What it looks like in Bangladesh's own numbers

Part II of three. A Bangladeshi manufacturer writes that one bad report on one factory tars them all. This is what being un-tarrable actually costs — measured in Bangladesh's own public register — and it turns out to be a question about money rather than about honesty.

In March 2025 a denim manufacturer in Chattogram wrote a sentence I have not been able to put down since: "It only takes a negative report on one factory for us all to be tarred with the same brush."

He is describing a market in which buyers cannot tell suppliers apart, so the whole group is priced on the worst thing anybody in it did. That is not a Bangladeshi condition and it is not a moral failing. It is what happened to Austria when a handful of people adulterated wine in 1985 and the country's exports fell from 48 million litres to four, taking seventeen years to recover — with ten thousand growers who had done nothing paying for it. It is what happened to Chinese dairy exporters in 2008, where economists measured the innocent firms bearing 64 per cent of the penalty the guilty ones bore. It happened to German car makers after 2015.

And the mechanism requires nobody to be dishonest. It only requires that the buyer cannot tell the difference at the moment he has to decide. That is the whole of it, and it is why this piece is not about anybody's character. (Part I sets out the economics — Akerlof's 1970 paper, and why the honest supplier is the one the mechanism selects against.)

I should say who is speaking. I ran garment factories in Bangladesh for seven years; I have been out for two; and I am now back in the same trade from the European end, which means the offers arrive at my desk instead of leaving it — and I cannot tell them apart either. Not because I think they are false. Because nothing in them lets me check.

So: what does it cost a factory to be checkable? Bangladesh publishes enough of its own data to answer that, and the answer surprised me.

The argument is not mine, and it is not new

The manufacturer above — Mostafiz Uddin — has been making it for years. In 2018 he wrote that "there do exist a small number of manufacturers who are trading improperly, and they are the ones tarnishing the reputation of the industry and Bangladesh as a whole," reaching for the English proverb about a rotten apple and a barrel. In 2019 Rubana Huq, then President of the BGMEA, put it in a sentence: "Sometimes a few bad examples tarnish the image of the sector and the entire industry suffers."

And it had already happened in the most literal way available. In 2007 a number of exporters were found to have faked GSP certificates of origin, and Germany responded with a 12 per cent security deposit on all Bangladeshi garment imports. Not on the offenders. On everybody. The Financial Express in Dhaka wrote that the offenders had "not only tainted Bangladesh's image abroad, it has also imperilled our trade relations with EU" — while insisting, correctly, that "Bangladesh cannot be singled out for such behaviour on the part of a handful of bad elements."

One more, because leaving it out would make this dishonest. The same writer made the identical argument in the opposite direction, about the people buying: "This handful of unscrupulous brands are also tarnishing the image of the whole gamut of buyers." He is right, and it is the same mechanism. A factory cannot verify a buyer's promise about next season's volumes any more than the buyer can verify the factory's promise about next month's delivery. Both ends of this trade are making claims the other end cannot check.

What a buyer actually does about it

He screens — before he starts, not after.

A researcher at Warwick, working from Bangladeshi customs records covering 2005 to 2012, put that question to the large buyers directly. Their answer, in her words: "the general screening of potential suppliers in aspects like productive capacity or minimum social compliance and quality standards is undertaken before the trading relation starts." Her data show the consequence — the size of a seller before a relationship begins does not predict whether that relationship survives. Capacity is not an advantage inside the door. It is the door.

And then the finding that decides everything. She observes that a supplier already trading with a large buyer can use that fact as a signal to other buyers, in a context — her words — "in which at least part of the payoff-relevant heterogeneity of the seller is not immediately observable." Trading with a known buyer "can be interpreted as a guarantee that the seller has passed a certain overall quality threshold."

So the cheapest proof that a factory is good is a buyer it already has. Which is precisely what a new factory does not have, and precisely what it is trying to get.

The instrument you can buy, and what it is priced in

That leaves the other instrument: the certificate. Bangladesh maps its own garment factories on a public register, and the interface behind it is open. I read all 3,320 records off it directly; between them they record 2,966,915 workers. A third of those factories — 1,126 — record no certification of any kind, and for 1,125 the entry is not a blank but the words "Do not have any certification." They employ 193,221 people, about one worker in fifteen on the register.

Before the table, the thing the table means, because it is easy to misread and the misreading is cruel: an audit costs money, staff days and a share of the year, continuously, and a floor of a certain size is simply what carries that overhead. What follows is not a measure of honesty. It is a price list.

Of the four hundred and twenty largest factories on the register, exactly one has nothing to show. Of the smallest five hundred and sixty-one, nine in ten have nothing to show.

The certificate is evidence of scale. What the table measures is who can afford to be checkable — and having nothing to show is a very different thing from having something to hide.

A quarter of the register — 816 factories, median workforce ninety — belongs to no manufacturers' association at all, and 84 per cent of those hold no certification. That matters more than it looks, because association membership is not only a badge: the declaration that lets a manufacturer import inputs duty-free for an export order is issued by the association, and the rule, in the words of the study that documents it, is that "a manufacturer can submit a UD application only if it is a regular member of the Association." For a large member that is an ordinary morning's paperwork. It is simply one more thing that is easier the larger you already are.

The trap, stated precisely

Newer factories can prove less than older ones. Of the 523 on the register founded since 2020, seven in ten have no certification and 85 per cent no third-party inspection, against 15 and 17 per cent for those founded before 2000.

Some of that is simply age — certification usually follows the first buyer who demands it — and some is survivorship, since the older cohort is the part that lasted. But look at what the concession concedes. The instruments of proof arrive after the order, and the order is what requires them. That is not a hole in the evidence. That is the trap.

Put the two halves together and the loop closes. The cheapest proof of quality is a buyer you already have. The proof you can buy instead is priced in employees, and a factory of ninety people cannot pay it. So the factory that most needs a way of being believed is the one with no way of being believed — and it is not being kept out by anything anybody did to it.

The debt

Here is where the cost lands, and it does not land on whoever caused it.

Somebody over-claims, wins an order he cannot hold, misses it, takes his money and is gone. The bill is paid by the next twenty people who write to that buyer, and the twenty after them — because the buyer does not conclude that supplier was unreliable. He concludes something cheaper to act on: that this is a risk he does not need. And then he stops answering.

So a factory that opened last year, that has never broken a promise, that has done nothing to anyone, begins its life already owing something. Its first letter is read — if it is read — by a person carrying a bad experience with somebody else entirely. Every letter it sends is paying interest on a debt it did not take out. The Chinese dairy data puts it plainly: it was the new firms that were hit hardest.

I am not saying the buyer is being unfair and should get over it. The buyer is behaving correctly. He cannot distinguish, so he prices for the average, and pricing for the average is the right response to a claim you have no way of checking.

Nor is most of what gets read as a lie a lie. "Yes — one hundred percent, on time, no problem" is usually a hope wearing the clothes of a fact, said by somebody for whom the honest answer — "I don't know yet" — loses the order today, and for whom there is no next month. A market that cannot check anything cannot separate a hope from a lie either. It files them together, and it is right to, because it holds no instrument that would tell them apart.

Which raises the only question worth asking next: what does build such an instrument? A good deal is known about that, including a large body of evidence about what has failed — and the honest answer is not the one I expected, and not the one that suits me. That is Part III.

— A.K. · TB Textile Sourcing · Slovenia, EU
(European who ran knit factories in Bangladesh for seven years, and is now on the buying side of the same asymmetry. Every figure above is published; sources are named in the text.)