Why the second order is the hard one

Part I of three. In Bangladesh the barrier to making a garment is capital and not much else. The barrier to being believed is something no single factory can buy. On a fifty-six-year-old piece of economics that explains the gap — and on what the same mechanism did to a European wine industry, a Chinese dairy industry and a German car industry.

Two complaints run through this trade, and I have never once heard anybody put them side by side.

The first is that making a garment in Bangladesh is easy. Not easy as in simple — easy as in the barrier is capital and very little else. Rent the floor, buy the machines, hire people who already know the work. I would like to give you the figure for what that costs and I have to tell you I could not find one: no development bank, no institute, no paper I could reach puts a defensible number on starting a small unit in Bangladesh, so I am not going to invent one. Take the entry evidence instead, which is measured. Between 2005 and 2011, against a standing population of around 6,500 garment exporters, roughly 1,700 firms started exporting every year and roughly 1,700 stopped. In men's woven suits, about half the exporters present in a given year were gone within a year, and only a quarter to a third were still there after three. That is not an industry with a wall around it. It is a revolving door.

The second complaint is that getting a buyer to come back is nearly impossible. Everybody says it. Nobody explains it.

Here the measurement is better, and it is Bangladesh's own. Working from Bangladeshi customs records covering 2005 to 2012 — a hundred thousand orders — a researcher at Warwick followed 1,362 new relationships between large foreign buyers and Bangladeshi manufacturers. The probability of one surviving its first year was 0.57. After that: 0.37, 0.32, 0.27, 0.21. Two in five new buyer relationships did not last a year, and about one in five was still alive in the fifth. I could not find a more recent Bangladesh-specific equivalent and I will not dress a 2012 figure as a 2026 one — but the direction holds in the most recent global number there is: in 2019, roughly seventy per cent of world garment trade took place between parties who had traded the year before. Close to thirty per cent of it turns over annually.

The prize for being kept is not small. On that same customs data, the net present value of supplying a buyer who sources by relationship rather than by tender came to at least thirty per cent of the yearly profits in that relationship, and buyers who source that way measurably pay more for it.

So entry is cheap, and what entry buys is close to worthless without a second thing that is not for sale. Production is the half capital can buy. Being believed is the half it cannot.

Where I am standing

I should say where I am standing, because it is not where I was.

For seven years I ran garment factories in Bangladesh. In that job I was on the other side of every sentence above. I was the man a European buyer had to take on trust — making claims about capacity and dates and quality that he had no way of checking from where he sat, and being believed or not believed for reasons that often had nothing to do with me. I did not think of it as an information problem at the time. I thought of it as my week.

I have been out of it for two years. I am back in the same trade now, from the European end, which means that for the first time the offers arrive at my desk instead of leaving it.

And I cannot tell them apart.

I want to be exact about that, because it is the whole point and it is easy to hear as an insult. I am not saying the offers I receive are false. I am saying I do not know, and that nothing in them lets me find out. Most are probably close to what they claim. Some are probably better than they claim and are underselling themselves. A few are probably not what they claim at all. I have no instrument that separates the three, and the sender has no instrument that would let him prove which one he is.

That is the one demonstration I can offer that somebody arguing this from a desk cannot. I ran those floors. I know the vocabulary, I know which numbers get rounded in which direction and why, I know what a sample tells you and what it does not. If I cannot tell, then a buyer in Copenhagen who has never been to Dhaka has no chance whatsoever — and he knows it, which is why he does not answer.

That is not a complaint about anybody's honesty. It is a measurement of how little information is actually travelling.

It happened to my neighbours

I write from Slovenia. Our northern neighbour is Austria, and in 1985 it emerged that diethylene glycol had been added to Austrian wine to make thin stuff taste like a late harvest. A handful of people did it. Austria's own wine body still records the episode in one flat sentence — the adulteration led to "what is referred to as the 'wine scandal'" — and notes that exports fell almost to zero.

They did, and the figures deserve to be read slowly. Austria exported 48 million litres of wine in 1984. In 1986 it exported four. Not four per cent less. Four million litres against forty-eight. And it stayed there: five million in 1987, four in 1988, five in 1989. Austrian wine exports did not pass their 1984 level again until 2001. Seventeen years.

Austria has about ten thousand wine producers. Essentially every one of them, in 1985, had done nothing whatever. They lost the export market anyway, because a buyer in Hamburg holding an Austrian bottle had no way of telling which kind he was holding, and the cheapest answer to that uncertainty was to buy something else.

If you want the same thing with a number on the innocent, the cleanest measurement comes from China. When melamine was found in Chinese milk in 2008, three economists tracked the consequences through firm-product-level customs data. The contaminated firms lost 84 per cent of their export revenue against the industry trend — no surprise. The finding that matters is the other one: the firms that had done nothing suffered 64 per cent of what the guilty firms suffered. Two further results belong in any honest discussion of this subject. Firms that government inspectors examined and declared innocent did no better than firms nobody inspected at all. And new firms were hit harder than established ones.

It is not confined to poor countries or to food. When Volkswagen's emissions software became public in 2015, the other German car makers — who had done nothing — lost an estimated 34.6 per cent of annual sales and about two thousand dollars a vehicle in what buyers would pay.

Nobody thinks Austrians are dishonest. Nobody thinks German engineering is a fraud. That is the point, and it is why those cases come first. What happened in each was not a verdict on a national character. It is what happens to any market in which the claims stop being checkable.

The mechanism has a name

In 1970 an economist named George Akerlof published thirteen pages in the Quarterly Journal of Economics under the title "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism". In 2001 it won him a share of the Nobel Memorial Prize, awarded — in the committee's words — "for their analyses of markets with asymmetric information."

The argument is about used cars. The owner has learned what his car is worth; the buyer cannot tell, and cannot find out at the moment he must decide. Good cars and bad must therefore sell at one price, because nothing distinguishes them to the person paying. That price sits near the average. But at the average, the owner of a genuinely good car will not sell — he would be giving it away — so he keeps it, or never enters. Which lowers the average. Which lowers the price. Which drives out the next tier. Akerlof follows it down: the bad drives out the not-so-bad, the not-so-bad the medium, until it is possible that no market exists at all.

He notes that Gresham's law has "made a modified reappearance", then, carefully, that the analogy "is not quite complete" — under Gresham's law both parties can tell good money from bad, and bad money circulates because the exchange rate is fixed. Here they cannot tell. The difference is information, and that difference is the whole of this argument.

Two things about the mechanism matter more than the mechanism.

The first: nobody has to be dishonest for it to run. Not one seller in the model is required to lie. All that is required is that the buyer cannot tell the difference. That is why this argument can be made about a country without being an accusation about that country, and it is the only reason I am willing to make it.

The second: who pays. Akerlof is blunt — "dishonest dealings tend to drive honest dealings out of the market" — and he makes an accounting point most people who quote the paper leave out. The cost of dishonesty is not merely the amount by which some buyer is cheated. It must also include the loss from driving legitimate business out of existence. The party who loses most is the seller with the good car, whose quality is real, unpaid for, and priced as the average of himself and somebody he has never met.

Twenty-six years later Jean Tirole formalised why that penalty attaches to a whole group and outlives the people who earned it. When an individual's record is observed only with noise, the group's record enters the price he is offered, and therefore his incentives. His phrase for what this does to somebody who has just arrived is exact: new members "may suffer from an original sin of their elders long after the latter are gone."

And in 2015 two economists wrote the version that is almost this argument's thesis in a sentence. Studying exports where "buyers cannot observe quality prior to purchase", they found that a range of genuinely high-quality firms are "permanently kept out of the market by the informational friction", so that "countries with bad quality reputation can therefore be locked into exporting low-quality, low-cost goods."

The honest objection

There is a study that finds the opposite, and it should be met rather than avoided.

Two economists examined what French importers actually did after the Rana Plaza collapse in 2013. Their finding: French textile imports from Bangladesh rose continuously after the disaster, and the retailers who had sourced from Rana Plaza showed no drop in overall imports. What they did find was a relative decline for the retailers publicly named as having sourced there.

So the punishment landed on identified brands, not on an origin — the exact reverse of what happened to Austrian wine and Chinese milk. Anyone who tells you collective reputation always bites is overstating it, and I am not going to.

But look at what separates the cases. In Austria and in China the thing that had gone wrong was in the product, and no buyer could tell one bottle or one carton from another. After Rana Plaza the thing that had gone wrong was in a building, and the party that could be identified — and was, by name, in every newspaper — was the brand. The collective penalty attaches where the buyer cannot tell the units apart. Where somebody can be named, the market names them and stops there.

Which is a hopeful finding, if you follow it. The Chinese dairy study says the same from the other side: the spillover onto innocent firms was smaller in destinations where buyers had better information about who was actually involved. Distinguishability is the variable. It is also the thing that can be built.

What this leaves unanswered

So the buyer prices for the average because he cannot do anything else, and the honest supplier pays for it. That is the mechanism, and it is fifty-six years old.

It leaves the question I actually care about, and it is not a theoretical one. If a factory wanted to be distinguishable — if it wanted to be the one the buyer could check rather than the one he had to guess about — what would that cost it, and could it afford it?

Bangladesh publishes enough of its own data to answer that precisely. The answer is not the one I expected, and it is not about honesty at all. That is Part II.

— 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.)