Independent research · 2015 to 2025 · EPQ
Sources: Thailand Ministry of Tourism and Sports · Vietnam National Statistics Office
China sends more tourists abroad than any other country,8 and where they go is not purely a matter of taste. Governments can shape those flows: group tour approvals, travel advisories, safety messaging, aviation agreements, campaigns to keep spending at home. When those flows move, they move fast, and entire local economies move with them.
My question: how can changes in Chinese outbound tourism affect countries, and the people who depend on those tourists? I study it at two levels, the flows and the people, and I built the evidence for both.
Every politically motivated restriction on Chinese outbound tourism from 2015 to 2025, separated from safety-driven measures that use the same instruments.
No existing dataset covered this, so I compiled one: sixteen episodes, each coded on fourteen fields from tier 1 reporting and official notices.
A framework for restriction strength (channel control, advisory, caution) and for shadow versus formal measures.
Each episode scored on political result, operational effect, and backlash, then read against the target's exposure to Chinese arrivals.
Online interviews with businesses in Mae Sot, Thailand, in Summer 2026, to see what the aggregate numbers could not show.
Scored tourism against trade in goods on seven criteria of coercive usefulness, to place it in the economic statecraft literature.
The Author
I am Thiranart (Winn) Suksriwong, Year 13 at Winchester College in the UK, from Bangkok, Thailand. My interests are political economy, global studies, geography and public policy, and my subject is the region I come from, studied from the school that taught me how.
This project began as an independent question: how do states use economic tools, trade, investment, and increasingly tourists, as instruments of power, and what does that mean for the countries on the receiving end? It did not stay a reading exercise. I built a ten-year database of Chinese tourism restrictions, tested it against sixteen episodes, then interviewed businesses in Thailand to see what the numbers miss.
If you would like to talk about the research, write to me. I read everything.
The Data
Every politically motivated restriction China placed on its outbound tourists, 2015 to 2025. Collected, coded, and scored by me.
There is no official list of these events. China has never formally admitted to using tourism as a coercive tool, so most episodes exist only as travel advisories, embassy notices, or quiet instructions to travel agencies that were reported at the time and then forgotten. I built this database because I could not find one anywhere else.
It is the spine of everything on this page: sixteen episodes, each coded and scored across fourteen dimensions, from the instrument Beijing used to the target's vulnerability and the outcome. Every chart, pattern, and success rate that follows is computed directly from these rows.26
The full interactive database · three sheets
Evidence basis: tier 1 media and official government notices only. Politically motivated episodes are separated from safety-driven ones, which use the same instruments but are protection, not pressure.
Data: Thailand Ministry of Tourism and Sports, compiled by year.1 2020 to 2022 Chinese figures are near zero due to border closures.
From 2015 to 2019 Chinese arrivals grew from 7.9 to 11.1 million while holding a remarkably stable 27 to 28 per cent of all arrivals: more than one in four visitors to Thailand was Chinese, year after year. COVID took that to effectively zero. The recovery has been partial and is now going backwards: 6.7 million in 2024, then roughly 4.5 million in 2025, with the Chinese share nearly halved to 13.6 per cent, the year Malaysia overtook China as Thailand's largest source market.1,2
The important part: Thailand's 2025 collapse involved no Chinese government instrument at all. No advisory, no ban. It was driven by safety narratives on Chinese social media after the kidnapping of a Chinese actor near the Thai border,18,19 alongside prices and competition from Vietnam and Japan. A country can suffer restriction-scale damage without ever being restricted, which is exactly why the distinction between what Beijing does and what Beijing merely benefits from runs through this whole project.
Each entry records the date, target country, and cause; the goal Beijing was pursuing; the instrument used; the target's vulnerability, measured as the Chinese share of its inbound arrivals at the time; the impact on arrivals and the estimated economic cost in dollars; the destination country's response; and an effectiveness score from 1 to 9 for each side.
Restriction strength grades the instrument: A is hard channel control (tour bans, permit suspensions), B is an official advisory, C is a weak consular caution. Shadow versus formal records whether the measure was ever officially announced, or only observable in its effects.
When China controls the distribution channel, the agencies and permits, it works. When it only issues words, it usually does not.
Fourteen of sixteen episodes fall in 2015 to 2020, then four years of silence while China's own outbound flow recovered from COVID, then a resurgence in 2025. You cannot switch off a tap that is not running.
Every episode scoring 7 or higher hit a country where Chinese visitors were over 20 per cent of arrivals: South Korea, Taiwan twice, Palau, Japan. Every episode against a low-exposure country, under 5 per cent, scored 4 or lower regardless of instrument: the United States three times, Canada, Sweden, India, Turkey. The strongest evidence is Japan in November 2025: a mere advisory, the same instrument that failed repeatedly against the US, scored 7, with around 491,000 flight tickets cancelled within days, because Chinese visitors were over a fifth of Japan's arrivals.13,4 Who you target predicts success better than what you use.
Only four episodes involved hard channel controls, all between 2017 and 2020. Twelve of sixteen were shadow measures: deniable advisories framed as safety concerns. Both 2025 episodes were purely shadow. China's method has evolved toward instruments it can always deny, and the Japan case shows deniable instruments now achieve what formal bans used to.
Measuring effectiveness only by lost tourists misses the point: coercion is supposed to change another government's behaviour. So I evaluated each episode against three criteria: did China achieve a clear, intended political outcome; did the operation cause the intended fall in tourism and economic loss; and did China face meaningful backlash or countermeasures that undermined the gain. I excluded the four 2020 episodes, since COVID border closures make tourism effects impossible to isolate, leaving twelve.
South Korea 2017 only: billions in losses followed by real diplomatic concessions. The only case where the target conceded.
Taiwan 2016, Palau 2017, Taiwan 2019, Japan 2025. Severe economic pain, but the political demand was refused.
Turkey, India, Sweden, Canada, and the United States three times. Advisories against diversified, low-exposure economies produced noise, not leverage.
That gives a clear success rate of 8.3 per cent, a weighted rate of 25 per cent counting mixed cases as half, and an operational effectiveness rate, meaning the tourism damage itself landed as intended, of 41.7 per cent. The gap between those last two numbers is the finding: China is good at causing pain and bad at converting pain into policy change. Pain is necessary but not sufficient.
The only complete success. A de facto group-tour ban cut Chinese arrivals by around 48 per cent across the year, with monthly falls exceeding 60 per cent at the trough. Tourism losses of roughly $6.8 billion; whole-economy estimates ran to $15.6 billion. Recovery came only after political concessions.5,6
When Chinese agencies stopped selling Palau packages, arrivals fell from a 2015 peak of 91,000 toward 28,000 by 2019, charter bookings halved, and the economy contracted. Yet Palau never switched recognition.7,17
Individual travel permits suspended for 47 mainland cities, timed ahead of an election; an industry estimate put the cost at roughly 700,000 fewer arrivals over six months and about $900 million in lost revenue. Taiwan subsidised, diversified, and did not bend.16
A shadow advisory after the prime minister's Taiwan remarks: around 491,000 tickets cancelled within three days, December arrivals down 45 per cent year on year, roughly $1.2 billion in spending lost by year end, and projections near $14 billion annually if sustained. Advisory-level deniability with ban-level damage, because the target was exposed.4,13,14,15
The database shows coercion works on exposed countries, so the obvious question is who is exposed now. I ranked the ten most China-dependent tourism markets by the Chinese share of their 2025 arrivals, then assessed how much total leverage Beijing holds over each, adding trade dependence, debt, and investment.
Compiled from national statistics for the 2025 periods recorded in the leverage sheet10,26
Arrivals share alone understates the picture: Laos ranks only fifth on tourism yet first on my overall leverage index, because around half its debt service is owed to China.9 Thailand sits tenth at 13.6 per cent, its dependence nearly halved since 2019 by the collapse itself, and diluted further by diversified trade with the US, Japan and the EU.1 The uncomfortable version of that sentence: Thailand's declining vulnerability was not achieved by policy. It was inflicted.
There is a large academic literature on states weaponising trade in goods: Hirschman's National Power and the Structure of Foreign Trade argued in 1945 that asymmetric trade dependence is political power, and Farrell and Newman's "weaponized interdependence" extends this to modern economic networks.23,24 Tourism barely appears in that literature. This section is my argument that it should.
First the raw numbers. For my ten countries I compared the estimated money Chinese tourists bring in against the value of commodity imports from China, calculated from 17 commodity categories in 2024 trade data.11,26 Commodities are larger in eight of ten cases; only in Palau and the Maldives does tourism money exceed commodity flows.
| Country | Chinese tourist spend | Commodity imports from China | Larger channel |
|---|---|---|---|
| Mongolia | $456M | $990M | Commodity |
| Palau | $29.6M | $14.4M | Tourism |
| South Korea | $7.47B | $19.9B | Commodity |
| Vietnam | $2.44B | $23.6B | Commodity |
| Laos | $548M | $639M | Commodity |
| Cambodia | $461M | $2.36B | Commodity |
| Japan | $11.1B | $15.9B | Commodity |
| Singapore | $5.87B | $12.4B | Commodity |
| Maldives | $579M | $93.1M | Tourism |
| Thailand | $10.2B | $11.9B | Commodity |
Seven criteria, weighted by importance in the EPQ · ○ marks a criterion that depends on conditions rather than favouring one channel
By value, then, tourism looks like the weaker lever. But I scored both channels against seven criteria of coercive usefulness, weighted by importance: substitutability, controllability, speed of impact, cost to the sender, deniability, targeting precision, and perception effects. Tourism wins four of the seven. It acts faster, since bookings can collapse overnight through a travel warning amplified on social media, where commodity flows have contracts and buffers; it is cheaper for the sender, hurting Chinese travellers' holiday plans rather than Chinese manufacturers' reputations as reliable suppliers; it is far more deniable, because a safety advisory never has to admit political intent, while an export restriction is visibly a state act; and it shapes perception, quietly rebranding the target as dangerous in the minds of millions. Commodities win on controllability and precision: goods move through customs, people make their own decisions.
Tourism is not a smaller version of trade coercion. It is a different instrument with a different comparative advantage: speed, deniability, and narrative.
That is why Beijing reaches for it first, and why the shift from formal bans to shadow advisories is not a softening but a refinement.
Everything in this chapter treats countries as units and tourists as flows. But a 48 per cent fall in arrivals is not experienced by a country; it is experienced by a hotel, a gem trader, a street vendor. The next chapter goes to ground level.
The Fieldwork
Fieldwork evidence · three online interviews · Mae Sot, Tak Province · Summer 2026 · Yaowarat is the next site
The database treats countries as units. This chapter is what one of those numbers, a collapse in Chinese demand, looks like from inside a hotel lobby, a restaurant floor, and a bakery kitchen in one border town.
Mae Sot is not a tourist town, and that is exactly why it is useful. The Chinese presence here was never sightseeing: it was casino visitors crossing to Myanmar, traders, and businessmen, layered on an economy that runs on border trade. When national headlines talk about Chinese money, scam centres, and safety fears in Thailand, this is the place those stories physically touch. It is a contrast case: somewhere Chinese demand shapes livelihoods through channels that arrival statistics never capture.
Semi-structured interviews, conducted online and in Thai, following a ten-question guide built around five core questions: what drives the business, how it compares with the Chinese peak, why they think the numbers fell, how Thailand is seen from abroad, and whether Thailand should depend less on Chinese visitors. Every respondent heard the same consent script. All three asked for their identities to be kept private, and I have honoured that: respondents are described by role rather than name, identifying details are withheld, and quotations are lightly reworded from my interview notes and left unattributed.25
I interviewed people whose businesses sit at different distances from Chinese demand: a hotel that once filled half its rooms with Chinese guests, a hotel and restaurant that lived on high-spending visitors, and a bakery that never depended on them at all. That spread is deliberate. If exposure determines damage, these businesses should have lived through the same years very differently. They did.25
Before COVID, half his guests were Chinese, most of them crossing to the casino on the Myanmar side, with the rest split between Thai and Myanmar visitors. Today the Chinese guests are gone: those who still come pass through without staying, avoiding the name registration a hotel record requires. Room rates rose from 800 to 1,200 baht as the hotel repositioned toward nature tourism instead of casino traffic.
Chinese share of guests · ~50% then · ~0% nowAt the peak, Chinese guests were 40 per cent of customers and by far the biggest spenders: two Chinese diners might spend 3,000 to 4,000 baht where two Thai diners spend 500. The share fell to 30 per cent during COVID, 20 per cent as the Myanmar conflict escalated, and under 5 per cent today. Staff kept their salaries, but the service charge that once topped up incomes collapsed with the revenue it was based on.
Chinese share of customers · ~40% then · <5% nowOpened months before COVID, closed for two years, reopened, and has grown every year since: from working alone to ten employees, and from 100,000 to 200,000 baht a month to 600,000 to 800,000. Chinese customers were only ever a tenth of her sales, ordering in bulk through intermediaries for delivery into Myanmar. Her explanation for her own growth is simple: the bakery never relied on Chinese customers, so their disappearance never touched it.
Direct Chinese share · ~10% then · Indirect only nowThis is the micro version of the finding that runs through my database: exposure determines damage. Nationally, coercion works on countries where Chinese visitors are a large share of arrivals. Inside one town, the same rule holds business by business. The hotel built on Chinese casino traffic lost that market entirely; the bakery that never had it grew fourfold through the exact same years.
What follows is three businesses' accounts as I recorded them, and my reading of them. It is not a survey of the town.25
Disproportionate to their headcount, because of how they spent. Respondents consistently described Chinese customers as the highest spenders: a dinner bill of 3,000 to 4,000 baht for two, against 500 baht for two Thai customers, and bulk bakery orders of 4,000 to 5,000 baht at a time. Losing a Chinese customer was never losing one customer's worth of revenue. One hotel reported over two thousand customers in a recent month, of whom ten were Chinese.
Not only the number of Chinese visitors but the kind. Older sightseeing and casino-bound tourists have been replaced by young businessmen trading with Myanmar, and by people passing through without staying the night, some deliberately avoiding hotel registration records. Respondents dated the turn precisely: the Myanmar conflict from 2020 and 2021, then a sharp cut about two years ago when the scam-centre stories and the kidnapping of a Chinese actor made national news18 and enforcement visibly tightened.
The damage arrives through pay packets before it arrives through job losses. Hotel staff kept their base salaries, but the service charge, the revenue-linked bonus shared among staff, collapsed from a peak of around 20,000 baht a month to between 2,000 and 5,000, taking a worker's total income from roughly 45,000 baht to 35,000. Around town, the losses concentrated in the most tourism-dependent businesses: nightclubs and bars closed while everyday commerce carried on.
Faster and more concretely than national policy debates suggest. The restaurant cut prices and switched to a volume strategy, joined Grab delivery, began promoting to Bangkok customers, and prices differently for businessmen and students. The hotel repositioned its marketing from casino traffic to waterfalls and nature reserves, and raised room rates for a different clientele. The bakery sells through Facebook and pays its Karen and Myanmar staff above what it paid at opening. Myanmar customers, described repeatedly as high-spending, have partly replaced the Chinese across all of them.
Not one respondent mentioned Chinese government policy. One said directly they were unaware of policies on either side. The explanations given were: safety news travelling at speed through Chinese social media, the kidnapping of the actor, high-profile shootings, a Russian tourist killed; the Myanmar conflict next door; Thailand's scam-centre reputation, which attaches to Mae Sot by name; dual pricing that charges foreigners more; and a generational shift toward independent travel. The strongest formulation I heard: the biggest reason is news, because it is the first thing a Chinese traveller learns about Thailand, and it changes the decision before price or beauty ever enter it. For a database full of deniable shadow advisories, this matters: on the ground, the instrument is invisible. What people experience is the information environment it moves through.
Here the ground disagrees with the policy textbooks. Diversification is the standard prescription, and it is what my own database implies. But the people absorbing the losses did not want less dependence; they wanted the Chinese back, because nobody else spends like them. What they asked for instead was the state doing its job: punishing the people who scam tourists, guaranteeing safety and being seen to guarantee it, promoting a new picture of Thailand, and developing tourism beyond the same few hotspots. Nobody blamed the tourists.
Chinese visitors only pass through now. They do not want a record of staying in Mae Sot, because a registered name makes coming in again harder.
Two Chinese customers at dinner would spend three or four thousand baht. Two Thai customers spend five hundred.
The salary never changed. What disappeared was the service charge. The bonus used to reach twenty thousand baht in a good month. Now it is two to five.
Before, Thailand was the land of smiles, safe. When a small number of people do these things, the whole country carries the reputation.
The biggest reason is news. It is the first thing they learn about us, and it changes the decision before they ever look at prices.
You cannot blame the Chinese tourists. Punish the people here who trick them. Information is a double-edged sword.
Quotations are reworded from interview notes taken in Thai, and are intentionally not attributed to individual respondents.25
Two findings from this fieldwork feed straight back into the macro project: exposure determines damage at every scale, and the instrument is invisible from the ground. The next chapter puts the two layers side by side.
The Synthesis
The database describes a mechanism operating at the level of states. The interviews describe what that mechanism feels like at the level of a shopfront. This chapter tests whether the second is a good description of the first, and asks what Thailand can actually do about it.
Everything in this project sits somewhere on one causal chain:
The value of doing both halves of the research is being able to see which links actually hold, and which break. Every one of them can:
An instrument can fire without moving flows. China issued advisories against the United States three times and against Canada, Sweden and India. Every one produced noise rather than leverage, because Chinese visitors were under 5 per cent of arrivals in each case.26
Flows can move without any instrument at all. Thailand lost roughly a third of its Chinese arrivals in 2025 with no advisory and no ban, driven by safety stories on Chinese social media.2,18 The chain can start at link three.
A fall in flows does not reach every business. In one town over the same years, a hotel lost essentially all of its Chinese guests while a bakery with no Chinese dependence grew fourfold. National averages hide who is actually standing under the falling number.25
Revenue loss arrives through pay packets before job losses. Hotel salaries held, but the revenue-linked service charge collapsed from around 20,000 baht a month to between 2,000 and 5,000, cutting a worker's income by roughly a fifth without a single dismissal.25
Putting the database and the interviews side by side produces three convergences and one honest disagreement.
The same rule predicts which countries coercion works on and which businesses on one street get hurt.
Every episode scoring 7+ hit a country above 20 per cent Chinese arrival share; every target under 5 per cent shrugged it off.
The hotel at ~50 per cent Chinese guests lost the market entirely; the bakery at ~10 per cent grew through the same years.
Twelve of sixteen episodes were deniable shadow measures, and the ground shows why they work.
Japan 2025: a mere advisory, 491,000 tickets cancelled in days, because the story travelled faster than any policy.13
No respondent mentioned Chinese policy. What they cited was news: the kidnapping, the shootings, the scam reputation, spreading at speed on Chinese social media.
Causing damage and converting it into change are different skills, at both levels.
Operational effectiveness 41.7 per cent, but clear political success 8.3 per cent. Palau bled and never switched recognition.7
Businesses did not surrender; they adapted. Volume pricing, Grab delivery, Bangkok promotion, a pivot to high-spending Myanmar customers.
Economic dependence on Chinese tourists would produce a desire for diversification. That is the standard prescription, and it is what my own database implies.
The businesses absorbing the losses wanted the Chinese back, because nobody else spends 3,000 baht on dinner for two. Their ask was safety and law enforcement, not new markets.25
The policy question moved from "how do destinations escape dependence?" to "how can destinations reduce vulnerability without destroying the benefits of the relationship?"
Four directions, stated as options with their costs rather than as recommendations. Each draws on the database, the fieldwork, and arguments made in Thai public debate by tourism academics since the 2025 collapse.20,21
Spread exposure so no single government's advisory can move a third of your arrivals. It is already happening regionally: Cambodia has courted Indian tourists as Chinese numbers fell,12 and in my database, diversification was the standard response of every targeted country, from Korea's Southeast Asia marketing push to Taiwan's source-market widening. In Thailand's case the collapse has done part of it involuntarily, with Malaysia overtaking China as the top source market in 2025.2
The catch: in the database, diversification scored consistently mediocre as a response, because replacement is slow and no market replicates Chinese spending. The fieldwork put a number on that: 3,000 to 4,000 baht at dinner against 500. Diversifying arrivals is not the same as diversifying revenue.
The fieldwork's clearest message: the biggest reason is news, and marketing cannot outrun a story that is substantially true. That points at enforcement, not advertising. Crack down visibly on the scam networks and call centres, with Chinese cooperation where it is offered, because visible enforcement is itself the message: when South Korea cracked down on zero-dollar tours, it made national news in China.21,22 End dual pricing so tourists pay what locals pay. Punish the people who trick tourists, which was the direct request of my own respondents. And take on the corruption underneath it, since, as Thai tourism academics have argued bluntly, foreign crime in Thailand does not operate without local help.21
The catch: this is the slowest and politically hardest option, and it costs real revenue in the short term, because profiting from dual pricing and looking away from grey money are entrenched. It is also the only option that addresses the cause rather than the symptom.
Three specific reforms recur in Thai debate.20,21 First, reorganise the free-visa programme: intended to attract tourists, it backfired in Chinese perception by appearing to open the door to grey business, so screening and enforcement need to be visible parts of the offer. Second, build the capacity to respond at narrative speed: the 2025 damage travelled through Chinese social media faster than any official channel could answer, so Thailand needs content and crisis response that actually reach Chinese platforms, not press releases that reach nobody. Third, build standing bilateral machinery with China, a hotline and a joint working group on tourist safety, on the model of the structures created during the US-China trade disputes, so problems are handled before they become stories.
The catch: narrative-speed response only works if there is substance behind it, which loops back to enforcement. And bilateral machinery requires a government willing to raise problems with Beijing directly, which Thai commentators note has historically taken more nerve than Bangkok has shown.
The uncomfortable option, because Thailand is already living it: Chinese dependence fell from 28 per cent of arrivals to 13.6 per cent between 2019 and 2025.1 But that resilience was inflicted by collapse, not achieved by policy, and my fieldwork shows the ground actively resents it: businesses want the dependence back because it paid. Genuine de-risking would mean deciding what level of exposure is acceptable and building toward it: renewing the tourism product beyond the same few hotspots, as Vietnam has with new attractions and routes;3 spreading development more evenly, which my respondents asked for; and pursuing higher-spending visitors, which every analyst recommends and every operator finds hard.
The catch: reducing dependence deliberately means accepting less income now for less vulnerability later, and the people who pay the "less income now" are the hotel workers whose bonuses already vanished. Any honest version of this policy has to say who bears that cost.
I started with a pattern nobody had catalogued: a decade of Chinese outbound tourism restrictions. Collecting, coding, and scoring 16 episodes turned an anecdote into a dataset, and the dataset into findings about instruments, vulnerability, and success rates.
The data showed tourism behaving like a tool of statecraft: clustered around disputes, deniable by design, effective against the exposed. Comparing it with commodity coercion placed tourism inside the economic statecraft literature, where it has barely been studied.
Then I went to where the numbers land: fieldwork on the Thai-Myanmar border, interviewing the people underneath the arrival statistics. The interviews confirmed the vulnerability rule at street level, and corrected the policy debate in ways only fieldwork can.
These limitations define what I can reasonably claim, and they are part of why the research combines case evidence, quantitative data, documentary evidence, and fieldwork.
A small-N study. It can show patterns, such as the link between exposure and effect, but not statistical proof.
Shadow measures are deniable by design, so the episodes that can be identified may under-represent the ones that occurred.
Arrivals move for many reasons: exchange rates, flights, safety news, a pandemic. No single episode isolates Beijing's action; the four 2020 cases were excluded for exactly this reason.
Economic-loss figures are ranges drawn from reporting and my own calculations, not audited accounts.
Individual accounts from one border town chosen as a contrast case, not a sample of the destination economy.
What respondents said is reported as said. Why the numbers fell is their reading and mine, and it is marked as such.
The database publishes in full alongside my EPQ, and the fieldwork continues in Yaowarat, Bangkok's Chinatown, where the relationship between Chinese demand and Thai livelihoods is the conventional one. The step I cannot yet take is the comparison case: the same questions, asked of Vietnamese businesses on the receiving end of the same shift. This site will grow as the work does.
Sources
Every externally verifiable claim on this page is numbered to an entry below; the superscripts in the text link here, and each entry names what it supports. Findings from the interviews rest on my own notes and are marked as fieldwork evidence.