Carol Shiue
Market Arbitrage and Transport Routes: Evidence from 18th Century China

Carol H. Shiue
University of Texas
Department of Economics
Austin, TX 78712 USA
(512) 475-8513

1. Introduction

This paper presents empirical evidence that domestic markets of 18th century China were much more integrated than previously understood. The study employs a new data set drawn from previously unpublished documents in the Beijing Number One Historical Archives and also from published statistics, and assesses the degree to which markets for rice were inter-regionally and inter-temporally integrated in the 121 prefectures of 10 provinces of southeast and central China, from 1742-1795. The analysis utilizes a benchmark model of competitive equilibrium and builds on the idea that both inter-regional trade as well as inter-temporal storage activities can reduce the impact supply fluctuations have on the stability of the path of consumption. Important insights about the spatial extent of trade can therefore be found in analysis of its substitutes: the impact of current local harvest shocks on current local prices and regional changes in storage levels.

When there is independent underlying randomness in production across regions, trade can serve to pool the effects of fluctuations in supply associated with uncertain states of nature, thereby smoothing consumption for all trading regions as a whole. If in addition storage is possible, aggregate fluctuations may be reduced further: grain storage smooths consumption, but unlike trade, it does so by spreading the effects of current local harvests across several years. This analysis indicates that domestic markets in the 18th century were integrated up to an average distance of approximately 150 kilometers from major water-routes. Out of the 121 prefectures considered, 86 prefectures exhibit evidence that inter-regional trade was the predominate method used to smooth supply shocks.

In inter-regionally integrated regions, and more notably, in inland areas, current local prices were not significantly linked to current bad local supply shocks, indicating that by the 18th century China was for the most part not vulnerable to the effects of severe crop failures. The evidence presented here strongly suggests that despite the relative lack of regional trade in inland areas, market incentives were responsible for driving inter-temporal arbitrage opportunities. For the 35 inland prefectures considered in this study, inter-temporal trade was effective in its purpose of pooling markets over time.

The data set includes bi-annually tabulated rice prices ({Gongzhong zhupi zouzhe, nongye lei, liangjia qingdan }[Grain Price Lists in the Agricultural Section of the Vermillion Rescripts in the Palace Archives]), annual weather data (State Meteorological Society, 1981. {Zhongguo jin wubai nien hanlao fenbu tuji,} [Collected maps of droughts and floods in China in the past five hundred years]), and a measurement of the kilometer distance between all prefectures which utilizes longitudinal and latitudinal points of the historical locations of each prefectural city.

The ten provinces studied here constitute an area that, according to official Ming and Qing land tax records, was somewhere between 50 and 60 percent of the total registered land acreage in the 18th century; approximately 60 percent of the total population of China resided here as of 1787 (Ho, 1967). This region encompasses areas which are historically known to be the most urban and commercial regions of China--the southern coast, the Yangzi Delta region and the middle Yangzi region--as well as regions located far away from these urban centers.

2. Tranport costs

Studies of market integration in developing economies frequently focus on grain markets because grain is the major commodity of trade. This was true also of China in the 18th century. Although cereals accounted for a negligible portion of foreign shipments, grain was the most important domestically traded good, accounting for 42 percent of the total value of seven major commodities that entered into inter-regional trade prior to the 1850s (Wu, 1985). Most of this inter-regional trade took place via water-routes.

The limitations of available transport technology, however, constrained both the volume as well as the spatial pattern of trade. Costs of transportation over land were considerably higher than costs of transportation via waterways, the main route through which grain could be moved at low cost. Junk transport had a 9-10 fold cost advantage over human carriage and approximately a 5 fold cost advantage over animal carriage (see paper for derivation of estimate). The pronounced difference in transport costs implies quite different degrees of market development across different regions and absolute limits on the movement of grain (see Elvin, 1973; Evans, 1984; Schran,1978).

For large areas over the empire however, data on the extent of trade through time are not available; the scarcity of data on private trade in turn yields large margins of error in estimates on the volume of inter-provincial trade (Perkins, 1969). Moreover, quantitative estimates on actual trade volumes give limited insight into the degree to which markets are integrated. In the limit, with zero transport costs and perfectly mobile consumers, the potential of trade alone can equalize prices. Even if observed trade volumes are low, one cannot rule out the possibility that markets are integrated. More generally, trade volumes do not have to be large for commodity prices to be similar across regions. At the same time, since trade volumes are related to demand, higher absolute volumes of trade may be observed between less integrated markets which are more densely populated than between more integrated markets which are less populous. For these reasons, an examination of market integration using price data rather than trade volumes will be more informative.

3. Testing for Patterns of Market Arbitrage

Prices were in general free market prices which reflected the supply and demand for grain. When the government attempted to influence the price of grain in a region, it did so by selling official stocks of grain at reduced prices or through buying and selling grain from different regions, rather than setting a price ceiling over the market.

The first part of the analysis determines the interaction between the spatial extent of market integration and storage in a way which separately identifies the influence of transport costs and supply shocks that arise from weather conditions. This allows for a more comprehensive analysis which takes into account the interaction of both trade and storage. Year-to-year harvest shocks must be absorbed somewhere in the economic system: the effects will appear in the form of local price shocks, or in changes in either storage levels or trade volumes. Thus, an investigation into howthe underlying randomness of exogenous shocks is dispensed with may also reveal why there are regional variations in the extent of storage in these economies.

The benchmark model of competitive equilibrium is used to guide the empirical analysis (Williams and Wright, 1991). The structure of the model is simple but captures the implications of how trade and storage interacts in market arbitration. Although there existed potentially significant barriers in China's political economy which could have severely constrained trade and other market activities, it is not obvious how important these factors were or how localized were their effects. The key question is whether these barriers constricted trade over a large geographic area; if they did not, then barriers themselves may have had a limited impact on market activities.

I argue that if price correlations among different markets are related in a systematic way to the costs of trade between them, then this relationship is not a mere coincidence, but is rather driven by the realization of arbitrage opportunities. I therefore divide my sample of 121 prefectures into three groups based upon the distance of the prefecture to a major waterway. I hypothesize that prefectures located nearest to the waterways form the most inter-regionally integrated group while prefectures located furthest from the waterways are least integrated.

The first question of interest is whether price correlations among markets are being driven by a common supply component. High price correlations may be evidence of the presence of trading markets, or it may be evidence that common weather shocks are responsible for coordinating supply fluctuations which in turn drive common price movements but have nothing to do with trade or the potential for trade. These are important questions because common weather patterns are geographically localized phenomena, as is, in general, trade in the presence of transport costs.

Therefore the second question of interest is how decreasing transport costs between regions determine the relationship between price correlations and local weather shocks. Assuming that each prefecture receives an i.i.d. weather shock, prefectures that are located in an area where transport costs are low can stabilize the price effect of a negative (positive) weather shock by trading with a prefecture that received a positive (negative) shock. This should de-link the relationship between prices and weather patterns in trading (i.e. low transport cost) regions to a greater degree than in non-trading (i.e. high transport cost) regions.

Rather than imposing a particular structure between cities, I use a resampling approach, confronting the model with many randomly drawn subsamples. The method has the advantage of allowing considerable flexibility in establishing the relationship between the variables of interest because it is not based upon a specific parametric model. It is also robust in the sense that it prevents a few isolated observations (outliers) from driving the central results.

Using resampling techniques, I show that among potential trade partners in the most integrated group, price correlation is significantly related to distance between partners, but not to current period supply shocks. In contrast, among inland prefectures, price correlation is significantly related to common supply shock patterns, but not to transport costs, Hence, integrated trading regions exist along rivers and coasts in geographical space next to regions resembling insular storage economies.

4. Vulnerability to Bad Harvest Shocks

Regions which could not trade cheaply may have relied upon autarkic methods such as storage to smooth fluctuations in harvest yields (Hsiao, 1960; Will and Wong, 1991). Although there is evidence that grain was stored in both public granaries and private households, the relative proportion of public versus private stores is unknown. It is also uncertain whether the amount stored was sufficient to deflect the consequences of crop failures.

The second part of the analysis therefore examines the effects of local harvest shocks on local market prices across all 121 prefectures. Regression methods allow us to examine the relationship between carryovers and supply shocks. As suggested by McCloskey and Nash (1984), if carryovers were large and routine, then this year's supply is unlikely to strongly impact the following year's prices; whereas if carryovers were unusual and sporadic, then the effects of a very good or very bad crop year can be expected to depress or raise prices in the following year. Regression analysis provides support for the notion that inland prefectures are in fact not vulnerable to supply fluctuations. Rather, these regions relied on inter-temporal arbitrage to a greater extent than river and coastal regions.

Furthermore, year-to-year prices are relatively more stable the further distant a region is from access to cheap transport route (less regionally integrated) while within-year prices are relatively more stable the more integrated a region is with a large trading network. Both of these findings are also consistent with the hypothesis that prefectures in the 18th century in China used primarily the method of consumption stabilization in which they have a comparative advantage: trade in river and coastal regions, and storage in inland regions.

5. Direct Evidence: Share of Private and Public Grain Storage

In the third part of the paper I derive an estimate of the share of private and government grain storage. The role of the state in stabilizing prices and helping regions overcome local disasters is potentially very important because the Qing government stored grain for the purpose of civilian famine relief. A priori it is possible that government response in the form of disaster relief, not market activities in arbitrage, was the primary mechanism responsible for the delinking of local prices from local supply shocks. The fact that the available evidence on storage levels in government civilian granaries does not correspond as strongly across regions and over time to variations in the degree of market integration as does the levels of storage implied by the indirect evidence from market prices suggests that significant private storage may have existed in addition to government storage.

The differential regional impact of government disaster relief is the subject in a separate paper (Shiue, 1998). While there are no direct records of private storage available, my estimates using historical accounts of famine relief policy indicates that the share of private grain storage in total storage was overwhelmingly greater than the government's share (see paper for derivation of estimate).

6. Conclusion

The paper assesses the degree to which markets for rice were inter-regionally and inter-temporally integrated in 121 prefectures of 10 provinces of southeast and central China in the 18th century. By doing so it focuses on the degree of market integration in China--and the functioning of markets in regions which were trading least with other regions during this period. Within the framework of a competitive equilibrium model for trade and storage, I demonstrate that there was a significant inter-regionally integrated market for rice already in the late 18th century. Furthermore, I show that regions which were not spatially trading with each other instead relied on inter-temporal forms of arbitrage and were therefore inter-temporally integrated. The differences across prefectures in these patterns of trade and storage are consistent with the predictions of a model based on competitive market incentives in the presence of transport and storage costs.

These results support the hypothesis that existing market failures were not sufficient to prevent arbitrage opportunities from being realized even in the relatively uncommercialized areas. Moreover, while it is generally agreed that the private actors predominated in the markets for inter-regional trade, preliminary estimates on actual storage suggests that the overwhelming share of stored grain was held in the hands of private agents, and not the government. The evidence points to the need to re-evaluate the source of modern economic growth in China in the 19th century and the functioning of private markets and institutions in the 18th century.


Elvin, Mark. 1973. The Pattern of the Chinese Past. Stanford: Stanford University Press.

Evans, Laurence. 1984. "Junks, Rice, and the Empire: Civil Logistics and the Mandate of Heaven." Historical Reflections (11)3: 271-313.Gongzhong zhupi zouzhe, nongye lei, liangjia qingdan (Grain Price Lists in the Agricultural Section of the Vermillion Rescripts in the Palace Archives). 1742-1795. Beijing: Number One Historical Archives.

Ho, Ping-ti. 1967. Studies on the Population of China, 1368-1953. Cambridge: Harvard University Press.

Hsiao, Kung-chuan. 1960. Rural China: Imperial Control in the Nineteenth Century. Seattle: University of Washington Press.

McCloskey, Donald, and Nash, John. 1984. "Corn at Interest: The Extent and Cost of Grain Storage in Medieval England." American Economic Review 74(1): 174-187.

Perkins, Dwight. 1969. Agricultural Development in China, 1368-1968. Chicago: Aldine Publishing Company.

Schran, Peter. 1978. "A Reassessment of Inland Communications in Late Ch'ing China." Ch'ing-shih wen-t'I 3(10):28-48.

Shiue, Carol. 1997. "Integration of Rice Markets in the Yongzheng Period: 1723-1742". University of Texas-Austin. Mimeograph.

Shiue, Carol. 1998. "Public Provisions for Disaster Relief in 18th Century China". University of Texas-Austin. Mimeograph.

State Meteorological Society. 1981. Zhongguo jin wubai nien hanlao fenbu tuji, [Collected maps of droughts and floods in China in the past five hundred years]. Beijing: Ditu chuban she.

Will, Pierre-'Etienne and Wong, R. Bin. 1991. Nourish the People. Ann Arbor: University of Michigan Press.

Williams, Jeffrey C., and Wright, Brian D. 1991. Storage and Commodity Markets. New York: Cambridge University Press.

Wu Chengming. 1985. Zhongguo ziben zhuyi yu guonei shichang (Capitalism in China and the National Market). Beijing: Zhongguo shehui chuban she.