0026ldquo;Bedtime procrastination: introducing a new area of procrastination,\u0026rdquo; Frontiers in Psychology 5:611 (2014), open access; and Kroese, Evers, Adriaanse \u0026amp; De Ridder, Journal of Health Psychology 21:853–862 (2014). Defined as needlessly and voluntarily delaying bedtime despite expecting to be worse off; conceptualised as self-regulation failure and associated with low trait self-control - converging with the \u0026ldquo;self-control problems\u0026rdquo; construct in Allcott, Gentzkow \u0026amp; Song (2022) from a separate literature. Depletion predicts it. \u0026ldquo;Too Depleted to Turn In: The Relevance of End-of-the-Day Resource Depletion for Reducing Bedtime Procrastination,\u0026rdquo; Frontiers in Psychology 9:252 (2018), open access. Daily-diary design: participants who attempted to resist more desires during the day were more likely to miss their intended bedtime. Correlational within-person, not experimental. Bidirectionality. Longitudinal follow-up study of college students, Frontiers in Psychology (2025), doi 10.3389/fpsyg.2025.1690574. Mobile phone use and bedtime procrastination predict each other across waves - phone use depletes self-control resources; lost sleep prompts compensatory phone use. \u0026ldquo;Sleep is our competition.\u0026rdquo; ✓ Netflix CEO Reed Hastings, April 2017, widely reported. School phone bans: test scores +~1 week of schooling, gains concentrated among low achievers. ~ Beland \u0026amp; Murphy, \u0026ldquo;Ill Communication,\u0026rdquo; Labour Economics 2016. Norway: bans improved girls\u0026rsquo; GPA and mental health, reduced bullying. ~ Abrahamsson, Norwegian School of Economics working paper, 2024. Technoference; conversational turns; still-face paradigm. ~ McDaniel \u0026amp; Radesky, \u0026ldquo;Technoference,\u0026rdquo; Child Development 2018; Radesky et al., Pediatrics 2014; Romeo et al., Psychological Science 2018; Tronick et al., 1978. Platform monetisation of €0.05–0.20 per user-hour. F - author\u0026rsquo;s calculation, not a published figure. Method: platform advertising revenue divided by total user-hours. Meta alone yields roughly €0.18/hour (~$160bn annual ad revenue across ~3.3bn daily users at ~0.5–1h/day). Global social advertising revenue divided by DataReportal\u0026rsquo;s aggregate of 15bn user-hours per day yields nearer €0.05/hour. The range reflects genuine variation across platforms and revenue definitions. Fermi estimate: 1.7T regretted hours ≈ €3.5–8T/yr; platform monetization €0.05–0.20/user-hour. F Inputs: 31% regretted share (item 8); ~2.66h/day average social media use across ~5.66B users (DataReportal 2026), caveat some users may be double counted; time valuation €2–5/h assumed. The feed that ate the evening\nConsumption share of GDP stable in streaming era. ✓ World Bank/OECD national accounts. Facebook rollout \u0026amp; broadband natural experiments. ~ Braghieri, Levy \u0026amp; Makarin, AER 2022; Falck, Gold \u0026amp; Heblich, AER 2014; Campante, Durante \u0026amp; Sobbrio, JEEA 2018; Guriev, Melnikov \u0026amp; Zhuravskaya, QJE 2021. Retail media ≈ $178B in 2025, surpassing total TV advertising for the first time. ✓ WPP Media, This Year Next Year, Dec 2025; WARC\u0026rsquo;s independent estimate $174.9B consistent. The crossover holds under WPP Media\u0026rsquo;s and WARC\u0026rsquo;s definitions, in which \u0026ldquo;total TV\u0026rdquo; includes linear and streaming. Trackers define retail/commerce media slightly differently, so the exact crossover year is definition-dependent. Social commerce approaching $1T globally. ✓ Industry estimates (Statista, McKinsey) place global social commerce GMV in the $700B–$1T+ range for 2024–25. Face-to-face socializing: adults −\u0026gt;20% since 2003; teens ~−45–50%. ✓ American Time Use Survey (BLS), analyzed in Derek Thompson, \u0026ldquo;The Anti-Social Century,\u0026rdquo; The Atlantic, Feb 2025. UK nightclubs: more than half lost since 2013. ✓ NTIA/CGA-NIQ: ~1,700 clubs (2013) → 787 (2024). Harvard Study of Adult Development: relationships the strongest predictor. ~ Waldinger \u0026amp; Schulz, The Good Life, 2023. Lacking social connection ≈ 15 cigarettes/day mortality risk. ~ Holt-Lunstad, Smith \u0026amp; Layton, \u0026ldquo;Social Relationships and Mortality Risk: A Meta-analytic Review,\u0026rdquo; PLoS Medicine 2010 (DOI 10.1371/journal.pmed.1000316); 148 studies, 308,849 participants. The study measures social relationships broadly - objective social integration plus perceived support - not loneliness as a subjective state; the two are distinct constructs and Holt-Lunstad has publicly corrected the common conflation. The fifteen-cigarette figure is a benchmark comparison of odds ratios across meta-analyses, not a direct measurement. For loneliness specifically, see Holt-Lunstad et al., Perspectives on Psychological Science 10(2), 2015. The broken dashboard\nKuznets\u0026rsquo;s 1934 warning. ~ Kuznets, National Income, 1929–1932, report to the US Senate, 1934. Stiglitz–Sen–Fitoussi; NZ Wellbeing Budget; Wales Future Generations Act killed a motorway. ✓ Commission report 2009; NZ Wellbeing Budget 2019; Well-being of Future Generations (Wales) Act 2015. The pledge we already made\nMilitary spending $2.7T in 2024. ✓ SIPRI, April 2025. Extreme poverty gap \u0026lt;$300B/yr; hunger ~$93B/yr. ✓ UN Secretary-General\u0026rsquo;s report The Security We Need (Sept 2025). Cash transfers: rigorous evidence, \u0026gt;80% reaches recipients. ~ Haushofer \u0026amp; Shapiro (QJE 2016); GiveDirectly Kenya long-term UBI study; GiveDirectly reports ~80–83 cents per donated dollar delivered. 0.7% GNI pledge: UN GA Resolution 2626 (1970), reaffirmed 2005/2015. ✓ OECD/ONE Campaign documentation. 2024: exactly four countries met 0.7%. ✓ OECD preliminary 2024 ODA data: Norway 1.02%, Luxembourg 1.00%, Sweden 0.79%, Denmark 0.71%. Germany: 0.7%+ from 2020–2023, 0.68% (2024); the 2025 coalition agreement commits to lowering the ODA ratio. ✓ OECD DAC1 (grant equivalent), live data September 2026 (earlier vintages, e.g. OECD Development Co-operation Profiles 2025, gave 0.67%); Eurostat sdg_17_10 agrees. Koalitionsvertrag 2025 (CDU/CSU/SPD), section 5.1: \u0026ldquo;Aufgrund der Notwendigkeit, den Haushalt zu konsolidieren, muss eine angemessene Absenkung der ODA-Quote erfolgen\u0026rdquo;, and in the savings list \u0026ldquo;Einsparung bei der ODA-Quote\u0026rdquo;. https://www.spd.de/fileadmin/Dokumente/Koalitionsvertrag2025_bf.pdf Czechia: ODA cut ~29% in 2024; ~quarter of the target. ✓ OECD preliminary 2024 data. 0.7% kept by all ≈ +$231B/yr ≈ nearly the poverty gap. ✓ ONE Campaign ODA analysis. The ~$231bn shortfall (ONE Campaign) and the \u0026lt;$300bn poverty gap (UN) are computed by different bodies on different bases and are not directly commensurable. Hence \u0026ldquo;Nearly close\u0026rdquo; is a defensible formulation. What can actually be done\nGerman Autobahn advertising ban. ✓ §33 StVO. Czech motorway billboard ban 2017. ✓ Act 196/2012 amending the Roads Act, effective 1 Sept 2017. São Paulo Lei Cidade Limpa: ~15,000 billboards, ~70% approval. ~ Enacted 2006, effective 2007. Tobacco: comprehensive ad bans cut consumption; partial bans leak. Saffer \u0026amp; Chaloupka, J. Health Economics 2000 - 22 OECD countries; comprehensive bans reduce consumption, limited sets have little or no effect. Their model estimated the forthcoming EU directive\u0026rsquo;s effect at 6.9%. Retrospective: the directive as implemented (2003/33/EC) was not comprehensive - the ECJ annulled the 1998 version, limiting the EU to advertising with a cross-border dimension, so billboards, cinema and point-of-sale stayed national. EU27 prevalence fell 31%→24% (2006–2023), but taxes, smoke-free laws and the 2014 Tobacco Products Directive arrived in the same window and no evaluation isolates the advertising ban. Quebec child-advertising ban lowered fast-food consumption. ~ Consumer Protection Act 1980; Dhar \u0026amp; Baylis, Journal of Marketing Research 2011. UK junk-food ad ban live. ✓ HFSS pre-21:00 TV ban + paid online restrictions, in force 1 Oct 2025. DSA: manipulative-interface ban, audits, researcher access, 6% fines. ✓ Regulation (EU) 2022/2065, Arts. 25, 37, 40, 74. TikTok preliminary DSA findings on addictive design, Feb 2026. ✓ European Commission preliminary findings. US jury verdicts vs. Meta and YouTube on addictive design, Mar 2026. ✓ JCCP social-media personal-injury litigation. Digital Fairness Act proposal expected Q4 2026; heavy lobbying documented. ✓ Commission work programme 2026; Corporate Europe Observatory report, Feb 2026. Australia under-16 social media ban in force Dec 2025. ✓ Online Safety Amendment (Social Media Minimum Age) Act 2024. Right to disconnect: France 2017, Portugal 2021. ✓ Loi Travail (El Khomri); Portugal Law 83/2021. DSTs: France 3% (5% increase repeatedly proposed), UK 2%, Austria 5%, Maryland progressive 2.5–10%. ✓ France DST (2019, 3%; 5% rate voted for 2025, revisited again in 2025/26 budget cycles); UK DST (2020, 2%); Austria Digitalsteuergesetz (2020, 5%); Maryland Digital Advertising Gross Revenues Tax (2021), tiered 2.5%/5%/7.5%/10% by revenue thresholds - upheld on its substance in Chamber of Commerce v. Lierman (4th Cir., Aug 2025), only the invoice-disclosure-gag provision struck down. Romer\u0026rsquo;s progressive targeted-ad tax. ~ Paul Romer, \u0026ldquo;A Tax That Could Fix Big Tech,\u0026rdquo; New York Times, May 2019. 55a. The tax does not discipline addictive interface design. A platform\u0026rsquo;s incentive to maximize engagement comes from selling ad impressions, not from the targeted/contextual pricing distinction; taxing targeted revenue changes ad pricing, not the payoff from maximizing time-on-app. The tax\u0026rsquo;s justification is data-exploitation harm and revenue-raising; mechanism bans (items 48–49) are the load-bearing instrument against addictive design. 55b. Burden of proof for \u0026ldquo;targeted\u0026rdquo; classification: GDPR/ePrivacy consent-logging as existing infrastructure. ✓ The IAB Europe Transparency and Consent Framework already logs, per impression, whether personalized or contextual-only consent applied - reusable for tax classification via a rebuttable presumption, analogous to the documentary burden transfer-pricing rules place on firms holding the underlying data. 55c. Viability risk: third-party cookie deprecation and first-party/walled-garden consolidation. ~ A tax base defined around third-party tracking risks taxing a shrinking category while first-party targeting becomes the dominant mode. Open design risk, not resolved here. Wikipedia: top-visited site, ~$180M budget. ✓ Wikimedia Foundation annual expenses ~$178M (FY2023–24). UPI \u0026gt; Visa + Mastercard combined within India. ✓ NPCI: UPI ~172B transactions FY2024–25. Google ruled an illegal monopoly in search. ✓ US v. Google, D.D.C., Judge Mehta, 5 Aug 2024. Search-quality decline account: Ed Zitron (2024), journalistic reconstruction, disputed by Google. Rundfunkbeitrag ~€220/yr; efficiency criticism. ✓ €18.36/month = €220.32/yr; criticism incl. the Zukunftsrat reform recommendations, 2024. BBC comparison is opinion, flagged as such in the essay. Cap-and-trade on engagement-hours: no prior formal proposal found. ✓ Searched; none located. Not included in the essay body - conflates the two now-separated instruments (bans on mechanism, tax on ad revenue) and deserves separate treatment. Choosing what to optimize for\nAll synthesis claims trace to items above; the tobacco-profits point is item 5. ","permalink":"https://wondering-wanderer.eu/posts/26-harvested-hour/","summary":"\u003cp\u003eThere is a room in my flat with more toys than my two boys could meaningfully play with in a year. They are not spoiled by any usual standards - this is what childhood in a developed country looks like. The toys mostly sit there. The scarce ingredient of their play was never the toy. It was attention, the unhurried hour, the absence of stimulus, to be bored long enough for imagination to kick in. Many adults I know live in a scaled-up version of the same room - closets full of gear, gadgets, subscriptions. We own more than we can attend to.\u003c/p\u003e","title":"The Harvested Hour: What We Optimize For, and What It Costs Us"}]