How 4,000 year old business letters helped economists find 11 lost cities
Thousands of 4,000-year-old clay tablets, mostly merchants’ receipts and complaints about dangerous roads, helped economists estimate where eleven lost Bronze Age cities once stood. Their 2019 study used a standard trade model and found that geogr...
Turkey tablets: How 4000-year old business letters helped economist in discovering lost cities. (AI generated image)
Nearly 4,000 years ago, a merchant in what is now central Turkey pressed a few lines into wet clay. He was holding a shipment back, he told a partner, because the roads were dangerous. Once the first caravan got through safely, silver would follow.
That tablet is one of roughly 12,000 deciphered texts used in a paper by historian Gojko Barjamovic and economists Thomas Chaney, Kerem Coşar and Ali Hortaçsu. It first appeared as an NBER working paper in November 2017 and was published in The Quarterly Journal of Economics in 2019. Most texts come from Kültepe, ancient Kanesh, in Turkey’s Kayseri province, where the paper says around 23,500 tablets have been excavated.
Donkey bills and delay notices, not poetry
Nobody wrote these for posterity. They are business letters, shipment records, expense accounts and contracts, found in merchants’ own houses. The traders came from Aššur, a city on the Tigris about 100 km south of today’s Mosul, and settled in Kanesh and other Anatolian towns. According to the paper, they carried luxury fabrics and tin, which originally came from Central Asia, and took back silver and gold.
One expense note quoted in the paper is almost painfully ordinary. A merchant records paying 2 shekels of silver and 2 of tin to hire a donkey for one stretch. For the next leg, he pays 4.5 of each for a donkey and a packer. That’s it. A transport bill.
Here is the problem the researchers faced. The tablets name dozens of places, and some have never been pinned to a map. Historians argue over them using clues like rivers and travel directions.
ADVERTISEMENT
The economists tried something else. They started from the gravity model of trade: places that trade more tend to be closer together. They searched the texts and found 2,806 tablets naming at least two cities. They read those one by one, and 198 described actual journeys. That gave them 391 trips across 26 cities, 15 with known locations and 11 lost. Feed in the known cities, let the model sort out the rest.
Where it worked, and where it didn’t
For several cities, it landed close. For Durhumit, the paper’s table puts the estimate about 30 km from one historian’s proposed site and 201 km from the other’s. For Washaniya, it was 7 km from one proposal. The authors say the estimates for 7 of the 11 cities carry a precision of under 30 km.
It also failed, and the authors say so. Purushaddum came out far from where either historian puts it, and the team itself calls that estimate unlikely to be right. Three other cities drifted into the Taurus mountains, because the model has no idea that snow-covered ranges are a bad place to build a town. As a check, they pretended Kanesh itself was lost. The model put it 12 km from the real site.
ADVERTISEMENT
The small town the model thought was huge
The model decided that Qattara, a small polity in northern Iraq, was enormous. The paper’s explanation: Qattara sat on the first stretch of the route from Aššur to Anatolia, and nearly every caravan passed through it, even though the Assyrians weren’t allowed to trade there. The model counted every passing caravan as business with Qattara.
Anyone who has looked at shipping data from a chokepoint today will recognise the trap. That comparison is ours, not the paper’s.
Why a 4,000 year old trade route still matters for investors
Here’s what makes it feel oddly modern. The authors describe merchants chasing profit and taking real risks, backed by financial contracts and a court system that actually worked. Aššur offered clear legal procedures and transparent taxes. Local rulers protected the roads and bridges and charged tolls for it. Swap the clay for paper, and it’s a list of what investors still ask for.
Then there’s the long game. The authors found that their estimates of ancient city size line up with modern population and night-light data in the same locations. They argue that geography, not farmland, explains it: cities that sit where natural routes cross keep winning. Kanesh sat at such a crossing, and so does Kayseri, 20 km away.
A caveat: the authors admit their estimates are not statistically sharp, and they count shipments, not their value. What this gives archaeologists are search areas, not a spade-ready address.
Nobody planned this. The merchant with the delayed shipment wasn’t writing to historians. He was managing a risky delivery. Four millennia later, his complaint about dangerous roads is data. Maybe that’s the odd lesson here: the records that survive are the dull ones.
