Cheap Truth

Text from GitHub, as it stood on 2026-11-04. Source: https://github.com/SimonKreis/verification-loop/blob/HEAD/cheap-truth.md

Updated October 2026. Facts are stated as of mid-September 2026.

A companion essay to The Verification Loop, which carries the full argument, the thirty cases, and the source ledger. Licensed under CC BY 4.0.

The state that houses Singapore would starve it if it ran the supermarkets. In 2025, 77.2 percent of Singapore’s resident households lived in flats built by the Housing and Development Board (HDB), the government agency that plans the estates, allocates the units, and sells them (SingStat, 2025). The same state, with the same civil service, the same incorruptibility, the same planners, and the same authority, would fail at groceries within a decade, and the failure would take the form every state grocer has taken: shelves that empty by mid-morning, a queue that becomes the true price, a second market in the alley, and a ministry issuing statistics that the queue refutes.

That claim is a prediction, and it has been tested more often than almost any prediction in political economy. The Soviet shop and its queue were the daily verdict on Gosplan for two generations. Venezuela built Mercal, a state-subsidized retail chain launched in 2003 that grew to more than 16,000 outlets in the years of the oil boom (Wikipedia, Mission Mercal); it sold controlled-price goods until it could not stock them, and the country’s GDP per head then fell by roughly three quarters between 2013 and the early 2020s (Economics Observatory, 2024). No state that wrote its own shelf prices has run a grocery store well. Several have run housing, railways, vaccination campaigns, and power grids well. Ownership is constant across those outcomes, so ownership does not explain them.

Neither does competence. Singapore’s planners are as good as planners get, and the failure would not come from stupidity or theft. It would come from the removal of the mechanism that tells a grocer, every hour, which of ten thousand items is overstocked, underpriced, spoiling, or wanted, and that punishes the grocer who ignores the message.

The public versus private axis is dead as a predictor. It predicts nothing inside a country, which is where all the interesting variation lives. The United States, the most private of the rich health systems, runs the most expensive and least verifiable one. In the other direction, France’s state-owned nuclear fleet and its state-run high-speed rail perform to a standard its state-led industrial policy never approached. Nations that abolished markets did fail, but they failed because abolishing markets killed a verifier, not because a ministry held the title deed, and the same verifier can be killed under private title. A classification of nations, even one as serious as Acemoglu and Robinson’s, cannot say why HDB works and a Singaporean state grocer would not, inside the same institutions, run by the same people, in the same year. The variable that does the work varies by sector, not by country, and it is the cost of finding out whether the output is any good.

Every system that sustains performance runs on the same machine, and the machine has three parts. The first is the generator: whatever produces the output, a firm, a ministry, a hospital, a planning bureau, a model. The word is deliberately indifferent to ownership and to ideology, because the machine is. The second is the verifier: an evaluation mechanism external to the generator, cheap enough to run continuously, and hard for the generator to capture. A generator’s own quality department is not a verifier, because it is internal; a regulator whose staff are hired from and return to the industry is a verifier that has been captured, and a captured verifier is a generator wearing a badge. The third part is the binding consequence: the verdict must change the allocation of resources. A verdict that changes nothing is a comment. A verifier whose verdict can be overridden at will by the generator or its patron is a decorative verifier, and decoration is the most common state of oversight in the world.

The health of a system is the product of the three, not the sum. A generator with a live verifier and no consequence learns exactly what it is doing wrong and keeps doing it. A generator with a binding consequence and no verifier is punished at random, which teaches nothing and selects for luck. When all three are present and independent the loop is closed, and closed loops improve without anyone intending improvement, because every bad output is found and costs somebody something. Remove one part and the loop is open: generation continues, feedback dies, complexity accumulates in the space where nothing is checked, and output decays at a rate set by how long the generator can go on describing its own work.

HDB is a closed loop and the state grocer is an open one, and the difference lies entirely in the second and third parts. HDB does not build the flats. It plans the estate, sets the specification, and buys construction through open tender, a price verifier it does not control. The output itself is cheap to verify: a flat exists or does not, was delivered on the announced date or was not, cost what the winning bid said or cost more. The state operates the layer whose output is cheap to verify and buys the layer whose output is expensive to verify through a verifier it cannot rig.

The state grocer inverts every one of those choices. It fixes prices, so the one verifier that runs on every shelf every hour is switched off and replaced by a number chosen in an office. It forbids or crowds out rivals, so the shopper’s exit has nowhere to go except the illegal market, where it becomes a crime instead of a signal. It covers the losses, so the shop that loses money does not close. And it reports its own performance, so the statistic and the queue diverge until the queue wins. Groceries are the easiest output in the world to verify, and the result is not a bad grocer but no grocer, with a building where one used to be.

The reader will meet eight verifiers. Price on a shelf is the first; a tender, which prices a contract among rival bidders, is the second. The shopper who walks to the next store is exit, Hirschman’s word. A published record compared across years or across agencies is a benchmark. A shop that closes when it loses money is bankruptcy. Elections and a press work only as a pair and count as one verifier, election/press. Export verifies against buyers no ministry can lobby; referendum verifies a completed statute ex post, which is exactly what separates it from a veto point. They are instruments, each with a cost to run, each capturable in a known way, and each attachable only to some outputs and not others.

That last clause is the axis on which the whole argument turns, and the reader can carry it in the form of a hand tool. For any system, the three questions are who verifies, what it costs them, and what happens when the verdict is bad. Applied to a bakery, the answers are every customer, nothing, and the customer leaves. Applied to a hospital, the answers are a regulator and, at scale, mortality statistics; a great deal, because a bad outcome is hard to distinguish from a bad patient; and usually nothing, because the ministry covers the deficit. Applied to a national industrial strategy, the answers are nobody, because there is no output to inspect for a decade; an amount nobody has ever paid; and nothing, because by the time the verdict arrives the decision-makers have retired. The three answers, not ownership, predict which of them will still be performing in twenty years.

The third question has a name in the literature. János Kornai called a consequence that will predictably be cancelled a soft budget constraint, shortened here to soft constraint. Softness is not the bailout; it is the expectation of the bailout, and it operates before any money moves. A firm that expects its losses to be covered behaves as if price had no force, and the aggregate of many such firms is the shortage economy, in which the queue and the empty shelf are the equilibrium rather than the accident. The mechanism has no ideology. A bank too large to be allowed to fail is under a soft constraint; so is any generator whose patron cannot commit to walking away, which Kornai, Maskin, and Roland later showed is a problem of commitment, not of virtue: once the money is sunk, refinancing is rational, and everyone downstream knows it (Kornai, Maskin, and Roland, 2003).

The thesis can now be stated in full. Every high-performing economic, political, or organizational system runs on the same machine: a loop in which a generator produces, an external verifier evaluates the output cheaply and continuously, and a binding consequence follows from the verdict. Markets succeed because price is the cheapest, most continuous, least corruptible verifier ever discovered, not because of private ownership per se. States succeed as operators exactly where they preserve verifiers they do not control and consequences they cannot cancel, and fail wherever they destroy or capture them. Market sectors fail by the identical mechanism. The variable that predicts performance is not public versus private, left versus right, big state versus small state; it is the cost, independence, and bindingness of verification, a continuous axis measurable sector by sector inside a single country. That is why the same state can run world-class public housing and would run a catastrophic grocery store.

The paragraph compresses to one sentence. System performance is determined by the cost, independence, and bindingness of verification. Cost is a property of the output, and it is the reason a country’s performance varies by sector. Independence is a property of the verifier’s relation to the generator, and it is the reason a regulator and a world price are not interchangeable. Bindingness is a property of the consequence, and it is Kornai’s leg.

The first of the book’s seven propositions, P1, is that markets are verification machines. Two more follow directly, and they have the same shape. The first of the pair, P2, is that market failure is verification failure. The purest instance on record is the credit rating agency before 2008. The agencies were verifiers by design: their entire product was a verdict, on securities they did not issue, and they were paid by the issuers whose securities they rated. A verifier whose income depends on the generator’s satisfaction with the verdict is a captured verifier by construction, and the issuer-pays model remains dominant (SEC, 2024). The second of the pair, P3, is that state failure is the same failure. Gosplan set targets in physical units; the enterprise reported fulfillment; the ministry that set the target received the report, covered the losses when the report was false, and had every reason not to look. Issuer-paid ratings and Gosplan’s self-reported fulfillment are one mechanism: the generator wrote its own verdict, and the consequence that should have followed a bad verdict was cancelled by a patron who could not afford to let it bind. Diagnose broken loops, not bad souls; the souls are the same everywhere, and the loops are not.

Five cases

The book runs the loop through thirty cases in one fixed form; five of them carry the argument here.

Switzerland. Switzerland is a closed loop, stable, with a slow ex ante creep at the edge: the country runs two verifiers on its own government that almost no other polity runs at all, and both fire after the output exists. The Swiss federal state operates little and structures nearly everything: a constitution that submits every federal statute to a possible popular vote, and a fiscal order in which 26 cantons set their own tax rates in public view and a resident or a firm can move across a cantonal line at the cost of a commute. The system’s own logic is that the people who pay for a law are the only verifier of it that cannot be lobbied in aggregate, and that a government small enough to leave keeps itself honest without having to be virtuous.

The first live verifier is the referendum, in its ex post form: the optional referendum, which any 50,000 citizens can trigger within 100 days of a statute’s passage, has let 58 percent of the challenged laws stand, whereas the popular initiative, which asks voters to generate constitutional text themselves, has passed 11 percent of the time (Le News, 2025; admin.ch). The spread between those two figures is the difference between a verifier and a generator: the electorate is good at judging what parliament made and reluctant to make law itself. The second is the benchmark, in its fiscal form: the combined corporate rate runs from 11.85 percent in Zug to 20.54 in Bern (KPMG, 2025), and every canton can read its neighbors’ rates and its own net migration of firms and taxpayers, which is exit converted into a published number.

The referendum binds absolutely: a statute that loses the vote is dead, and parliament’s only recourse is to draft again, which is iteration, not override. The soft spot is procedural: the pre-parliamentary consultation, the Vernehmlassung, in which the cantons, the parties, and the umbrella associations are invited to comment on a bill before the Federal Council submits it (Swiss Federal Assembly, 2017). An association that can collect 50,000 signatures is consulted before generation so that it will not fire the verifier after it. That is rational for the generator and corrosive for the loop, since it converts an ex post verifier into an ex ante veto point, and a veto point, unlike a verifier, leaves no record of the laws it prevented.

The component is that the referendum is ex post: the verifier evaluates a finished statute, not a document about it, the verdict binds, and a rejected law returns in another draft. The book puts the distinction in one sentence, “The Swiss referendum is ex post; CEQA is ex ante,” CEQA being the California review that judges a forecast of a project rather than the project. If the cantonal benchmark were destroyed by federal harmonization of tax rates, expect intercantonal relocation of firms to fall toward zero and cantonal spending per head to converge upward within ten years, because a verifier that fires once a federal statute exists cannot select among 26 cantonal budgets.

Singapore. Singapore is a closed loop in every sector where the state operates or invests, stable, with one verifier decorative by design: the election/press pair does not verify the state’s conduct of itself, and the system runs instead on export, benchmark, tender, and exit. The state operates housing, land, ports, the airline, and the sovereign portfolios, and it structures the rest. Every HDB building contract is procured by open tender on the government’s GeBIZ portal from bidders registered with the Building and Construction Authority (HDB, 2026); the builder is a private generator that wins on price, is paid on delivery, and loses the next job by failing this one. The system’s own logic is that a city-state with no hinterland lives on what the world will pay for, so the state’s task is to keep every domestic generator exposed to that verdict and to keep politics from softening it. On that logic one-party continuity is not the suppression of a verifier but the protection of the others from electoral softening.

Price runs in the resale market next door: the resale price of an HDB flat is a public number that verifies the subsidy, the location, and the build, with no ministry able to set it. Exit runs in the grocery sector, which the state pointedly does not operate. Export runs on the Temasek companies, and the losses are booked where they occur: Temasek reported a negative one-year return for the year to March 2023 and wrote off its USD 275 million stake in FTX with a cut to the compensation of the team responsible (CNBC, 2023). The captured and decorative verifier is election/press. The press operates under a licensing act that places management shares in approved hands (Wikipedia, Newspaper and Printing Presses Act); the governing party has won every general election since 1959 and took 65.6 percent of the vote and 87 of 97 seats in May 2025 (Wikipedia, 2025 Singaporean general election). The election fires, the press reports its firing, and neither passes a verdict the state has not already priced in.

The benchmark on Temasek binds less than it looks, because 50 percent of the portfolio is unlisted (Temasek Review, 2026), so half of the reported value is marked by the generator being judged, and this is where a Singapore failure would first be hidden. The decorative press means that the one output the pair alone verifies, the state’s use of its own discretion, has no verifier and no consequence, so that a bad verdict on it would arrive last, from outside, and at once.

The component is that HDB is the developer-client and never the builder. That is the operator rule (P4): a state may operate where verification is cheap and consequences still bind, and elsewhere it must structure loops. The same state that houses more than three quarters of its population does not sell them groceries and does not build its own flats. If HDB internalized construction, replacing the tender with an in-house public works builder, expect unit cost to rise and completion times to lengthen within one build cycle of about five years.

The USSR. The Soviet loop was open by design from the first five-year plan in 1928 and stayed open for six decades; its trajectory was decay, masked first by mobilization and then by oil, and the dissolution of December 1991 was overdetermined, because every one of the eight verifiers had been dead at once for a generation. The state operated everything and structured nothing: Gosplan set targets in physical units, Gossnab assigned each enterprise its suppliers and its customers, and the State Committee on Prices fixed the price of every good, the retail price of bread holding from 1955 (Christian Science Monitor, 1985). The plan’s own logic was coordination without the waste of duplication and speculation, and mobilization of a peasant country into an industrial one within a decade rather than a century; and it delivered, where it was pointed: the Urals steel base, the tanks of the war, the first satellite in October 1957. Every item on that record is a count. The plan succeeded exactly where output could be counted by someone other than the generator, and it failed wherever quality was not a count: a shoe that fit, a harvest, a machine tool anyone wanted. The record has been read for seventy years as evidence that planning works in emergencies. It is evidence that counting works, in emergencies and out of them.

None of the eight was live for the output the plan existed to make. Price was administered, so the tag on a Soviet shelf carried a number and no verdict. Tender was absent, one assigned supplier per input. Exit was forbidden, and ran into the second economy, where it was a crime. Export was partially live: oil, gas, gold, and timber sold at world prices, so the state knew precisely what its raw materials were worth and nothing about the value of anything it manufactured, fuel and raw materials making up about four-fifths of Soviet exports in the 1980s (Carnegie Endowment, 2017). Election/press was decorative, a single candidate on the ballot and Glavlit reading every page before it was printed. Benchmark was captured, because enterprises were compared on figures they had themselves reported. Referendum was held once, nine months before the Union ceased to exist. Bankruptcy was nonexistent; a loss-making plant was a plant with a subsidy.

The soft constraint was not a defect of the Soviet enterprise; it was the enterprise’s charter. Next year’s target was set from this year’s reported output plus an increment, the ratchet, so a manager who reported true capacity was punished with an impossible target, and the rational manager underreported capacity, hoarded labor and inputs, and stormed at the end of each quarter to hit the number, and nobody above him checked. The price on the Soviet shelf was fixed, sometimes for decades, so the true price was the queue, paid in hours; and the queue is a verifier that fired every day in every town, with a verdict that landed on the shopper and never on the shop. A queue is exit with the door locked and price with the number removed.

The component is that every one of the eight verifiers was dead at once, which is why the collapse was overdetermined: not one autopsy finding but eight. When glasnost freed the press and semi-competitive elections attached a consequence, sixty years of suppressed verdicts arrived within five, and the system dissolved rather than reformed, because a loop that has been open for a generation has no tissue left that can read a verdict, only tissue that can suffer one. If the election/press verifier is switched on in a system whose other verifiers are dead and whose consequences are soft (Cuba and North Korea are the clearest surviving replicas), expect dissolution rather than reform within five years, on the pattern of 1986 to 1991.

US health insurance. American health insurance is an open loop, decaying, and opened by intermediation rather than by ownership: the most private of the rich health systems is the one in which the patient reads no price, cannot leave, and cannot tell whether the care was good. The state operates almost nothing and structures the rest through administered prices, a tax exclusion, and licensing. The steel-man is real. Insurance exists to remove price at the moment of illness, because a person having a stroke cannot shop and a person with cancer cannot pay from income; the employer group solves the adverse selection that kills individual insurance markets and negotiates with hospitals on the employee’s behalf in a way no individual could. The tax exclusion that binds coverage to the job was an accident of wartime wage controls, and the accident became the largest single tax expenditure in the federal budget (Congressional Research Service, 2010). The stack is five generators deep before a patient meets a physician, and not one of them produces an output the patient can read, compare, or refuse.

Price at the point of consumption is dead, not weak. Out-of-pocket spending was USD 556.6 billion of USD 5.3 trillion in 2024, 11 percent (CMS, 2024 data), so the party who could verify reads a number for one dollar in nine and reads it after the fact. Exit is disabled twice: the patient cannot leave the insurer without leaving the job, and cannot leave the provider without leaving the network. Benchmark is the one verifier that works, at scale and on mortality: Medicare’s published readmission rates, with a penalty attached since 2012, are a benchmark with a consequence, and readmission rates fell after the penalty attached (New England Journal of Medicine, 2016). Where insurance does not reach, price reappears: the Surgery Center of Oklahoma posts all-inclusive prices for named procedures on its website and has done so since 2009 (Smith, 2013), not because its surgeons are better but because a patient who can read the number can also refuse it. Same physicians, same licensing, same malpractice law, same state: the only variable that differs is whether a verifier stands between the patient and the generator.

The consequences bind on the two parties who generate nothing, the patient and the employer. The patient bears the deductible, the balance bill, and the medical debt, which is the sector’s bankruptcy verifier misdirected: bankruptcy is supposed to end the generator that produced the bad output, and here it ends the household that consumed it. The generators sit under a soft constraint, and the insurer’s margin scales with spending.

The component is that the loop was opened by intermediation, not by ownership: price dies as easily under contract as under decree, and a stack of private generators with no verifier between them and the consumer is Gosplan with a billing department. If the claims-based actual prices and executive attestation required from April 2026 (CMS, 2025; Dentons, 2026) are enforced to the point that a patient can compare the negotiated rate for a knee replacement across the hospitals of one metropolitan area, expect within five years a measurable compression of price dispersion for those procedures; if the rule is met with uninterpretable files, expect no change, because a verifier that is published but not read is decorative.

China. China is a partial loop: closed on two verifiers from roughly 1980 to 2008, decaying from 2008 as the first of them was captured, and opening deliberately since 2012 as the remaining ones were silenced. It was one of six cases chosen in advance to kill the theory, and its bar was fixed before it was read: benchmark plus export must be accepted as sufficient verifiers with the press dead, and the post-2012 deceleration must follow their decay. The party’s own logic is the best steel-man in the book. A continental country with no functioning price system in 1978 could verify nothing through markets, so it built a verifier out of the one instrument it possessed, personnel control: ranking provincial leaders against each other on growth and promoting the winners converts a hierarchy into a tournament, and opening the coast to foreign buyers imports a second verifier the party cannot lobby. Li and Zhou found that the probability of a provincial leader’s promotion rose, and of termination fell, with provincial GDP growth (Li and Zhou, 2005): a benchmark verifier in its purest political form. After accession to the World Trade Organization in December 2001 (WTO), a Guangdong factory’s output was judged by a buyer in Ohio who could not be promoted, demoted, or reported to.

Export stayed live and incorruptible for the whole period; it is the only verifier the state never touched, because it could not. Election/press was dead throughout, and bankruptcy was decorative for state enterprises and local governments. The benchmark was live until about 2008 and captured after, and the capture was structural rather than accidental, because the metric on which provinces were ranked was compiled by the provinces. A binding consequence attached to a falsifiable verdict does not select for output; it selects for falsification.

The component is that the benchmark was a verifier the generator could falsify, so its decay was predicted by the theory rather than discovered by it. The National Bureau of Statistics had been correcting the sum of provincial figures downward by about 5 percent of GDP since the mid-2000s, which is the verifier working; after 2008 the adjustment did not keep pace, and national growth from 2010 to 2016 ran about 1.8 percentage points a year below the reported figure (Chen, Chen, Hsieh, and Song, 2019). The forensic figure is contested at the decimal (Federal Reserve, 2025), and the case rests on the 2008 date rather than the number. Retirement followed: the youth unemployment series was suspended in August 2023 after a record 21.3 percent and resumed redefined (Time, 2023), and hundreds of other series have since stopped (China Digital Times, 2025). Growth fell toward the 5 percent target in the same years (Bloomberg, 2024), with demography and the exhaustion of easy catch-up acknowledged as residuals. A verifier the generator can write is a verifier with a scheduled death. If the benchmark is rebuilt on data the provinces cannot touch, expect official growth, partner-reported trade, and tax-implied output to reconverge within three years.

Two stress tests

Six cases were chosen to kill the theory, with the bar for each fixed before any case was read. None did. None survived cleanly either; China survived by the thinnest margin, and India and Japan show what the distinctions are for.

India survives, on two distinctions of unequal cleanliness. The puzzle is a democracy that stagnated for four decades; the resolution is that a general election is a noisy verifier of licensing policy and that the License Raj was vetocracy, ex ante blocking with no ex post feedback. A noisy verifier is a verifier that fires, but whose signal is too weak, slow, or aggregated to select on the output in question. The bar was that both distinctions had to have been defined before India and not for it. Vetocracy meets the bar without remainder: the book built it on CEQA and the Swiss referendum, and the License Raj enters only as an application after the definition. The noisy verifier is less clean, and the book said so in advance: it was defined eight chapters before the case, but on the case’s material. What rescues it is the stricter test set for exactly that situation, a concept defined on one case and predicting in several, and the noisy verifier does work where there is no India. It is Madison’s reason for “auxiliary precautions,” and it is why a Greek election reset the deficit figure at each changeover and could not verify a quarterly number. The component is P6: the licensing office grew as a generator nobody could verify, and it was removed in 1991 by demolition rather than verdict because a veto point has no verifier of its own. One fact tempers the margin: the bankruptcy verifier of 2016 is decaying on its timing leg, closed cases averaging 853 days against a statutory 330 (IBBI via IBC Laws, 2025). The closing loop is not closed.

Japan strengthens the theory, and it is the only one of the six that does. The bar was that bindingness be shown to be a separate leg: a country with live verifiers that refused the consequence and rotted anyway, as against a refusal that was a verifier failure in disguise. The finding is one sentence: “The consequence leg failed alone, and it failed in the private sector first.” Every verifier fired, the electorate removed a government in 1993, and the share of listed firms on subsidized credit rose from between 5 and 15 percent until 1993 to above 25 percent in every year after 1994 (Caballero, Hoshi, and Kashyap, 2008). The disguise objection is that bankruptcy, one of the eight, was dead. The answer is in the mechanics: a bankruptcy court rules only if someone files, and the party who declined to file was the party holding the verdict, the bank with the loss in hand. That is a consequence cancelled, not a verdict falsified. A system “can verify perfectly and refuse to act, and when it does, it rots on the same schedule and with the same symptoms” as one that never verified at all. The margin is the widest of the six.

What the lens does not claim

The asymmetric law (P7) sets the limit: dead verifiers predict failure with near certainty, while live verifiers enable success and do not guarantee it, which makes this first a theory of failure and only then a theory of what performance requires. The claim is not that verification explains everything. The lens does not explain trust, geography, culture, scale, or oil, and it does not try; each is a multiplier on a closed loop and a multiplier of zero on an open one. The claim is that no system has sustained performance with its verifiers dead, and that every autopsy of a collapsed system finds the verifiers killed first.

The best verifier ever found is already in the reader’s hand. Walk into any store and look at the small card on the edge of the shelf. It was set by someone who will lose money if it is wrong in either direction, it is read by every person who passes, it is compared without effort against the card in the next aisle and the next store, and it changes when the thing behind it stops selling. No committee approved it, no statistic reports it, and no one can order it to lie for long. A price tag on a shelf is the smallest verifier in the world.

Sources

1. SingStat, Resident Households, 2025: https://www.singstat.gov.sg/find-data/explore-data-themes/households/resident-households/latest-news-data
2. Wikipedia, Mission Mercal: https://en.wikipedia.org/wiki/Mission_Mercal
3. Economics Observatory, 2024, Why did Venezuela’s economy collapse: https://www.economicsobservatory.com/why-did-venezuelas-economy-collapse
4. Kornai, Maskin, and Roland, 2003, Understanding the Soft Budget Constraint (Journal of Economic Literature): https://www.ias.edu/sites/default/files/sss/papers/econpaper19.pdf
5. SEC, 2024, Office of Credit Ratings staff report: https://www.sec.gov/files/feb-2024-ocr-staff-report.pdf
6. Le News, 2025, Swiss votes, initiatives and referenda, a guide: https://lenews.ch/2025/02/14/swiss-votes-initiatives-and-referenda-a-guide/
7. admin.ch, Popular votes: https://www.admin.ch/en/popular-votes
8. KPMG, 2025, Clarity on Swiss Taxes: https://kpmg.com/ch/en/media/press-releases/2025/05/clarity-swiss-taxes.html
9. Swiss Federal Assembly, 2017: https://www.parlament.ch/centers/documents/en/asgp-dhaka-schwab-communication-2017-e.pdf
10. HDB, 2026, Tender eligibility: https://www.hdb.gov.sg/business/tenderers/hdb-tender-opportunities/eligibility
11. CNBC, 2023, Temasek 2023 report: https://www.cnbc.com/2023/07/11/singapore-temasek-2023-report.html
12. Wikipedia, Newspaper and Printing Presses Act: https://en.wikipedia.org/wiki/Newspaper_and_Printing_Presses_Act
13. Wikipedia, 2025 Singaporean general election: https://en.wikipedia.org/wiki/2025_Singaporean_general_election
14. Temasek Review, 2026, Performance and portfolio: https://www.temasekreview.com.sg/performance-and-portfolio.html
15. Christian Science Monitor, 1985: https://www.csmonitor.com/1985/0916/obread.html
16. Carnegie Endowment, 2017, The formation and evolution of the Soviet Union’s oil and gas dependence: https://carnegieendowment.org/posts/2017/03/the-formation-and-evolution-of-the-soviet-unions-oil-and-gas-dependence
17. Congressional Research Service, 2010, RL34767: https://www.everycrsreport.com/reports/RL34767.html
18. CMS, 2024 data, National Health Expenditure fact sheet: https://www.cms.gov/data-research/statistics-trends-and-reports/national-health-expenditure-data/nhe-fact-sheet
19. New England Journal of Medicine, 2016: https://www.nejm.org/doi/full/10.1056/NEJMsa1513024
20. Smith, 2013, The Surgery Center of Oklahoma Story: https://surgerycenterok.com/blog/the-surgery-center-of-oklahoma-story/
21. CMS, 2025, CY 2026 OPPS final rule, hospital price transparency policy changes: https://www.cms.gov/newsroom/fact-sheets/cy-2026-opps-ambulatory-surgical-center-final-rule-hospital-price-transparency-policy-changes
22. Dentons, 2026, CMS enforcement of updated hospital price transparency requirements to begin on April 1, 2026: https://www.dentonshealthlaw.com/cms-enforcement-of-updated-hospital-price-transparency-requirements-to-begin-on-april-1-2026/
23. Li and Zhou, 2005, Political Turnover and Economic Performance: The Incentive Role of Personnel Control in China (Journal of Public Economics): https://www.sciencedirect.com/science/article/abs/pii/S0047272704001355
24. WTO, China and the WTO: https://www.wto.org/english/thewto_e/countries_e/china_e.htm
25. Chen, Chen, Hsieh, and Song, 2019 (NBER Working Paper 25754): https://www.nber.org/papers/w25754
26. Federal Reserve, 2025, Is China really growing at 5 percent: https://www.federalreserve.gov/econres/notes/feds-notes/is-china-really-growing-at-5-percent-20250606.html
27. Time, 2023, China stops publishing youth unemployment statistics: https://time.com/6304881/china-stops-publishing-youth-unemployment-statistics/
28. China Digital Times, 2025, Censored statistics, deleted data muddy the waters: https://chinadigitaltimes.net/2025/05/censored-statistics-deleted-data-muddy-the-waters/
29. Bloomberg, 2024, China sets 2024 GDP target at around 5 percent: https://www.bloomberg.com/news/articles/2024-03-05/china-sets-2024-gdp-target-at-around-5-npc-highlights
30. IBBI via IBC Laws, 2025, Recovery rate through the IBC 2016 and SARFAESI in financial year 2025: https://ibclaw.in/recovery-rate-through-the-ibc-2016-and-sarfaesi-in-financial-year-2025-performance-of-the-insolvency-code/
31. Caballero, Hoshi, and Kashyap, 2008, Zombie Lending and Depressed Restructuring in Japan (American Economic Review): https://www.aeaweb.org/articles?id=10.1257%2Faer.98.5.1943