Four
Grain Markets and Food Supplies in Eighteenth-Century Hunan

R. Bin Wong and Peter C. Perdue

By the sixteenth century, recent immigrants to Hunan had begun to open new lands, from which rice surpluses were shipped to the growing metropolis of Hankou and the handicraft centers of the Lower Yangzi.^[1] The late Ming proverb "When Huguang [Hunan and Hubei] harvests are plentiful, all under Heaven are fed" demonstrates that Hunan had become an important producer of rice. During the Qing dynasty, the province's rice exports grew. The importance of the Hunan rice trade was stressed some 30 years ago by Abe Takeo in his classic study of food supplies in the Yongzheng period (1723–35).^[2] More recent studies of empirewide grain movements have confirmed the importance of the province's exports, the volume of which totaled 8 million (plus or minus 2 million) shi in normal harvest years.^[3] The strong demand for Hunan rice in eighteenth-century China forged important commercial ties between the province and other parts of the empire. But what about the impact of Hunan's rice export trade on the province itself? Earlier studies of Hunan's rice export trade offer some answers by focusing princi-

pally on two topics: (1) the institutions of the export trade and (2) the distribution of benefits from the trade among different groups of merchants, landlords, and peasants.^[4] From this work we know that it was often outside merchants who bought Hunan's surplus rice, that landlords could be major suppliers to local markets, and that all producers of surpluses could benefit from the expansion of the trade. We need, however, a sharper picture of the spatial dimensions of the trade in order to push forward our understanding of this trade's impact on the province more generally.

This paper reconstructs the spatial structure of rice marketing within Hunan Province. Which parts of the province were linked together by the rice export trade? How large were these areas—did they form narrow bands along trade routes or were the hinterlands of the trade routes also part of the market? What about rice commerce and markets in areas outside the export zone? Answers to these questions help determine the significance of market integration to an agrarian economy. Our approach combines qualitative and quantitative analyses of market integration. In the first section of this paper we present qualitative evidence on commercial rice circulation in Hunan. In the second, our discussion shifts to an analysis of rice prices. Finally, in the third section we conclude with some thoughts on market integration. We shall discover that analysis of either high prices or low prices by themselves provides an incomplete guide to market integration, since there are no a priori reasons to argue that either set should represent market integration better than the other. For Hunan, separate analyses of high and low prices reveal roughly similar pictures of market integration. We shall show that analysis must include price relationships that are not addressed in other chapters in this volume; even if some of the price relationships are not intuitively obvious, we must consider relationships among high and low prices of different prefectures in order to demonstrate that the prices reflect not just similar but related pictures of the rice export market.

Spatial Structures of the Rice Trade: Qualitative Evidence

Hunan's eighteenth-century commerce largely followed the province's river systems. Each of the four major rivers—Xiang, Zi, Yuan, and Li—flowed into Dongting Lake, located in the northeastern part of the province. The Xiang River was by far the most important river, draining nearly half the province's land area. Not surprisingly, merchants along this river collected and shipped considerable amounts of rice destined for export out of the province. A 1753 investigation of specialized rice markets, covering 49 of Hunan's 56 counties, found 16 counties in which major rice markets existed and three additional counties where minor rice markets also served the export trade.^[5] These 19 counties—Anxiang, Baling, Chaling, Changsha, Hengshan, Hengyang, Huarong, Linxiang, Liuyang, Longyang, Shanhua, Taoyuan, Wuling, Xiangtan, Xiangxiang, Xiangyin, Yiyang, Youxian, and Yuanjiang—were all located in five of Hunan's 13 prefectures; three of the prefectures (Changde, Lizhou, and Yuezhou) bordered Dongting Lake, while the remaining two (Changsha and Hengzhou) were south of Dongting Lake on the Xiang River. Major markets contained the facilities to accommodate boats, store grain, and arrange transactions; minor markets sent rice to these major markets for export. Map 4.1 displays the 16 counties with specialized rice markets and the three additional counties with export surpluses.

By adding information from other sources, we can piece together commercial rice flows in other parts of the province. Small quantities sometimes moved along Hunan's other three major rivers and their tributaries. Along the Zi River, rice moved from Wugang to Shaoyang and from this point to Xinhua; other shipments of rice reached the tea-producing county of Anhua further downstream. These shipments along the Zi River probably did not reach Dongting Lake. Protests against the trade in the early nineteenth century make clear, however, that even these small, short-distance, nonexport shipments were significant sources of food to the people who depended on them.^[6] In contrast, gazetteers provide no evidence of trade crossing county borders anywhere along the Li River.^[7]

[Full Size]

Map 4.2.
Correlations of Annual Price Differences for High-Grade Rice in Qing Dynasty Hunan.
(See p. 138 for Map 4.1.)

Along the Yuan River available sources allow us to reconstruct a more complex situation. Although they inform us that people in Chenzhou depended on imports from other counties, they do not tell us where the boats came from. The neighboring upstream counties of Luxi, Chenxi, and Xupu were unlikely suppliers of rice because of their own food supply limitations.^[8] It appears, therefore, that the boats came upstream from the fertile paddy areas near the lake, apparently carrying rice that would otherwise flow out of the province with the export trade. Additional trade, including sales to Huitong, flourished along the upper reaches of the Yuan and its tributaries in Yuanzhou and Jianyang. These movements in southwestern Hunan, which also carried grain across the Guizhou border, were physically separate from the movements in downstream areas between the lake and Chenzhou.^[9]

In southern Hunan small amounts of rice appear to have moved across prefectural boundaries. For instance, rice grown in Lanshan and Xintian fed miners in Guiyang. Sources also provide early eighteenth-century evidence of shipments of rice from Yongzhou to Hengzhou.^[10] The southern mountain region within which rice was sold to feed miners also, at least for a while, sent rice into the Xiang River export trade. Map 4.1 also shows those counties outside the export zone in which rice trade is noted in gazetteer sources.

In summary, the export trade dominated rice movements in Hunan above the local level but coexisted with spatially separate movements along the Zi and Yuan rivers and in the southern mountains. Qualitative data indicate some rice trade crossing county borders in ten of Hunan's 13 prefectures, but only half of those ten participated directly in the province's Yangzi River export trade. The evidence we have just reviewed establishes the outlines of separate patterns of grain trade within the province. But this information cannot answer two important questions. Were the physically distinct movements of rice outside the export zone economically independent of the interprovincial trade? What degrees of market integration were achieved within the export zone and in areas beyond it? Our analysis of grain prices provides some answers.

Price Data and the Dimensions of Market Integration

From qualitative evidence we have shown a kind of market integration defined by physical movements of rice. We do not have much sense of the size of these rice shipments; we know only that some amounts of rice moved between various points within the province and to places beyond Hunan. The size of shipments, however, is not crucial to the reasoning about market integration we now develop on the basis of price data. When prices from two areas move in related ways over time, we believe the markets of these areas are integrated. There need not be very much trade between two points to cause related price movements. As long as grain merchants have enough information about prices in another area to cause them to adjust the volumes and prices of their purchases and sales accordingly, the two regions are economically integrated, regardless of the absolute size of the trade. Conversely, evidence of physical movements of grain may not signify market integration if changes in the prices do not reflect those movements. This would indicate that even though grain flows between two regions, the trade is too sporadic or localized to have a significant impact on price movements. Qualitative evidence of physical movements and quantitative price information generally complement each other, but they do not always agree completely.

The price data we use come from the monthly provincial reports to the central government of the highest and lowest county-level prices within each prefecture.^[11] We have created annual series of the high prices and low prices for the most commercialized grain, high-grade rice, which we have adjusted for missing monthly data, in each of 13 prefectures reporting this information in the 63 of 68 years between 1738 and 1805 for which we have at least some observations.^[12] The quantitative indicator we have chosen to guide our discussion of market integration is the bivariate correlation (Pearson's r ) between differences of annual price averages. We consider any correlation between two sets of price differences that exceeds 0.65 to indicate a market relationship and thus integration of markets.^[13]

With the prefecture as the unit of observation, two kinds of market integration are easily conceptualized: (1) inter prefectural integration, indicated by much of our qualitative information and by analyses of high prices and low prices separately, and (2) intra prefectural integration, indicated by the correlation of annual price differences of the high prices and low prices for each prefecture.^[14] Obviously, market integration within each of two prefectures says nothing about integration between them. Likewise, if less obviously, measures of integration between two prefectures also say nothing by themselves about integration within each prefecture. Two separate indicators of market relationships, the correlation of differences in annual high prices and the correlation of differences in annual low prices, need not themselves be necessarily related. After examining interprefectural market integration indicated by high prices and by low prices, we shall explore the relationships between highs and lows across prefectures to demonstrate a market integration more complex than that suggested by evaluations of interprefectural correlations of highs and of lows and by intraprefectural correlations. These two independent measures are unable to capture the reality of market integration. Indeed, evaluated in light of qualitative evidence, these measures present us, in the Hunan case at least, with a serious puzzle.

Spatial Patterns of High and Low Prices

The proposition that related changes in annual price differences reflect market factors assumes that other forces are not driving the observed movements of prices. If, for instance, there was a strong and sharp trend in annual prices, due perhaps to inflation, the changes in annual price differences could not reasonably be taken as indicators of market integration. We therefore first look at an adjusted annual series of high and low prices to see what kind of long-term trend there might be. The provincial averages are displayed in Figure 4.1. Despite some fluctuations in individual years, for most of the eighteenth century prices remained between 1.1 and 1.3 taels; prices rose only modestly. In most prefectures, therefore, the rates of increase cannot help to explain the relationships among annual price changes to be analyzed below.^[15]

Rice prices throughout Hunan also displayed remarkably similar seasonal patterns, compared, for instance, to those for Gansu millet examined by

[Full Size]

Fig. 4.1.
Adjusted Annual Averages of High and Low Prices for High-Grade Rice in Hunan Province, 1738–1858 (taels per shi )

Note: There is a substantial amount of missing data after 1798.

[Full Size]

Fig. 4.2.
Seasonal Variation in Prices of High-Grade Rice in Selected Hunan Prefectures, 1738–1858
(difference from January price, in taels per shi )

[Full Size]

Peter C. Perdue in his essay in this volume. Figure 4.2 displays seasonal variations for four of Hunan's 19 prefectures, two representing exporting regions and two selected from the 15 non-exporting prefectures.^[16] The lowest prices of the year come in December and the highest in June and July. Prices are generally low in the winter months, rise steeply in the spring, and plummet between August and September. Since these roughly symmetrical curves are similar in areas linked by rice trade as well as those that are not, trade alone cannot explain the similarities. Similar schedules of planting and harvesting are the more general reasons for the seasonal price patterns.

To use our price data to study market integration, we must remove the shared price behavior due to common annual trends and monthly variations. We obtained coefficients for each of 11 monthly dummy variables by fitting a regression equation to the price series for each prefecture, then recalculated annual averages after subtracting the coefficients from each monthly price.^[17] Because the correlations discussed are calculated from the annual differences, the effect of the annual trend is nearly completely removed. The remaining correlations are a minimal set representing those price connections that are most likely caused by trade relations and not by common annual or seasonal patterns.

For the high prices in the years for which we have data between 1738 and 1805, 14 relationships show correlations of annual price differences exceeding 0.65, of which the 13 displayed in Map 4.2 confirm the basic outlines of the rice export network in eighteenth-century Hunan.^[18] Nine of these links con-

nected the major rice-exporting prefectures near Dongting Lake (Changde, Lizhou, and Yuezhou) and along the Xiang River (Changsha and Hengzhou). Four of these links (Changsha-Hengzhou, Changsha-Yuezhou, Hengzhou-Yuezhou, Changde-Yuezhou) lay along paths on which rice physically moved, while the other five links represent indirect price relationships without direct physical movements of rice between them. Changde's prices, for example, were connected indirectly with Hengzhou's via links to Changsha, even though no grain flowed directly from Changde to Hengzhou. Together, the nine links outline the integrated export market.

Four other links connect Chenzhou to three of the five rice-exporting prefectures and to Yuanzhou. The price data confirm the market relationship with Changde described by gazetteers. Since prices were higher in Chenzhou than in Changde, rice must have moved upstream, allowing Chenzhou to tap the export trade. Chenzhou's links to Changsha and Hengzhou further support the notion that it had ties to the export zone by showing market relationships without the connection of direct physical trade. In addition, Chenzhou prices are related to Yuanzhou prices, but Yuanzhou prices are lower, suggesting that Yuanzhou rice supplemented shipments from Changde to Chenzhou.

Completely separate from the export network is the price relationship between Guiyang and Chen. The generally higher prices in Chen suggest some small-scale trade going from Guiyang to Chen. The scanty qualitative information on this region for the eighteenth century provides no clear indication of this trade. In fact, this price relationship may not be a genuine economic link but may only represent changes due to harvest fluctuations in adjacent areas subject to similar weather conditions.

The failure of quantitative data to reveal movements of grain along the upper reaches of the Xiang, Zi, and Yuan rivers suggests that small amounts of grain can cross prefectural borders without clearly influencing price relationships. As we said above, small amounts of physical trade need not create strong economic relationships. Also, as is likely in the case of the Xiang River trade flows between Yongzhou and Hengzhou, the trade that existed in the early 1700s may have nearly disappeared by the second half of the century.

[Full Size]

Map 4.1.
Qing Dynasty Hunan. (See p. 129 for Map 4.2.)

Data on low prices again show a concentration of relationships among prefectures within the rice export zone. Map 4.2 also displays nine significant correlations of annual price differences for the low prices of high-grade rice.^[19] Six of the links connect the five export prefectures, while two others connect Chenzhou to the export zone. Only the one between Yongzhou and Chen is completely separate from the export prefectures; here again, as in the Guiyang-Chen high-price relationship, a combination of trade and common weather conditions may have caused the observed connection.

A comparison of high-price correlations and low-price correlations reveals strong similarities. Both high and low prices connect the export prefectures. Six of the nine low-price relationships parallel high-price links (Changde-Lizhou, Changde-Yuezhou, Lizhou-Changsha, Changde-Changsha, Yuezhou-Changsha, Changde-Chenzhou). The Hengzhou high-price links that are lacking for low prices account for most of the differences between the high-price and the low-price relationships.

Why is Hengzhou different from the other export prefectures? Before finding an answer to this question we must first confirm that interprefectural highs and lows both represent the export market. We can generally attribute the differences between highs and lows to transport costs in the same way as we treat differences between prefectures as the product of transport cost differences. But this is not enough. For the highs and lows to be related, we must find correlations between them. The most obvious place to look for such relations with prefectural-level data is in intraprefectural correlations.

Our sources provide the highest and lowest prices within each prefecture without giving the counties these prices come from, which can in principle vary each month. Still, if the highs and lows are closely correlated, all the counties in the prefecture likely follow similar patterns. In other words, a relationship between the high and low prices in the same prefecture indicates market integration within the area, including counties reporting prices between the high and the low. The precise counties reporting the high and the low each month need not therefore necessarily be the same; we can still observe more general features of the rice market in the prefectures. We consider first the correlations of annual price changes for high and low prices

within each prefecture (with the number of counties under its jurisdiction shown in parentheses):

Seven of the 13 prefectures—Hengzhou, Yongzhou, Baoqing, Chenzhou, Yongshun, Yuanzhou, Chen—have no statistically significant correlations. Two of the six with significant correlations are small southern prefectures (Jingzhou and Guiyang), where short distances and low production levels allow greater impact from common weather patterns. Four export prefectures (Changsha, Yuezhou, Changde, and Lizhou) are the other four cases with statistically significant correlations, confirming the 1753 report's identification of major rice markets in these prefectures.^[20] But the correlations for Changsha and Yuezhou are well below the 0.65 cutoff we use to identify strong indications of market integration. We therefore have a problem. Interprefectural highs and interprefectural lows each outline market integration for the rice export trade in a roughly similar manner (Hengzhou being the important difference), but we cannot establish convincing relations among the highs and lows by looking simply at intraprefectural relations. Further analysis of the price data is necessary.

Thus far we have considered correlations of annual price differences for high prices and low prices between and within prefectures. In other words, for prefectures A and B, we have considered those correlations of highs and lows depicted in Figures 4.3a and b. Together our measures represent the combination of intraprefectural and interprefectural market integration depicted in Figure 4.3c.

The organization of data by prefecture leads analysts to concentrate on the relations making the boxlike Figure 4.3c. But market relationships do not map neatly onto politically defined space. In economic terms, there is no reason to expect the market(s) providing the low prices in one prefecture to be more related to the market(s) providing the high prices in that prefecture than to the market(s) providing the high prices in some nearby prefecture. Market placements and transportation networks could easily link the highs of one prefecture with the lows of another. Unfortunately, since we do not know the county-level locations of the prefectural highs and lows, we can

[Full Size]

Fig. 4.3.
Ways to Measure Correlations of High (H) and Low (L) Prices in Prefectures A and B

only propose this scenario as a reasonable explanation of the correlations we shall examine in a moment. When related "cross" prices, shown in Figure 4.3d, are combined with the relationships between highs and lows of two prefectures, we can argue that the two interprefectural relationships are themselves related (Fig. 4.3e); in other words, we can establish market integration between highs and lows, even in the absence of strong intraprefectural correlations.

For Changsha and Yuezhou, the two export prefectures with low intraprefectural correlations, we can demonstrate the integration of their high and low prices into a common rice export network by observing in each case their cross-price relationships with nearby Lizhou shown in Figures 4.4a and 4.4b. The correlations of high Lizhou prices with low Changsha prices (0.70) and with low Yuezhou prices (0.68) reveal a market integration by cross prices that is much broader and stronger than an examination of high and of low prices between and within prefectures would have mistakenly suggested.

The number of strong relationships between rice-exporting prefectures varies. The extreme case is Changde and Lizhou (Fig. 4.4c), where all possible relationships among high and low prices are strong, thus suggesting widespread market integration within and between prefectures. For other pairs of export prefectures, some combination of strong and weak correlations can be found. The larger the number of strong correlations, the broader

[Full Size]

Fig. 4.4.
Correlations of High (H) and Low (L) Rice Prices in Selected Rice-Exporting Prefectures in Hunan,
1739–1805

the likely degree of market integration spanning the two prefectures. This kind of analysis confirms the integration of highs and lows in the export zone and asserts that the political boundaries of prefectures are not an important guide to the possible lines of economic integration represented by cross prices. The export market encompassed more than a narrow string of places along a single major transportation route; it spanned large portions of all but one of the export prefectures.

Hengzhou is the exception. The absence of low-price correlations between Hengzhou and other prefectures is shown most clearly for the Hengzhou price relationships displayed in Figure 4.4d. We first observe, in dramatic contrast to the high-price relationships, that the low prices have no relationship. Second, in contrast to the strong connection between high Hengzhou prices and low Changde prices, low Hengzhou prices have no link to high Changde prices. Finally, low and high Hengzhou prices have no relation to each other. Figure 4.4d displays this right triangle of strong

relationships (high Hengzhou prices–high Changde prices–low Changde prices), showing how high cross-price correlations strengthen our sense of highly integrated markets. The low and high prices from Hengzhou together suggest that the prefecture embraces integrated zones along the Xiang River and isolated zones toward the province's hilly eastern border. The examination of high and low prices together reveals more completely the degree to which rice-exporting prefectures are integrated into a larger marketing network than does examination of either highs or lows by themselves. Prefectures outside this network have fewer strong relationships among their prices.

Market Integration

We have analyzed qualitative and quantitative data on Hunan's rice trade. Our findings delineate Hunan's rice export trade and distinguish it from rice trade patterns in other parts of the province. First, price data generally confirm the outlines of the export trade based on qualitative information. The rice market within the export zone, with the important exception of Hengzhou, appears highly integrated both between prefectures and within prefectures. Second, price data reveal very little market integration outside the export zone—only a few thin connections in southern Hunan. The price data clarify movements along the Yuan River by suggesting amounts of some size going upstream from Changde and downstream from Yuanzhou. For the Zi River trade, however, Baoqing prices correlate poorly with other prefectures; this suggests that the trade identified by qualititative evidence may have been largely limited to movements within Baoqing Prefecture.

In Hunan's periphery, no rice markets, with the important exception of those in Chenzhou, were tied to the export trade. The case of Chenzhou, which tapped part of the Changde trade, demonstrates how an export market can include prefectural exports inside as well as outside the province. As is true for interprefectural integration, intraprefectural integration, except for two small isolated prefectures (Guiyang and Jingzhou), is high only in commercialized prefectures connected to interregional trade (Changde and Lizhou).

In sum, our analysis confirms the importance of commerce to eighteenth-century China's agrarian economy by combining qualitative and quantitative analyses of the spatial reach of China's grain markets within one province. The two approaches complement and correct each other: gazetteer evidence of small amounts of trade outside Hunan's export zone need not mean clear-cut market integration, while prices in export prefectures could be tied together without direct trading links. Persistent images of rural isolation and "natural" rather than commercial economy in Qing China will ultimately have to yield before the mounting evidence of active, integrated mar-

kets. But the presence of areas outside the export zone unaffected by rice market integration reminds us that any picture of China as a market society is also incomplete.

Let us not, therefore, jump to the conclusion that the kind of market activity we find in this agrarian Chinese setting implies everything that market activity does in other times and places. True, eighteenth-century China displayed some features of modern economies, but which ones? Even though additional research is clearly needed, we can provisionally propose that market integration in eighteenth-century China, as in modern economies, promoted regional specialization. Price signals informing people of the most profitable production strategy to pursue stimulated specialization according to comparative advantage then as now. But the presence of economically "rational" behavior does not necessarily mean that the resulting economic activity as a whole was "modern." Lacking were the forces that promote modern economic expansion—big capital investments, major technological advances, and new expanding markets. These forces work through an integrated market economy. Integrated markets may be highly desirable for economic advance, but they hardly create such developments by themselves. The Hunan rice case, like other Qing dynasty grain price examples, may represent a kind of market integration achieved without the familiar dynamics of modern economic change.