CCP:GR-RI/04/2

Table of Contents

I. INTRODUCTION

II. FOOD CEREAL SITUATION IN CHINA

A. cereal food consumption might be lower than estimated by fao

B. WHEAT CONSUMPTION BELOW FAO/GIEWS AND DECLINING FAST

C. rice consumption below fao/giews data and exceeds wheat

D. maize consumption slightly higher than shown by fao/giews

III. FAO/GIEWS FEED USAGE PROBABLY UNDERESTIMATED

A. FEED REQUIREMENT ESTIMATES

B. feed availability estimates

IV. CEREAL STOCKS

A. GOVERNMENT STOCKS

B. on-farm stocks

V. IMPLICATIONS FOR CHINA AND WORLD

VI. CONCLUSIONS

I. INTRODUCTION

1. The purpose of this paper is to shed some light on the prevailing cereal market situation in China¹. The study considers alternative estimates for consumption and stocks in China to those currently reported by FAO/GIEWS². While the paper will discuss some of the implications of the continuing draw down of stocks in China for world markets, the main focus is on China itself³.

2. For more than two decades, cereal policies in China, the world’s most populous country, were largely aimed at increasing output by giving priority to self-sufficiency and this, over time, resulted in a large buildup of stocks – both on farm as well as in Government warehouses. Since 1999, however, a sizable contraction in production, associated with new stances in cereal policies, has been the main factor underlying a significant draw down in stocks.

3. In recent years it has become increasingly evident that the lack of accurate information, especially with respect to the level of stocks undermined not only the accuracy of FAO’s assessment of the situation and outlook in China but also the analysis of international trade and more importantly, of world food security. Calculating the “true” level of stocks in China has thus become an important concern.

4. Back in 2000, FAO confronted this problem when it recognized that the sheer scale of declines in stocks in China could only be explained if the absolute size of stocks was considerably larger than was carried in China’s cereal balances. In other words, this implied that FAO’s previous estimates for stocks in China were significantly under-estimated. This concern eventually led to a major revision of the FAO/GIEWS cereal balances for China, which ultimately resulted in a 10 fold upward adjustment in China’s cereal stocks and a doubling of FAO/GIEWS estimates of cereal stock at the world levels. FAO reported the revised estimates for the first time in the February 2001 issue of Food Outlook.

5. In recent years, China has been the main contributing factor underlying a steady decline in FAO/GIEWS estimate of world cereal stocks. Of the overall decrease of 220 million tonnes in global stocks between 1999 and 2003, China alone accounts for almost 80 percent. These substantial reductions have given rise to a number of concerns not only about available supplies in China but also regarding the real state of the world grain economy. Most importantly:

• Are stocks really falling, or are these reductions only a statistical illusion? If stocks have been falling so rapidly, how long such reductions can continue?
• What consequences will a continuation of recent developments have on both China’s and international cereal markets?
• Are current stocks in China an adequate buffer against a sudden or even medium-term shortfall in its domestic production?

6. To start addressing at least some of these concerns, FAO has put its cereal balance for China under scrutiny again and this paper was prepared as a first step. It should be noted that cereal production and trade data in FAO/GIEWS balances for China mostly reflect official figures. As a result, the focus of this study is on evaluating the accuracy of the current FAO/GIEWS estimates for cereal food consumption, animal feed use and stocks. The study draws from extensive research into official and unofficial data that compare and contrast those elements of the balance sheet.

II. FOOD CEREAL SITUATION IN CHINA

7. The primary cereal varieties in China include wheat, paddy rice, maize, barley, rye, oats, millet and sorghum. In China, cereals are referred to as grains, but the “grain” aggregate also includes soybeans and roots crops (mainly sweet potatoes). Wheat, rice and maize constitute the three main cereal crops in China since, based on the official statistics, their production amounts to roughly 97 percent of the total cereal production (or 85 percent of the total grain production – using the national grain definition) ⁴. Wheat and rice are mainly used for human consumption. Generally speaking, wheat and wheat-related products are the staple food in the North while the population in the South consumes predominantly rice. By contrast, maize is used primarily as feed, except in some of the poorest regions, where it can be an important food.

8. Only more accurate estimates on cereal utilization in China could settle the debate about whether the recent sharp drops in cereal inventories have actually taken place or not. The scale of the data uncertainties related to the level of cereal stocks in China makes a strong case for a closer scrutiny of the available statistics on food consumption and feed uses.

A. CEREAL FOOD CONSUMPTION MIGHT BE LOWER THAN ESTIMATED BY FAO

9. For a comprehensive examination of food consumption over time, regular surveys with large sample sizes and consistent methodology are needed. The Chinese National Bureau of Statistics (NBS), formerly the State Statistical Bureau (SSB), conducts such kinds of surveys for rural as well as urban households. The General Team of Rural Survey of NBS is responsible for the rural household survey. Detailed results are published regularly in “China Rural Household Survey Yearbook” and the aggregate results are reported in the “China Statistical Yearbook”. For rural food consumption, NBS reports data for wheat (unprocessed) and rice (paddy rice, unprocessed) from 1981 to 2001 and, on occasional instances, also for maize (for example 1993-1996, 2001). NBS also carries out urban consumption surveys, although these are based on expenditure as opposed to consumption, and publishes estimates for wheat and rice related products (Yearbook of China Prices and Survey of Urban Household Expenses). Unlike in the rural survey, urban wheat consumption is expressed in flour equivalent terms and includes standard flour, bread and other manufactured products. Detailed data for rice exist for the period 1992 to 2001, expressed in milled terms.

10. Overall, many factors seem to have contributed to reductions in grain-based food consumption in China. With the rise in purchasing power, eating habits have significantly diversified since the mid-1980s. According to the official statistics, between 1980 and 1990, the share of rural population declined from 81 percent to 74 percent, while between 1991 and 2001, with urbanization accelerating, this share dropped even faster, from 73 percent to 62 percent. Indeed, the phenomenon of large-scale urban migration seems to have affected overall food consumption patterns since the beginning of the 1990s, as evidenced also in the annual surveys by the NBS.

11. However, while NBS provides detailed data for consumption, these may not fully be consistent with those used in FAO/GIEWS cereal balances⁵. One important difference is that FAO/GIEWS utilization estimates correspond to “apparent” consumption, whereas, NBS numbers tend to reflect the “true” level of cereal consumption. In order to reconcile the data, therefore, some adjustment is in order. Attempts have been made to adjust NBS consumption numbers to represent more accurate consumption levels using various coefficients.⁶ These are reported in Table 1.

Table 1: Estimated coefficients used to adjust household grain consumption levels (based on NBS surveys)

Sources: Based on unofficial estimates, see footnote 6 for details.

B. WHEAT CONSUMPTION BELOW FAO/GIEWS AND DECLINING FAST

12. Based on the historical review of NBS consumption data, China’s wheat per capita consumption of rural population increased substantially between 1987 and 1993: from 76 kg/p.a. to 85 kg/p.a. However, it started to decrease after 1993 and, by 2001, wheat consumption approached the levels of the mid-1980s. In fact, per capita wheat consumption by urban residents had started declining already by the mid-1980s: from 61 kg/p.a. in 1987 to roughly 40 kg/p.a. in 2001 (in unprocessed terms), with the sharpest decrease occurring between 1993 and 1995. The sharp drop in consumption in urban areas has been triggered, in part, by a notable shift away from grain-based food to meat-based products. Overall, the increase in grain selling prices (since 1993) also negatively affected consumption levels across the country.

13. In order to compare the NBS reported survey consumption data and FAO/GIEWS apparent consumption estimates, the NBS data have been adjusted upward, using the coefficients (percentages) shown in Table 1⁷. Results are summarized in Table 2. Using NBS consumption estimates for rural and urban areas (weighted by their corresponding population shares), the adjusted NBS consumption numbers for the total population in recent years would be 11 million tonnes lower than those shown by FAO/GIEWS, also reflecting some 8 kg difference in per capita levels. In addition, the pace of decline in NBS wheat consumption seems faster than indicated by FAO/GIEWS figures.

Table 2: Wheat Food Consumption

*Adjusted NBS consumption data – applying the upward adjustment coefficients listed in Table 1.

C. RICE CONSUMPTION BELOW FAO/GIEWS DATA AND EXCEEDS WHEAT

14. The NBS rice consumption numbers, at both rural and urban levels, demonstrate significant declines since the mid-1980s. In rural areas, rice consumption fell from roughly 90 kg/p.a. in 1987 to 84 kg/p.a. in 2001; and for the urban population, from 57 kg/p.a to 45 kg/p.a. In spite of the declines, however, it appears that rice consumption has not experienced as fast a decrease as wheat. One explanation may be that for China as a whole, consumers maintain their preference for rice over wheat. While total per capita consumption of rice (milled terms) declined to nearly 70 kg/p.a. in recent years, it still outweighs wheat consumption by about 10 kg.

15. Nonetheless, as in the case of wheat, the adjusted NBS rice consumption data would still fall short of those estimated by FAO/GIEWS. As shown in Table 3, between 1995 and 2002, the adjusted NBS numbers would be some 3 to 15 million tonnes, or at least 10 kg per capita per year, lower than consumption figures estimated by FAO/GIEWS.

Table 3: Rice food consumption

* Adjusted NBS, using the coefficients shown in Table 1

D. MAIZE CONSUMPTION SLIGHTLY HIGHER THAN SHOWN BY FAO/GIEWS

16. Even though maize is mainly used for animal feed, many people in China consume maize as food, especially in the poor and remote rural areas. Maize consumption by rural habitants in 2001 averaged around 18kg/p.a. (China Rural Household Survey Yearbook 2002) while urban residents consumed less than 3 kg/p.a. (Yearbook of China Prices and Survey of Urban Household Expenses 2002). After the adjustments to the NBS numbers, average per capita maize consumption reaches 14-17 kg/p.a. or about 4-6 kg/p.a. higher than FAO estimates – Table 4.

Table 4: Maize food consumption

* Adjusted NBS, using the coefficients shown in Table 1
 

III. FAO/GIEWS FEED USAGE PROBABLY UNDERESTIMATED

17. FAO/GIEWS estimates of total cereal feed use hovers around 100 million tonnes in the mid-1990s, rising to roughly 110 million tonnes in more recent years. Among the cereal varieties in China, coarse grains account for 86-89 percent (of which, maize alone 83-84 percent), rice 8-9 percent and wheat 4-6 percent of total feed use. Estimating the feed use of grains is even more complex than calculating China’s food use⁸. The difficulty in evaluating feed grain usage arises from several factors, including 1) the different types of animal rearing that exist in China; 2) the uncertainty regarding the conversion rates, especially for backyard operations; and 3) the suspected overestimation of official meat production statistics.

18. It is important to underline that a comparison between meat consumption and production data already point to large discrepancies, with consumption much lower than official production data. Many experts have analysed this issue and considered the discrepancies to have originated from the practice of over-reporting meat production⁹. The first Chinese Agricultural Census (conducted in 1997) provides some indications. The Census results indicate that the numbers of slaughtered animals previously reported in official statistics were overestimated by a factor of 25-30 percent, thus implying production numbers were overestimated by a similar factor. Such evidence convinced the former SSB to revise down their statistics on meat production as of 1996. Unfortunately, overstated meat production for all the years prior to 1996 has yet to be corrected by the statistical authorities.

19. Broadly speaking, there are two approaches to estimate feed grain utilization. One is the “supply” or “availability” approach (i.e. the feed use = total grain production-food use-seed use-waste and other uses). Another is the “utilization” approach (i.e. the feed use = meat production multiplied by corresponding overall feed conversion factor, which are intended to reflect both the intensity and the technical efficiency of the livestock system). Therefore, the first calculation stems from the feed supply side, the second from the feed demand side. The ideal situation is when the feed supply equals to the feed requirements (assuming that the meat production data is accurate).

A. FEED REQUIREMENT ESTIMATES

20. Notwithstanding the difficulties in obtaining more accurate animal production numbers and determining the appropriate production distribution among the various livestock feeding practices, the general view shared by researchers is that at least 70 percent of total pork production is produced by China’s backyard feeding operations (less than 5 pigs). For the backyard feeding operations, the grain share in the feed composition is generally much lower than for other feeding operations (roughly 20 percent below the specialized producers): the small pig-raisers use less grain, but more bran (from wheat or rice processing) and green fodder.

21. In recent years, researchers and livestock specialists have made various attempts at determining more suitable animal conversion rates in order to improve the estimation of feed requirements in China. Different conversion rates lead to different feed requirement levels. Unfortunately, a detailed review of different conversions and the methodologies applied are beyond the scope of this paper. However, the findings point to some interesting observations. It seems that, on average, grains account for more than 70 percent of the fine feed fed to pigs and between 60 to 70 percent for hens and broilers (table 5). Based on these preliminary results, three different feed requirement scenarios could be considered for the period 1996 to 2001, which are summarized in Table 6. Under the low-level scenario case, the total feed grain estimates range from 167 to 186 million tonnes; 202-216 million tonnes in the middle scenario; and as high as 210-233 million tonnes in the high scenario.

Table 5: Grain conversion factor by product and by feeding type, 1997-2001

Note: The conversion ratios (grains/ live weight gain) are calculated from “Compilation of Cost and Benefit Data for Agricultural Products in China”, 1997-2002.

Table 6: Feed grain estimation in China, 1996-2001 (million tonnes)

B. FEED AVAILABILITY ESTIMATES

22. How much grains are available for livestock production? One major difficulty in comparing FAO/GIEWS feed supply numbers with those derived from national sources is in the Chinese definition of feed grains. In order to provide a comparison, at least for years when production numbers deem more accurate, i.e. after 1996, attempts have been made to adjust the FAO/GIEWS feed grain estimates to make them more consistent with the Chinese definition of feed grains (i.e. this is contrary to the approach used for adjusting/comparing the food estimates discussed earlier).

23. Therefore, the FAO/GIEWS feed estimates were raised by about 20 million tonnes for roots crops (potatoes), by 20 million tonnes for post retail waste of grain (assuming that wastage is totally re-utilized as feed), and by around 10-30 million tonnes for rice and wheat (considered as over-estimated human consumption). Based on the FAO/GIEWS adjusted numbers, total amount of grains used as feed would be in the range of 150-180 million tonnes, which would be closer to the grain feed requirements under the low scenario (Table 6).

24. To put the FAO/GIEWS adjusted numbers in perspective, some studies indicate that during the late 1990s, out of the grain available for feed, Chinese farmers contributed about 70-100 million tonnes and the milling enterprises provided another 40-50 million tonnes (in fact, the compound feed production rose from 56 million tonnes in 1996 to 74 million tonnes in 2000). In addition, feed use by larger scale livestock producers (and milling companies) need also to be taken into consideration since they purchase feed ingredients and produce their own compound feed (the National Feed Association does not appear to record this part of feed uses).

25. Higher feed use is also supported by other studies, which confirm the fact that farmers tend to use more paddy and wheat for animal feed at times when prevailing market prices are not favourable. For example, a field survey of 60 rural households in the province of Sichuan showed that almost 20 percent of the recorded human consumption of rice was actually used as feed. This is compared to a range of 10-15 percent estimated for China (Yang, 1999, pp.198-199).

IV. CEREAL STOCKS

26. In China, cereal stocks are generally classified as Government or on-farm stocks. China does not publish stock numbers and most information about stocks comes from unofficial sources, and is often limited to national totals, with no information on the volume of stocks by single cereal types.

A. GOVERNMENT STOCKS

27. Broadly speaking, government stocks fall into 3 categories: State (central government) Special Grain Reserves (SAGR), established in 1990; local government stocks; and “business” stocks held by the State Owned Enterprises (SOE). Most analyses confirm that total Government stock levels were exceptionally high during the 1990s, particularly in 1999, when China harvested a record crop.

28. While a detailed review of the literature on this subject is beyond the scope of this paper, references are made to some of the more recent studies, all pointing to large scale stock holding practices by the Government. One important study (Liu Yunzhi, 2000) showed that Government stocks had reached 130-140 million tonnes by the mid-1990s. In one detailed presentation on the state of grains stocks (Chen, 2001), the level of government stocks in 2000 was estimated at nearly 250 million tonnes. Another study (Li Chenggui, 2001) showed government stocks at around 220 million tonnes in 2002, with the SAGR holding at about 50 million tonnes (in July 2001, the State Council raised the SAGR to 75 million tonnes).

29. Table 7 presents the estimates for total Government cereal stocks between 1995 and 2002. The figures listed in this table reflect derived estimates based on various sources. However, in the absence of any official confirmation of the “real” size of Government stocks, these estimates only provide a weak indication of the possible size of stocks under some sort of direct Government control. For example, by the end of 2001/02 season, stocks held by the Government were put at 204 million tonnes, almost double the estimated on-farm levels for the same period.

Table 7: Ending season stocks – adjusted NBS numbers and FAO/GIEWS

B. ON-FARM STOCKS

30. As for on-farm stocks, there are two sources for information in China. One is the Research Center for Rural Economy (RCRE) of the Ministry of Agriculture (MoA). Another is the NBS (General Team of Rural Survey). In their surveys, both institutions use similar questionnaires. However, the RCRE survey, which shows lower stocks than the NBS, is based on a much shorter time span (from 1995 to 2001 as compared to 1981-2002 by NBS) and a much smaller sample size (15000 households compared to 67000 used by NBS). Consequently, the NBS surveys are regarded as more informative for comparison with FAO/GIEWS data.

31. Again, because of several important methodological differences about estimating stock levels, FAO/GIEWS and NBS stock data are not compatible. Above all, the NBS uses the concept of “year-end stocks” and in China, the grain statistical year is from April to March, which is different from its grain marketing year (or cropping year). For example, wheat harvest is mainly in June, early rice in late July, and maize at the end of September or early October (Northeast in November). In FAO Balance Sheets, stocks are carryovers only, which are defined as all cereals still available for domestic utilization on the first day of the new marketing year, hence the term carryover.

32. Therefore, to make the NBS on-farm grain stock data comparable with FAO data, some adjustments are necessary. From the beginning of the year to the following crop harvesting season, farmers need to use their stocks for various uses, including a portion for their own consumption, some for feed usage (farmer themselves often raise 1 to 3 pigs and 10-20 chickens or ducks during the year) and some for sales to the market¹⁰.

33. It is assumed that, on average, the time from the beginning of the calendar year to the next crop harvest is about 7 months (or roughly 60% of total consumption in any particular season, i.e. farmers still have to cover about 7 months of their food and feed requirements from the supplies held by the beginning of the year)¹¹. As for grain sales by farmers in this period, some additional explanations are called for. The NBS survey collects information on annual total sales and quota sales. Normally, farmers deliver the grains quota to the state after the crop harvest, mainly between August and November, while sales to the market usually takes place before and after the Chinese New Year, which usually falls in January/February. Thus, the total amount of sales by farmers needs to be deducted from the carryovers into the following season.

34. After deducting the farmer’s own food, feed use and sales from opening stocks, per capita carryovers and total carryovers for every year can be estimated. Based on this, total cereal carryover stocks (on-farm) were estimated to have reached their highest level in 1999, at around 210 million tonnes (with rice in milled equivalent).

35. Against these estimations, it seems that by end of crops seasons in 2002, cereal carryover stocks (including both government and on-farm) in China was still relatively high, of the order of 309 million tonnes. Adding government stocks to on-farm carry-over stocks, aggregate stock levels in the mid-1990s amounted to 300 million tonnes and by 1999, stocks could have hit a record, at some 469 million tonnes. Over the past few years, however, with consecutive declines in cereal output, stocks have also fallen, perhaps as much as 160 million tonnes between 1999 and 2002. Overall, as shown in Table 7, the adjusted stock estimates would still be well above FAO’s.

V. IMPLICATIONS FOR CHINA AND WORLD

36. Table 8 reflects the current FAO/GIEWS cereal balance for China from 1995/96 through 2001/02 (based on available information as of September 2003). Table 9 shows the balance after adjusting the food and feed use along the lines presented in the previous section. Since the changes basically entailed a redistribution of cereal utilization between food and feed, they do not substantially affect total usage and, hence, alter only marginally the estimates for carryover stocks¹². Thus, the resulting stock estimates would still remain well below those derived in the earlier section from national sources, as shown in Table 7. While this may indicate that FAO/GIEWS stocks are still underestimated, the Secretariat has refrained from raising current FAO/GIEWS estimates until a more in-depth review of all the other main variables in China’s cereal balances, including production statistics, is undertaken.

Table 8: Current FAO/GIEWS Cereal Balance*

* As of September 2003
 

Table 9: Alternative FAO/GIEWS Cereal Balance based on adjusted utilization data

37. Despite a widening gap between production and utilization (Chart 1) for the previous 6 years and up until the start of the current season (2003/04), China’s grain trade pattern (such as continued cereal exports) and prices have not signalled any impending shortages or supply tightness in China. While the existence of a large stock buffer probably contributed to the stability of the market, it is legitimate to question the sustainability of the stock draw downs in coming years. Based on FAO/GIEWS estimates, the stock in 2001/02 covered around 70 percent of total utilization (or 80 percent using stocks estimates derived from national sources). This share was forecast to plunge to only 37 percent by the end of the current marketing season in 2004.

38. Indeed, if stocks in China have dwindled as rapidly as it now appears, the chances for China to emerge as a significant cereal importer must have also increased enormously. Chart 1 and 2 further illustrate that if we were to trend-out cereal production and utilization to 2008, the gap between production and utilization would continue to widen.

39. Assuming no exports and imports, if China were to solely use its cereal stocks to close this gap, then China’s stocks could fall to almost nil by 2007. Even so, this situation may still not occur since China can turn to international markets to fill the gap and/or change its domestic production policy and raise cereal output.

40. Nonetheless, questions pertaining to how much China will import and when, still much depend on the actual level of its stocks. In terms of individual cereals, China has recently started to export wheat, but the prospects for it to remain a large wheat exporter are weak. Instead, China’s growing appetite for high quality wheat and constraints associated with larger domestic wheat production could eventually result in the country becoming a net-importer. As feed demand is increasing rapidly and, unless domestic maize production expands much faster in the coming years, China could also emerge as a net importer of maize. In recent years, China has joined the league of the world’s leading maize supplier, but future exports would certainly decrease. International prices would not be neutral if China were to reduce or curtail its maize exports and would increase even more sharply if it turned to being a sizable importer. As for rice, China is still a sizable net rice exporter but demand for high quality rice is on the rise. Thus, it seems unlikely that China can sustain the current level of low rice imports and high exports unless it engages in more supportive production policies.

VI. CONCLUSIONS

41. Much of the drawdown in world stocks has been due to a sharp decline in China’s cereal inventories. In spite of its discernible statistical shortcomings, the findings presented in this paper support the view that stocks in China must have fallen substantially in recent years. However, it remains a difficult task, if not an impossible one, to substantiate what may or may not be the “true” level of stocks in China. This is an important limitation since as long as FAO does not have reliable and consistent information about the overall level of stocks in China, the Organisation’s assessment of the global cereal market situation will not be fully reliable.

42. The Group might therefore wish to encourage the Secretariat to explore the possibility of improving further these estimates in collaboration with the interested parties. The Group may also want to keep under scrutiny the evolution of the global stock situation, and recommend a close follow up of stock developments also in other countries. Moreover, it might wish to call for a more active role by FAO in informing the Member countries of available risk management strategies that can be used in the event of unexpected large variations in international cereal prices.

________________________________

¹ Excluding Taiwan Province of China, Hong Kong SAR and Macao SAR.

² Two sets of cereal supply and demand balances are maintained in FAO, which are, by and large similar: one set of balances are in public domain as part of FAOSTAT (Food Balance Sheets); another set is used for the purpose of Global Information and Early Warning System (GIEWS) analysis. The discussions in this paper refer solely to the latter since cereal balances in GIEWS are based on marketing seasons and provide estimates for carryover stocks; the balances in FAOSTAT are constructed on calendar year basis and contain only changes in stocks.

³ The Secretariat gratefully acknowledges research contributions made in support of this paper by Professor Xiande Li, Chinese Academy of Agricultural Sciences.

⁴ In the government official statistics, production and yield are recorded in a systematic way only for paddy rice, wheat, maize, soybean and potatoes (in starch equivalent).

⁵ The data are not always compatible with those reported by FAO at the national per capita levels. Broadly speaking there are at least three problems with NBS data: 1) rural grain consumption is the actual quantity consumed, while urban grain consumption data is the quantity bought; 2) rural consumption data are in unprocessed terms (i.e. paddy rice) but urban consumption data are in processed terms (wheat flour, milled rice); and 3) the quantity of each grain variety in rural or urban grain consumption is not available.

⁶ The method applied in calculating these coefficients is based on several research findings. One study (Lu) examined the out-of-home consumption of meat, egg and seafood products and estimated that the quantity consumed of those products in the out-of-home occasions accounted for 8 percent of the total consumption for urban dwellers and 6 percent for rural habitants. It is generally agreed that people tend to consume more animal products and less grain in restaurant meals than in an ordinary home meal. In the absence of any empirical estimates, it is also assumed that the waste rate of 4 percent is constant over time (1987 to 2002) and the same for both the rural as well as the urban sectors. However, urban estimates are higher than rural ones, not least because they are based on expenditure rather than food consumption data, they are also more likely to have increased in line with the strong rise in income.

⁷ Because of limited space, consumption data based on NBS surveys are not presented in this paper.

⁸ In China, feeds could be generally classified into fine feed, gross feed, green feed, animal protein feed. Fine feed (jing siliao) includes grains (cereals+ potatoes); cakes (soybean meals, cotton meals, rapeseed cakes, sesame meals, etc.); bran (rice bran, wheat bran, etc.). They constitute almost the totality of the ingredients of compound feed. Gross feed (chu siliao) includes straws (rice straw, wheat straw, maize straw, etc.); tigers of potatoes. Green feed (qing lu siliao) includes green manure, vegetable leaves, tree leaves, water grass, etc. Animal protein feed: fish powder, silkworm chrysalis, earthworm, etc. (Cf. Editing Team of the Livestock Sector in Modern China, 1991).

⁹ See Aubert (2000), Colby et al., (1999), Fuller et al. (2000), Lu (1998) and Zhong (1997).

¹⁰ For example, Zhang Xiaohui decomposed 2001 ending stock levels. Of the 1305 kg stocks (or equivalent to 58% of production), he identified 50% for human consumption, 20% for seeds and feeds, and the remaining 490kg for risk reserve - or equivalent nearly 6 months of ration for human consumption (Zhang, 2002).

¹¹ In North China (wheat growing region), farmers harvest wheat in May and June (they sell normally several weeks late) and farmers in the South (rice growing region) begin to harvest the early rice in July (they sell normally several weeks late, say in August).

¹² For other usages of cereals such as seed and industrial use, the difference between FAO/GIEWS estimates and national data are relatively small.