In August 2024 the water had not reached central Thailand or Bangkok. The damage had. One listed company wrote it into its filing to the Thai SEC, in its own words:
Those customers who deferred were sitting in places the water never reached. This damage appears in no aerial photograph, in no property loss report; nobody filed for relief on it, and it was never counted in the event’s damage figure — even though it happened, and was written down in a legally binding document.
1The blind spot in damage assessment
When a flood hits, what gets counted fastest is what can be seen: how many roads, how many houses, how many rai of farmland, how many factories. Those figures are necessary. But they are first-order damage only — the things the water physically touched.
What escapes the count is the second wave, which travels along the economy’s arteries rather than along the water line:
- Orders that vanished because a distribution centre could not deliver into the region — the loss lands at a warehouse in an entirely different province.
- Advertising budgets cut across a whole region — one of the fastest-reacting indicators of local economic conditions, and nobody collects it.
- Journeys that did not happen, surfacing as lower fuel sales.
- Investment deferred in provinces the water never reached, purely out of fear that it would spread.
- Residual effects across fiscal years, reappearing in the following year’s accounts long after the relief centres closed.
Nobody files for relief on this wave, because most of those affected are not inside a declared disaster zone. They were not “damaged” in any sense a government form recognises — and yet, in their financial statements, the numbers really did fall.
Why it matters: if the assessment instrument sees only assets inside the disaster zone, the damage figure will be systematically too low — and any measure designed from that figure will be systematically too small and aimed at the wrong place. This is not random error. It is bias with a direction.
Three examples of the invisible wave
All three passages below come from documents the companies filed with the Thai SEC themselves. Click “Open the source filing” to check any of them now.
2We read what the companies wrote, not an estimate
Every listed company is already legally obliged to explain its own results and risk factors to the Thai SEC. Inside those documents are sentences that say plainly what hit the business, how hard, and through which channel. The data is public, has been for years, and has almost never been read across the whole market at once — because it is scattered through tens of thousands of filings.
Boom Leverage Terminal reads the entire market in one pass — searching by meaning rather than by literal string, in Thai and in English — and returns the real passage with its page number and a link to the source:
There is exactly one screening criterion, and it is strict: a passage must have been verified as matching a real page in the source document. Anything that failed verification was dropped entirely. Nothing is paraphrased, nothing is rewritten, and no figure on this page was computed by a model. Every line opens back to its page in the original filing.
From model validation: this is exactly where a tool like this parts company with AI in general. A language model can write “the floods hurt northern retail” without ever having seen a document. But an answer you cannot check is useless in policy work — because when someone in the room asks “where did that number come from”, you have to open the page, not say the model said so.
Why this dataset actually works for policy
- It adds no reporting burden to the private sector, not one line — these are documents companies must file by law. No new questionnaire, no request for cooperation, no waiting for a survey cycle.
- The picture is visible within one quarter of the event — faster than fieldwork, and covering the whole country simultaneously.
- Every line is traceable — each passage carries the company, the fiscal year, the page number and a link to the source filing at the SEC. Evidence that survives the meeting, not output from a black box.
3Where the artery narrows — the transmission chain
The sequence below runs from the place the water actually was, outward to businesses that were never in the region at all. Read down it and you can watch the damage travel out of the disaster zone, and see how far it goes. Every passage is the company’s own wording, as filed.
4What the north shows more clearly than other events
The six observations below appear more sharply in the northern data than elsewhere, and each bears directly on how a measure should be designed — particularly the last two, which warn against treating every business in the region as equally damaged.
5What the private sector already did, before the state spends over it
The question rarely asked before a measure is designed is “who is already doing what?” Part of the answer is already sitting in the companies’ own filings — because a measure a company actually carried out affects its financial statements, and therefore has to be written into the report.
So what, for whoever designs the measure: if you know the lenders in the area have already suspended interest, a state debt-moratorium landing on the same group adds no relief — it adds budget cost. Surveying what is already being done before designing a measure is what puts the resource into the real gap, instead of layering it over something the market has already handled.
Capacity that already exists, and can be mobilised
Among the companies that discuss flooding, one group treats flooding as its line of business rather than a risk to dodge. That group is real, mobilisable capacity already listed on the Thai market. The list comes from companies describing that kind of work in their own filings — it is not a ranking, and it is not an endorsement of quality.
In the north specifically, one contractor reported revenue from tunnelling and underground structures up by THB 607.90 million on the Den Chai–Chiang Rai–Chiang Khong rail project — which means heavy construction capacity is already working in this area and does not have to be brought in from elsewhere.
6Precision policy — putting the money where it lands
What follows comes directly out of the evidence above. These are observations about how a measure is designed, not budget proposals, and they should be read alongside the data held by the responsible agencies. They all reduce to one point: blanket relief is expensive because we cannot see where to aim. Once you can see, the same money does more work.
-
Communicating the boundary of the affected area has direct economic value
One company states that its customers deferred investment from August 2024 because they were worried the water would spread to central Thailand and Bangkok. That damage occurred in places the water never reached, and was produced by uncertainty alone. Communication that defines the at-risk boundary clearly and quickly therefore reduces this damage without any construction budget at all — the cheapest measure of everything this evidence points to.
-
A one-quarter assessment window is too short
The impact still shows up in the following year’s accounts — both as reduced profit share from businesses in the region, and as a distorted comparison base created by stockpiling during the event. A recovery figure that looks healthy in the next quarter can therefore turn negative the following year with no new cause at all. An evaluation that closes its books too early will get the answer wrong in both directions.
-
The event is linked to a phenomenon that can be forecast in advance
One company states the cause explicitly: the onset of La Niña conditions in 2024, which brought sustained heavy rain and flooding across many areas. That link opens the door to using seasonal forecasting as an advance preparation signal — and preparing before the water arrives is always cheaper than paying for the damage after it recedes.
-
There is a way to measure a measure already in force, without a new survey
One company records a northern tourism stimulus covering 17 provinces in its own filing. When the private sector writes a state measure into a legally required report, the measure can be tracked through the operators’ own numbers. That is a feedback loop that already exists and costs nothing extra.
-
The impact spreads outside the disaster zone systematically
Those affected span out-of-home advertising, nationwide wholesale and retail, fuel stations, construction materials and online hotel booking. A damage assessment that counts only assets inside the disaster zone is therefore systematically below reality, and a measure aimed only at declared disaster areas misses an entire class of affected parties.
-
The private sector has already begun mitigating — design to complement, not to overlap
There are lenders waiving interest for customers in the disaster area under their own criteria, and reinsurers who have already absorbed part of the loss into the system. Surveying which private measures are already running, before designing a state one, is what puts the resource into the real gap.
7Every region, for context
For context, the table below shows every region the companies named in their own text. The figures are the number of passages and the number of companies mentioning it, not the size of the damage. A higher figure does not mean a region was hurt more.
| Region | Passages | Companies | Note |
|---|
Mentions of the northern floods, by fiscal year
8What this dataset cannot do
This section is necessary before anyone uses the work. Here, plainly, is what this dataset cannot do — because credibility comes from stating the weaknesses, not from hiding them.
Flood-specific cost cannot be isolated from public disclosure
Insurers report claims from several events in a single sentence — one reinsurance result refers to “damage from both the northern and southern floods in Q4 2024” together, and several others fold in the earthquake as well. The cost attributable to the northern floods alone therefore cannot be extracted. A separable figure would have to be requested directly from the insurance regulator.
We still cannot summarise the preventive measures the private sector actually uses
We tried to compile the mitigation measures companies state they actually carried out, and failed — so we are not presenting it. Almost everything found was not a company’s own measure but its line of business: manufacturers selling flood barriers, insurers paying claims, contractors building floodgates. Presenting that as “how the private sector protects itself” would distort the evidence. The cause is that the searchable set today is management discussion, not the risk-factor section, which is where companies actually write their mitigation. That section is still being processed.
- It tells you who reported what. It does not tell you where the water will go. Use it alongside the hydrological and warning data of the responsible agencies, never instead of them.
- It covers listed companies only, so it cannot see SMEs, smallholder farmers or households — who are most of the affected population in the north. That limit matters especially here, because the northern economy leans on small operators and agriculture more than the stock market can reflect. This dataset is therefore a direction indicator for the shock wave, not a headcount of the people hurt.
- Mention count is not risk level or damage size. The companies that disclose in detail are the companies that report well. Nothing here ranks a company or a region.
- The impact is not uniform. One hospital group states explicitly that the northern floods did not materially affect its operations. Assuming every business in the area was equally damaged is simply wrong.