Everything is certified. Nothing is checked.

The environmental impact assessment for a 512-hectare development at the foot of Hải Vân Pass was prepared by a consultancy called Thành Tâm. The firm was incorporated on 13 September 2024. It filed the completed assessment roughly two months later. RFA, which reviewed the document, reports the company has four staff and a registered address in a Hanoi residential building. The site is 78.56 percent forestland. The assessment graded the environmental consequences as low-level and short-term.

That document had been on file for nine months when, on 2 September 2025, tens of thousands of people camped overnight on the streets of Hanoi to hold a place for the A80 parade at Ba Đình Square. Sixteen thousand soldiers marched. Reporting at the time described it as the largest public celebration the country had staged. Four months before that, hundreds of thousands had filled the streets of Ho Chi Minh City for A50, some of them camping out as well. State media covered the enthusiasm approvingly and made a specific claim about it: this patriotism, Vietnam News wrote, "doesn't feel imposed; it feels chosen."

Both things were true at once, and that is the piece.

Two failures on one day, and a third behind them

On 2 July, the Ministry of Education and Training opened a review into mathematics results at an examination centre in Tuyên Quang. The same day, police announced charges against the former director of PNJ Laboratory over a ring alleged to have moved more than 28,000 smuggled diamonds into the country. Three weeks later, scrutiny reached the assessment behind the cleared hillside at Hải Vân.

None of these is a story about a defective thing. Vietnamese students are not bad at mathematics, the diamonds were real diamonds, and the mountain is a real mountain. In all three cases the defective object was the certificate: the score, the gem report, the impact assessment. Each exists so that a party can act on information it cannot check for itself. A university admits a student it has never met on the strength of a number. A buyer pays a premium for a stone they cannot grade. A ministry authorises the clearing of hundreds of hectares of forest on the strength of an impact finding.

That is the whole function of a certificate, and it is infrastructure in the same sense a port is. Vietnam's growth model assumes it works.

The case that the system works

The strongest version of the official position first, because it is better than it usually gets credit for.

All three were detected, and quickly. Eleven days separated the first anomaly in the examination results from nineteen people being charged, with a teacher in emergency detention on day four. A second case in Quảng Trị produced 26 more prosecutions, and the Deputy Prime Minister ordered a nationwide review of 2026 results. The diamond investigation charged 22 people and reached Saigon Jewelry Company as well, where a further 3,400 stones worth more than 500 billion đồng entered the retail network between 2022 and 2024. Market discipline was immediate: PNJ went limit-down on 3 July, the maximum single-session fall the exchange permits, with no bids at the floor.

Read generously, this is a system with functioning antibodies. Eleven days from anomaly to nineteen charges is not the behaviour of a state protecting its own, and the current anti-corruption campaign is catching people who would previously have been untouchable.

Nobody with standing was looking

Here is what the detection story cannot account for.

Tuyên Quang was found by analysts reading the published score distribution, who noticed that 147 of the province's 154 perfect mathematics scores sat inside a single block of 328 candidate numbers, of whom 299 had scored between nine and ten. That is visible to anyone with the results file and an afternoon. The invigilators, the examination board, the provincial department, and the ministry's own controls all passed it.

The diamond fraud was found because police were investigating smuggling, which is a customs matter. Nobody was auditing the laboratory. PNJ Lab certifies more than 10,000 items a month and holds around 70 percent of the domestic market. Its former director is alleged to have bought stones whose physical characteristics did not match their GIA paperwork, removed the GIA laser inscription, and issued his own report in its place, across 141 shipments beginning in 2024. PNJ maintains the diamonds sold to its customers were not among the smuggled stones, and the charge is against an individual rather than the company.

Note what the alleged mechanism concedes. The GIA inscription had to be physically removed, which means an independent verification layer existed and had to be defeated. The capability to catch this was never the problem. What was missing was any party with a mandate to look, because GIA has no standing in the domestic retail market and no Vietnamese body was auditing the auditor.

The Hải Vân assessment was found by people looking at photographs of a bare hillside, having already cleared the process. RFA's review found blank sections where community consultation should be, an incomplete biodiversity offset plan, and a development described as a tourism and resort complex that on inspection contains roughly 6,633 residential units and a planned population of 19,000. Those defects were on the page the entire time.

So the pattern is not that Vietnam fails to punish certification fraud, and it is not that the frauds were caught by luck. Published score distributions and GIA laser inscriptions are deliberate mechanisms, working exactly as designed. The pattern is that in every case the party that noticed had no standing to compel anything: outside analysts, a police unit looking for something else, members of the public with photographs. Detection happened after issuance, from outside, unpaid, by people who then had to hope somebody official would act. What gets caught under those conditions is the clumsy version, the one that bends a distribution curve badly enough to be visible from the street.

Why the certifiers all failed the same way

The three cases share a structure, and it is not incompetence.

The developer commissions and pays the consultancy that assesses its own project; Thành Tâm was the consulting unit on the Làng Vân filing, roughly two months after being incorporated. The jeweller owns the laboratory that certifies stones for the market it leads; PNJ Lab is a wholly owned subsidiary. The examination centre is evaluated on the results it produces, which is the pressure Nikkei identified in its reporting on the case and which Nguyễn Đắc Vinh described from the other direction on 14 July, when he called the irregularities extremely serious in open session and pointed at temporary staff and absent independent oversight.

In every case the certifier is paid, directly or structurally, by the party that benefits from a favourable certificate. This is issuer-pays, and readers of this publication have a fully formed prior about where issuer-pays certification leads. Nobody needs 2008 explained to them.

The distinction matters because it changes what kind of problem this is. An auditing problem is solved by hiring: more inspectors, better training, tighter procedures. That is roughly what the official response has been. An incentive problem is not solved by hiring, because every additional inspector is hired into the same structure and paid, eventually, out of the same pocket. Vietnam has responded to all three cases with enforcement, and enforcement is the correct response to the individuals. It does nothing to the arrangement that produced them.

It also supplies a selection rule, which is more useful than a general anxiety about Vietnamese institutions. The exposure is not everywhere. It is specifically wherever the certifier is paid by the certified: environmental and technical assessments commissioned by developers, product certification held in-house by the manufacturer or retailer, quality and safety audits procured by the audited, and examination administration devolved to the institution being measured. That is bounded and it is testable. It does not require believing anything about a national failure rate.

What the anger is and is not evidence of

The visible reaction has been online, and volume is worthless as evidence. Vietnamese platforms in July were saturated with outrage about all three cases, and a large share of that is engagement farming, recycled screenshots, and accounts that would be furious about anything. Duration is different: how long a grievance stays live is much harder to manufacture than a spike, and it is what the account below predicts, since anger with a private exit available should fade faster than anger without one. I cannot run that comparison here. The exam case broke on 2 July and the assessment came under scrutiny three weeks later, which is not enough elapsed time to observe decay in either. So it stays a prediction, and the test is at the end of this piece rather than claimed as a finding.

What is measurable now is behaviour. All 322 parents and all six independent candidates whose children sat the Tuyên Quang examination backed a full mathematics re-sit, including the families of the 147. That unanimity makes sense once you read the charges, which describe conduct entirely on the staff side. Jewellery shops across Ho Chi Minh City suspended trading in July and some closed permanently, while holders tried to sell certified stones into a market that had stopped believing its own paperwork. These are costly actions by people with money at stake.

The exit was already underway

Every exit number in this piece predates the scandals. The record 250,000 Vietnamese enrolled abroad, led by South Korea, Japan, and Australia, is a 2025 figure. The finding that roughly 80 percent of self-funded students do not return was reported in 2024. The locality survey showing that of 8,850 students abroad, 1,160 came home in 2022 and 543 in 2023 covers a period ending three years before any of this surfaced. Household preference for gold over instruments that require trusting a document is decades old, and the State Bank puts the hoard at 300 to 500 tonnes sitting outside the banking system.

None of that is a response to July 2026. It cannot be. And that is the finding rather than a hole in it.

Behaviour read for years as risk aversion, or family parochialism, or a cultural attachment to physical assets, turns out to have been tracking something real. Households that stayed in bullion rather than trading up into certified stones were not being unsophisticated. They were declining to buy the one component of the product that has now been charged as counterfeit. Families sending children abroad were not merely chasing salaries. They were routing around an allocation mechanism whose failure became a criminal matter three years later. Read that way, 2026 is not the cause of the exit. It is the year the evidence caught up with a judgement the public had already made.

A caveat the argument needs. This generation is not one actor. The parade crowds, the students abroad, and the households holding gold are different populations inside a cohort of some thirty million, and nothing here shows the same individuals doing all three things. What it shows is separate groups reaching a compatible conclusion without coordinating, which is weaker as a story and stronger as evidence.

It also explains why the exit takes almost no organised form. For two of the three failures a private remedy exists that works immediately, and the collective remedy does not exist at all. If your child's examination was corrupted there is no class action, no ombudsman with teeth, and no realistic path to a repaired national examination system on any timescale relevant to a seventeen-year-old. If you hold a certified stone that may not be what its report says, there is no meaningful consumer redress. In both cases the return on voice is close to zero and the return on exit is immediate and private. Exit is not resignation. It is the correct answer to the incentives, which is Hirschman's original point and the reason the framework is worth borrowing rather than gesturing at.

None of which is contradicted by the parade. Sincere national pride and no confidence whatever in a provincial examination board are different beliefs, and there is nothing odd about holding both. It is worth saying because the standing read of Vietnamese under-thirties is political apathy, a framing New Naratif used as an article title. Apathy is the wrong word for a population that camps overnight for a parade and separately moves a quarter of a million of its children offshore. Both are engagement, pointed at different things.

The last exit

There is a third rung, and it is the one that does not reverse. You can sell the gold. You can come home. You cannot un-decline a child at thirty-eight. It is tempting to read that decision as the longest-duration bet available on a country's institutions, eighteen years to an examination that turned out to be for sale and twenty-two to a diploma the labour market already discounts. Treat that as an intuition rather than a finding, for reasons I will come to.

Vietnam's fertility rate fell from 2.11 in 2021 to 1.91 in 2024, a record low and the third consecutive year below the replacement level of 2.1. The state has an answer and it deserves stating: the number has since risen to 1.93 in 2025 and an estimated 1.96 across the first half of 2026. The two-child policy was scrapped in June 2025, and a new population law took effect on 1 July 2026 with cash bonuses, maternity leave extended from six to seven months for a second child, housing preferences, and subsidised screening. Ho Chi Minh City reports a recovery from 1.43 to 1.51, though those are city health department figures rather than the national series and the two are not strictly comparable. On the official reading, the intervention is working.

There is a larger piece of counter-evidence, and it is a month old. On 1 July 2026 the World Bank reclassified Vietnam as an upper-middle-income economy, its Atlas gross national income per capita having risen from $4,490 to $4,970 against a threshold of $4,636. That arrived four years ahead of the schedule set at the 13th and 14th Party Congresses. A country that clears an income threshold early, on 10 percent average gross national income growth across 2021 to 2025, is not obviously a country whose people have stopped believing in it.

The regional evidence cuts the other way on the fertility question. South Korea's rate also went up, from 0.72 in 2023 to 0.75 in 2024, on the largest jump in marriages since 1970, and nobody believes South Korea has solved anything. What that country has is a generation that gave up on dating, then marriage, then children, in that order, and grants above $1,400 plus monthly allowances that did not stop it. Japan pays roughly $3,200 a child and Singapore up to S$13,000, and both have at best slowed the decline. Against that, Vietnam is offering a one-off payment of up to $228, in a country where raising a child is estimated to cost between $380 and $760 a month, more than the average income. The bonus does not cover a single month. The Health Ministry has also proposed formally recognising and honouring couples who have two children in ten localities, which is to say that one official response to a collapse in the value of certificates is to issue a certificate.

A word on what this is not. China's vocabulary for the same mood, involution and lying flat, describes an urban, educated, largely male refusal to keep competing, and it does not transfer. Vietnam's own figure is 1.6 million people aged 15 to 24 not in education, employment, or training, which is 11.4 percent and sits below the Southeast Asian average of 16.3 percent. The composition points away from withdrawal: rural rates run more than four points above urban and women's nearly three points above men's, which looks like unpaid care work and a skills mismatch rather than anyone opting out. What is worth watching is not the level but the speed. That number rose by 173,000 in a single quarter, up 212,500 year on year, from roughly 10 percent three months earlier. A regionally unremarkable rate moving 1.4 points in one quarter is the alarming part.

The strongest objection to this whole section is that none of it needs a trust variable. Falling fertility tracks urbanisation and female educational attainment almost everywhere it has been measured, and those two forces plausibly account for most of Vietnam's level and most of its trend without any reference to examinations or gem certificates. I think that is largely right, and I am not going to claim the certification failures are driving the birth rate. There is a second objection that runs the same way: a shrinking cohort mechanically raises income per head for any given output, so the demographic decline I am treating as a cost also accelerates the per capita convergence I am about to measure against. Both are real. What survives them is narrower and still binds, which is that the fiscal and productivity consequences arrive regardless of cause, and that the state's chosen remedy is failing for reasons the demographic literature already predicts.

Which brings the arithmetic, and the part that genuinely does not hold still. Vietnam's stated goal is gross national income per capita of $13,000 to $15,000 by 2045. But the high-income threshold is redefined every July in current dollars, and it drifts up with global inflation at something like 2 percent a year. The FY2027 threshold is $14,375. Carried forward on that path, the bar Vietnam has to clear in 2044 is somewhere near $21,000, call it $19,000 if escalation runs at 1.5 percent and $25,000 if it runs at 3. On the middle estimate the top of Vietnam's own target range falls about a quarter short of the finish line and the bottom about two-fifths short. The country could hit its stated number in full and still not be classified high-income.

That is a certification failure of the same species as the rest of this piece. A target was issued, published, and repeated for years, and nobody with standing rechecked it against the moving benchmark it is meant to track.

The clock is the part that does not negotiate. Vietnam crosses into an aged society around 2036, nine years before the 2045 target, on a transition variously estimated at eighteen to twenty-five years against Japan's twenty-four. That is what old before rich means in practice: not a slogan, but the bill for the first arriving well before the revenue from the second.

It also reframes the growth target adopted at the January Party Congress, of more than 10 percent a year from 2026 to 2030, against 8.18 percent in H1. Published estimates put the growth actually required for 2045 at around 5 percent annually, and those assumed a higher bar than the one above, so the true requirement is lower still, somewhere in the range of 4 to 5 percent depending on the escalation path. The conversion is rough, because real output growth does not map cleanly onto dollar income per head; population, the deflator, and the exchange rate all sit in between. But the order of magnitude holds, and it makes the interesting question the opposite of the one I started with. The target is not falling short of what 2045 demands. It is set at roughly twice it. Either that is deliberate headroom against the demographic drag described above, or it is a number chosen for reasons other than arithmetic.

Where exit is not available

There is one grievance in this piece with no private remedy at all. You can hold gold instead of certified stones and you can put a child on a plane, but you cannot emigrate from your own coastline. The Làng Vân site runs to 512 hectares at the foot of one of the most recognisable landscapes in the country, cut down from an original 1,565 hectares after protected forest was carved out over years of boundary adjustment. The images of the cleared slope travelled because the loss is not substitutable.

Two other explanations compete with mine and both are serious. Land and environment is the most tolerated register of public complaint in Vietnam, so grievances find that channel because it is open. And a photograph of a stripped hillside is shareable in a way a score distribution is not. Either would predict a large reaction without any of my machinery, and on the evidence available in August I cannot rule either out.

The project deserves the same care as PNJ. Vinpearl broke ground in June 2025 with a Permanent Deputy Prime Minister present, on a site the state had been trying to develop since 2002, which is the reason it was available at all and why this is not an arbitrary target.

What it costs

There are three bills.

The allocation bill. Covered above, and it is the largest. A manufactured score does not merely advantage one student; it displaces a real one, and the misallocation compounds across a working life, out of a cohort that is contracting at the same time.

The trade-up bill. This is the one for equity investors. The Vietnamese consumer thesis is an escalator: incomes rise, households move from unbranded to branded, informal to formal, uncertified to certified, and pay a margin for the privilege. PNJ is that thesis in listed form. Roughly 10 trillion đồng of market value came off the company by 20 July, about 30 percent. The entire alleged smuggling operation, across 141 shipments and more than 28,000 stones, is put by police at 280 billion đồng of turnover. The market removed something like thirty-six times the ring's gross revenue from a company that denies its own customers received any of the stones. Whatever that is pricing, it is not the fraud. It is consistent with the escalator itself being repriced, and the exposure is not every consumer name but the ones where authentication is done in-house by the party selling the product. The same logic reaches project finance, where international lenders read assessment credibility as a country-level input rather than a project-level one.

The legitimacy bill. A50 and A80 showed something real: reserves of public goodwill at a scale most governments would envy. This is a constraint rather than advice. The 2026 to 2030 agenda requires land reform, state enterprise restructuring, and a real tax base, and reform of that kind runs on public tolerance for short-term cost. That tolerance is finite, and the demographic numbers are what it looks like when a population stops extending it.

The diagnosis

Vietnam does not have a corruption problem in these three cases so much as a market design problem. The corruption was punished, quickly and publicly, and the punishment was appropriate. What was never in place, and still is not, is a certifier whose income does not depend on the party being certified.

Cheating at an examination centre was caught by strangers reading a spreadsheet. The largest gem laboratory in the country was policed by a smuggling investigation looking for something else. A 512-hectare forest clearance was assessed by a firm two months old and reviewed, in effect, by members of the public with photographs. In each case verification was performed after the fact, from outside, unpaid, by people with no power to act on what they found.

The bill for that does not arrive as scandal. It arrives as the people who concluded, quietly and early, that the certificate was not worth acquiring, and it is already arriving.

None of which makes the outcome settled, and it would be lazy to write as though it were. This is a state that scrapped a population policy it had held for decades once the evidence turned, that moved from an anomaly in a spreadsheet to nineteen charges in eleven days, and that has repeatedly done difficult things faster than outsiders expected. The capacity is not in question. What has not been tried is the structural fix, which is cheaper than the enforcement campaigns and considerably less popular, because it means taking certification away from the people currently paying for it. The demographic clock sets the deadline: an aged society around 2036, a high-income target in 2045, and a window in between that does not reopen. Whether the response moves from prosecuting certifiers to restructuring who pays them is the thing worth watching, and Vietnam has surprised on the upside before.

What to watch

• Does any certifier get separated from its payer? The highest-value signal here. An independent gem laboratory with no retail parent, an environmental consultancy panel developers cannot select from, examination administration moved off the institution being measured. Any one is a structural fix. More inspectors is not.
• Duration, not volume. Measured as weekly counts of Vietnamese-language news items and search interest for each case, indexed to its own July peak. If Hải Vân is still running materially above its baseline in October while the exam case has returned to it, the exit account survives its own test. If both have decayed together, shareability explains the July pattern and mine adds nothing.
• The fertility print for 2027, not 2026. The national cash regime began on 1 July 2026, so no birth recorded this year was conceived under it and the 2026 number cannot test the policy either way. The first informative print covers 2027 and lands in 2028. Anyone citing a 2026 recovery as evidence the incentives worked, this piece included if it had been sloppier, is reading a number that predates the treatment.
• Does the Tuyên Quang retest happen, and what does it show? The distance between the centre's original distribution and a clean re-sit is the closest thing to a direct measurement of fraud magnitude available, though it would capture only what a re-sit can capture.
• Returnee numbers for 2025 and 2026. The fall from 1,160 to 543 across 2022 and 2023 is the tightest available proxy for exit. Another halving is a break, not a cycle.

Trung Truong is an investor and researcher based in Ho Chi Minh City. This diagnosis represents the author's independent assessment. It is not financial advice. The author holds no position in PNJ, VIC, VHM, or VPL.

Corrected 2 August 2026, shortly after publication. The original gave the 2045 high-income target as a static threshold of about $14,000 and did not note that the World Bank reclassified Vietnam as upper-middle-income on 1 July 2026. Both are now addressed, and the correction widens the gap the section describes. The NEET discussion has been rebuilt on the quarterly change rather than the level, after the level turned out to sit below the regional average, and the comparison to China's lying-flat phenomenon has been withdrawn because Vietnam's NEET population is rural and female where China's is urban and male. An unsourced figure for Japan's income at aged-society entry has been removed.