PART ONE
PRICE BEHAVIOR

One
Secular Trends of Rice Prices in the Yangzi Delta, 1638–1935

Yeh-chien Wang

In an agrarian society like Qing China, grains are the most important commodities in domestic trade, and food consumption makes up more than half of the average household budget.^[1] Grain prices are therefore the leading indicator in the market; the direction and the magnitude of their movement generally reflect conditions of inflation, deflation, or crises of major proportion. Moreover, persistent changes in grain prices relative to prices of other commodities give rise to a process of income redistribution affecting the welfare of virtually all groups of people and eventually the social and political stability of a country. As such, a clear knowledge of the trends of grain prices will provide not only a key to understanding the state of economy and society but also a basis for further research in real wages, the standard of living, and many other areas once data on other economic indicators are uncovered.

[1] See Wu Chengming, Zhongguo ziben zhuyi yu guonei shichang (Beijing, 1985), p. 253; John Lossing Buck, Chinese Farm Economy (Chicago, 1930), pp. 361–64, 386; Sidney Gamble, Ting Hsien: A North China Rural Community (Stanford, 1968), p.118.

Empirical studies in price history for imperial China are still in their infancy because of scarcity of data,^[2] but the gradual opening of archival resources in both Beijing and Taibei offers us rich mines for historical exploration. What I am attempting to do in this paper is to delineate broadly the secular trends of the prices of rice, the single most important staple food of the Chinese people, in the Yangzi Delta for three centuries prior to World War II and to suggest some tentative explanations for the trends observed.

The Yangzi Delta is chosen as the focus of observation for two main reasons. First, price data for the area are more abundant and, by and large, of better quality than price data for other areas. I am thus able to construct a price series extending over three centuries. Second, because of its economic centrality, prices in the area reflected conditions of demand and supply in the national, not just regional, market. In late imperial China most of the long-distance trade used the waterways, of which the Yangzi River, the Grand Canal, and the sea route along the coast were by far the most important. Linking the eastern coast with the interior, the Yangzi River flows through China's most productive regions. Together with its tributaries and connecting lakes it provided the most efficient network of inland transportation. The Grand Canal joined the capital region to the resource-rich South, while the sea route tied together all of the coastal provinces from Hainan Island to the Liaodong Peninsula. Only the northwestern region and the southwestern corner of the empire remained relatively isolated. Before the Opium War (1840–42) there were, according to one study, more than 200,000 junks plying these waterways and other smaller rivers with a total carrying capacity amounting to 4–5 million tons.^[3] Strategically situated at the focal point where the three principal arteries converged, the Yangazi Delta thus became the hub of interregional trade (see Map 1.1).

In addition to the advantage it possessed in geographic position, the industrial structure of the delta further enhanced its economic significance. It was, on the one hand, the center of the textile industry. On the other hand, its agriculture was unable to produce sufficient food to feed its inhabitants because it had the highest density of population in the country and much of its cultivated acreage was occupied by cash crops, such as cotton and mulberries.^[4] These structural features of the delta economy gave rise to a

[2] A few works may be cited: Han-sheng Chuan and Richard A. Kraus, Mid-Ch'ing Rice Markets and Trade: An Essay in Price History (Cambridge, Mass., 1975); Hwang Kuo-shu and Yeh-chien Wang, "Qingdai liangjia de changji biandong, 1763–1910," Jingji lunwen 9, no. 1 (March 1981): 1–27; Yeh-chien Wang, "Food Supply in Eighteenth-Century Fukien," Late Imperial China 7, no. 2 (December 1986): 80–117.

[3] Fan Baichuan, Zhongguo lunchuan hangyunye de xingqi (Chengdu, 1985), pp. 35–83.

[4] Cf. Yeh-chien Wang, "Food Supply and Grain Prices in the Yangtze Delta in the Eighteenth Century," Proceedings of the Second Conference on Modern Chinese Economic History (Taibei: Academia Sinica, 1989), pp. 424–27.

Map 1.1.
Grain Trade Routes in Qing China.

growing two-way traffic in which cotton cloth and silk, the staple products of the delta, were distributed to the rest of the country while surplus food from inland and the newly developed areas came to the delta for local consumption or for transshipment to other areas where food was also in short supply. In the latter part of the eighteenth century the annual volume of long-distance trade in rice down the Yangzi River to the delta was probably between 15 million and 20 million shi , of which 5–6 million was transshipped to North China and the southeast coast (including 3 million shi as grain tribute to the capital). In addition, around 15 million shi of soybeans, bean products, and a variety of grains and fruits was transported from Manchuria and North China to the delta via the coastal waters and the Grand Canal. Beyond this, grain trade across provincial borders in the rest of the country was, quantitatively speaking, insignificant.^[5] It must be noted, furthermore, that grains and textiles formed an overwhelming proportion of commercial cargoes carried across provinces. Before the middle of the nineteenth century, those two categories, as estimated by Wu Chengming, accounted for 42 and 31 percent, respectively, of the total value of the seven major commodities that entered interregional trade.^[6] As the principal supplier of textiles to, and the consumer of most of the surplus food from, other parts of the country, the delta inevitably assumed the central role in the domestic market.

In a study on food supply in the delta in the eighteenth century, I selected for observation and analysis rice prices for 1738–89 in Suzhou and Hangzhou, the most flourishing prefectures in the delta and two principal prefectures in the Lower Yangzi Region, in conjunction with Quanzhou Prefecture of the Southeast Coast, Hanyang Prefecture of the Middle Yangzi Region, Huaian Prefecture in North China, and Guangdong Province in the Lingnan Region. My findings lend strong support to the proposition that the delta had economic centrality. First, prices show a remarkable degree of synchronized movement across all of the five macroregions linked by the three major waterways. Second, a Pearson correlation analysis of the deseasonalized, decycled, and detrended prices in these regions gives coefficients that are all positive, most of them of relatively high value (0.6 and over), and degrees of association between Suzhou and the rest that are the most pronounced. Although the data are far from complete and perfect, this survey of the grain trade nationwide and of grain price movements in a large part of the country does indicate the central position Suzhou occupied in the country's grain market.^[7]

Not only was Suzhou the national market for grain and textiles, it was also

[5] Wang Yejian (Wang Yeh-chien) and Hwang Guoshu (Hwang Kuo-shu), "Shiba shiji Zhongguo liangshi gongxu de kaocha" (Paper read at the Symposium on Rural Economy in Modern China, Taibei, Institute of Modern History, Academia Sinica, 1989).

[6] Wu Chengming, Zhongguo ziben , pp. 247–51.

[7] Wang, "Food Supply and Grain Prices," pp. 444–51.

the foremost emporium for many other commodities. In 1756, for example, Governor Gao Jin said in a memorial to the emperior that tung oil and black plums were produced in Huguang, white wax in Hunan and Guizhou, copper in Yunnan and Guangdong, coir fiber in Huguang, Jiangxi, and Zhejiang, and rattan in Guangdong. But all of these products were, he pointed out, shipped to Suzhou for distribution to other parts of the empire.^[8]

In Table 1.1, I have compiled an annual price series for the delta from 1638 through 1935 by combining four shorter series as follows: a Shanghai series for 1638–95, a series for Suzhou City (the capital city of Suzhou Prefecture) covering 1696–1740, a Suzhou Prefecture series for 1741–1910, and a Shanghai series for 1911–35. There are, however, a number of years for which price data are missing. In such cases, I have filled out the missing data by extrapolation (marked with an asterisk in the third column); for the years 1862–64, when Suzhou was occupied by the Taiping rebels, I have used Shanghai prices.

The core of these data is the 170-year-long Suzhou Prefecture series plus the preceding Suzhou City series. Combined, these two series cover the entire Qing period except for the beginning decades. The data for the Suzhou City series are obtained from reports of governors and imperial commissioners of silk works residing in the city. Since the city was then the largest grain market in the country, to which early Manchu emperors paid close attention, more price reports came from there than elsewhere in the country. But it was not until the establishment of a nationwide grain-price-reporting system in the late 1730s that reports became regularly required of local administrations. Under this system, provincial authorities throughout the country were required to submit to the throne monthly reports on prices of major grains in every prefecture under their jurisdiction.^[9] The Suzhou Prefecture series is based on these reports. My colleague and I have gathered 1,632 monthly reports for this period (1740–1910), of which 96 years are complete with 12 months of data, another 25 years with 11 months of data, and only 6 years without data at all (see column 3). Nonetheless, how reliable are these official data? Obviously we cannot proceed with our research unless we have some degree of confidence that they provide a good approximation of market prices.

For the present purpose of trend observation I shall employ two kinds of tests to evaluate the official data, first, to observe whether excessively high prices occur in, or are preceded by, years of major natural or man-made calamities in the area or other parts of the country and, second, to see whether the secular movements of prices as manifested in the official series

[8] Gongzhongdang Qianlongchao zouzhe (Taibei, 1979), 15:431.

[9] For monthly grain price reports, see James Lee, Cameron Campbell, and Guofu Tan, "Infanticide and Family Planning," in this collection.

(Table continued on next page)

(Table continued on next page)

(Table continued on next page)

(Table continued on next page)

(Table continued on next page)

(Table continued on next page)

(Table continued on next page)

Fig. 1.1.
Rice Prices in the Yangzi Delta, 1638–1935 (taels per shi )

are in line with those exhibited in other price series for the area that are derived from different and yet reputable sources. As indicated in Table 1.1 and Figure 1.1, rice prices peaked in 1756, 1786, 1814–15, and the early 1860s. In 1755 heavy flooding afflicted the Lower Yangzi and Huai River valleys, and in 1785 a dreadful drought blanketed the Lower and Middle Yangzi Valley, the North China Plain, and Manchuria. In 1814 the country was hit by both—flooding in Zhejiang, Fujian, and Jiangxi and drought in Jiangsu, Anhui, Henan, Sichuan, and Shanxi.^[10] Finally, it was at the height of the Taiping Rebellion (1850–64) that prices skyrocketed as never before in the Qing period.

In the fourth column of Table 1.1, I have assembled three more series of rice prices of shorter duration for Xiaoshan (a county in the neighborhood of Hangzhou), Changshu (a county in Suzhou Prefecture), and Shanghai. The Xiaoshan series for 1684–1802 is derived from the records of the Lai clan in the county; the Changshu series for 1836–1860 from casual notes of a local scholar; and the Shanghai series for 1862–1910 from the decennial reports of China's Maritime Customs. As shown in Figure 1.1, prices of these shorter series move mostly in step with those of the annual series. Moreover, the linear trends fitted respectively for the annual series and for the Xiaoshan series for a whole century (1684–1788) turn out to be virtually the same, that is, an increase at the rate of 0.0131 tael per year for the former series and 0.0136 for the latter. The same can be said of the annual series and the Maritime Customs series for the last three decades of the Qing period, with respective increases of 0.0848 and 0.0860 tael. There is, then, good reason to believe the general validity of official price data for the Qing so far as the secular trend is concerned.

For the last dozen years of the eighteenth century, however, the annual series and the Xiaoshan series diverge significantly from each other (see Fig. 1.1). I suspect that something went wrong with the price reports from Suzhou Prefecture for those years because they show a sudden and drastic decrease in the extent of price fluctuations compared with any period before or after. On the other hand, not only for Suzhou but also for many other prefectures in and out of Jiangsu, the movement of prices represents a V-shape between the mid-1780s and the early years of the nineteenth century and the trough in the Suzhou series appears to be among the deepest. I cannot at the moment give a satisfactory explanation for the divergence between the two series for this short period. Perhaps a real picture will emerge with the discovery of new data.

Of all the data in the annual series, those for 1911–35 are unquestionably

[10] Zhongguo jin wubainian hanlao fenbu tuji , comp. Zhongyang Qixiangju Qixiang Kexue Yenjiuyuan (Institute of Meteorology, the Central Bureau of Meteorology; Beijing; 1981), pp. 148, 163, 178.

Fig. 1.2.
Thirty-one-Year Moving Average of Rice Prices in the Yangzi Delta, 1638–1935 (taels per shi )

the most solid because they are compiled on the basis of price quotations in the Shen Bao and Xinwen Bao , two widely read newspapers in prewar Shanghai. The most rudimentary data are those for the seventeenth century, which are derived from random notes left by two contemporary local scholars. These notes do not have price entries for every year, and price data therein lack statistical uniformity in terms of product and date. Nonetheless, prices recorded by one scholar are highly consistent with those by the other. Moreover, virtually all peak prices noted by either of them coincide with years of calamities or come immediately afterward. All in all, while the quality of the four sets of data that form the annual series is uneven, the data are good enough to be taken as a general indicator of price trends in the market.

To better observe the broad trends of prices, I have, as did W. G. Hoskins for wheat prices in England, smoothed the annual series by using 31-year moving averages.^[11] The results are rendered in column 5 of Table 1.1 and plotted in ratio scale in Figure 1.2. The smoothed averages fit the original series rather well. In view of the probable defects of the series for the last dozen years of the eighteenth century, which may have exerted an undue influence over the projection of the price trend, the shallow trough in the mid-1780s may well be nothing more than a statistical illusion. I have therefore smoothed it out freehand with a broken line. So modified, the figure shows prices to have moved in two broad swings over the three centuries. Beginning in the 1640s, the first swing headed steeply downward, reaching bottom in the early 1680s, then changed direction, rising at a modest rate for over a century to its peak around 1820. Then came the second swing. Prices descended gradually through the early 1850s and, after experiencing drastic fluctuations caused by the Taiping Rebellion and its aftermath in the third quarter of the nineteenth century, shot up in the early 1880s and kept rising through the 1920s.

As pointed out before, prices moved generally in a synchronized fashion in the Yangzi Delta and most other regions. Indeed, we can find few exceptions to the long inflationary trend in the eighteenth century, the slight downturn in the second quarter of the nineteenth century, and the half-century upswing before the Great Depression.^[12] It may be worth noting that while broad trends in grain prices are also recognizable in premodern Europe, the turning points are much less uniform from one region to another. There were the well-observed Price Revolution in the sixteenth century, the fall of prices in the seventeenth, renewed inflation in the eighteenth. Nonetheless, the sixteenth-century inflationary period drew to an end between 1590 and 1600 in the south but between 1620 and 1640 in the north; and the seventeenth-

[11] W. G. Hoskins, "Harvest Fluctuations and English Economic History, 1480–1619," Agricultural History Review 12 (1964): 28–46.

[12] See also the other papers in Part 1 of this volume.

century deflationary period reached its nadir in the 1690s in Würzburg and Vienna but not until the 1730s and 1740s in England.

More interesting is the contrast in regional price differentials between the two continents. In the last half of the fifteenth century the western Mediterranean region led Europe with the highest prices, eastern Europe stood at the other end, and the northern and Atlantic regions fell in between. The ratio of wheat prices then was about 6 or 7 to 1 between the two extremes. A century later the gap was narrowed to 4 to 1 because of the growth of the grain trade on the Baltic, which brought more and more Polish wheat to the Mediterranean region. Even more remarkable was the development of the Atlantic trade in the next one and a half centuries. By 1700 the north (England, France, the Low Countries) had taken the economic leadership from the Mediterranean and turned itself into the region of expensive wheat. By the middle of the eighteenth century prices in most regions had, by and large, merged into one another, with a somewhat higher level in the north.^[13] Europe thus became a highly integrated economy before the advent of the Industrial Revolution.

Early eighteenth-century China was, on the whole, comparable with Europe in terms of market integration. As shown in Map 1.2, in South China, where most people lived and where rice was the staple food, the Yangzi Delta (Suzhou and Hangzhou) had the highest prices, that is, about 1.5 taels per shi , around 1740. Next came neighboring Anhui and the southeastern coast of Fujian and Guangdong, where rice prices stayed at 1.0 taels or somewhat higher. The third region was the vast food-producing area of Sichuan, Huguang, and Guangxi, where rice was sold at around 0.8–0.9 tael per shi , the cheapest of all. The ratio between the highest and the lowest prices is 2 to 1 in the South. North China produced little rice, and most people there ate wheat, millet, and kaoliang instead. Almost all of the rice that the well-to-do consumed in the North was shipped from the South; its price was therefore much higher than in the South. Should the North be added as the fourth region for observation, the ratio between the highest and the lowest prices would increase to 3 to 1 (note that rice prices in the North vary from 1.8 taels in Xi'an to 2.4 taels in Jinan and Chengde).^[14]

In the latter part of the nineteenth century, modern transportation, such as steamships and trains, was introduced into China. A narrowing of price

[13] For European prices, cf. F. P. Braudel and F. Spooner, "Prices in Europe from 1450 to 1750," The Cambridge Economic History of Europe , vol. 4, ed. E. E. Rich and C. H. Wilson (Cambridge, 1967), pp. 378–486.

[14] All price data come from the First Historical Archives in Beijing and the National Palace Museum in Taibei. The 1909 price for Guangzhou (2.30 taels per shi ) is apparently unreliable. Because Guangdong had become the province with the severest shortage in food supply, the Guangzhou price should be at least as high as the Quanzhou price. The 1909 price for Tianjin is unavailable; instead I give the price of top-grade rice in Beijing for the same year.

Map 1.2.
Rice Prices of Selected Prefectures in China, 1738–1740 and 1909 (taels per shi ).

differentials is observable, though the degree of uniformity was much less pronounced than in Europe on the eve of its industrial takeoff. Although in the latter half of the nineteenth century the highest prices still prevailed in the North, much change had occurred in the South. The Yangzi Delta, still the country's center of economic gravity, was superseded by the two southern coastal provinces of Fujian and Guangdong as the region of dear rice because the Taiping Rebellion took a heavy toll of lives in the delta. In the interior, Sichuan and Hubei became overpopulated, raising the price of rice, while Hunan, Anhui, and Jiangxi remained an area with large amounts of food for export. The ratio of prices between the two extremes stood at 1.5 to 1 in the South in 1909, or 2.3 to 1 if the North is included; both ratios are lower than around 1740. What this brief survey of price history suggests is that, given the inexpensive network of water transportation that radiated from the Yangzi Delta, China had a more integrated economy than Europe did in most of the seventeenth century. However, the emergence of the North Atlantic economy moved Europe along at an accelerated pace. By the middle of the eighteenth century the position of the two continents had been unequivocally reversed.

Economic historians consider population and the quantity of money to be the prime factors affecting the secular trend in prices. In addition, some scholars note the causal relationship between climatic changes and food supply, which may produce a significant impact on prices. We may observe these relationships with prices by the use of the well-known equation of exchange

where P stands for the price level, M for the stock of money, T for the volume of transactions, Y for real output, and V_t and V_y for transaction velocity and income velocity respectively. If T or Y increases with no change in the money stock or in velocity, P will decline. On the other hand, given T or Y , if MV_t or MV_y increases, P will rise.

What follows in this section is a historical review of the major variables—population, the stock of silver (the primary component of money for most of the period), and climate cycles—that, individually or in combination, may in large measure account for the two long price swings noted above. However, our present state of knowledge on any of the variables is so incomplete that most estimates are subject to a wide margin of error. Accordingly, any conclusions we may reach cannot but be tentative.

Population affects prices in a variety of ways. Malthusians have long stressed the imbalance between population growth and food supply. Since cultivated land cannot be expanded indefinitely, population growth will inevitably result in the decrease of per capita acreage. Increased input of labor and capital may raise the yields of land per hectare and thus compen-

sate for its shortage. Given the state of technology, however, the marginal productivity of land will eventually diminish. From then on, the food supply will not increase in proportion to population growth. Food prices cannot but rise; the level of general prices must rise too. On the other hand, not only does population increase mean more labor, but it may also contribute to capital formation and technological progress. At the same time, higher population density may stimulate an intensification of the marketing system, leading to crop specialization and division of labor and thereby raising agricultural productivity. Higher productivity means more abundant goods and services, which may in turn bring down the general level of prices.

Nonetheless, in a historical study of English population and prices Peter H. Lindert does find a strong correlation between the two prior to 1815, that is, before industrialization. Specifically, he observes that prices rose greatly in periods of rapid population growth (1526–1603 and 1760–1801) but climbed less or fell "in the period of less dramatic population increase between the early seventeenth century and the middle of the eighteenth." He offers two theories to explain how population growth affects prices. According to the first theory, rapid population growth may bring about a higher ratio of children to adults. With more children to support, a household generally has fewer savings. Lower savings relative to household income implies a greater demand for consumer goods and services relative to demand for money holdings, raising the price level. The second theory is what Lindert calls the "Goldstone variation." According to Jack A. Goldstone, population growth brings increased population density, urbanization, and specialization within agriculture. Monetized transactions will increase at the expense of production for home consumption. In an economy where the marketplace is underdeveloped, more frequent and smaller individual transactions should bring economies in the holding of cash, leading to a rise in the velocity of monetary circulation and higher prices.^[15]

Despite the pioneer works of Ping-ti Ho and Dwight H. Perkins, the size of China's population before the 1953 census is still a matter of much debate. At the end of the sixteenth century the total number of people in the country was, according to Ho's estimate, around 150 million.^[16] After studying the data on famine relief in Henan for 1593–94, Shu-yuen Yim concludes that official statistics on population are substantially understated and that the total population in the empire was no less than 200 million in 1600.^[17] Mass uprisings and unprecedented droughts hit the greater part of the country in

[15] Peter H. Lindert, "English Population, Wages, and Prices, 1541–1913," Journal of Interdisciplinary History 15, no. 4 (Spring 1985): 609–34.

[16] Ping-ti Ho, Studies on the Population of China, 1368–1953 (Cambridge, Mass., 1959), p. 264.

[17] Shu-yuen Yim, "Famine Relief Statistics as a Guide to the Population of Sixteenth-Century China: A Case Study of Honan Province," Ch'ing-shih Wen-t'i 3, no. 9 (November 1978): 1–30.

the second quarter of the seventeenth century, causing a drastic decrease in population. Scholars are still far apart as to the extent of loss in lives during the time of the Ming-Qing transition. Estimates of the number of people at the beginning of the Manchu period vary from 70 million to 100–150 million,^[18] though I am inclined to think that the upper range, or, say, 120 million, is probably a better approximation.

Following the Manchu conquest in 1644, political and economic order gradually returned. Before the pacification of the Three Feudatories and Taiwan in the early 1680s, however, bitter fighting continued in one part of the country after another because of Chinese resistance to foreign rule. Therefore, economic recovery proceeded slowly in the latter part of the seventeenth century. By 1700 China's population was most likely still quite below the level attained a century before, and perhaps something around 150 million is not far off the mark.

The eighteenth century is one of a few periods in Chinese history in which the country enjoyed prolonged peace and prosperity, and its population grew as never before. The baojia system for annual registration of population started in 1741 and was to last until 1850. Nevertheless, the statistics before 1776 are incomplete, and some of those close to the mid-nineteenth century are believed to be much inflated. We are left with only those for the last quarter of the eighteenth century and perhaps the first two decades of the nineteenth century that can be considered relatively reliable. According to official reports cited by Ho, the number of people totaled 353 million in 1820. By 1850 it had risen to 430 million. But, after a thorough investigation of the county data of Sichuan and a survey of apparently exaggerated figures for several other provinces, G. William Skinner concludes that China's population then was around 380 million instead.^[19]

In the third quarter of the nineteenth century probably as many as 40 million people perished in the Taiping Rebellion and other social upheavals that devastated a large part of the empire and particularly the populous Lower Yangzi Valley. However, by the end of the century China had apparently regained the population lost several decades before. Despite political instability that characterized the last years of Manchu rule and the early Republican period, the country witnessed an expansion of population. In the early 1930s the total reached 500 million.^[20]

It is interesting to note that grain price movements and population move-

[18] Dwight H. Perkins, Agricultural Development in China, 1368–1968 (Chicago, 1969), p. 216; Kang Chao, Man and Land in Chinese History (Stanford, 1986), p. 40.

[19] Ho, Studies on Population , pp. 281–82; G. William Skinner, "Sichuan's Population in the Nineteenth Century: Lessons from Disaggregated Data," Late Imperial China 8, no. 1 (June 1987): 68–76. In the light of Perkins's and Skinner's evaluations, official population data for the early nineteenth century are in all probability exaggerated.

[20] For the 1933 population, see Perkins, Agricultural Development , p. 16.

ments in China appear to be in line with what Lindert finds in the English case. As shown in Table 1.2 and Figure 1.3, a fairly strong correlation between population and prices is observable over nearly three centuries. Moreover, periods of rising prices happened to be periods of relatively rapid population growth (the 1700s and 1880–1930), whereas downward movements in prices were mostly accompanied by slower or negative growth in population (1650–80, 1820–80).

China's monetary system in the Qing period was bimetallic, with silver and copper cash. Paper notes issued by native banks (qianzhuang ), pawnshops, and other merchants appeared in the late eighteenth century. After the Treaty of Nanjing (1842) foreign banknotes began to circulate in treaty ports. China established its first modern bank half a century later. Not until the early twentieth century did banknotes and demand deposits constitute the largest component of the money supply. To observe the possible relationship between the stock of money and prices, I have made an estimate of the total stock of silver, the most important monetary metal in China, for a number of years in the three centuries under study.

Although China did not possess rich silver mines, it was able to cope with the growing demand for specie by importing silver from abroad for most of the period. The arrival of Westerners in the Indian and Pacific oceans following the Great Discoveries opened a new era in relations between the East and

Fig. 1.3.
Trends of Population, Silver Stocks, and Rice Prices in China, 1650–1930

the West. Several Chinese products, such as silk, tea, and porcelain, were in great demand in Europe, America, and Japan, though manufactures from abroad struck little enthusiasm in the Middle Kingdom. Nor could the products—spices, ivory, sandalwood, etc.—from Southeast Asia, then under the domination of Western powers, offset the trade deficit incurred. To settle the balance, foreign merchants could offer only one thing, silver, that hardly any Chinese would refuse to accept. Meanwhile the discovery of rich silver mines in Japan, Peru, and Mexico provided a timely means of payment. Portuguese, Spanish, Dutch, British, and American, as well as Chinese, merchants all took part in this thriving trade, and large quantities of silver were thus brought into the country.

From the latter part of the nineteenth century China's economy was more closely integrated with the world market. When prices of silver were lower in the world market than in China, large amounts of specie would find their way through arbitrage into its trading ports, and vice versa. Since other countries had demonetized silver one after another, while China continued to use it as a medium of exchange and means of payment until 1935, it commanded a premium in the country for most of the period. The result is that China continued to be one of the largest recipients of silver even though it suffered a perennial trade deficit from the 1870s on.

How much silver did China possess over the three centuries under discussion? Let us begin with the most recent holdings. Eduard Kann, an expert on China's finance, estimated in early 1931 that there was about 2.5 billion ounces of silver in the country, including 1.7 billion in the form of silver dollars, sycee, and subsidiary coins and 0.8 billion in ornaments and household objects.^[21] At the rate of 1.30 silver dollars or yuan to an ounce this would amount to 3,250 million yuan , which accords well with the Bank of China estimate in the early 1930s.^[22] During the next five years, however, China incurred a net loss of 650–700 million yuan , largely because of the United States Silver Purchase Act. On the eve of China's monetary reform in early November 1935 its total silver stocks probably stood in the neighborhood of 2,600 million yuan .^[23]

Between 1893 and 1930, according to Maritime Customs statistics, China

[21] Eduard Kann, "How Much Silver Is There in China?" Chinese Economic Journal 8, no. 4 (April 1931): 410–20.

[22] See Thomas G. Rawski, Economic Growth in Prewar China (Berkeley and Los Angeles, 1989), pp. 363–65.

[23] The net outflow through Maritime Customs amounted to 292 million Haikwan taels, or 456 million silver dollars, in 1931–35. In addition, somewhere between 225 million and 250 million yuan was smuggled out of the country in 1934–35. Liang-lin Hsiao, China's Foreign Trade Statistics, 1864–1949 (Cambridge, 1974), p. 129; Lin Weiying, Zhongguo zhi xinhuobi zhidu (Changsha, 1939), pp. 4–5; Yu Jieqiong, 1700–1937 nian Zhongguo yinhuo shuchuru de yige guji (Changsha, 1940), p. 3.

received from abroad 1,284 million silver yuan (823 million Haikwan taels).^[24] Such being the case, it may not be far off to put the accumulated amount of specie at 1,900–2,000 million yuan at the end of the nineteenth century.

For earlier periods we are on a far less certain ground. In 1643 Jiang Chen, a scholar interested in fiscal and monetary matters, submitted to the falling Ming government a proposal for the adoption of paper money. He figured the total silver stocks then held in the private sector to be around 250 million taels, or 350 million yuan .^[25] In an essay on the crisis of silver drain in the Daoguang period (1821–50), Taiping Shanren maintained that before the crisis unfolded, the country must have possessed something more than a billion silver yuan .^[26] Both figures appear to be plausible.

In a broad survey of the influx of silver between the 1570s and 1830s Sun Yutang gives the following breakdown: 100 million silver yuan from the Spanish Philippines (1571–1821), over 500 million from England and other European countries (1700–1828), 100 million from the United States (1784–1833), and 140 million from Japan (seventeenth through early nineteenth centuries).^[27] It should be pointed out, however, that the second item most probably includes amounts imported from the Philippines and the United States,^[28] and so it should be revised downward to 300 million. According to Ch'üan Han-sheng, however, the total shipment from the Philippines during that same period was likely to have been 200 million or more instead of 100 million.^[29] Moreover, there were fairly regular shipments of Japanese bullion in the latter part of the sixteenth century. A. Kobata notes, for instance, that the Portuguese carried 500,000–600,000 taels out of Japan annually to finance their trade with China.^[30] Therefore, the silver inflow may reasonably be assumed to be around 800 million for the two and a half centuries in question.

Of the 800 million, how much was imported before the fall of the Ming dynasty in 1644? Liang Fangzhong estimates the amount at more than 100 million.^[31] But Arai Hakuseki's study shows that Japan alone exported over

[24] Hsiao, China's Foreign Trade Statistics , pp. 128–29.

[25] Cited in Peng Xinwei, Zhongguo huobishi (Shanghai, 1965), p. 736.

[26] Taiping Shanren, "Daoguangchao yinhuang wenti," in Zhongguo jin sanbainian shehui jingjishi lunji (Hong Kong, 1974), 5:41–45.

[27] Sun Yutang, "Ming Qing shidai de baiyin neiliu yu fengjian shehui," Jinbu ribao , Feb. 3, 1951.

[28] Momose Hiromu, "Shindai ni okeru Supein doru no ryutsu," Shakai keizai shigaku 6, no. 2 (May 1936): 22.

[29] Quan Hansheng (Ch'üan Han-sheng), "Ming Qing jian Meizhou baiyin de shuru Zhongguo," in Zhongguo jingjishi luncong (Hong Kong, 1972), 1:435–50.

[30] A. Kobata, "The Production and Uses of Gold and Silver in Sixteenth- and Seventeenth-Century Japan," Economic History Review , 2d ser., 18, no. 2 (Nov. 1965): 245–66.

[31] Liang Fangzhong, "Mingdai guoji maoyi yu baiyin di shuchuru," Zhongguo shehui jingjishi jikan 6, no. 2 (Dec. 1939): 324.

100 million (748,000 kan ) in the first half of the seventeenth century, most of which ended up in China.^[32] When the shipments from the Philippines between 1571 and 1644 and those from Japan before 1600 are added, the amount should, in the light of the foregoing discussion, be around 200 million instead.

While domestic production of silver played a secondary role in the supply of monetary metals, silver mining in Yunnan was in operation for most of the Ming-Qing period. Wei Yuan, one of the foremost scholars in the statecraft school and a keen observer of the nation's economy, wrote in 1842 that "of all silver stocks [in the country] 30–40 percent comes from mining and 60–70 percent from foreign vessels."^[33] On the basis of his observation and the quantities of silver inflow, we may readily estimate the amount of bullion held in China to be in the range of 290–330 million silver yuan in the mid-seventeenth century and 1,140–1,330 million in the 1820s.

In the second quarter of the nineteenth century the balance of trade turned against China primarily because of the growing opium trade. According to H. B. Morse, there occurred a net drain of silver to the amount of 200 million yuan over the period.^[34] On the eve of the Taiping Rebellion the country's holdings probably fell to the level of 900–1,100 million. In the latter half of the century, nonetheless, China resumed importing bullion from abroad. J. Laurence Laughlin notes that between 1852 and 1875 alone, at least 1 billion yuan of silver had been shipped from England and Mediterranean ports to India and the East.^[35] So vast was the influx that China found its stocks of silver doubling in the half-century.

However rudimentary the estimate of silver stocks, the direction of change in the amount of bullion that China held over the three centuries is quite clear. We may, in light of the data presented in Table 1.2 and plotted in Figure 1.3, make two observations. First, not unlike prices and population, silver stocks in China generally followed a rising trend. Second, during the latter part of the seventeenth century and the second quarter of the nineteenth century, when the country's holdings of the white metal changed little or decreased, prices were on the decline. It appears, then, that silver stocks can serve as a crude indicator of the secular movements of prices.

To be sure, the stock of money included not only silver but also copper cash, paper notes, and, later, bank deposits. In the twentieth century, in particular, banknotes and bank deposits soon overtook silver as the dominant components of money. Nor did all silver stocks circulate as money; a

[32] Cited in Otake Fumio, "Min Shin jidai ni okeru gaikoku gin no ryunyu," in Kinsei Shina keizaishi kenkyu (Tokyo, 1942), p. 57.

[33] Wei Yuan, Shengwuji (1927), 14:33.

[34] Cited in Taiping Shanren, "Daoguangchao," pp. 45–46.

[35] J. Laurence Laughlin, The History of Bimetallism in the United States (New York, 1897), p. 125.

good part was hoarded in safes or underground or used for jewelry by the well-to-do. Therefore, other components of money as well as the velocity of circulation should, if possible, be taken into consideration when specific periods are examined.

Climatic conditions significantly affect the state of harvest and food supply. Various empirical studies show that relatively warmer temperature increases crop output by lengthening the growing season and by making uplands cultivable, and vice versa. Referring to sixteenth- and seventeenth-century Europe, for example, Andrew B. Appleby, G. Parker, and L. M. Smith suggest that a one-degree-centigrade decline in temperature will shorten the growing season by three to four weeks and is equivalent to raising land elevation by 500 feet.^[36] A recent study by Wang Shaowu indicates, moreover, that low summer temperatures in a number of years between 1954 and 1976 reduced harvest yields by a third in Manchuria.^[37] It was precisely because of concern for food supply in the empire that the Quing government required all local officials to submit regular weather reports along with reports on grain prices.

The degree of correlation between harvest yields and grain prices is nevertheless not so clear, because prices also depend on such other factors as access to food supply from other areas or from other parts of the world, grain storage, speculation, and government policy. To observe the possible relationships between climate, harvest yields, and secular prices, let us first look into climate cycles and harvest fluctuations in the delta over the centuries in question. Two groups of scientists in China have conducted original research on the long-term temperature changes in the Middle and Lower Yangzi Valley. One group, under the leadership of Zhang Peiyuan, has focused its attention on phenodata for the Qing period; the other, led by Yan Jiyuan, has made a historical investigation of the years when the Huangpu River and Lake Tai froze. They derived virtually the same climate cycles for the area.^[38] Between the mid-1600s and the 1970s the country went through three long cycles. The cold period of the first cycle lasted through the first decade of the eighteenth century; the warm period that followed prevailed until 1780. The second cycle extended for almost a century (1781–1882), with 1830 as the turning point between the cold and warm periods. In the early 1870s another cycle started. It was also evenly split, with the 1920s as the transition years

[36] Andrew B. Appleby, "Epidemics and Famine in the Little Ice Age," Journal of Interdisciplinary History 10 (Spring 1980): 658; G. Parker and L. M. Smith, ed., The General Crisis of the Seventeenth Century (London, 1978), cited in Patrick G. Galloway, "Long-term Fluctuations in Climate and Population in the Preindustrial Era" (Paper read at the Ninth International Economic History Congress, Bern, Switzerland, 1986).

[37] Wang Shaowu, "Jin sibainian dongya de lengxia" (unpublished manuscript, 1988).

[38] Zhang Peiyuan, "Qingdai hannuan bianhua ji qi dui nongye de yingxiang" (Paper read at the Workshop on Qing Population History, Aug. 1985, California Institute of Technology, Pasadena, Calif.); Yan Jiyuan et al., "Changjiang sanjiaozhou de lengnuan tedian yu qushi zhanwang," in Quanguo qihou bianhua xueshu taolunhui wenji (Beijing, 1981), pp. 71–77.

between the cold and warm periods (see Table 1.3). It is worthy of note that in the nineteenth and early twentieth centuries secular prices moved downward in the warm period (1831–72) but upward in the cold periods (1780–1830, 1873–1920), whereas the opposite is true in the seventeenth and eighteenth centuries.

I have not yet compiled data on harvest yields. In my study of grain prices in the Yangzi Delta, however, I do find a very strong association between harvest yields and grain prices. Over the half-century between 1738 and 1789 eleven cycles in both rice prices and wheat prices can be readily identified, and all peaks of these cycles occurred in years of crop failure or poor harvests caused by floods, droughts, or epidemics in the delta or in a large part of the country in the same year or the year before.^[39] It is, however, another matter whether fluctuations in local harvests affect the secular movement of prices in an area with easy access to the food-exporting provinces of the country. We may take the deviations of annual prices from their trend values (31-year moving averages) as a measure of harvest fluctuations in the delta and then compare them with changes in secular prices. Assuming that years of normal harvests are those with rice prices within plus or minus 10 percent of the trend, that years of good harvests are those with prices 10 percent or more below the trend, and that years of deficient harvests are those with prices 10 percent or more above the trend, we find that years of good and normal

[39] Wang, "Food Supply and Grain Prices."

harvests outnumber years of poor harvests almost 4 to 1 and that average deviations are within the range for normal harvests for all periods (see the last four columns in Table 1.3). Accordingly, cyclical fluctuations in local harvests did not, in my judgment, have a significant impact on the long-term trends of prices, although grain prices were very sensitive to fluctuations in local harvests in the short run.

After reviewing the literature on population, silver stocks, and climate cycles, I can now offer some tentative explanations of the long-term trends of rice prices for the three centuries in question. The high level of prices at the time of the Ming-Qing transition clearly reflects the inflation and the shortage of food consequent to large-scale warfare and a series of natural catastrophes.^[40] After the establishment of Manchu rule in 1644, peace and order were gradually restored in the country. Economic reconstruction followed, albeit slowly, with a concomitant improvement in the man-land ratio (depopulation had been drastic between 1600 and 1650). The supply of grain became plentiful relative to the demand for food, despite the cold climate that prevailed in the latter part of the seventeenth century.

During the same time the increase in the stock of money was at best moderate because the new Manchu regime imposed a total ban on coastal trade until the pacification of Taiwan in 1683. The shortage of exchange media received some relief, for the new government issued 1–2 million strings of copper cash annually between 1647–57; but the issuance was later progressively reduced to 200,000–300,000 strings a year.^[41] The transaction velocity of money most probably fell after the decades-long inflationary spiral attending the demise of the Ming dynasty. When MV either decreased or remained rather stable and T expanded, prices could not but decline.

In the next phase of the price swing—from the early 1680s to 1820—silver stocks increased at an annual rate of about 0.9 percent, while the central government resumed the policy of expansion in the issuance of copper cash, which was facilitated by rising copper output from Yunnan. In 1724 the production from Yunnan copper mines was about 1 million catties; from the early 1740s through the first decade of the next century the annual output always went beyond the 10 million mark. Meanwhile, the quantity of cash issued by the two mints in the capital jumped from less than 0.5 million strings to well over 1 million a year, and counterfeit coins flooded a greater part of the country.^[42] Furthermore, to put down the White Lotus Rebellion (1796–1804) the government spent 200 million taels, an amount equivalent

[40] Zhang Xiangong, "Zhongguo dongbanbu jin wubainian ganhan zhishu de fenxi," in Quanguo qihou bianhua , p. 48.

[41] Wang Yeh-chien, Zhongguo jindai huobi yu yinhang de yanjin, 1644–1937 (Taibei, 1981), p. 25.

[42] See Yeh-chien Wang, "Evolution of the Chinese Monetary System, 1644–1850," in Modern Chinese Economic History , ed. Chi-ming Hou and Tzong-shian Yu (Taibei, 1979), p. 442.

to its budget expenditures for six or seven years in normal times, which added greater inflationary pressure to the economy.^[43]

During the period 1680–1820 China's population grew at 0.72 percent a year, the highest rate in the three centuries under discussion (see Table 1.2). Moreover, eighteenth-century China was a time not only of rapid population growth but also of unprecedented commercialization. Cash crops, such as cotton, mulberries, sugar cane, and tobacco, were planted more widely in various provinces. Markets and towns proliferated as never before.^[44] These developments would most likely, in view of Lindert's analysis, noted above, lead to a rise in the velocity of money. Thus both the growth of the money supply and the growth of population were building up strong pressures for inflation. At the same time the cultivated acreage expanded by a half or so; new food crops (sweet potatoes, corn, and peanuts) were spreading, and double-cropping was increasingly practiced.^[45] Given generally warm climate the agriculture sector also experienced rather vigorous growth. Without large increases of real output the rate of the price rise would have been much higher during the upswing, which lasted for more than a century.

The second price swing followed a generally downward trend from about 1820, but turned sharply upward in the early 1880s and continued to rise through the 1920s. We may divide the half-century of downswing into two subperiods, with 1850 at the dividing line. Before 1850 warmer climate in combination with the silver drain had the effect of depressing prices. But the money supply may not have decreased because of the multiplication of paper notes issued privately by local banks, money shops, and various commercial stores and because of the increase of copper cash by debasement and counterfeiting.^[46] What was more likely to bring prices down was, besides warm climate, the possible decline in the velocity of monetary circulation caused by the apparent drop in the rate of population growth and by the increased holding of silver by the public. Before 1850 the value of silver appreciated steadily in relation to commodities and copper cash.^[47] In the expectation that its value was to rise, "wealthy people and rich merchants are striving to hoard silver"; thus observed Bao Shichen, a contemporary scholar with a great interest in political economy.^[48] Not only did the silver crisis slow down the velocity of circulation, but it also severely strained the issue of private notes. (Without central banking, private notes were backed by the credit of the individual issuing banks or by the merchants themselves.)

[43] Wei Yuan, Shengwuji , 10:16.

[44] See, for example, Liu Shiji, Ming Qing shidai Jiangnan shizhen yanjiu (Beijing, 1987).

[45] Perkins, Agricultural Development , chaps. 2–3; Yeh-chien Wang, Land Taxation in Imperial China, 1750–1911 (Cambridge, Mass., 1973), p. 7.

[46] See Wang, "Evolution of the Chinese Monetary System," pp. 425–52.

[47] Ibid.

[48] Bao Shichen, Anwu sizhong (1872), 26:8.

In the third quarter of the century mass uprisings took place in many parts of the country. Most virulent was the Taiping Rebellion, which lasted 15 years and devastated a large part of the Yangzi Valley, particularly the delta area. It cost the government more than 400 million taels to bring these peasant and minority revolts to an end.^[49] Hyperinflation followed immediately, as is shown distinctly in Figure 1.2. But the war-induced inflation proved to be just an aberration, for the level of prices sank as quickly as it rose following the return of peace and order in one part of the country after another.

As in the late seventeenth century, economic reconstruction proceeded with an improved man-land ratio, but the pace of economic recovery was faster. For one thing, however ruinous were the mid-nineteenth-century uprisings, they were not accompanied by the natural calamities and the consequent famine that attended the fall of the Ming dynasty (the droughts that hit China four years in a row in 1641–44 were the worst in the last 500 years). For another, continued warm climate further aided agricultural production. Prices were going down as more and more land was brought back under cultivation and more irrigation works were restored or built.

Extending from the 1880s to 1920s, the second upswing was the most inflationary period of the three centuries. While rice prices rose annually at 0.7 percent in the first upswing, they ascended at the rate of 3.3 percent a year in the second (see Table 1.2). During the half-century before the Great Depression the money supply grew at a very rapid rate. In 1930 the total stock of money (currency and deposits) amounted to about 6 billion yuan , of which one sixth were metal currencies.^[50] How much money was in circulation 50 years earlier? We can only hazard a guess. China's silver stocks then stood probably around 1.5–1.6 billion yuan (see Table 1.2). Assuming that a quarter of the silver was then in circulation and that the ratios of circulating silver stocks to copper cash and to paper notes were 2 to 1 and 5 to 1 respectively, the total amount of money in circulation would be in the range of 600–700 million silver yuan . It was probably not more than 800 million.^[51] Accordingly, the money supply may have increased by eight to ten times, or at an annual rate of 4.2 percent to 4.7 percent in the half-century. Without doubt the accelerating growth of the money supply contributed significantly to the price inflation in the period.

On the other hand, the volume of transactions also grew at an unprecedented rate. The pace of commercialization sped up as the country was increasingly integrated with the world economy from the latter part of the

[49] Peng Zeyi, Shijiu shiji houban de Zhongguo caizheng yu jingji (Beijing, 1983), p. 136.

[50] Rawski, Economic Growth , p. 163.

[51] The ratios between the three components of money are quite close to Peng Xinwei's estimate for the last decades of the Qing period; see Peng Xinwei, Zhongguo huobishi , pp. 888–89. For the ratio of silver in circulation to its total stock, see Wang, Zhongguo jindai huobi , p. 35n.57.

nineteenth century on. Many local products, such as soybeans, straw mats, tung oil, and bristles, found their way into the world market. The introduction of steamships and railroads lowered the cost of transportation and also contributed greatly to market expansion. Consequently, China's trade, external as well as internal, experienced a rapid expansion. The Nankai indexes show that both imports and exports tripled in volume between 1881 and 1930.^[52] The interprovincial trade in real terms was, according to Perkins's estimate, more than three times in the 1920s what it was in the late nineteenth century.^[53] In view of these facts, a rate of growth in the volume of transactions of 2 percent a year is probably a good approximation. Then, the transaction velocity of money must, by implication, have risen annually by around 0.5–1.0 percent over the period.^[54]

There is good reason to believe that the velocity of transactions rose in the latest phase of secular price movements. First, population increased almost as fast as in the eighteenth century. Since the proportion of self-consumed farm output still approached nearly a half of farm output in prewar China,^[55] Goldstone's thesis is applicable, perhaps more than ever before. Second, the development of telegraph communication and modern banking, as well as the continued growth of native banks, greatly facilitated business transactions and thereby had the effect of allowing merchants to conduct their business with reduced holdings of cash.

To recapitulate, the secular movements of rice prices in the Yangzi Delta exhibit two long swings over the three centuries prior to World War II. Starting in the 1640s, prices followed a steep downward trend until the 1680s; afterward they moved up gradually, at 0.7 percent a year, to peak around 1820. The second swing went downward at first, for more than half a century. At the beginning of the 1880s it once again shifted direction, rising swiftly, at the rate of more than 3 percent a year, through 1930.

Given the fact that the delta occupied a position of economic centrality in the country, it is reasonable to regard the trends exhibited by the price series as reflecting supply and demand forces at work in the country. Population, the stock of money, and climate cycles are the three principal variables affecting the long-term trends. I found, among other things, that inflation was

[52] Hsiao, China's Foreign Trade Statistics , pp. 274–75.

[53] Perkins, Agricultural Development , pp. 119–24.

[54] According to Thomas G. Rawski, the income velocity declined by 32–46 percent between 1914–18 and 1934. I consider the transaction velocity a better measure of monetary circulation for the reason that nearly one-half of farm output was still self-consumed in prewar China. But as Michael D. Bordo and Lars Jonung point out in their study of secular trends in the velocity of circulation in various countries, a fall in income velocity may coincide with a rise in transaction velocity. See Rawski, Economic Growth , pp. 161–65; Bordo and Jonung, The Long-term Behavior of the Velocity of Circulation (Cambridge, 1987), chaps. 2–3.

[55] Buck, Chinese Farm Economy , p. 199.

nearly always accompanied by substantial expansion of the money supply, deflation by stagnation in monetary growth. The same was true for population. Rapid population growth appears to have contributed to inflation most likely because of changes in the age structure and because of the population-induced rise in the velocity of money. On the other hand, a decrease in population or a slowing down in its growth rate correlates strongly with deflation. Gradual changes in the climate probably helped moderate price inflation in the eighteenth century, pushing prices down in the nineteenth century and up in between; but the impact of falling temperatures on market transactions and hence on prices was neutralized or overbalanced by other factors in the latter part of the seventeenth century.

Two
Grain Prices in Zhili Province, 1736–1911: a Preliminary Study

Lillian M. Li

The recent availability of grain price data for the Qing period now provides historians with an unprecedented opportunity to develop their understanding of the agricultural economy of every region of China. This study represents a preliminary attempt to apply the Qing period grain price data to an ongoing study of agriculture, food crises, natural disasters, and government relief in North China.^[1] This rich series of data allows us to gain insight into the nature of the agricultural regime in the North, the long-term trends in the agricultural economy, the nature of short-term changes, the economic and social impact of natural disasters, and the extent of market integration and development. Because this is a first attempt, the methodology employed is exploratory, and the conclusions drawn should be regarded as tentative. In the future, completion of the data set and refinement of the methodology may

[1] This essay is part of a projected book on this region, Flood and Famine in China: State Policy and Ecological Disaster in the Hai River Basin, 1690s–1990s . The present study should be regarded as preliminary in part because the set of grain price data that I have collected thus far, though extensive, is still incomplete.

alter the conclusions, but in the interim these tentative results will, hopefully, generate some hypotheses for further analysis and research.

North China is the birthplace of China's civilization and the location of its imperial capitals throughout most of its history. Yet in recent centuries North China has also suffered unfavorable natural conditions and economic hardships. Unlike the fertile rice-growing regions of South China, the North practices mixed cultivation of dry-land grains under climatic conditions that have made the fate of each year's crop extremely uncertain.

Zhili Province (roughly equivalent to modern-day Hebei Province), the subject of this article, best embodies these contradictory characteristics. As the site of the imperial capital of Beijing since the thirteenth century, Zhili had economic advantages that derived more from its political centrality than from its natural resources. The emperors of the Qing dynasty (1644–1911), like their predecessors of the Yuan and Ming dynasties, sought to bring some of the material benefits of the South to the North. The Grand Canal was maintained specifically for the purpose of transporting rice and other products from the South to the capital. The grain tribute was intended to feed the court and officialdom, but it had the unintended consequence of linking the economies of North and South. It was the only significant North-South thoroughfare until railroads were introduced in the twentieth century. The strategic importance of Beijing also dictated that the court pay special attention to stocking both civilian and military granaries in the capital area as well as to the maintenance of the waterways in the province.

The care and protection of the Qing imperial court and its retainers took place, however, in the context of natural conditions that seem to have deteriorated through the centuries. The river system in particular has been a source of ongoing headaches. In ancient times many rivers flowed down from the Taihang mountain range, which forms a boundary between Zhili and Shanxi, its neighbor to the west. But the construction of the Yuan dynasty Grand Canal severed the normal channels of these waterways to the ocean, forcing them to flow into the canal itself. From that time on, the only outlet of five major waterways—the Bei, Yongding, Daqing, Ziya, and Nanyun rivers—was a single, short channel flowing from Tianjin to the ocean, known as the Hai River. This entire river system is known today as the Hai River Basin and is one of China's major river conservancy concerns.

Although smaller in scale, the Hai River Basin has many of the same characteristics as the larger Yellow River Basin, with which it became more closely linked after the Yellow River changed course in the 1850s to the north side of Shandong peninsula. Both have extremely shallow beds, which over the centuries have risen with increased siltation. In recent times deforestation in the mountains has accelerated the silting process. The control of these rivers requires intensive dredging, diking, and other engineering efforts. Over time, the vulnerability of these rivers to flooding has greatly increased, as has

Map 2.1.
Zhili Province in the Qing Period. Adapted from the end-map in Qingdai Haihe Luanhe honglao dang'an shiliao
(Beijing, 1981).

their tendency to waterlog the soil, particularly given the flat, even concave, topography of the land in their lower reaches.

The uneven pattern of rainfall in North China further contributes to the danger of flooding. Most of the annual rainfall in the region is concentrated in the summer months of July, August, and September. It does not take very much rain to cause the rivers to overflow. Ironically the danger of flooding occurs in the context of the overall scarcity of rainfall in the North. As a rule, drought is a more pervasive and potentially serious problem than flood, but it is the river system, the nature of the topography, and the soil that make flooding such a frequent danger and waterlogging a near-chronic condition in some low-lying areas.

Throughout the Qing dynasty, the court and bureaucracy paid close attention to the dual threats of flood and drought. In part this reflected a general bureaucratic concern with disaster prevention that was empirewide, but Zhili Province clearly posed special problems because of its political importance. The bureaucratic record reflects the tremendous concern of the court and officials with river conservancy and with the stocking of granaries to guard against shortages. In the eighteenth century these efforts appear to have been quite successful in both the prevention and relief of famines, but vigilance was constantly required. During the nineteenth century, however, especially from mid-century, either the efforts were less successful or nature was harder to control. Starting with the great drought of the 1870s, North China was periodically beset by one disaster after another until the 1960s. In the 1890s there was hardly a year in which flooding did not occur somewhere in this region. In 1917 there was a massive, provincewide flood, followed closely by the North China drought of 1920–21. Since 1949 the government of the People's Republic has assigned high priority to the management of land and water resources in the area. The sinking of tube wells for irrigation in many parts of the province has proceeded together with engineering projects to prevent the recurrence of major floods, such as that of 1963.

Despite this somewhat unstable context, North China in general, and Zhili/Hebei in particular, have been able to sustain a large population increase in the last two centuries. In 1749 the population of Zhili was reported to be about 14 million.^[2] In 1790 the population was recorded as 23.5 million.

[2] This 1749 figure almost certainly represents an underestimate. Before the baojia system of population registration was reformed in 1775, underestimation was common. See Ping-ti Ho, Studies on the Population of China, 1368–1953 (Cambridge, Mass., 1959), pp. 36–48. Ho concludes that the population figures of the 1741–1775 period were on the average underestimated by 20 percent. By that formula, Zhili's 1749 population may have been close to 16.8 million. Zhili did not submit its first detailed population return under the reformed system until 1778. If Zhili's population was 16.8 million in 1749 and 23.5 million in 1790, it experienced a 40 percent increase over 41 years.

The rate of growth slackened in the next half-century; in 1850 the reported population was only 23.4 million. In 1933 Hebei's population was about 38.4 million, and in 1953, 46.6 million. In 1982 the population of the province, together with that of the independent municipalities of Beijing and Tianjin, was about 70 million.^[3]

The history of the Zhili/Hebei area of North China poses questions of enormous consequence. How could this area maintain its political centrality for so many centuries despite a relatively weak economic base? How could substantial population growth be sustained in the face of what appear in the historical record to be frequent and regular occurrences of drought and flood?

The Price Data

During the Qing dynasty each provincial governor was required to submit to the throne a monthly report of grain prices in his province. This became a regular bureaucratic practice by the beginning of the Qianlong period in 1736. The Qing archives in Beijing and Taibei have a rather complete set of reports from Zhili for the eighteenth century and a more scattered sampling from the nineteenth century. I have collected approximately 609 of these monthly lists, including 233 monthly lists for 1738–64, 171 lists for 1765–95, and 205 lists for 1796–1910.^[4]

In the first subperiod, 1738–64, the lists give the low and high prices of seven types of grain from each prefecture (fu ) or independent department (zhilizhou ) in the province: rice (daomi ), high-grade millet (shang sumi ), ordinary millet (cisumi or zhongsumi ), white wheat (baimai ), red wheat (hongmai ), black beans (heidou ), and sorghum, or kaoliang (gaoliang ). After 1765 only five grains were reported: millet (sumi ), sorghum (gaoliang ), a type of panicum millet (nimi ), wheat (mai ), and black beans (heidou ).^[5]

Prices from seventeen prefectures or independent departments were reported by the governor-general of Zhili, although not all were reported in every period. Shuntian Prefecture (where Beijing was located) was not included until 1771. Chengde Prefecture was not included in the reports until

[3] The Qing figures are taken from Philip C. C. Huang, The Peasant Economy and Social Change in North China (Stanford, 1985), p. 322. The 1982 figure is reported in Judith Banister, China's Changing Population (Stanford, 1987), pp. 298–99, among other places.

[4] I am indebted to the staff of the Ming-Qing archives of the National Palace Museum in Taibei and the First Historical Archives in Beijing for allowing me access to these grain price lists.

[5] Sumi was Setaria italica , sometimes called foxtail millet, which was the most common type of millet grown in north China. Nimi was Panicum milaceum , sometimes called broomcorn millet. Heidou , lit. "black bean," was a type of soybean. See Francesca Bray, Agriculture , vol. 6, pt. 2 of Joseph Needham et al., Science and Civilization in China (Cambridge, Eng., 1984), pp. 434–48.

1778. From 1736–68, the prices for Baoding Prefecture, the location of the provincial capital, were reported a month in advance of the other provinces.

These grain price reports were submitted monthly, according to the Chinese lunar calendar—with intercalary months ("leap months") added from time to time to make the lunar year catch up to the solar year. Since any given lunar month might lag behind its corresponding solar month by up to two months, solar months might be more appropriate to use in studying the agricultural cycle. In this study lunar-month prices have been used where aggregated data for year or multiyear periods would cancel out the variations in the months. However, where seasonality is an important concern, data are converted to correspond with the solar months.

Agriculture and Food Availability in Zhili

As the above lists suggest, many grain crops were grown in Zhili. Although rice was the subject of experimentation in the early eighteenth century, it was never very widely grown.^[6] Wheat was the luxury grain in the North. Planted in the fall, it was harvested the following summer. Millets of various types were the staple of the poor people's diet. Like sorghum, millet was planted in late spring and harvested in the fall. It was hardy, having a tolerance for heat and drought. Sorghum, on the other hand, was more flood-resistant. Less desirable than millet as a food, sorghum was also used in making wine, and its stalks were burned for fuel.

Other crops were important too. Black beans, reported in the Qing grain price reports, were used both as a feedgrain for horses and as food for humans. They also became a cash crop, used in the production of oil. In the twentieth century corn became a major food crop, but it is not at all clear how extensively it was grown in the Qing period. Finally, cotton was the most important nonfood commercial crop in Zhili in Qing and later times, but the extent of its cultivation before the twentieth century is a matter of some uncertainty.^[7]

There were numerous cropping systems in North China, with great variation within regions. One system was a three crop rotation over two years. As Philip C. C. Huang describes it, sorghum and millet were planted in May or June and harvested in September or October. Wheat was planted in the fall and harvested in July, too late for the planting of sorghum or millet, so soy-

[6] See Timothy Brook, "The Spread of Rice Cultivation and Rice Technology into the Hebei Region in the Ming and Qing," in Explorations in the History of Science and Technology in China (Shanghai, 1982), pp. 659–89, for an exhaustive study of experimentation in rice cultivation in North China.

[7] Philip C. C. Huang, pp. 111–14, asserts that cotton cultivation was widespread in Zhili by the late Ming period, but others have disputed this. See, for example, Loren Brandt's review of Huang, Peasant Economy , in Economic Development and Cultural Change 35 (April 1987):670–82.

beans were planted for harvesting in October and November, after which the land would be left fallow. Other systems involved interplanting.^[8]

It is not until the twentieth century that we have some idea of the acreage devoted to each of these major crops. John Lossing Buck's well-known 1929–33 farm study reports that in the winter wheat—kaoliang region, which included Hebei, wheat accounted for 45.5 percent of crop area, millet for 23.1 percent, sorghum for 18.5, corn for 16.3, soybeans for 13.4, and cotton for 8.6.^[9] A survey conducted by Zhang Xinyi reports the following crop areas for Hebei in the 1930s: wheat, 31.3 million mu (28 percent), millet, 24.3 million mu (22 percent), sorghum, 21.7 million mu (20 percent), and corn, 15.5 million mu (14 percent). Beans (dadou ) accounted for 9.8 million mu (9 percent), and cotton for 8.1 million mu (7 percent).^[10]

These two estimates are similar in that they show the primary importance of wheat, millet, and sorghum, in that order. There is, of course, every reason to believe that the situation in the Qing period differed in significant ways. Corn almost certainly played a lesser role, and perhaps the proportions of wheat, millet, and sorghum were different from the twentieth century. Unless further research uncovers new sources, it is unlikely that we shall ever have an exact picture of the production of these crops during the Qing period, but this general picture of the relative importance of these crops is unlikely to be changed.

This study analyzes the prices of three of the grains reported in the Qing memorials—wheat, millet (sumi or setaria millet), and sorghum—because of their centrality in the agriculture and the diet of the North. Black beans probably did not constitute a significant portion of the caloric content of the average diet. Panicum millet and rice were unlikely to have been of critical importance either.

Two additional factors probably influenced the price structure, although they were exogenous to Zhili's agricultural production. First, a significant portion of the grain consumed in Zhili during the Qing was not grown in the province but was imported from the South through the grain tribute system. During the Qing, 3–4 million shi of grain were transported annually to the metropolitan area.^[11] Most of this was destined for the consumption of the court, bannermen, and soldiers stationed in the province. But it is quite likely

[8] Philip C. C. Huang, Peasant Economy , p. 61.

[9] John Lossing Buck, Land Utilization in China (Chicago and Nanking, 1937; repr. New York, 1956), 1:211–12. Because there was some double cropping, the percentages exceed 100.

[10] Cited in Shina nogyo kiso tokei shiryo , comp. comp. Toa kenkyujo (Shanghai, 1941), 1:41–43.

[11] The shi was a measure of grain volume. According to an authoritative estimate, "the likely weight of an imperial shih [shi ] of milled rice in the eighteenth century was about 185 pounds, with a margin of error unlikely to have been more than 5 percent either way (that is, the likely range was roughly 175 to 195 pounds)." Han-sheng Chuan and Richard A. Kraus, Mid-Ch'ing Rice Markets and Trade: An Essay in Price History (Cambridge, Mass., 1975), p. 98.

that some quantity of grain reached the market, through either direct or indirect sales. Pierre-Etienne Will has estimated that 0.5 million shi a year was not used for direct consumption in Beijing.^[12] Whether this amount was sufficient to have an impact on general grain price levels is not clear, and is a matter that deserves further investigation.

The second factor was an extensive state granary system, which flourished in the eighteenth century. The system included three types of granaries: the "ever-normal" granaries, the community granaries, and the charity granaries. During the eighteenth century these granaries were well stocked. In 1767, for example, the governor-general of the province reported that 3,534,536 shi of grain were actually stored in the province, 2,549,566 shi of which were in the ever-normal granaries.^[13] By the nineteenth century, however, all granary holdings were down, especially those of the ever-normal granaries. In 1833, for example, the governor-general reported holdings of only about 616,000 shi , of which 275,719 shi were held in ever-normal granaries.^[14]

Long-Term Trends

The Zhili price series affords an important opportunity to learn about the long-term behavior of prices over two centuries. This is important not only because of its relevance to a study of Zhili's economy in particular but also because it can serve as a general indicator of overall economic trends in North China. The price trend of these three major grains has several major characteristics. As Figure 2.1 shows, the overall price rise from 1738 to 1910 was not steep. During the eighteenth century prices rose very gradually. In the early part of the nineteenth century, there was a sharp increase in the price of wheat, followed by rises in the prices of millet and sorghum. From the 1830s to 1850, roughly during the Daoguang reign, prices fell precipitously, only to climb back up by 1870 and fall again. From 1890, prices rose steeply and steadily until the end of the dynasty. When these same prices, again grouped by four-year averages, are indexed to their base-period (1738–41) prices, the trends can be seen more clearly, as in Figure 2.2. At the end of the eighteenth century, the prices of wheat, millet, and sorghum were respectively only 134 percent, 122 percent, and 133 percent of the base-period prices. By the end of the dynasty, the three grains had risen to 258 percent, 243 percent, and 272

[12] Pierre-Etienne Will, Bureaucratie et famine en Chine au 18e siècle (Paris, 1980), pp. 241–44. Will points out that tribute grain surpluses were rarely used outside Zhili.

[13] Gongzhongdang, Palace Memorial Archives (Taibei), Qianlong 023616, 1767/12/12. These figures represent the actual holdings at the time of the report; the theoretical holdings, which took into account amounts loaned out but not yet paid back to the granaries, were larger.

[14] Junjidang, Grand Council Archives (Beijing), Daoguang 63339, 1833/4. These figures represent actual, not theoretical, holdings.

Fig. 2.1.
Grain Prices in Zhili Province, 1738–1910 (four-year averages, in tales per shi )
Note: See text for a description of the data used in this study. In Figs. 2.1, 2.2, and 2.4, there is a substantial amount of
missing data for the nineteenth century.

Fig. 2.2
Grain Prices in Zhili Province, 1738–1910: Indexed to the Base Period (four year averages; 1738–1741 = 100)

Fig. 2.3.
Annual Grain Prices in Zhili Province, 1738–1806 (taels per shi )

percent of the base-period prices respectively, but if the final two decades of the dynasty are omitted, the increases are only 146 percent, 155 percent, and 165 percent respectively of the base-period price.

To show more clearly what the annual trends looked like, in Figure 2.3 I present the annual averages (unadjusted) for 1738–1806.^[15] Here once again it is apparent that the trend for the eighteenth century was not very steep and that there seem to be cycles of about four to five years for millet and sorghum and possibly longer cycles in the price of wheat. Further, it appears that the prices of all three grains followed roughly the same trend, but there was at least one period—in the 1740s—in which the rise in the price of wheat was not accompanied by a rise in the prices of the coarse grains. A contrary trend is seen in the early 1760s, when a rise in the prices of the coarse grains was not accompanied by a continuous rise in the price of wheat.

[15] Only the eighteenth century is presented because that is the century for which the data are most complete. In this essay I have used the mean of the high and low prices reported from each prefecture each month. In analyzing the high and low prices separately, I found that their behavior generally followed the same pattern. This was true of all three grains. There seemed, then, little point in studying either high or low prices separately, particularly in a preliminary study.

The adjustment of annual averages to account for the seasonal variations in the missing data presents a greater challenge. Although it would be preferable to adjust the annual averages, in fact each possible procedure for adjustment produces its own problems. One way is to estimate, or "predict," missing data by using all the known data in combination with the seasonal coefficients produced by the regression equation discussed in the next section. Another method is to adjust each known piece of data by its relevant monthly coefficient. This creates not an annual average but an estimated January price based on all known data. Until a more accurate method of adjusting data can be found, however, leaving the data unadjusted does not, I believe, produce large distortions, because the annual seasonal range of prices in Zhili was not more than 0.14 tael in the extreme case of wheat (from a high of +0.08 taels in April to a low of -0.06 taels in September in the regression on data excluding Baoding, as presented in Table 2.1). For example, in 1740, a year for which there are no missing data, the mean price was 1.38 taels. If data were missing for the three low-price months of July, August, and September, the unadjusted mean of the other nine months of data would be 1.40, a distortion of only 0.02 tael, or 1.4 percent. If I use the regression coefficients to estimate the missing data, the annual mean would come to 1.39. The difference between adjusting or not adjusting the data is only 0.01 tael, or less than 1 percent of the actual mean price of 1.38. This example presents one of the worst possible cases since the mean price was rather low in comparison to the rest of the period 1738–1910. In a year when the prices were relatively high—over 2.50 taels, for example—the use of a seasonal coefficient would tend to underestimate the missing data (e.g., 0.06 tael would make less difference). The only time that a larger distortion might be introduced by not adjusting the data would be in a crisis year for which we have only one month of data in the early part of the year not yet affected by the crisis. In such a case, the price level would be very much underestimated. If, instead of "predicting" the price where data are missing, we use the seasonal coefficients to adjust the known data, the result in this example would be 1.37, only 0.01 less than the actual mean price. The fact that this hypothetical January price is close to the mean is somewhat accidental; the actual January price for that year was 1.42, which reflects the fact that the previous year was a crisis year (see n. 21 below) and that the price of wheat was abnormally high in the winter months.

Fig. 2.4.
Prices of Coarse Grains Relative to Wheat in Zhili Province, 1738–1910 (5-year averages; price = % of wheat price in
concurrent period)

On the whole these graphs show that the prices of millet and sorghum followed each other very closely, with sorghum consistently being the cheaper of the two. Although millet was a less desirable grain than wheat, the figures show that there were years when its price approached that of wheat (and even exceeded it in the early 1760s, early 1780s, and early 1800s). Figure 2.4 charts the five-year averages of millet, sorghum, and black bean prices as a percentage of wheat price averages for the concurrent periods. This shows more clearly the few periods in which millet prices exceeded those of wheat (around 1760, 1800, 1870, and 1890). And it also shows that while millet, sorghum, and black bean prices maintained a spread in the eighteenth century, during the next century the price of black beans relative to wheat and millet steadily increased.

These trends suggest certain hypotheses about the roles played by these commodities in Zhili's agricultural markets. First, the generally higher price of wheat confirms that it was the luxury grain and suggests the possibility that its markets were relatively well developed. Second, the price trends of millet and sorghum seem to parallel each other, suggesting that they were responding to the same cropping cycle and weather patterns. On the other hand, millet's consistently higher price confirms our impressions from the twentieth century that millet was the staple grain and was valued more highly than sorghum. However, the fact that millet prices occasionally reached, and even surpassed, those of wheat raises the question of the extent to which it was a commercialized agricultural commodity and the extent to which it was considered a substitute for wheat. Finally, the price of black beans rose relative to wheat prices in a clear secular trend, probably because of a growing nationwide, perhaps even international, market for bean oil and other bean products produced primarily in Manchuria but also in North China.^[16]

Regression Analysis

At any given month, the price of grain may have reflected not only the long-term price trend but also a place in the seasonal cycle, the impact of any irregularities in the weather that would affect output, as well as any exogenous factors affecting demand. To permit us to estimate, and separate out, the effects of time, seasonal variation, and crisis years on grain prices, a multiple regression of prices was run using the following equation:

[16] The extent to which Zhili/Hebei was involved in this industry is a subject that awaits further research. Certainly in the twentieth century one of Manchuria's principal industries and export commodities was soybean oil and other soybean products. See, for example, Lien-en Tsao, "The Marketing of Soya Beans and Bean Oil," Chinese Economic Journal 7, no. 3 (Sept. 1930): 941–71.

Here the price in any given month (P ) is measured by a constant (a ), plus a coefficient (b ) multiplied by time (T ), plus dummy variables (d ) that are created for every month (M ) except January and another dummy variable created for crisis years (C ). The dummy variable C is entered as a l in a crisis year and as a 0 in noncrisis years.^[17]

Since one of the purposes of this regression is to measure the effect of seasonality, all data used in the analysis were converted to solar prices using a method that weights the price by the number of appropriate lunar days in each solar month.^[18] Since prices for Baoding, the seat of the provincial capital, were routinely reported a month in advance of prices from other prefectures during 1736–68, a period that represents 44 percent of my data (268 out of 609 monthly reports), I also ran the regression with Baoding prices excluded lest the months for which there are only Baoding prices unduly affect the results.^[19]

The most problematic decision in setting up this regression concerned the crisis years. Although it would be ideal to use meteorological data (temperature and rainfall) to code years of crisis, I do not yet have access to such data. In their absence, I am forced to rely on reports of crises found in the historical records. Since such data represent the human (social and political) impact of natural disasters as seen through administrative lenses, they must necessarily have their limitations. The historical record itself is necessarily subjective, and our access to it is also incomplete, a function of what documents have survived in which collections and archives. In this regression analysis I selected as crisis years those years recorded in Qingdai Haihe Luanhe honglao dang'an shiliao^[20] as having floods that affected 50 counties (xian ) or more. Although this compilation itself may be flawed, it is drawn from a survey of local gazetteers. In addition I selected from miscellaneous records at my

[17] This regression equation was also run with T as a variable to see if there were perhaps a nonlinear relationship between time and price. The results were not statistically significant (t < 2), and overall did not produce a better R , and so we concluded that there is not a second-order relationship between time and price.

[18] I am very grateful to Peter C. Perdue for providing the complex formula and table by which the lunar data have been converted to solar data and to Keith Head, who wrote the program that adapted this table to my data.

[19] For example, the price list submitted by the Zhili governor-general reporting prices for the third lunar month included the fourth-month prices for Baoding Prefecture, presumably because these local prices were already available to him by the time he received the third-month prices from the outlying prefectures. Consequently, if I have the third-month provincial report but am missing the fourth-month provincial report, the Baoding price is the only one I have for the fourth month and the provincial average in fact represents only Baoding. Because there is reason to think that prices behaved differently in the capital, I have thought it wise to run the regression both with and without Baoding prices.

[20] Qingdai Haihe Luanhe honglao dang'an shiliao (Beijing, 1981).

disposal years in which 25 or more counties were affected by drought. This combined method yielded 54 years that were coded as crisis years.^[21]

When the regression was run for 1738–1910, with Baoding prices excluded, the results (see Table 2.1) showed that the effect of time was 0.0059, 0.0061, and 0.0045 taels for wheat, millet, and sorghum respectively, meaning that for each year, the price increased by this amount for each grain. These increases confirm the impression, derived from Figure 2.1, that the inflationary trend was not very steep or significant over this long time period.

The regression results show the monthly variation of prices, using January as the base month. The results, graphed in Figure 2.5, show the effect of the multiple-crop system of Zhili. Wheat prices reached their peak in April but did not reach their lowest point until September—somewhat surprisingly, since the wheat harvest took place around July. Millet and sorghum prices, however, peaked in June, July, and August, from which point they fell until

[21] 1738, 1739, 1743, 1744, 1747, 1750, 1759, 1761, 1762, 1771, 1780, 1790, 1794, 1801, 1806, 1808, 1810, 1813, 1814, 1816, 1819, 1820, 1822, 1823, 1830, 1832, 1834, 1835, 1839, 1840, 1855, 1871–73, 1876, 1877, 1879, 1882, 1883, 1886–90, 1892–1900, and 1908. Of course, I lack price data for many of these years.

Fig. 2.5.
Seasonal Variation of Grain Prices in Zhili Province (Excluding Baoding), 1738–1910 (difference from January
price, in taels per shi )

the end of the year, with a brief jump in October. The prices of all three grains rose steadily through the early months of the year. But if Baoding is included, as in Figure 2.6, there is a smoother descent for wheat in the fall, but in October, wheat, millet, and sorghum prices peak briefly. If Baoding is included and only eighteenth-century data are analyzed, the shape of the seasonal curves is smoother, with the annual low price for wheat and the high prices for millet and sorghum all occurring in August. Preliminary analysis of price data separately for each prefecture shows that the exact patterns of seasonality vary slightly from place to place but fall within a general pattern.

The regression coefficients also vary when either eighteenth-century data or nineteenth-century data are used alone. Although the greater quantity and likely higher quality of the eighteenth-century data would suggest that its regression results would be more significant, in fact the R² —a measure of the amount of variation in price explained by the regression as a whole—is considerably higher when both centuries of data are used together.^[22] When a Chow test—a measure of the extent to which two samples may affect the regression results—was run on these regressions, no significant difference was found between the two centuries of data.^[23] Consequently, the results of the regression run on both centuries of data are used throughout this paper.

These sets of monthly coefficients present several puzzles. Why does the annual low price for wheat come so much later than its presumed harvest-time? Why do the prices of sorghum and millet jump around in the autumn and early winter? Why does the seasonal pattern seem to vary depending on the time period and the number of prefectures included?

One possible explanation may be that seasonal patterns do differ from one area to another. In Buck's survey, the price of wheat in the wheat-kaoliang area was highest in January–February, and lowest in May–June. Millet had its high price in May–June and its low price in November.^[24] But another 1930s study of prices in Zhengding, Hebei, found that the annual high price for wheat was in April, and the low price followed immediately in May, while the high price of millet was in June–July, and the low immediately afterward, in August–September.^[25]

[22] Running the regression on either eighteenth- or nineteenth-century data alone produced much lower R s than using two centuries of data together. For wheat, for example, the R is 0.4859 for both centuries together, 0.2029 for the eighteenth century, and 0.1662 for the nineteenth century. These differences are undoubtedly due to the strength of the time trend, which is by far the most significant variable in any of these regressions.

[23] The Chow test was developed by Gregory C. Chow, "Tests of Equality between Sets of Coefficients in Two Linear Regressions," Econometrica 28, no. 3 (July 1960):591–605. I am grateful to F. Michael Scherer, formerly of the Economics Department at Swarthmore College, for his help with this test. In F-ratio was insighnificant for both wheat and millet (0.83 and 0.63, respectively) and marginally significant for sorghum (1.80).

[24] Buck, Land Utilization in China , 1:335–36.

[25] "The Seasonal Variation of Prices for Farm Products and the Profitability of Storage," Economic Facts , no. 7 (October 1937):319–42.

Fig. 2.6.
Seasonal Variation of Grain Prices in Zhili Province (Including Baoding), 1738–1910 (difference from January price,
in taels per shi )

The irregularities in the autumn prices could also be explained by deliveries of grain tribute to the province, which occurred before winter, or alternatively by the restocking of granaries, which was also done in the autumn. Since it was mostly unhusked millet that was stored (because wheat spoils too quickly), post-harvest prices may have risen a bit before falling to their annual low at the end of the year.^[26]

Although the precise effects of seasonality seem to differ according to the way the regression is run, the overall pattern is clear. Moreover, the regression analysis reveals that the effect of seasonality was less important than the effect of crises.^[27] The spread of prices within each year was less than 0.14 tael for wheat and less than 0.11 tael for millet and sorghum, no matter which way the regression is run.

Although regression analysis is a sophisticated tool for measuring the separate effects of different variables on price, it has its limitations. For the purposes of this grain price analysis, it can give us a good measure of the factors in price variability, but it is sensitive to differences in time and space and therefore cannot be absolutely accurate. It must be used as an approximate tool, not a precise measure. In future work, more accurate specification of crisis years, more complete data, and work with disaggregated, prefectural data may produce more satisfactory results. Still, these preliminary results do inspire some confidence. Although the t -value for each of the monthly coefficients tends to be under 2, in using descriptive data, as opposed to sampling, t -values are not relevant. Moreover, the R² s achieved here are not poor, and the F -values are so large that the overall strength of the regression variables can be seen to be very substantial.^[28]

Crises

One purpose of the regression analysis is to try to measure the effect of crises on price levels. The regression analysis for 1738–1910 suggests that the effect of a crisis year was, on the whole, not very dramatic. The inclusion of Baoding data raises the effect slightly (see Table 2.1). A separate regression run solely on Baoding data results in higher crisis coefficients. Moreover, if the

[26] Pierre-Etienne Will and R. Bin Wong, Nourish the People: The State Civilian Granary System in China, 1650–1850 (Ann Arbor, 1991), manuscript pp. 5, 132. Thirty percent of the granary stock was supposed to be sold off each year and replaced.

[27] This impression was confirmed when the regression was run with only the monthly dummies as variables. The resulting R s were astonishingly low: 0.0099, 0.0129, and 0.0189 for the three grains, respectively. Running the regression with interaction variables (C * M , C *T , etc.) also produced very low R s showing that a nonlinear relationship was not a better explanation for the behavior of prices.

[28] Concerning these and other statistical problems, I am grateful to Gudmund Iversen, professor of statistics at Swarthmore College, for his judgment and invaluable suggestions.

monthly price data are divided into a "crop year," from July to the end of June, the effect of crises is seen to be far higher for the coarse grains but slightly lower for wheat. Table 2.2 summarizes the coefficients generated in these various ways. In all cases the R² s are very similar, as are the coefficients for time. The monthly coefficients vary, but are within two standard errors of zero.

Dividing the year at July generates a higher coefficient for the crisis variable, because a crop year more closely approximates the seasonal weather pattern in the North. Since July and August are the months of greatest annual rainfall, droughts and floods have their first impact in the second half of the year and the next winter. They affect the coarse grains first. Although sorghum is known to be flood-resistant, its price seems to be the most differentially affected by the use of this technique. Overall, however, the price of millet was the most affected in a year of crisis.

The fact that Baoding prices were more affected by crises than prices in the rest of the province raises some interesting questions. As the seat of the provincial capital, Baoding Prefecture might be expected to have higher prices than outlying prefectures because of a stronger demand for grains. Its wheat prices were much more affected by crises than wheat prices in the rest of the province, suggesting a stronger demand for the luxury grain in the capital than elsewhere.

Tables 2.3, 2.4, and 2.5 attempt to evaluate the effect of natural disasters on grain prices in three crisis periods.^[29] The tables compare the actual prices of the three grains with their predicted prices (for a noncrisis year), calculated with the regression coefficients presented in Table 2.1, using data for 1738–1910 but excluding Baoding.

[29] The period 1743–44 was chosen because it is a well-studied drought, the periods 1762–63 and 1775–76 because the data were relatively complete. Although 1775–76 is identified in documentary sources as a drought period, these years are not coded as crisis years in the regression analysis (see n. 21 above) because fewer than 25 counties were apparently affected.

Time is the year minus 1738. Each case study covers the two calendar years that include the period of crisis.

In 1743–44 Zhili experienced a drought. As Table 2.3 shows, the actual prices of all three grains at the beginning of 1743 were below the predicted prices, particularly so for millet and sorghum. However, by May and June, when wheat prices should have fallen, they instead continued to rise, reflecting a poor harvest or an impending harvest. Although the data for the key months of the crisis, from summer 1743 to summer 1744, are missing, we can see that by July 1744 the prices of wheat and millet had risen 0.20 tael above the predicted price, and the price of sorghum, 0.15 tael. By the end of the calendar year, however, the price of wheat was almost down to the predicted level, and the prices of millet and sorghum had fallen to well below the predicted prices.

Although the actual prices at their peak reflected the impact of the drought more than the crisis coefficients of 0.06, 0.06, and 0.01 predicted they would, nevertheless it can certainly be concluded that the overall impact of the drought was rather limited in magnitude and duration. This substantiates to a considerable degree the picture of this crisis drawn by Pierre-Etienne Will.^[30] First, the impact of the drought, according to the historical record, was limited to 27 counties, primarily in four prefectures. Second, these counties were the recipients of massive amounts of government relief, deployed from several sources, most notably the state granaries in Tongzhou. The grain prices in these prefectures taken separately show no difference from the provincial trends. The famine relief campaign mounted by the government was truly a model effort, and the price history seems to show that the efforts of the Qing officials were well rewarded.

Table 2.4 presents the predicted and actual prices during and after a flood in 1775, which had been preceded by a drought in 1774, to show what the conditions were in a crisis where the government may have played a less active role. In this case, it appears that wheat prices were the most seriously affected. By September 1775 wheat prices were 0.20 tael above the predicted price, and prices stayed high until June, when an apparently successful harvest sent prices tumbling down to well under their predicted or normal levels. Millet prices were close to their normal levels in the spring and summer of 1775, while sorghum prices rose 0.10 tael or slightly more. But after the fall 1775 harvest, which does not seem to have been much affected by the flood, both millet and sorghum prices fell and stayed below their predicted levels in 1776. So the main impact of the crisis fell on wheat, and as in 1743–44, the impact was limited in magnitude and duration.

[30] See Will, Bureaucratie et Famine .

In Table 2.5 the data for the latter part of the 1761–63 flood crisis are presented. According to documentary evidence, this flood affected 53 counties in the Hai River Basin. By the beginning of 1762, prices were considerably higher than normal for all three grains. With the exception of a brief dip in June 1762, wheat prices kept climbing, reaching a level about 0.21 tael above normal in March–April 1763, after which prices began to come down, reaching their normal levels by the end of the year. Millet was the most severely affected of all the grains. Its price kept climbing until winter 1762–63, when it was more than 0.60 above normal. Its prices remained high during the year. Sorghum prices also reached a level of about 0.40–0.46 above their predicted prices in January–April 1763.

As both Table 2.5 and Figure 2.3 show, this flood marked one of the few times in the Qing period when the price of millet actually exceeded the price of wheat. In an extensive flood, property is damaged and recovery may take a longer time than after a drought. The data in this case certainly show a greater impact and a much slower recovery than do the data in the two cases involving drought. These results suggest a hypothesis for future investigation, namely, that floods in the Qing period had a more pronounced effect on prices than droughts did.

On the whole, however, these three cases show that the impact of crises in the eighteenth century was relatively moderate, especially in comparison with the staggering price increases of other well-documented world famines, or even late-nineteenth- or twentieth-century famines in China. According to Andrew B. Appleby, grain prices in French subsistence crises of the seventeenth and early eighteenth centuries rose to three or four times their normal levels.^[31] In Zhili, in the eighteenth century at least, prices were not generally affected more than 10–20 percent; prices rose just over 40 percent for millet in 1762–63, the worst case seen so far.

Appleby also argues that the English mixed farming system, with animal husbandry and multiple grains, worked to minimize the effects of shortages because people could choose to eat inferior grains, usually reserved for livestock, instead of wheat, the preferred grain—eating down the food chain, so to speak. In Zhili the grain prices maintained a separation from each other, except in the 1761–63 flood, when millet prices reached and then exceeded those of wheat. Separation of prices suggests either that the markets were indeed separate, and there was little substitutability in crisis times, or else that there really was not a crisis, because there was no need for substitution. In a real crisis people become unable, or unwilling, to pay the exorbitant price of an expensive grain and therefore substitute an inferior grain, which in turn drives up the price of the second grain. Consequently, the separation

[31] Andrew B. Appleby, "Grain Prices and Subsistence Crises in England and France, 1590–1740," Journal of Economic History 39, no. 4 (Dec. 1979):865–86.

of prices (their nonconvergence) in 1743–44 and 1775–76 suggests that there was no real crisis in those two instances, while the convergence of millet and wheat prices in 1762–63 suggests that there was a real crisis in that situation. Again, this idea, like others in this paper, is advanced as a hypothesis that must be tested, particularly with prefectural data that will focus on the particular parts of the province affected in a particular crisis.

Regional Variation

A central question in the study of these grain prices is the extent to which they varied within the province. In a well-developed market system, the correlation of prefectural prices ought to have been high, and price variation ought to have been low. Price variation within the province should also tell us something about the impact of crises. In a well-developed market system, the impact of natural disasters ought to have been cushioned, since grain would have been able to flow from unaffected regions to the affected ones. Of course, the same effects—strong price correlations and low price variation—might also have been achieved if the granary system was highly effective in its functions of price stabilization and famine relief or if the entire province had identical weather and other environmental conditions.

In approaching the question of market integration, we first employed the statistical measure called the coefficient of variation. The coefficient of variation is the standard deviation divided by the mean, multiplied by 100. It is a measure of the extent to which prices varied among prefectures during a given period. If market integration increased over time, then the coefficient of variation should decline. If, on the other hand, markets deteriorated, then the coefficient of variation should increase. We calculated the coefficients of variation of prefectural prices for each year from 1738 to 1910 for which we had data, omitting Xuanhua and Chengde, which were in the northern sections of the province. Then we did a regression analysis of the coefficients of variation with year (T ) and crisis year (C ) as variables. The resulting regression coefficients for year were 0.018, 0.041, and 0.013 for wheat, millet, and sorghum respectively (with t -values of 3.8, 7.8, and 2.6). In other words, the coefficients of variation for all three grains increased over time, and millet prices experienced significantly greater increases in regional variation than the other two grains.

These results are contrary to the expectation that over time markets should have become more integrated; if so, the coefficient of variation should have decreased. They also draw attention to, and invite explanation for, the different price behavior of millet, which was the staple grain for most people and which might perhaps have been more sensitive to population growth and to crises.

In the same regression analysis, however, the regression coefficients for

crisis years were all negative, with t -values under –2.0. That crisis years would cause variation to decline seems counterintuitive. When the various components of the coefficient of variation are separated out, however, a more plausible picture emerges. As Table 2.6 shows, the coefficient of variation declines for crisis years only because the standard deviation does not increase as much as the mean price. In other words, the lower coefficient of variation for crisis years is more a reflection of a higher mean price, totally to be expected, than diminished variation among prefectural prices. The table not only shows that the standard deviation did not increase very significantly in crisis years but also shows that the increase in the mean price was greater for millet—0.16 taels on the average—than it was for wheat or sorghum—0.13 and 0.10 taels respectively.^[32] On the whole, this analysis confirms the impressions derived from the earlier case studies in supporting the view that the impact of crises, both on the level of prices and on their variation across regions, was relatively limited.

As a second step, we used Pearson correlations to study regional integration. The correlation coefficients measure the relationship between grain

[32] The careful reader may note that the differences between the mean prices of crisis and noncrisis years differ somewhat from the crisis variables presented in Table 2.2. They are, however, roughly of the same order of magnitude as the last set, representing "crop year, Baoding only." Once again, it should be emphasized that regression analysis cannot produce precise results.

prices in pairs of prefectures, with 1.0 being a perfect correlation. The comparison of annual prices for the 17 prefectures of the province revealed astonishingly high degrees of correlation for most pairs of prefectures, with the unsurprising exceptions of Xuanhua and Chengde prefectures. Pearson correlations for wheat prices were mostly over 0.70, sometimes over 0.90, often above 0.80. Similar generalizations can be made for millet and sorghum.

When correlations are calculated separately for eighteenth- and nineteenth-century prices, no clear trend can be confirmed. While wheat correlations between most paris of prefectures seem to rise, for millet the picture is more mixed. Excluding Chengde and Xuanhua, correlations for the remaining 15 prefectures show that in 59 of 105 cases, correlations declined from the eighteenth to the nineteenth centuries, while in 46 cases they increased. This partly supports the picture presented by the study of coefficients of variation, and continues to make interesting the hypothesis that millet prices experienced greater spatial variation in the nineteenth century than the eighteenth century. Clearly, the next step in pursuing this question is to study the province on a region by region basis to see which regions shared this experience and which did not.

Since the correlation of prices themselves incorporates a similar time trend for all of the prefectures, it exaggerates the extent to which prices were actually correlated. Studying the first difference of these prices (the difference between the price in a given year and the price in the previous year) omits the common time element and provides a better measure of correlation. A preliminary analysis of annual price differences for the eighteenth century reveals that for wheat and millet about two thirds of the possible correlations between prefectures were 0.60 or over; for sorghum over 55 percent of the correlations were over 0.60. These results also suggest a reasonably high degree of correlation, but understanding their significance must await further work, and comparing them to nineteenth-century correlations must await a more complete set of nineteenth-century data.

Taken together, these preliminary attempts to study price variation in Zhili suggest a relatively high degree of integration within Zhili Province south of the Great Wall, but an integration that may have declined in some sections of the province, particularly in the case of millet, from the eighteenth to the nineteenth centuries. It should not be assumed, however, that integration was necessarily the result of a well-developed market system. The relative behavior of wheat and millet prices suggests that the market for wheat, a commercial product, may have been well developed and may have become more integrated in the nineteenth century. Millet, on the other hand, was the staple grain for ordinary people. In the eighteenth century, its apparent high degree of price correlation may have been due to the fact that the Zhili granaries predominantly stocked millet. With the decline of the granary system in the nineteenth century, price correlation may also have declined, because

the eighteenth-century integration of prices was probably more a function of the granary system than of the market system. Until some estimate can be made of the output and supply of each of these grains in the Qing period, and until the nineteenth-century data are more complete, these hypotheses must await testing.

Conclusions

While indicating directions for future work, this preliminary study of grain prices in Zhili Province during the Qing period also permits us to draw certain tentative conclusions. In the most general terms, this study suggests, first, that there was a relatively low inflationary trend for the Qianlong through the Xuantong reign periods (1736–1911), particularly so when the last two decades are omitted. Second, although there were distinct seasonal patterns in grain prices, they were not great in magnitude, and were offset by the multicrop system, particularly by the planting of winter wheat.

Third, crises were seen in several contexts to have been moderate in their impact, at least in the eighteenth century. Regression analysis showed the impact of crises to have been relatively contained. Crises caused the mean price of grain to rise somewhat but did not cause a very great increase in regional variation. The three eighteenth-century case studies confirm these general impressions. In fact, while these periods were deemed crises in administrative terms, and while they were triggered by natural crises, perhaps they were not really food crises if food crises are defined as periods of abnormally high food prices.^[33]

Fourth, the multicrop system appears to have been a significant factor in mitigating the effect of crises, particularly because of the different seasonal patterns of the crops. Multicropping certainly helped to offset the disadvantages presented by weather and geography, and has perhaps been overlooked in previous discussions. Appleby's study concludes that in England and France "the evidence suggests that a symmetrical price structure and subsistence crises went hand in hand. When all grains were costly at the same time, food shortage had an impact on both mortality and fertility; when one or another grain remained cheap, the demographic aftereffects were absent."^[34] It is too soon to conclude that such a generalization could be made for Zhili, but certainly this suggests a framework for future investigation. For North China, Buck observed in the 1920s and 1930s that farmers sold the higher-priced grain and ate inferior grain. Farmers in the wheat

[33] I have explored these issues more fully in "Using Grain Prices to Measure Food Crises: Chihli Province in the Mid-Ch'ing Period," The Second Conference on Modern Chinese Economic History (Taibei, 1989), II, pp. 467–509.

[34] Appleby, "Grain Prices," p. 882.

region, he said, sold half the wheat they grew, and purchased inferior grain, an estimated one quarter of their food needs, from the market.^[35]

The long-term behavior of millet prices seems to be the outcome of either deteriorating economic conditions (population pressure, etc.) or the diminished role of government intervention in the grain market or, more likely, both. The greater volatility of millet in some crises, its increasing coefficients of variation over time, and its lower price correlations in some regions in the nineteenth century all suggest certain long-term changes in its role in the food supply of Zhili. On the other hand, price analysis suggests that the role of sorghum did not experience a similar change. Francesca Bray has suggested that sorghum probably became more important in the nineteenth century with increasing population pressure.^[36] However, the long-term trends shown in Figure 2.4 do not support this view, since the price of sorghum continued to remain a constant percentage of the price of wheat and always remained separate from that of the higher-priced millet.

Finally, future work may confirm that prices throughout the region, with the exception of the two outlying prefectures, were remarkably well correlated, suggesting the powerful interaction of granaries and markets.

To understand better the nature of these secular changes, we need to have more complete price data for the nineteenth century and better weather data. This will permit us to understand more about long-term trends over both centuries. We also need to know more about the interaction between the grain tribute system, the granaries, and the grain markets in the determination of prices. Only then will it be possible to grasp why a region so indifferently and curiously endowed by nature could play such a large political role over centuries of history.

[35] Buck, Land Utilization in China , 1:416.

[36] Francesca Bray, Agriculture , vol. 6, pt. 2, of Joseph Needham, ed., Science and Civilization in China (Cambridge, 1984), pp. 434, 464, 451–52.

Three
The Qing State and the Gansu Grain Market, 1739–1864

Peter C. Perdue

Most Chinese celebrate the eighteenth-century Qing empire for two achievements: the expansion by conquest of China's territory to unprecedented size and the growth of its population to become the largest in the world. Two major institutions of the Qing state made these achievements possible: the military supply system and the granary and famine relief systems. Both of these institutions depended heavily on extensive private grain markets. New data from the price memorials in the Qing archives allow us to examine the degree of integration of these grain markets in Gansu Province in the eighteenth and nineteenth centuries. Scholars have recognized that market exchange of basic commodities was spreading widely in the Chinese empire in the eighteenth century, especially in the densely populated rice paddy belts of the Lower, Middle, and Upper Yangzi, the Pearl River Basin in South China, and along the Grand Canal leading to the North. Nevertheless, few have yet studied the substantial progress of trade in the far northwestern periphery of Han China. I shall argue that Gansu had achieved a considerable degree of integration of its grain markets by the eighteenth century and that the Qing state, through its military-provisioning and granary systems, indirectly promoted this process of commercialization.^[1]

The impact of the military was much greater in Gansu than in Coastal, Central, South, and Southwest China, because of Gansu's strategic location on the supply route to the garrisons occupying Central Asia. Continual military demands placed great stress on a fragile agrarian regime. But the efficient transport system developed by Qing governors of the province for military

[1] I would like to acknowledge the invaluable help of my research assistants, Jiang Xiaohong and Ren Jingzhen, in preparing the statistics for this paper. The M.I.T. Provost Fund and Metcalfe Fellowship supported the data collection and analysis.

supply and the injection of cash into the economy, combined with efforts to prevent military demands from excessively burdening the local population, stimulated market exchange. By linking Gansu to its neighbors, Shaanxi to the east and Sichuan to the south, the Qing could both maintain the local population and send support through the Gansu corridor to the large military establishment on the frontiers. A full analysis of the Northwest should include its core area in Shaanxi Province, but lacking the price data for Shaanxi at present, I shall only discuss Gansu itself.

Gansu's land area in the Qing was over 600,000 square kilometers, half again as large as California or Japan today. It was one of the largest provinces in the empire, smaller only than Yunnan and Sichuan among the 18 provinces of interior China. (During the Qing dynasty Gansu also included within its borders present-day Ningxia Autonomous Region and Xining Prefecture in present-day Qinghai.) Its population density, however, only 25 per square kilometer in 1957, makes it the most sparsely populated province of Han China.^[2]

Although the population was sparse, the cultivated acreage was also low, giving Gansu a ratio of cultivated land per capita close to the national average. In 1887 less than 3 percent of the total area of the province was cultivated land, the lowest in the empire next to Yunnan, Guizhou, and Guangxi in the Southwest.^[3] Wheat and millet were the major crops in the province, with acreages of 16.4 million and 14.9 million mu respectively in 1931–37. Miscellaneous crops (field peas, broad beans, oats, buckwheat) accounted for 6.2 million mu , corn for 2.2 million mu , barley for 1.9 million mu , and sorghum for 1.1 million mu . Only very small amounts of rice were grown. This was a very stable mix of crops, typical of regions with very low rainfall, little irrigation, and frequent droughts.^[4]

Dwight H. Perkins estimates Gansu's grain yields as the lowest of the 18 provinces, an estimate confirmed by Governor Huang Tinggui, who complained in 1744 that harvests were poor in Gansu because the local people failed to use manure properly or to plow deeply.^[5] Private grain storage was so low, claimed Governor Wu Dashan, that the people obtained half of their

[2] Population figures for Gansu are highly suspect: if the population was only 12.8 million in 1957, it is hard to believe that it could have attained the official figure of 15.2 million given for 1787. On the other hand, the 1749 figure of 5.71 million is a clear underestimate. (Dwight H. Perkins, Agricultural Development in China, 1368–1968 (Chicago, 1969), pp. 207–8; Yeh-chien Wang, Land Taxation in Imperial China, 1750–1911 (Cambridge, 1973), p. 87; Qingchao wenxian tongkao (hereafter WXTK ) 36/5195, 37/5205-6.

[3] Perkins, Agricultural Development , pp. 223, 236; Ping-ti Ho, Studies on the Population of China, 1368–1953 (Cambridge, 1959), pp. 124–25. The reported area was, of course, much less than the actual cultivated area, but the relative standing of the provinces is roughly accurate.

[4] Perkins, Agricultural Development , pp. 249–58.

[5] Ibid., p. 19; Zhupi zouzhe (hereafter ZPZZ ), tunken gengzuo 1744.3.25.

seed for planting each year from government granary loans.^[6] Within the province, the districts west of the Yellow River suffered colder weather and later harvests than those to the east. Not only frequent floods and droughts but also hail, wind, sandstorms, insect plagues, frost, and snow could easily ruin a crop.^[7]

Besides its low productivity, Gansu is noteworthy for its high percentage of tuntian , or former military garrison lands, most of which were cultivated by civilians in the Qing. Tuntian formed over 37 percent of the registered land area in 1753.^[8] By 1772, the Qing had assigned 27 percent of the registered population, or perhaps three million people, to cultivate these fields. The Qing rulers relied heavily on the agricultural output of Gansu to feed not just local military forces but also the armies stationed in Xinjiang.^[9]

The military presence both in Gansu and in Xinjiang strongly affected state demands on Gansu's resources. Regular garrisons in the northern, eastern, and southern military districts of Xinjiang numbered at least 25,000 men by the early nineteenth century. Including their dependents, this meant an army of at least 125,000 people to be supported by a combination of garrison lands, cash stipends, and grain shipments from the interior. Joseph Fletcher estimates that of the annual military pay in Xinjiang of 3 million taels in silver, Han China supplied 1.2 million taels.^[10] Gansu was not the only source of subventions for Xinjiang, but it was the route through which all cash, grain, horses, and clothing reached the frontier.

Given the heavy demands by the state on Gansu and its low level of agrarian output, it is not surprising that the tax accounts of the province were almost always in arrears. Even though its tax quota was only 250,000 taels in 1725, it still owed the central government an unpaid deficit of 290,000 taels.^[11] The tax reforms of the Yongzheng reign (1723–35) imposed on Gansu additional annual demands of over 75,000 taels to provide for the "nourishing-virtue" supplements to magistrates' salaries, for which the ordinary source used in other provinces—meltage fees on silver collection—was insufficient. Gansu used 20,000 taels from surplus collections on frontier trade duties and 11,900 taels in customary fees from the sale of merchant licenses, but still had to request a special transfer of surpluses from Shaanxi.^[12] Gansu's inability to

[6] ZPZZ, zhenji 1759.3.10, Wu Dashan.

[7] ZPZZ, tunken gengzuo , 1742.2.2, Huang Tinggui.

[8] Yeh-chien Wang, An Estimate of the Land-Tax Collection in China, 1753 and 1908 (Cambridge, 1973), table 24.

[9] Jiaqing chongxiu Daqing yitongzhi, juan 251–73. Although Xinjiang was not established as a province until 1884, for convenience I use this term to refer to Chinese Central Asia in the eighteenth century.

[10] Joseph Fletcher, "Ch'ing Inner Asia," in John K. Fairbank, ed., The Cambridge History of China , vol. 10, Late Ch'ing, 1800–1911 , pt. 1 (Cambridge, 1978), p. 61.

[11] Madeleine Zelin, The Magistrate's Tael: Rationalizing Fiscal Reform in Eighteenth-Century Ch'ing China (Berkeley and Los Angeles, 1984), p. 312n.13.

extract sufficient tax income from agriculture led it to rely on more imaginative and less orthodox methods, especially taxes and contributions by merchants for official degrees. The central government, primarily interested in grain supplies for the military, seems to have been slow to realize Gansu's limitations. The state lowered its demands for cash by decreasing the surcharge for salary supplements from 30 percent to 15 percent, but it maintained high demands for grain. Gansu's tax quota in cash in 1745, 299,000 taels, was the lowest in the empire except for Yunnan and Guizhou, but its quota in grain (508,000 shi ) was the sixth highest.^[13] In fact, Gansu paid less than its quota in grain and more in cash, relying on an annual income of 32,000 taels from the tea and horse trades.^[14]

Gansu, like the other peripheral western and southwestern provinces (Shaanxi, Sichuan, Yunnan, Guizhou) was on the whole a low-revenue, deficit tax collection area, heavily dependent on subsidies from interior provinces.^[15] Within the province, however, rates of collection varied widely. In 1908 the average collection per county (xian ) was 3,600 taels of silver and 5,300 shi of grain, but Zhangye (the Ganzhou prefectural capital) and Wuwei (the Liangzhou prefectural capital) paid the enormous sums of 101,800 and 94,400 taels respectively.^[16] Such regions relied heavily on merchant taxes, especially on salt shipments, to meet these demands.

Gansu, then, was one of the poorest of the Han-dominated regions of the empire, comparable in remoteness, sparseness of population, and low productivity to the recently settled Southwest. Unlike the Southwest, however, Gansu occupied a strategic military position guarding the corridor leading to Central Asia, where the Qing rulers conducted their most expansive military campaigns. Unlike the Southwest, too, it had no major mining resources, and its native non-Han peoples—Muslims, Tibetans, Mongols—were assimilated far less willingly to Chinese culture than the native peoples of the Southwest. Persistent tension, sometimes leading toward accommodation, sometimes toward violent revolt, characterized Gansu's social fabric throughout the nineteenth and twentieth centuries.^[17] Economically, however, Gansu increased its ties to interior China from the eighteenth century on. Shaanxi merchants controlled much of the province's internal trade. Goods from Hebei, Shanxi, Sichuan, and Henan supplied the civilian and military needs of the population. Gansu merchants, in turn, sold furs as far south as Hunan.^[18] Private markets had much to do with drawing the regions together,

[12] Ibid., p. 140.

[13] Yeh-chien Wang, Estimate , tables 26, 27.

[14] Yeh-chien Wang, Land Taxation , p. 71; Yongzheng zhupi yuzhi (hereafter ZPYZ ) 4.4.92b–93 (Gansu governor's report of 1725).

[15] Yeh-chien Wang, Land Taxation , p. 101.

[16] Ibid., p. 59.

[17] Jonathan Lipman, "The Border World of Gansu" (Ph.D. diss., Stanford University, 1981).

[18] Ningxiang xianzhi , 1816/8/8.

but the institutions established by the Qing founders and perfected in the eighteenth century, interacting with the market economy, played an important role. The institution with the greatest influence on the agrarian sector was the national granary system.

Granary Reserves and Food Supply

The nationwide granary system of the Qing stored large amounts of grain for leveling annual price fluctuations. "Ever-normal granaries" (changpingcang ) in each county built up their reserves during the eighteenth century using funds and grain obtained from a combination of regular state revenues, contributions for degrees, and transfers from surplus provinces and areas along the Grand Canal.^[19] Local elites in many provinces also supplied and managed community granaries (shecang ) and charity granaries (yicang ). This extraordinary grain storage system, whose total reserves far surpassed the holdings of any other premodern state, did succeed for a while in amassing large amounts of grain and in using these reserves to level price fluctuations. Reserve holdings rose to a peak near the end of the eighteenth century, rising from 30 million to 45 million shi . This was a volume of 31 to 46 million hectoliters, equivalent to a weight of milled rice of 261 to 391 million metric tons.^[20] Along with this growth in holdings, however, appeared many signs of corrupt management, rotting of grain, and ineffective use of grain for price relief. Although the official level of grain stores dropped back to 30 million shi in the nineteenth century, the real level declined even more rapidly. Furthermore, granary reserves were increasingly diverted to other uses. By the mid-nineteenth century, the use of granary reserves for military supplies had become a very common cause of depletion of the system.

The system functioned well in the eighteenth century as long as officials maintained adequate supervision over granary accounts, took care to prevent spoilage by turning over the stocks, and used grain only for price leveling. Gansu is an example of a province where latent destructive forces of the granary system appeared very early. Supplying military demands, in particular, was explicitly recognized as one of the functions of Gansu reserves. This made it all the more difficult to maintain high levels of reserves, despite the very great demands placed on Gansu by the center.

Since other studies have described the general functioning of the granary

[19] This discussion relies on references provided in Pierre-Etienne Will and R. Bin Wong, Nourish the People: The State Civilian Granary System in China, 1650–1850 (Ann Arbor, 1991).

[20] The shi was a volume measure of grain, equivalent in the Qing dynasty to 2.94 U.S. bushels, or 103.5 liters. Its weight varied by type of grain, but one shi of milled rice, the most common granary holding in south China, roughly equaled 185 pounds, or 84.1 kilograms. See Han-sheng Chuan and Richard A. Kraus, Mid-Ch'ing Rice Markets and Trade: An Essay in Price History (Cambridge, 1975), pp. 79–98.

system, here I shall only discuss certain aspects of the system that are peculiar to Gansu. The Qing empire demanded extraordinarily high amounts of grain storage from this poor province. Nearly all grain stored in Gansu was either wheat or millet. The Yongzheng emperor (1723–1735) set no fixed quotas for Gansu granaries, but in 1735 Gansu reported reserves of 750,000 shi .^[21] In 1748, however, when the Qianlong emperor assigned new targets to all the provinces, he gave Gansu the very high total of 2.29 million shi .^[22] By 1763 it had actually achieved only 1.28 million shi . After 1748 the emperor allowed most provinces to reduce their target levels, but he increased Gansu's targets because frontier military garrisons needed provisions.^[23] By 1789, he had raised Gansu's target to 3.31 million shi , but it actually reported only 2.2 million shi (all figures here are given in Qing imperial units, or cangshi ). Gansu, a poor province whose population at best amounted to a mere 2 to 5 percent of the national total, was expected to store about 10 percent of the national aggregate in both 1748 and 1789. In fact, it accumulated stockpiles that accounted for 3.9 percent of national reserves in 1767 and 7.4 percent in 1789. This substantial rise in Gansu grain holdings raised it from thirteenth to fourth in size of holdings among the nineteen provinces storing grain.^[24]

The target figures for 1748 and 1789, however, did not genuinely represent the required amount of stored grain in each province. At best, they revealed the expected role of each province in the granary system. It is unlikely that any Qing official seriously believed that Gansu would be able to collect over 3 million shi in 1789. Gansu's "deficit" of 1.1 million shi in this year did not necessarily signify severe inadequacies in its granary administration, but only the great limitations on the province's ability to extract large amounts of grain from a poor population. Since the northwestern provinces always maintained above-average levels of per capita grain reserves, they may well have stored enough grain to carry out their primary function of price leveling.^[25]

Building up and maintaining such large reserves was always a difficult problem. Gansu had three important sources for grain besides the local agri-

[21] ZPYZ 53.44a, 45b.

[22] The grain measure used in Gansu, the jingshi , was equal to 0.7 cangshi , the standard granary measure used in the rest of the empire. Memorials in Gongzhongdang archive, Palace Museum, Taibei, Qianlong reign (hereafter GZD-QL ), 2.25, 1751/11/22. All the grain figures given here have been converted to cangshi . Prices, however, are given here in taels per jingshi . For comparison with other provinces, these prices should be raised by 14 percent.

[23] Daqing gaozong chunhuangdi shilu (Qianlong) (hereafter QSL-QL ), j.330.33-35 (1749/1/30).

[24] For 1748 figures, see Qingchao wenxian tongkao , 36.5195, 37.5205–5206, cited in Pierre-Etienne Will, Bureaucracy and Famine in Eighteenth-Century China , translated by Elborg Foster (Stanford, 1990), pp. 193, 196. For 1766 and 1789, see the tables in Will and Wong, Nourish the People .

[25] Maps in Will and Wong, Nourish the People .

cultural population: special allocations from the central government, transfers from Shaanxi, and merchant contributions for degrees. Central government officials, when they realized that especially large deficits and restocking problems plagued the Northwest, allocated large amounts of money to the region. They gave Gansu 3 million taels to restock its granaries in 1766, a year of good harvests, and 8,000 more taels in 1767.^[26] These large transfers, however, were extraordinary, one-time measures. Shaanxi, a much more productive province because of the fertile land along the Wei River, could also supplement Gansu's supply in good years. In 1756, a bumper crop year, the emperor allowed the Shaanxi governor to buy more grain than his normal quota and send it to Gansu.^[27] The governor could not, however, transfer grain regularly to Gansu, for fear of straining Shaanxi's supplies. Other Shaanxi governors blamed excess government demand from Gansu for driving up prices on the Wei River by causing competition between official and merchant purchasers.^[28] "Contributions" (juan )—fees paid by local merchants and others to obtain lower-level examination degrees, bypassing the first level of the examination system—became the most common regular source of grain for Gansu. From 1741 to 1745 Gansu reported collection of over 1 million shi of grain from contributions. On the other hand, this controversial policy did not always work. In 1747 Governor Huang Tinggui reported collection of only 43,900 shi from contributions, despite his exhortations.^[29] A 1766 edict prohibited reliance on merchant grain contributions, because officials feared embezzlement and the difficulty that grain purchases created for the local population.^[30] Gansu, however, soon resumed the practice, once local officials realized that the province had no other way to maintain its grain reserves.

Through merchant contributions for granary restocking, the Qing rulers used their monopoly authority over literati status to extract resources from the merchant community for the benefit of the rural population. The Qing state, in principle, derived most of its revenues from the land tax, levying very small taxes on internal trade. In practice, however, ties between the state and merchants were much closer than the formal fiscal structure suggests. For example, even though licensing fees for brokers in local markets formed a very small fraction of total revenue, the Qing rulers used these fees effectively to supervise local markets.^[31] The use of merchant contributions for

[26] Ibid., manuscript version, p. 72.

[27] GZD-QL 21/9/9.

[28] GZD-QL , Zhongyin, 1753/6/16, 1752/8/10, 1752/8/21; Yang Yingju 1763/6/11.

[29] ZPZZ, caizheng cangchu , 1747/4/11, cited in Will and Wong, Nourish the People , manuscript version, p. 84n.72.

[30] GZD-QL 13303; WXTK 37.5205, cited in Will and Wong, Nourish the People , manuscript version, p. 63n.23.

[31] Susan Mann, Local Merchants and the Chinese Bureaucracy, 1750–1950 (Stanford, 1987).

granary reserves shows that Qing officials not only accepted the presence of competitive markets but also recirculated commercial revenues back into agricultural subsidies. Price-leveling sales, loans to farmers, and relief distribution both stimulated agricultural production and facilitated the operations of the private grain market.

Allowing merchant contributions for granaries, however, led Gansu into one of the Qing dynasty's worst political scandals.^[32] Wang Danwang, taking over as provincial treasurer in 1774, illegally commuted contributions from grain into cash, siphoned a sizable fraction off into his own pockets, and wrote fraudulent reports to the central authorities about the actual reserves in Gansu. He demanded that subordinate officials submit false reports of disasters to obtain famine relief funds from the central government. Those officials who cooperated donated part of these fraudulently obtained funds to Wang himself and kept the rest as a reward for collusion. The vast dimensions of this scandal were only discovered in 1781, after Wang had left the province, taking several hundred donkey loads of loot with him.^[33] Grand Secretary Agui, sent to repress a Muslim rebellion there, exposed the huge deficits in granary accounts and set in motion an impeachment process that led to the execution of 56 officials and the banishment or flogging of more than 46. This scandal vividly illustrates the dangers of reliance on contributions and unorthodox methods to fill granary reserves. The early Qing suspicions of this method were justified: merchant contributions, in cash or grain, proved too tempting for wily and unscrupulous officials.^[34] Wang's scheme devastated Gansu's granary reserves from 1774 to 1781, but they recovered under close supervision in the following years. Whatever the bureaucratic repercussions of the scandal, the measurable effects on prices, grain supply, and the local economy were slight.

Memorial reports on actual granary reserves in Gansu, as opposed to official quotas, confirm the great difficulty of maintaining stable grain reserves in the province (see Figure 3.1). In the late fall of each year, after restocking from the fall harvest, every provincial governor reported the total reserves held in his granaries. As the graph shows, Gansu's annual holdings fluctuated greatly, more, in fact, than almost every other province in the empire.

[32] Sources are from Shangyudang, QSL-QL, GZD-QL , 1774–1781; Qinding Lanzhou jilue . A brief discussion is in Will and Wong, Nourish the People . The most complete discussion in English is now Muhammad Usiar Yang Huaizhong, "The Eighteenth Century Gansu Relief Fraud Scandal" (Paper presented to the conference "The Legacy of Islam in China: An International Symposium in Memory of Joseph F. Fletcher," Harvard University, Cambridge, Mass., April, 1989).

[33] Shangyudang , 1781/7/12, p. 139.

[34] Officials, however, disagreed on whether grain or cash was easier to steal. For discussion, see R. Bin Wong and Peter C. Perdue, "Famine's Foes in Ch'ing China," Harvard Journal of Asiatic Studies 43, no. 1 (June 1983):313–14.

Fig. 3.1.
Reported Grain Reserves in Gansu Province, 1740–1860 (millions of shi )
Note: "1748 Target" and "1789 Target" refer to quotas assigned to the province in these years. The bars give the actual holdings.
(See text.)

Although reserves built up over the long term, from the 1740s to their peak in the 1790s, the military campaigns of the 1750s and 1760s reduced granary holdings drastically by siphoning off much of the ever-normal granary reserves to feed the troops. By 1769 the granaries appear to have recovered their level of 1753, but Wang Danwang's reign of fraudulent reporting, beginning in 1774, makes the figures for the 1770s suspect. The exposure of the relief scandal revealed that in 1781 true reserves had dropped to less than 1.5 million shi . Gansu experienced a genuine recovery in the 1780s and 1790s, when its granary accounts were under close scrutiny. Figures for the nineteenth century, by contrast, show that Gansu's grain holdings plummeted to a stabler but much lower level.

Gansu's granary reserves follow, in exaggerated form, the pattern of the empire as a whole. Year-to-year fluctuations were higher in the eighteenth than in the nineteenth century. Nineteenth-century granary officials undertook sporadic, short-term rebuilding campaigns, as in Gansu during the 1830s, but these campaigns did not offset a longer-term trend toward declining reserves. The figures for the nineteenth century are also more suspect, because the institutional controls over corruption, spoilage, and false reporting were looser. Rather than interpreting this drop as evidence of general decline in Qing administration, we can also view it as a shift in Qing policies away from the difficult methods of reliance on storage in kind toward greater reliance on the private market. Gansu, in this sense, pioneered moves by the Qing state toward injection of money into the regional economy. Governor Nayancheng's campaign of 1810–11 to relieve a drought that struck 30 counties exhibited further moves toward money and away from grain. He distributed 571,900 taels of silver and only 95,600 shi of grain to feed over 2,777,000 people.^[35]

Any assessment of the Qing ever-normal granary system must distinguish between what the Qing officials expected it to do and how it really functioned. The Qing rulers designed the ever-normal granary system to serve only one goal: price leveling. In principle, each granary should have sustained itself. After an initial build-up period, during which reserves were increased through official purchases and grants, granary managers were expected to keep the granaries at a stable level without outside support. By selling at high prices in the spring and repurchasing at lower prices after the fall harvest, officials should have been able both to maintain reserves at constant levels and to use the profits to pay salaries and maintenance costs. Emperors frequently reminded local officials of their duties, required annual reports of them, and sometimes punished them for very small discrepancies in granary accounts. They expected regular and full restocking every fall. Of

[35] The principal source for this relief operation is Nayancheng, Zhenji (1813). Brief discussion given in Wong and Perdue, "Famine's Foes", pp. 304–9.

course, only perfect prediction of future market conditions could have maintained absolutely stable reserves, but regions with regular harvests could much more easily keep their granaries stocked than regions suffering from frequent, unpredictable disasters. Stability also required that grain stocks be used predominantly for price-leveling sales. If reserves were diverted to other uses, either as official levies or for sales below market prices, extra funds would be required to restock in the fall.

Despite these difficulties, many provinces did succeed in keeping reserve levels stable for considerable periods of time. This stability, however, reflected a variety of relationships between regional grain markets and official purchasing activity. In the Southwest, for example, extremely stable reserves reflected a very low rate of grain turnover in regions of relatively localized markets. In the Lower Yangzi, on the other hand, stability resulted from highly commercialized, well-integrated grain markets and low per capita reserve levels.

In this paper, I stress the very wide variability of Gansu's reserves and what it reveals about Gansu's grain market. Substantial diversion of grain for military use combined with frequent poor harvests produced widely varying annual reserve levels. Still, even though Gansu's granaries fell short of the ideal design of the Qing system, they had significant economic effects. In fact, high annual fluctuations could indicate that reserves were being used effectively to relieve harvest shortages. In a region of frequent disasters, the best way to use granaries would be to accept deficits in bad years and make them up in good years, balancing the reserves over a multiyear cycle. If we could be certain that most of Gansu's reserves were used in this manner, the fluctuation in reserves would indicate highly effective management.

They could, on the other hand, indicate widespread diversion and peculation. Some provinces reported the total amount of grain purchased and sold during the year, allowing us to calculate the turnover rate, equal to the total purchases and sales divided by the end-of-year stocks. Gansu, unfortunately, is not one of those provinces, so it is difficult to determine exactly how much grain was bought and sold on the market. A more detailed examination of famine relief distributions would help resolve this question. For now, we may say that price stability after the 1760s provides at least some evidence that granary reserves were used effectively during the late eighteenth century to relieve the impact of harvest disasters. Although Gansu did not meet its targets, its high per capita grain holdings meant that grain distributions did have a relatively strong effect on the local grain market.

Military Demands on the Grain Supply: the Campaigns of 1758–1761

The three great military campaigns of the early to middle years of the long Qianlong reign (1736–95) consolidated the Chinese hold on Central Asia,

eliminated the centuries-old Mongol military threat, and expanded China's territory to unprecedented size. Only enormous logistic support from the interior made these campaigns possible. All of China's northern and northwestern provinces—Zhili, Shanxi, Shaanxi, and Gansu—bore the brunt of supplying the troops on the frontier with animals, wagons, porters, food, straw, uniforms, and weapons, but Gansu suffered the most. During the campaign against the Eleuth Mongols, from 1758 to 1761, large numbers of soldiers marched through Gansu. Even though they carried part of their rations with them, their demands on local grain markets drove up prices to spectacular heights. Grain prices doubled or tripled, but the prices of other goods also increased. Officials in Gansu had to raise the price they paid for horses from 8 to 10 taels, for cattle from 4.4 to 8 taels, in order to meet the local market price.^[36]

Long pack trains marched through the province: 6,000 camels were sent from Zhili and Shanxi, many of which died of disease, requiring replacements; 12,000 horses were sent to Barkul (Balikun), of which the Eleuths stole 300.^[37] Neighboring provinces received allotments of 3 million taels for grain transport to Gansu, in addition to 3 million taels given to Governor Huang Tinggui for military supplies.^[38] The garrison in Hami, which increased from 10,000 to 20,000 in 1758, required at least 40,000 shi of grain per year.^[39] Surprisingly, only 20 percent of these supplies were in kind and 80 percent in cash.

The vast distances—850 kilometers in a straight line from Lanzhou to Anxi, 300 more kilometers from there to Hami—and the high cost of transport across the steppes made it impossible to provision the troops through military pack trains alone. Expansion of garrison lands in Hami and Turfan provided valuable supplements, but only enough for 9,000 men for seven months.^[40] Although Huang Tinggui instructed commanders to avoid purchases in Gansu by ordering soldiers to carry their own rations and even considered feeding the entire 20,000-man Hami garrison from Sichuan, he inevitably concluded that much of the grain had to be bought locally.^[41] Of necessity, the private grain market in Gansu supplied a large share of the rations for the troops fighting on the frontier. The increase in military demand, combined with the influx of silver from government purchases, drove up local prices relentlessly.

This campaign unfortunately coincided with several years of widespread drought throughout the Northwest. Ningxia, with its irrigated fields, did send Liangzhou and Ganzhou its surplus of 112,000 shi , which was soon

[36] QSL-QL 554.2b (1758/1).

[37] Ibid., 576.36a (1758/12), 554.21b (1758/1), 557.31b (1758/2), 556.15 (1758/2).

[38] Ibid., 512.25b (1756/5), 575.17b (1758/11).

[39] Ibid., 564.19a, 564.17 (1758/6).

[40] Ibid., 573.23a (1758/10).

[41] Ibid., 567.27a (1758/7), 565.13a (1758/6).

exhausted. Relief needs alone were estimated at 500,000 shi of grain and 300,000 taels of silver.^[42] Once again, the private grain market supplied much of the relief grain. In 1759 the emperor relaxed the usual rule of 50-50 distributions of relief in cash and kind to allow full cash relief where needed.^[43] Where prices were low, especially east of the Yellow River in the ninth and tenth months of the year, relief in silver allowed the poor to buy food. As winter shortages exhausted market supplies, relief shifted to grain alone.^[44] Besides direct distribution, the officials sold grain to reduce prices, but these sales only reduced prices by several tenths of a tael per shi .^[45] Sales of millet at 2.4 taels per shi did not prevent the price from rising to 3.5 taels in Lanzhou by the end of 1759. It peaked at 4.4 taels in mid-1760 before dropping in 1761. Refugees flocked to sheds built for them in the cities, and Shaanxi shipped in 1.02 million shi of grain simply to provide seed loans for spring planting.^[46] These are only a few signs of the major disaster inflicted on the province by the combination of the frontier military campaign and widespread drought.

In the short term, the campaigns of 1758–61 inflicted great suffering on Gansu's people, but in the long term they may have promoted Gansu's integration into the rest of the empire. Military salaries and relief allocations in cash poured large amounts of silver into the local economy. Despite its remoteness from both the copper mines of the Southwest and the silver imports on the coast, in 1761 Gansu's silver-copper ratio of 890 cash per tael was comparable with the ratio in the rest of the country.^[47] Gansu officials set up government moneychanging bureaus in 1761 to help currency exchange, and they tried to standardize currency within the province.^[48] We need more data on currency flows to confirm this hypothesis, but these military campaigns may well have contributed decisively to the monetization of the Northwest's economy.

The defeat of the Eleuths obviated the need for additional great military expeditions and brought relative peace to the Northwest. Of course, local unrest did not disappear. Tensions between Chinese Muslims and Han Chinese broke out in a short-lived revolt in 1781, but this had no effect on

[42] Ibid., 565.20b (1758/6); ZPZZ, zhenji , 1759/6/3, Yang Yingju.

[43] QSL-QL 567.12b (1758/7), 581.2a (1759/2).

[44] ZPZZ, zhenji , 1758/10/17, Huang Tinggui.

[45] QSL-QL 578.2a (1759/1).

[46] ZPZZ, zhenji , 1760/9/9, Wu Dashan.

[47] Hans-Ulrich Vogel, "Chinese Central Monetary Policy and Yunnan Copper Mining in the Early Qing (1644–1800)" (Ph.D. diss., University of Zurich, 1983); Hans-Ulrich Vogel, "Chinese Central Monetary Policy, 1644–1800," Late Imperial China 8, no.2 (December 1987): 27; Chen Chao-nan, Yongzheng Qianlong nianjian de yinqian bijia biandong (1723–1795) (Taibei, 1966).

[48] Proposals to standardize currency and increase the copper supply are in Gongzhongdang Yongzhengchao zouzhe (Taibei, 1977–79), vol. 5, p. 230 (1725/10/1); vol. 11, p. 782 (1728/11/16); QSL-QL 580.13a (1759/2).

price levels. Whatever the underlying ethnic tensions in the region, Gansu's markets functioned much more stably after 1761 than they had before.

Major Qing institutions affected the grain supply in Gansu, then. In feeding armies, collecting taxes, and stocking granaries, the Qing officials had to adapt to the limitations of agricultural output in a poor peripheral province. At the same time, they stimulated market exchange by injecting cash into the economy through grain purchases and soldiers' salaries. The collection of taxes in cash also stimulated market exchange by forcing peasants to sell their surplus crops for cash.^[49] The eighteenth-century conquests in Central Asia not only brought vast new territories under Chinese control; they also contributed to knitting together the interior regions of China by linking internal markets to the military supply route. The analysis of price data that follows examines the extent to which Gansu's prefectures were linked together by a system of grain markets extending from the core regions of the Southeast through the corridor into Central Asia.

Price Data and Market Integration

The 351 Gansu price memorials available to me cover the years 1739 to 1864. The eighteenth century is much more fully covered than the nineteenth: there are at least some data for 38 of the years 1739–99 but for only 21 of the years 1800–64. Monthly coverage is also much fuller for the eighteenth century than for the nineteenth. Some of the reports from the nineteenth century are too suspiciously constant, giving the same price three or four months in a row. My suspicions about some of the nineteenth-century data make me reluctant to draw conclusions about differences between the two centuries until more information becomes available. The Gansu price memorials report high and low prices for five major crops: millet (two varieties), wheat, beans, and barley. Here we shall analyze only the data for the primary millet crop (sumi ). The reports cover the 13 prefectural divisions of Gansu Province, plus Hami, located in Xinjiang, and, for some years, the Jingni garrison, 100 kilometers northwest of Anxi.

Although the memorials report the highest and lowest prices within each prefecture every month, they do not tell us the locations of each high and low price. The yearly average curves of high prices and low prices within each prefecture parallel each other quite closely, but the high prices in Gongchang, for example, show occasional sharp peaks not found in the low prices (Figure 3.2). The most likely explanation for this pattern is that low prices

[49] Compare with similar processes in England and France described by Rudolf Braun, "Taxation, Sociopolitical Structure, and State-Building: Great Britain and Brandenburg-Prussia," p. 318, and Charles Tilly, "Food Supply and Public Order in Modern Europe," pp. 380–455, both in Charles Tilly, ed., The Formation of National States in Western Europe (Princeton, 1975).

Fig. 3.2
Millet Prices in Gansu Province and Gongchang Prefecture, 1739–1864 (adjusted annual
averages, in taels per shi )

represent the stabler patterns found in prefectural and county capitals, which were more commercialized than remote parts of the prefecture, where sporadic periods of dearth caused the sharp peaks in the series of high prices. The limited amplitude of seasonal variation in low prices compared to high prices supports our interpretation of low prices as coming from commercialized capitals (see Figure 3.3). For these reasons, analysis of the correlation between low prices is the most appropriate way, I believe, to measure market integration in this region.

Figure 3.2 shows the trends of the adjusted annual averages of millet prices for the province and one prefecture. (On the adjustment of data here and in other figures, tables, and the map, see the appendix at the end of this chapter.) As Map 3.1 demonstrates, there was a clear gradient of average prices across the province, because high transport costs raised price levels as one moved from east to west. Millet in Hami, where the mean of annual prices averaged 2.40 taels per shi , cost nearly twice as much as in Lanzhou, and Anxi, at 1.50 to 2.19 taels, was over 40 percent higher than Lanzhou. Low-price prefectures included Qinzhou and Jiezhou, Qingyang, and Pingliang, in the eastern end of the province, with access to Sichuan and Shaanxi supplies and the irrigated fields of Ningxia.

Several factors combined to produce the variation in grain prices among prefectures. Long-term changes in the balance of population and agricultural production, or changes in monetary supply, affected the long-term trend of prices. The cropping cycles of each region determined seasonal fluctuations from month to month. The impact of the major drought and military campaigns of 1758–61 produced drastic price changes, which swamped the impact of seasonal and annual trends. I used a multiple regression equation to analyze the relative effect, on average, of these three factors on the price for any given month.

The equation at the foot of Table 3.1 describes the monthly price as a function of a constant, the annual trend, seasonal variation due to monthly fluctuations, the disaster years of 1759–60, and an error term indicating random variation. A dummy variable (Y ), whose value is 1 for the major drought years and 0 for all other years, measures how widely prices in these two years diverged from the long-term trend. The value of the coefficient (T) of this dummy variable given in the table shows that the combined stress of widespread drought and military campaigns in 1759 and 1760 raised the average high price by 2.10 taels, or 135 percent of the average price for the period. Prices more than doubled during these two years in nearly all prefectures, notably excepting Qinzhou and Jiezhou, in the southeast. The coefficients for both high and low prices of the dummy variable for disaster influence in Qinzhou range from 0.42 to 1.08, much less than the effect on the average of prices for the province. Jiezhou's disaster coefficient ranges from 0.22 to –0.04, showing that this prefecture, which belongs to the Sichuan

Map 3.1
Correlations and Variation of Low Millet Prices in Gansu Province, 1739–1864.

Note: Bivariate correlations of differences in adjusted annual averages are indicated by solid and dashed lines; solid lines represent r greater
than 0.8, while broken lines represent r greater than 0.7 and less than 0.8. Means of high and low prices are shown for each prefecture
(high prices first) in taels per shi .

watershed, was completely unaffected by the drought in Gansu. Drought and disaster hit hardest at the core prefecture of Lanzhou (G = 2.82) along with Liangzhou (G = 3.64) and Anxi (G = 2.55) on the corridor leading to Central Asia.

If we exclude the effect of these two disaster years, prices hardly rose at all over the long term. The value of a , the coefficient of the variable T for the year, gives the annual trend for each prefecture excluding the effect of the disaster years. This coefficient is negative for both high and low prices in Anxi, Ganzhou, Hami, Suzhou, and Xining and only slightly positive for the rest. (Jingni, the one exception, only provides data for a limited time period.) The ratio of a to K , the constant term, for the average for all low prices in the province, yields a long-term annual rate of increase of less than 0.1 percent per year. Gansu did not share in the rise of prices from 1780 to 1830 described by Yeh-chien Wang and Kuo-shu Hwang for the rest of the empire.^[50] Its prices followed a different rhythm, determined more by military operations than by flows of silver. The high annual fluctuations up to 1762 contrast markedly with great stability after the 1770s. The price data confirm that once the early Qianlong campaigns had brought peace to the turbulent Northwest, Gansu's grain markets proved comparatively immune to sporadic attacks of rebellion, famine, and drought.

Seasonal variation from month to month was also remarkably low. Each month in the equation, except for the first month, is represented by a dummy variable, whose value is 1 for the price in that month and 0 for all other months. For example, the value of M2 is 1 for February prices and 0 for all other prices, the value of M3 is 1 for March prices and 0 for all other prices, and so forth. The coefficients of these variables indicate how much, on average, each month contributed to a change in price from the first month of the year (see Table 3.1). Month-to-month variation rarely exceeds 10 percent of the average price in each prefecture. The especially small variation of the low prices implies that grain storage in the commercialized capital cities damped out nearly all monthly fluctuations.

The monthly variations suggest that most prefectures had two major annual harvests, with a spring crop bringing low prices in April, May, or June (sometimes February or March), while the main fall crop forced prices to their low point in October (see Figure 3.3). The shapes of the curves, however, are not uniform over the region, because cropping regimes varied widely. Gansu's mix of crops and weather differed markedly from the much more uniform seasonal patterns of a rice paddy region like Hunan (see the chapter by Wong and Perdue in this volume). In Ganzhou, Suzhou, and Anxi, at the western end of the province, the fall harvest came in later than

[50] Hwang Kuo-shu and Wang Yeh-chien, "Qingdai liangjia di changji biandong, 1763–1910," Jingji lunwen 9, no. 1 (1981).