The previous lesson ended with a puzzle: private long-distance trade looks thin in the documents, and yet Central Asia is full of Chinese silk, so somebody other than merchants was moving it.
The answer changes how the whole subject looks. Silk was not only a luxury fabric. In China it was money, in the strict sense of a standardised unit that things were priced in and debts were settled with, and the largest single holder of it was the state, which took it in tax and had to spend it somewhere. Where it spent much of it was the western frontier.
What money has to be
A commodity works as money if it is durable, portable, divisible, hard to counterfeit, widely wanted, and above all standardised, so that a quantity of it can be named in a contract without inspection. Cattle fail on divisibility, grain on durability, and most cloth on standardisation, since a length of homespun is worth whatever the buyer thinks of it.
Chinese silk was engineered past that last obstacle by the tax system. Tang law defined the bolt, pi, as 40 chi long by 1.8 chi wide. At a Tang chi of about 30 cm that is 12 m by 0.54 m, an area of 6.5 m², and at the weight of a plain tabby weave a bolt comes out at roughly a third of a kilogram. Every adult male taxpayer owed cloth to the state annually under the zu yong diao system, grain and labour service and cloth being the three components, and cloth delivered in payment of tax had to meet the statutory dimensions or be refused. A standard enforced on millions of households every year produces something a merchant in Kucha can accept sight unseen from a stranger.
Add the other property that made it necessary. China ran a chronic shortage of copper coin. Minting never kept pace with the economy, coin was heavy relative to its value, and it circulated unevenly, so large payments in cash were physically awkward and often impossible on the frontier. A currency that is also a valuable export commodity, and that weighs a third of a kilogram per unit rather than several kilograms, solves a real problem.
Example. Test the bolt against the properties above, and identify the one on which it is weakest.
It is durable, since silk stored dry lasts for centuries, and the Palmyra and Astana finds prove it. It is portable at an extreme: a camel-load of 150 kg is about 430 bolts, so a fortune moves on one animal. It is divisible in a rough way, since cloth can be cut, though a cut bolt is worth less than its share of a whole one, which is a genuine defect. It is hard to fake in the sense that weave quality is visible to anyone in the trade. It is universally wanted from the Pacific to the Mediterranean. Its weakness is supply control: unlike a coinage, its quantity is set by how much silk the empire's households happen to weave and how much the state chooses to take in tax, so it cannot be managed, and its value against grain and coin drifted considerably across the Tang. It is commodity money with all the strengths and the one central weakness of the type.
Now you. The Turfan and Dunhuang documents price small purchases in copper coin, medium ones in grain, and large ones in bolts of silk. Why would a single market use three currencies at once, and what does that tell you about the economy?
Answer
Because each is best over a different range and none dominates across the whole range. Coin is convenient for small sums and useless for large ones, since a major purchase would need a cart of it. Grain is what most people actually have, since most people farm, and it is the natural medium for a loan repaid at harvest, but it rots and it is bulky. Silk carries large value in small compass and stores indefinitely, so it is the medium for a horse, a house or a slave. A market that uses all three simultaneously is one where no issuer can supply enough of a single reliable medium, which is precisely the condition of a frontier prefecture at the end of a very long supply line from a state whose mints were always short. It also tells you that prices in these documents are not directly comparable without a conversion rate, and the rates moved, which is one reason economic history of the region is done in ranges.
What else was circulating
Silk was not alone, and the mixture in circulation is itself evidence about how the region worked.
Sasanian silver drachms, struck in Persia, circulated widely in Central Asia and turn up in hoards and in Chinese tombs, valued as bullion of known fineness rather than as the coin of any authority the holder recognised. Byzantine gold solidi reached China too, and here the finds are unusually informative: a substantial proportion of the pieces recovered from Chinese burials are imitations or thin one-sided copies, made to be buried rather than spent, which tells you the object had become a prestige token whose function was symbolic by the time it arrived. Chinese copper coin circulated in the Tarim wherever Chinese administration reached and thinned out sharply beyond it.
The Tang eventually produced an instrument that points at where all this was heading. Around the beginning of the ninth century merchants and provincial offices began using feiqian, flying cash: a merchant deposited coin in the capital, received a certificate, travelled, and drew the equivalent at the other end, so that the money made the journey and the metal did not. It is a draft rather than a banknote, since it represents a specific deposit, and it exists for exactly the reason this lesson has been describing, that moving value physically across a continent is expensive and dangerous. The Song turned the same principle into printed paper currency, and it was Yuan paper money that astonished Marco Polo enough for his description of it to become, six centuries later, one of the arguments that he had really been there.
What forty bolts bought
Two documents from Turfan put the currency to work, and they sit uncomfortably beside one another.
A contract of 731 records the sale of a Sogdian slave woman in the Turfan market. The price is 40 bolts of silk. The document is a routine commercial instrument: it names the parties, warrants that the seller has good title, provides for a guarantor, and follows the standard form of a sale of livestock or land, which is how the law classified what was being sold.
And 40 bolts of silk was also, in the same century, the standing rate for one horse from the Uighurs.
The equivalence is worth sitting with, because it is not a rhetorical flourish but a price. It also demonstrates the point of the section: bolts function as a genuine unit of account across categories of goods that have nothing else in common, which is what money does and what a mere luxury commodity does not.
At the labour valuation used earlier in this course, roughly fifteen labour-days to produce a bolt, 40 bolts is 600 labour-days, about two years of a working life. That is the order of magnitude of a horse, a person and a small farm all at once, and it tells you both how expensive horses were and how cheaply a human being could be bought at the edge of an empire.
The fiscal engine
Now the argument that reorganises the subject. The Tang held the Tarim Basin through the Anxi Protectorate, headquartered at Kucha, with garrisons at the Four Garrisons of Kucha, Khotan, Kashgar and Karashahr, and a standing force usually put at about 24,000 men. Those men had to be paid, fed and clothed at a distance of nearly four thousand kilometres from the capital, in a region that produced very little the army needed and no coin at all.
They were paid substantially in cloth. The Turfan documents show the machinery: government offices buying grain, hiring animals, settling accounts, issuing travel permits, all in a mixture of coin, grain and bolts, and the bolts arrive from the east because the state's tax system delivers them there.
Example. Suppose each of 24,000 soldiers received the equivalent of a dozen bolts a year in cloth, as pay and issue combined. How much silk is that, physically, and how does it compare with what private trade could plausibly have carried?
That is bolts, or kg, about 101 tonnes, which is camel-loads a year. Set that beside the private trade visible in the same documents, which consists of individuals moving with three or four animals and loans measured in single bolts. The state was putting several hundred camel-loads of currency-grade silk into the Tarim Basin every year as a routine administrative operation, and doing it whether or not any merchant sold anything. Whatever the private trade was, this was almost certainly larger, and unlike the private trade it is attested by the documents rather than inferred from luxury goods in graves. The assumption of twelve bolts is a guess and the conclusion is not sensitive to it: at six bolts a man the flow is still 336 camel-loads.
Now you. If the fiscal engine account is right, what should have happened to the oasis economies when the Tang lost control of the corridor after the An Lushan rebellion of 755, and is that a prediction you could test?
Answer
It should have produced a contraction that looks nothing like a decline in trade demand: not fewer exotic goods but less money. Garrisons unpaid or withdrawn, silk no longer arriving from the east, coin drying up, prices in the documents shifting away from cloth and coin towards grain and barter, and government paperwork thinning as the government that generated it stops functioning. That is testable in principle, because the Dunhuang and Turfan documents continue after 755, and it broadly holds: the Tibetans took the Hexi corridor and Dunhuang itself in the 780s, cutting the region off from Chang'an, and the character of the documents changes, with more barter and more local and monastic administration. The honest caveat is that the same period brings a change of ruler, of language of administration and of religion, so it is not a controlled experiment, and disentangling the loss of the subsidy from everything else that happened is not something the evidence supports doing cleanly.
Horses, and why China had to buy them
The other half of the fiscal engine is what the silk was for, and it is the most durable structural fact in the whole of Chinese frontier history.
China could not breed cavalry horses in the numbers it needed. The reason is agricultural: horses of military quality need extensive pasture, and Chinese land under the plough is worth far more in grain than in grass, so a horse raised in China is competing with wheat and losing. Add the disease environment of the wet south and the loss of the northern pastures whenever the frontier moved, and the arithmetic never improved. The one place with limitless pasture and expert horsemen was the steppe, occupied by the very people the cavalry was needed against.
So the pattern repeats for two thousand years: China buys horses from its enemies and pays in silk. The Han bought peace and horses with the heqin arrangement, and the Hanshu's record of the annual gifts to the Xiongnu, tabulated by Yu Ying-shih in Trade and Expansion in Han China (1967), shows the payment rising from around 8,000 rolls of silk in 51 BCE to around 30,000 in 1 BCE, a factor of nearly four in fifty years. Emperor Wu tried the alternative and sent armies to Ferghana in 104 and 102 BCE for the famous "heavenly horses", at a cost the Shiji makes clear was ruinous and a return of a few thousand animals, most of which died on the road. Buying proved cheaper than conquering, which is a conclusion the Tang reached independently.
Under the Tang the buyer's position got worse. The Uighurs had helped suppress the An Lushan rebellion and were owed for it, and in 773 they delivered 10,000 horses and demanded payment at 40 bolts a head, 400,000 bolts. The 754 census counted about 9.07 million registered households; if each owed something on the order of half a bolt a year, the empire's whole annual silk revenue was a few million bolts, so a single year's horse purchase absorbed something like a tenth of it. Tang officials wrote about it as extortion, and in a sense it was, but the court paid because the alternative was to face Turkic cavalry without cavalry of its own.
Example. The Han payments to the Xiongnu rose from around 8,000 rolls of silk in 51 BCE to around 30,000 in 1 BCE. What is the annual rate of increase, and what does the shape of the series suggest?
The ratio is over fifty years, so the annual growth rate is , about 2.7 per cent a year. That is a steady compounding escalation rather than a step change, which is the signature of a recurring negotiation in which one party can raise its demand a little at each renewal and the other pays rather than fight. Two readings are available and both are probably right in part. From the Han side, this is the cost of a policy that worked: fifty years of relative peace on the northern frontier bought at a price that grew slowly and predictably, which is cheap compared with the campaigns of Emperor Wu. From the Xiongnu side, it is a tribute stream extracted by a confederation whose leaders needed silk to distribute to their own subordinates, which means the demand grew with their internal political needs rather than with anything the Han did. The number to hold on to is the rate: a 2.7 per cent annual escalation is invisible in any one year and quadruples the bill in a lifetime.
Now you. Chinese officials repeatedly described these payments as tribute received from barbarians, while the recipients understood them as tribute paid by China. Which was it, and does the question have an answer?
Answer
Materially it was a transfer from China to the steppe, and the direction of the goods is not in dispute. Diplomatically both descriptions were maintained deliberately, and the ambiguity was the point. The Han court framed the arrangement as gifts bestowed on a submissive vassal who had come to offer horses, because the ideology of the Chinese state could not accommodate paying a foreign power, and the annals were written for a domestic audience that needed to read about submission. The Xiongnu and later the Uighurs framed the same exchange as payment owed, because their internal authority depended on being seen to extract it. Each side got the account it needed in the language it needed it in, and the goods moved regardless. This is worth generalising, because it is the standard structure of tributary relations across Eurasian history and it makes the sources treacherous: a Chinese record of tribute received may describe a payment made, and the way to tell is to follow the goods rather than the vocabulary.
What this does to the picture
Three consequences follow, and they should be held together.
The trade of Central Asia had a state at the middle of it. The silk that reached Sogdian merchants, Uighur khagans and eventually Persian and Byzantine markets was largely silk that had entered circulation as Chinese tax revenue paid out to soldiers and allies. Private commerce was real and was substantially downstream of a fiscal flow.
The relationship was not one-sided in the way the phrase "Chinese silk exports" suggests. China was the party with the structural weakness, buying a strategic necessity it could not produce from suppliers who could and did raise the price.
And the currency function explains the survival pattern of the evidence. Silk turns up in Central Asian graves, temple inventories and contracts not because Central Asians were unusually fond of Chinese fabric but because that is what money looked like there.
The limits are worth stating. This picture is built on two document hoards, Turfan and Dunhuang, both in the zone the Chinese state administered, so it may over-represent the state precisely because the state generated paper. The strongest form of the argument, that private long-distance trade barely mattered, goes further than the evidence can carry, and Hansen's critics are right that a trade conducted by illiterate carriers, settled in cash, and taxed by nobody would be invisible in exactly this way.
Still, somebody had to carry the goods, whoever was paying. The next lesson is about the people who did it: a single Iranian-speaking people from the valleys around Samarkand who ran the relay for six hundred years, and whose own letters, left in a mailbag near Dunhuang, show how a trading diaspora actually worked.