Are dragon-head companies heading agricultural development in China? The case of apple chains.

P. Moustier;

CIRAD, MOISA, , Corresponding author email: [email protected] Abstract: The purpose of this paper to empirically document the role and impact of agri-business development in the case of apple chains in China, and to investigate the role that the provision of services and marketing plays in this impact. We compare some performance indicators, including yields and profits between farmers belonging to chains with and without dragon-head companies, i.e. companies which invested in various infrastructures, and provide services to farmers including extension services, storage and input provisioning. The data is based on a survey of 355 apple farmers in Shandong and Shaanxi provinces. Dragon-head companies have no significant impact on the yield and profit obtained by farmers, in contrast with belonging to a . They do not have a significant impact on training relative to . They do not propose higher prices for farmers, even though farmers’ practices may be more eco-friendly and reduce the use of chemical pesticides. The conclusion is that the optimal conditions for agri-business development to be conducive to economic development as shown by the literature do not seem to be gathered in the apple sector of China. Acknowledegment: The research is part of the “Sustainapple” programme, coordinated by Jean-Marie Codron (INRA, France). It receives funding from the French National Research Agency (ANR). JEL Codes: O33, Q

#2066

Are dragon-head companies heading agricultural development in China? The case of apple chains.

Abstract

The purpose of this paper to empirically document the role and impact of agri-business development in the case of apple chains in China, and to investigate the role that the provision of services and marketing plays in this impact. We compare some performance indicators, including yields and profits between farmers belonging to chains with and without dragon-head companies, i.e. companies which invested in various infrastructures, and provide services to farmers including extension services, storage and input provisioning. The data is based on a survey of 355 apple farmers in Shandong and Shaanxi provinces. Dragon-head companies have no significant impact on the yield and profit obtained by farmers, in contrast with belonging to a cooperative. They do not have a significant impact on training relative to cooperatives. They do not propose higher prices for farmers, even though farmers’ practices may be more eco-friendly and reduce the use of chemical pesticides. The conclusion is that the optimal conditions for agri-business development to be conducive to economic development as shown by the literature do not seem to be gathered in the apple sector of China.

Key words Agri-business, China, apples, eco-friendly practices, supermarkets, governance, chains.

Introduction

Objective

It is commonly acknowledged that agri-business development and agro-industrialization are key for rural development, income generation and competitiveness (Swinnen and Maertens, 2009). Agri- business development is indeed actively promoted in China, as it is in most emerging and developing countries. Yet its impact on economic development and rural transformations is still the subject of debates and little documented by quantitative empirical data.

Agribusiness development is considered to combine private investments in infrastructures and technology with the development of contractual arrangements (Cook and Chaddad, 2000; Reardon, 2016). It is promoted by governments and donors in developing and emerging economies. Yet its spreading in as well as Africa is still limited. It is estimated that contracts represent less than five per cent of agricultural transactions in Asia and Africa (Minot and Sawyer, 2016).

Reviews of the impact of contractual arrangements on farmers’ economic results show a positive effect (Ton et al., 2018; Minot and Sawyer, 2016; Wang et al., 2014; Miyata et al., 2009. Yet, there may be a bias in publication in favour of positive impacts, and studies mostly measure short-term effects, in the first years of the setting of the contractual schemes. Besides, the level of positive impact depend on the type of services supplied by purchasers, and on the price premium that the purchaser can provide (Ton et al, 2018).

1

The purpose of this paper to empirically document the role and impact of agri-business development in the case of apple chains in China, and to investigate the role that the provision of services and marketing plays in this impact.

Context As poverty is a primary characteristic of the rural sector, any opportunity to raise farmer incomes is of major interest to government agencies and academics in China, including participation in modern, high-value food chains, even though empirical evidence is scarce (Wang et al., 2009). Our empirical case is the apple sector. China is number one in production and export of apples. Apple production developed quickly from the 1980s for both the export and domestic markets. The apple sector is considered as a brilliant example of how a myriad of small-scale farmers was successfully connected to various types of consumers, including affluent foreign consumers (Zhang et al., 2009). In China, poor farmers have a role to play in the development of the horticultural sector (Wang et al., 2009). Yet, many challenges are still ahead. Food safety is a particular issue because farmers have to deal with various insect and fungus attacks. In China, production is small-scale (0.5 ha on average for apples, 0.6 ha on average for agriculture) and trade is characterised by long chains of market intermediaries (including collectors, rural and urban wholesalers and retailers), which makes control of food safety especially complex. On the other hand, apple farmers have benefited from a fairly positive enabling environment. From 2001 onwards, in the framework of the Pollution-free Food Action Plan, various experiments and training sessions were conducted relating to the use of organic fertilizers and the biological control of pests, especially in Shandong Province (Zhang et al., 2009). In 1993, the practice of bagging apples was introduced, which protects the fruits from insects and fungi and reduces pesticide residues in the fruits. The number of private storage facilities also developed quickly. The majority of farmers grow the Fuji variety. ` From the early 2000s onwards, the Chinese government has been promoting agro-industrial enterprises, called ―dragon-head-driven‖ companies, for better coordination between individual farmers and the midstream and downstream segments through contract farming (Guo et al., 2007). These agri-business companies select farmers and provide them with seed, fertilizer and other inputs, as well as technical expertise. In exchange, the farmers sell their produce to these companies, giving them assurance of quality and quantity (World Bank, 2006). Farmer Specialized Associations (FSAs) and Farmer Specialized Cooperatives (FSCs) began to be established in the late 1980s. They were promoted in 2004 by the ―Law on Farmer Cooperative Economic Organizations‖. ―Farmer Professional Cooperatives‖ (FPCs) were recognized by the law in 2006. Since 2007, farmer cooperatives and organisations have been promoted by the government for collective action, increased bargaining power and a stronger farm-business linkage in both production and marketing. A survey of 189 FPCs showed that 18% of them reported that buyers insisted upon food safety and that they are more involved in ―modern‖ channels than others (including supermarkets and exports) (Jia et al., 2012). A large number of these initiatives were launched at the local level with the establishment of consolidated and aggregated production on a large scale, called Production Bases (PB). A production base is a special form of farm organization that coordinates the choice and timing of crops planted and utilizes uniform production technologies, farm management and marketing arrangements within a given location. The location, in turn, may consist of the farms in all villagers (or a subset thereof) from one or more villages or a land area leased by an outside agri-business firm. In the former case, the production base is run by the villagers themselves in a contract farming or out-grower scheme. In the latter case the base is run by the outside firm using either hired labour or through a sub-leasing or sharecropping arrangement with local residents or migrant farmers. In the meantime, market liberalisation through opening up to foreign investors further promotes the modernisation of agri-food systems.Large international retailers were the first foreign supermarkets established on Chinese soil. FDI flows to the supermarket sector are expected not only to increase the efficiency and competitiveness of the agri-food sector by bringing in another set of private players, but also to provide domestic actors with access to advanced retailer and supplier logistics and

2 management technologies from more experienced foreign retailers. Through home-office guidance for local branches of global chains, knowledge transfers along with imitation and innovation by domestic retailers, advanced logistics technology and inventory management methodologies have entered the country (Reardon et al. 2003). Seeing its potential to enhance food safety and improve supply-chain efficiency, the Chinese government launched initiatives with the intention of directly linking retailers and producers (Gale and Hu, 2012). At the end of 2008, the Chinese Ministries of Commerce and Agriculture jointly issued an ―Announcement to Initiate Direct Farm Pilot Programmes‖, encouraging supermarkets, agricultural enterprises and farmer cooperatives to engage in direct sales relationships. Adapting themselves to the Chinese business environment, large international retailers obtained experience and started to rely on market intermediaries to overcome the barriers to establish direct procurement from individual farmers or even farming communities. These intermediaries were heterogeneous, ranging from large, specialized service companies with significant capital investment and national coverage to more local companies that managed relationships with just one or two farm communities (Michelson et al., 2015).The combination of vendor consolidation, purchasing concentration and centralised distribution has led to significant changes in procurement of fresh produce by retailers, making it more flexible and contract-oriented.

Methodology Descriptive and multivariate analysis We investigate the relationship between the governance of the chain, in particular the presence of dragon-head companies, and the results for farmers in terms of the following indicators: yields, prices and profits. We also consider the impact of the governance on the changes in production practices, proxied by training, ecological practices and knowledge of pesticides. In terms of governance, we consider the nature of intermediaries between farmers and final retailers, in particular the presence of farm bases, dragon-head companies and supermarkets.

In estimating the impacts of dragon-head company on individual farmer training, our first model is:

Trainingijk = a0 +휃GOVij+ 훽HHij+µijk (1)

Trainingijk are the outcome variables and it includes two measurements that measure the training that farm i in farm base j got the long time training (k=1) and training in the field (k=2) during the last three years. Trainingij1 is a binary variable, which equals to 1 when the farm i got the training more than one time, 0 other-wise.Trainingij2 is a binary variable as well which equals to 1 when the farm got the training in the field, 0 other-wise. In estimating the impacts of dragon-head company on individual farmer yield ,net profit and weighted price, our second and third model are:

Productivityijq =a0 +휃GOVij+ 훽HHij+훿 INPUTijq +ρOTHERijq+µijq (2)

Priceij = a0 +휃GOVij+ 훽HHij +µij (3)

Productivityijq are the outcome variables for farmer’s production performance. It also includes two measurements, i.e. the apple yield per hectare (q=1) , the net profit per hectare (q=2). Priceij is one outcome variable that measures the weighted price farmers obtained during apple marketing.

Besides training, productivity and price, we also specify outcome variables for eco-friendly practices, (EcoPracticeij), and farms’ knowledge (Scoreij). EcoPracticeij is a binary outcome variable 3 representing the eco-friendly practices which equals to 1 when the farmer adopted the ecological practice, 0 other wise. Scoreij is the outcome variable that measure the knowledge of farm i in farm base j of the suitability of pesticides suitable for certain diseases and pests and it is scored against a full mark of 100.

EcoPracticeij = a0 +휃 GOVij+ 훽 HHij+µij (4)

Scoreij= a0 +휃 GOVij+ 훽 HHij+µij (5)

On the right side of our five regressions, GOVij is a set of dummy variables with a value of 1 or 0 that indicates the farm base’s governance chain. To compare with different governance schemes, we include SPM+Cooperative, SPM+Cooperative+Dragon, Cooperative, Cooperative+Dragon. Farmers that did not join any farm base are used as the reference (omitted) category in all the regressions. The coefficient of vector θ thus denotes the marginal difference in the various outcomes for different governance chain and those not, holding all else constant.

Other control variables include the vector HHij reflecting household characteristics, such as Farm size, measured in hectares per household; Farming experience of household head, measured in years; Education dummy measures the surveyed farmer obtaining formal education of more than nine years which equals to 1, 0 other-wise; Shandong province dummy variable to control for non-time varying unobservable regional differences. Moreover, in the model (2) we also include the input vector Input ij (such as labor, fertilizer, material fee) and vector OTHERijq (including age of apple tree measured in years and disaster dummy variable )to control the yield and profit different because of input use, fruit- age and disaster. We estimate the models by using Ordinary Least Square regressions (OLS) and LOGIT regressions. Data collection Throughout 2013 and 2014, the research team investigated the apple chain of two international retailers from downstream to upstream. For each of the retailers, the chief fresh produce procurement manager was interviewed about the general procurement of fresh produce. Then we asked the manager to provide us with a full list of their fresh apple suppliers. Through coordination by the manager and the department of procurement, we contacted all the suppliers and inquired about their procurement of apples from farm bases (FBs). The term of farm bases (called sheng-chan-ji-di or ji-di in Chinese) has been widely used in both scholarly documents and public narratives about agribusiness in China but is often vaguely defined. A farm base is considered here as a special form of organizing production and farm management within an identifiable and potentially traceable location, typically a village or contiguous group of villages (Ding et al., 2015). It can be managed by either the villagers themselves in an out-grower scheme or by outside firms using wage labor or sub-leasing contracts. For the purposes of this study, a farm base (FB) is defined by two criteria. First, it has an identifiable and potentially traceable location, typically a village or contiguous group of villages. Second, the vendor establishes a contractual relationship with the farming community or invested on its own. He participates in production via pre-planting planning and, in some cases, the provision of inputs, technical assistance and fixed investment. A face-to-face survey on farm bases was conducted in the field. We focused on farm bases in two provinces in China (Shandong and Shaanxi, Figure 1), where apple production accounts for 75% of the national portfolio. Two apple-dominated cities were selected from each study province. After the identification of the location of the farm bases (counties, townships, and villages), several counties were selected from each city. In total, the sample covered 8 counties in Shaanxi province (Figure 2) , and 5 counties in Shandong province (Figure 3). As part of the agreement with the retailers, the research team insisted on conducting an independent on-site survey with the farm base managers without the presence of other chain partners (such as vendors or supermarkets). During the survey, farm base managers were asked detailed information regarding the history, size, production,

4 marketing, farm management, decision-making and relationships between the farm base and the chain partners. We also surveyed farmers in FBs not supplying international supermarkets. We interviewed the officer(s) of the Agricultural Bureau at the county level and requested a list of cooperatives or production entities in the form of farm bases. After understanding the purpose of the study and the definition of farm bases, the officer shortlisted ideally three farm bases, from which we chose one that was not a neighbor of the chosen farm base supplying international retailers. It is unknown whether the additional identified farm bases supplied supermarkets or not. To verify this, an identical survey form of farm bases was used in a face-to-face interview. The farm base managers were surveyed about their sales to supermarkets in a previous year, the identity of the supermarkets and the marketing share for all the marketing channels. We also asked the farm base managers about their certification as compliant with any of the agricultural standards. The additional samples of farm bases allow us to mitigate estimation problems from possible selection bias related to the chain governance driven by the studied international supermarkets. A household survey was conducted on farmers who were members of the farm base sample and those who did not join any farm organization in the community. The farm base managers were asked to list the member farms; we randomly chose six of them. In the same village, we conducted a community survey by interviewing the village head and collected information about the village characteristics, its agricultural production and organizations. We also showed the village head the list of members provided to us by the farm base manager and, on the basis of this information, the village head helped to list apple farmers who did not participate in any farm organization in the village, from which we randomly chose four non-FB farmers. Finally, we constructed a hierarchical dataset consisting of 238 apple farmers who were identified as members by 42 farm bases—an average of six farmers for each of the farm bases —and 117 non-FB farmers in the same and neighboring communities. Among the 42 farm bases, 30 of them supplied supermarkets and the remaining 12 farm bases did not supply any supermarket either directly or through vendors. In total we had information for 355 farmers. We conducted a survey on the training obtained by farmers, on production practices (e.g., irrigation, apple bagging, fertilizer use, pesticide use, labor use, harvesting) and marketing of apples (e.g. the grade, the price of the grade, quantity of sales). As regards the assessment of training, we included two questions, the length of training and the training method (in-door or in the field). We focused on farmers obtaining the training (such as technology training) from the cooperative, farm bases or others. At the same time, we asked the farmers whether the training was only one time or the farmers can access to the training within the whole apple growth season. We defined the yield as total output (quantity) divided by sowing area and the weighted price as averaging the different prices relative to grades according to the percentage of each grade. At last, we defined the per-unit net profit as the total income (yield times weighted price) deducted the total cost (e.g. labor cost, material cost). We also conducted a survey of farmer knowledge regarding the suitability of pesticides. In each of the surveyed counties, the research team conducted a canvas survey by collaborating with local entomologists and extension experts in the field of plant protection. On the basis of their years of field experience, the local experts listed typical pest and disease problems and ten pesticides that were used widely in the local region. During the household survey, the enumerators presented the ten pesticides in the local area to the interviewed farmer, asking him to choose the disease(s) that the pesticide is supposed to control. On the basis of the answers, we generated the variable of farmer knowledge on pesticide suitability. Similar to the variables in the knowledge score of standards, we scaled the score of pesticide suitability into a full mark of 100 on the basis of the answers to the ten questions. As regards the assessment of ecological practices, we focused on farmer use of inputs (such as fertilizers and pesticides) and adoption of biological and ecological control for pests and diseases in the previous production season. For example, we asked whether farmers applied chemicals and composted manure in their production. For pesticide use, we included two variables, the number of applications and whether the farm applied pesticides within the fruit harvesting period. Lastly, we included a dummy variable that indicated whether the farm had used, during the previous harvest 5 season, any type of biological or physical pest control measure, such as a moth-killing lamp, pheromones, and natural enemies such as predatory mites.

Main results The agrifood chain governance To consolidate marketing and agricultural production, from upstream to the downstream end of the chain, FBs introduced various governance modes. The organizational forms of FB governance chain are quite complicated and heterogeneous. This complexity and heterogeneity is interesting as it suggests both that the dragon-head companies are playing multiple roles and that these roles are different depending on governance chain. In a first case, the supply chain consists of variable layers for supermarkets, dragon-head company, cooperatives that coordinate consolidated production, and farmers who are consolidated in FBs. Supermarkets, either international or domestic, have contracts with a number of dragon-head companies that have built suppliers’ networks for procuring fresh agricultural products. Then the dragon-head companies generally make some investment directly (such as cold storage, office equipment, packaging) or search for ideal regions where the quantity and quality of products meet with the minimum requirement of supermarkets. Very often, such an investment or searching process is intermediated by the local government who have strong interest in upgrading the local agri-food chain. One dragon-head company has multiple FBs. We call such a governance chain supermarket- 1 coordinated FBs with cooperative and dragon (SPM +Cooperative+dragon). In a second case, the supply chain consists of variable layers for supermarkets, cooperatives that coordinate consolidated production, and farmers who are consolidated in FBs. And the procurement is coordinated by supermarkets that purchase farm produces directly through farmers’ cooperatives. We call such a governance chain supermarket-coordinated FBs with cooperative (SPM+Cooperative). In a third case, FBs do not sell their apples to supermarkets, the supply chain consists of dragon-head company, cooperatives that coordinate consolidated production, and farmers who are consolidated in FBs. The most of dragon-head company role is the same as before, and it usually invests directly to set up its own cooperative or search for ideal cooperative where the quantity and quality of apples meet with its requirement. In the end the cooperative supply dragon-head company a certain proportion of produce. We call such a governance chain non supermarket-coordinated FBs with cooperative and dragon (Cooperative+dragon). The fourth type of chain is non-supermarket-coordinated FBs with cooperative (Cooperative). All these chains will be compared with chains when farmers are not organized into farm bases, they usually supply collectors and wholesale markets. The majority of farm bases belongs to the SPM+Cooperative+Dragon type. As shown in Table 1,for the 42 farm bases, 30 farm bases supply supermarkets, and 19 farm bases belong to the SPM+Cooperative+Dragon FBs (45%=19/42). Meanwhile the percentage of total farmers accounts for 31.8%. Only 4 farm bases belong to the Cooperative+ Dragon type, and this type only covers 18 households. Besides, The total number of SPM+ Cooperative FBs and Cooperative FBs are 11 and 8 , and their farmers account for 26.2% and 19.0%, respectively. Production and marketing performance Dragon-head companies have no significant impact on the yield and profit obtained by farmers, but belonging to a cooperative has. Farmers appear to have higher yield and more profit in apple production if they are the members of cooperative-coordinated FBs who are not coordinated by dragon-head companies. As shown in Table 2, the average yield is 41.39 and 44.62 for the cooperative-coordinated FBs who may or may not have supplied supermarkets, but the figure is only

1 SMP represents supermarket. 6

32.95 for the non-FB farmers (row 5 in Table 2). And the same situation occurs for the profit obtained by farmers. The yield and profit of the dragon-head company coordinated FBs farmers is much lower than the Cooperative-coordinated FBs farmers, and the figures are not significant higher than that of non-FB farmers. Over more than half farmers have access to the extension services, and this figure is particularly high for non-supermarket-coordinated FBs with cooperative farmers. On the average, the percentage of farmers with long term training is 54.17%, and the percentage of field training is 51.39%.These figures for FBs farmers who may nor may not supply supermarket is significantly higher than that in the non-FB farmers (row 2-3 in Table 2). By further breaking down farmers into supermarket- coordinated FBs and non-supermarket-coordinated FBs, we find that the percentage of non- supermarket-coordinated FBs farmers who sell their apples only through cooperative access to training is far better than non-PB farmers. For example, the average percentage for non-FB farmers having access to long training in the study area is only 38.09% and the figure nearly may double for the Cooperative farmers (Table 2). Moreover, dragon-head companies seem to be not effective on the weighted sale prices obtained by farmers. The average weighted price that the non-FBs farmer obtained is only 6.65, but the figure is not significant different from the FB farmers, be they in a chain driven by dragon-head companies or not (row 8 in Table 2). On the other hand, dragon-head companies do seem to be effective to urge farmer to adopt of ecological practice and improve farmer knowledge of pesticide suitability. The percentage of non-FB farmers adopting ecological practise is 25.7%, but this share is more than double for the supermarket- coordinated FBs with cooperative and dragon FB farmers. Farmer knowledge of using pesticides suitable for certain diseases and pests is limited. The average score for non-FB farmers is only 38.10 and the figure is not significantly different from FB farmers who may not have supplied supermarkets, but the score for supermarket-coordinated FBs with cooperative and dragon FB farmers is 46.31 which is slightly higher (row 10 in Table 2). All kinds of governance chain have positive effect on extension services relative to the chains without farm bases. The estimates coefficients for all types of chains are statistically significant and positive in all estimations, either long term training or training in the field (column 1-2 in Table3). Meanwhile, the coefficients of cooperatives in the two regressions are bigger than others which means the effects are even stronger (row 4 in Table2).This implies that holding other things constant, there is a significant difference between farmers who are members of FBs and non members. FBs farmers are more likely to have access to the trainings. It is observed that the regression results are similar to the statistical results before. The results show a limited impact (not significant) of having dragon-head companies as apple purchasers on yields and profits. As shown in table 4, nearly none of dragon-head companies coefficients is significant, either yield or net profit (column 1-2 in Table4). When viewing the cooperative terms, joining in the cooperative can increase yield by 8.13 ton per hectare and net profit by 79.28 thousand yuan per hectare (row 4, column 1-2 in Table4). This result is not difficult to understand because of the services brought by cooperatives in terms of extension services. Dragon-head companies have no significant impact on the prices obtained by farmers. As show in Table 4, the coefficients which are not significant suggest that dragon-head companies are not associated with the prices obtained by farmers (column 3). Indeed, farmers store apples in expectation of higher prices at the end of the main cropping season. They may be pushed to practice this due to the security of outlets provided by agri-business companies and supermarkets, a point to be investigated in further analysis. Dragon-head companies and supermarkets also drive farmers to grade the apples. This does not necessarily translate into higher prices considering the bulk of the apples that are sold. The descriptive analysis shows no significant difference in terms of weighted sale price between the different categories of farmers. The multi-variate analysis even shows a negative impact on sales to supermarkets on the weighted prices (row 2, column 3 in Table 4), which may be due to low prices offered for apples of a lower grades.

7

Surprisingly, while the different governance chains except belonging to a cooperative do not seem to yield beneficial effects on yield and profit, we do observe positive effects on the farmer adoption of ecological control in pest and disease management. The estimated coefficients are significantly positive (column 1, Table 5), showing increased use of ecological practices when farmers belong to farm bases that are coordinated by supermarket and dragon head company. However neither do the chain governance affect farmer knowledge in apple production. None of the estimated coefficients is significant and this implies that FBs farmers (selling to dragon-head companies or not) are not different from non-FB farmers in their knowledge of pesticide suitability. This can be explained by the fact that farm bases along with dragon-head company play an important role in delegating the purchase and application of pesticides (Ding et al., 2017).

Conclusion Apple chains are indeed in a process of transformation, with agri-business companies and supermarkets trying to control farmer decisions, but having little success so far. This is mostly due to little additional provision of technical advice, hence few changes in technology, yield and quality. The adoption of more eco-friendly practices is not translated into higher prices for farmers. The study shows that the optimal conditions for agri-business development to be conducive to economic development shown by the literature are not yet gathered in the apple sector of China. It paves the way for additional research in the following directions: (i) investigating consumers’ willingness to pay for apples grown with eco-friendly practices; (ii); understanding farmers’ storage strategies and their impact on selling prices; (iii) analysing the role of cooperatives in enhancing farmers’ yields and profits.

8

References

Cook, M. L., and Chaddad, F. R. (2000). Agroindustrialization of the global agrifood economy: bridging development economics and agribusiness research. Agricultural economics, 23(3), 207-218. Ding, J., Huang, J., Jia, X., Bai, J., Boucher, S., & Carter, M. (2015). Direct farm, production base, traceability and food safety in China. Journal of Integrative Agriculture, 14(11), 2380-2390 Ding, J., Moustier.,P., Ma,X., Huo,X.X., & Jia,X.P.(2017). Doing But Not Knowing: How Apple Farmers Comply with Standards in China. Paper presented at 15th European Association of Agricultural Economists (EAAE). Parma, Augus t 28th - September 1st. Fan, H., Ye, Z., Zhao, W., Tian, H., Qi, Y. & Busch, L. (2009). Agriculture and food quality and safety certification agencies in four Chinese cities. Food control , 20(7), 627-630. Gale, H.F., & Hu. D. (2012). Food Safety Pressures Push Integration in China’s Agricultural Sector. American Journal of Agricultural Economics 94(2), 483-488. Guo, H., Jolly, R. W., & Zhu, J. (2007). Contract Farming in China: Perspectives of Farm Households and Agribusiness Firms. Comparative Economic Studies, 49(2), 285-312 Huang, J., Rozelle, S., Dong, X. , Wu, Y., Zhi, H., Niu, X., & Huang, Z. (2007). Agrifood Sector Studies China: Restructuring agrifood markets in China - The horticulture sector Part A - meso-level study. Regoverning Market Program, IIED. Jia, X., Huang, J., & Zhigang, X. U. (2012). Marketing of farmer professional cooperatives in the wave of transformed agri-food market in China. China Economic Review, 23(3), 665-674. Michelson, H., Perez, F., Reardon, T., Huang, J., Jia X., Bai, J., Boucher, S., Carter, M., & Huang, X. (2015). Modernizing Markets, Small Farmers, and Food Security. ACES Food Security Symposium. Retrieved on line on05/08/2015/ at http://intlprograms.aces.illinois.edu/sites/intlprograms.aces.illinois.edu/files/Michelso n_ACES_FS.pdf. Minot, N., & Sawyer, B. (2016). Contract farming in developing countries: Theory, practice, and policy implications. Innovation for inclusive value-chain development: Successes and challenges, International Food Policy Research Institute (IFPRI), Washington, DC, 127-158. Miyata, S., Minot, N., & Hu, D. (2009). Impact of contract farming on income: linking small farmers, packers, and supermarkets in China. World Development, 37(11), 1781-1790. Reardon, T., & Barrett, C. B. (2000). Agroindustrialization, globalization, and international development: an overview of issues, patterns, and determinants. Agricultural economics, 23(3), 195-205. Reardon, T. (2015). The hidden middle: the quiet revolution in the midstream of agrifood value chains in developing countries. Oxford Review of Economic Policy, 31(1), 45- 63. Swinnen, J.F. & Maertens, M., (2007). Globalization, privatization, and vertical coordination in food value chains in developing and transition countries. Agricultural Economics ,37(1,89-102. 9

Ton, G., Vellema, W., Desiere, S., Weituschat, S., D'Haese, M. 2018. Contract farming for improving smallholder incomes: What can we learn from effectiveness studies? World Development, volume 104, pp. 46 – 64. Wang, H., Dong, X., Rozelle, S., Huang, J. & Reardon, T. (2009). Producing and Procuring Horticultural Crops with Chinese Characteristics: The Case of Northern China. World Development , 37(11, 1791-1801. Wang, H. H., Wang, Y., & Delgado, M. S. (2014). The transition to modern agriculture: Contract farming in developing economies. American Journal of Agricultural Economics, 96(5), 1257-1271. Zhang, X., Qiu, H. & Huang, Z. (2009). Linking small-scale farmers in China with the international markets: a case of apple export chains. International Food and Agribusiness Management Review, 12(3, 89-109.

10

Figure 1.The geographical distribution of the sample provinces in China, 2014.

11

Figure 2.The geographical distribution of the sample counties in Shaanxi province, 2014.

12

Figure 3.The geographical distribution of the sample counties in Shandong province, 2014.

13

Table 1. The governance chain of the fresh apple and description of sample. Number of FBs Number of HH FB 42 218 SPM+ Cooperative 11 65 SPM+ Cooperative+Dragon 19 103 Cooperative 8 32 Cooperative+Dragon 4 18 Non-FB n.a. 105 Note: n.a., not applicable. HH represents the household. FB represents farm base. SPM represents supermarket.

14

Table 2. The characteristics of governance chain of fresh apple in China, 2014. Supermarket-coordinated FBs Non-supermarket-coordinated FBs Total Non FB farmers Cooperative Cooperative & dragon Cooperative Cooperative &dragon (N=323) (N=65) (N=103) (N=32) (N=18) (N=105) % of farmers with Long term Training 54.17 60.00*** 57.28*** 75.00*** 72.22*** 38.09 Field training 51.39 56.92*** 52.43*** 75.00*** 66.67*** 37.14 Production performance Yield (ton/ha) 35.42 41.39*** 31.89 44.62*** 32.17 32.95 Profit(thousand yuan/ha) 140.59 147.31 125.33 228.90*** 125.88 128.39

Marketing performance

Weighted pricea (yuan/kg) 6.70 6.38 6.84 6.97 6.90 6.65 % of farmers with ecological 33.74 36.92 35.92 31.25 61.11*** 25.71 practices The average score of pesticide suitability 42.20 44.31 46.31*** 44.69 30.56 38.10 (Full mark=100) Note: a represents that weighted price is calculated by averaging the different prices relative to grades according to the percentage of each grade. T- tests are made by choosing ―non-FBs‖ as the reference. Source: Author’s survey.

15

Table 3. Multivariate analysis of relationship between chain governance and accessing to the extension services for Fuji apple farmers in 2014, Logit model. (1) (2) Long term Training Training in the field Agricultural organization of FBs (Yes=1; No=0) SPM+ Cooperative 0.21*** 0.20*** (2.96) (2.63) SPM+ Cooperative+ Dragon 0.17** 0.13* (2.56) (1.92) Cooperative 0.33*** 0.35*** (4.71) (4.80) Cooperative+ Dragon 0.29*** 0.27*** (3.37) (2.65) Control variables of households, individuals and regions Farm size (hectare) 0.01 0.01 (0.32) (0.37) Farming experience of household head 0.003 0.003 (0.76) (0.86) Education of household head (dummy 0.11 0.14 for middle school and above) (1.23) (1.62) Shandong province dummy -0.04 -0.06 (0.71) (0.90) Note: Absolute values of t-ratio in parentheses; *, **, *** indicate statistically significant at the 10%, 5%, and 1%, respectively. The sample size used in regression is 323.

16

Table 4. Multivariate analysis of relationship between chain governance and production, marketing performance for Fuji apple farmers in 2014, OLS model. (1) (2) (3) Net Profit Yield Weighted price (Thousand (Ton/ha) (yuan/kg) Yuan/ha) Agricultural organization of FBs (Yes=1; No=0) SPM+ Cooperative 2.63 -16.49 -0.51* (1.13) (0.77) (1.67) SPM+ Cooperative+Dragon -1.89 -7.84 0.14 (0.93) (0.42) (0.64) Cooperative 8.13*** 79.28*** 0.16 (2.77) (2.92) (0.33) Cooperative+ Dragon -2.65 -28.61 0.22 (0.71) (0.83) (0.57) Control variables of production, households, individuals and regions Labor use (ha) (production) 0.02*** -0.07* (3.39) (1.69) Fertilizer use (kg/ha) (production) 0.0004*** 0.002 (2.81) (1.69) Material fee use (thousand 0.47*** 2.54** yuan/kg) (production) (4.26) (2.51) Disaster (Yes=1;0=No) -1.86 -13.03 (production) (1.06) (0.80) Age of apple tree(year) (individual) 0.34*** 3.11*** 0.01 (3.04) (2.97) (0.54) Farm size (hectare) ( households) 0.17 -2.34 -0.06* (0.31) (0.47) (1.89) Farming experience of household -0.05 -0.29 0.00 head (individual) (0.46) (0.29) (0.01) Education of household head 2.79 10.04 -0.08 (dummy for middle school and (1.16) (0.45) (0.30) above) (individual) Shandong province dummy 8.02*** 85.13*** 1.14*** (regions) (4.39) (5.05) (5.59) Constant 5.66 15.14 6.01*** (1.41) (0.41) (13.38) R2 0.364 0.198 0.095 Note: Absolute values of t-ratio in parentheses; *, **, *** indicate statistically significant at the 10%, 5%, and 1%, respectively. The sample size used in regression is 323.

17

Table 5. Multivariate analysis of relationship between chain governance and food safety in 2014,Logit and OLS model. (1) (2) Use of ecological The score of practices of pest pesticide control suitability Agricultural organization of FBs (Yes=1; No=0) SPM+ Cooperative 0.18** 1.42 (2.12) (0.37) SPM+ Cooperative+Dragon 0.13* 5.43 (1.79) (1.62) Cooperative 0.10 3.48 (0.92) (0.72) Cooperative+ Dragon 0.39*** -8.35 (3.30) (1.37) Control variables of households, individuals and regions Farm size (hectare) 0.03 1.77** (1.52) (2.01) Farming experience of household head -0.001 0.46** (0.27) (2.57) Education of household head (dummy for -0.05 8.89** middle school and above) (0.63) (2.27) Shandong province dummy -0.22*** 17.15*** (3.83) (6.13) Constant 11.42** (2.18) R2 0.189 Note: Absolute values of t-ratio in parentheses; *, **, *** indicate statistically significant at the 10%, 5%, and 1%, respectively. The sample size used in regression is 323.

18