An exploration of livestock-development policies in western China

Food Policy 37 (2012) 12–20

Contents lists available at SciVerse ScienceDirect

Food Policy journal homepage: www.elsevier.com/locate/foodpol

An exploration of livestock-development policies in western China Adam M. Komarek a,⇑, Scott A. Waldron b, Colin G. Brown b a b

School of Natural Sciences, University of Western Sydney, Locked Bag 1797, Penrith NSW 2751, Australia School of Agriculture and Food Sciences, The University of Queensland, Brisbane Qld 4072, Australia

a r t i c l e

i n f o

Article history: Received 20 May 2011 Received in revised form 30 September 2011 Accepted 4 October 2011 Available online 9 November 2011 Keywords: Policy scenarios China Livestock Household incomes Heterogeneous-agent model

a b s t r a c t Limited agricultural land potential, poorly integrated markets and low rural incomes are all interrelated problems for agricultural households in western China. Within the wide range of development options available, this paper focuses on the impact of forage-growing subsidies and reducing livestock feed trade barriers on agricultural household incomes and enterprise mixes in the Qingyang Prefecture of Gansu Province, China. A heterogeneous-agent model is used to assess the local consequences of these two policies. Data from a survey conducted in 2009 are used as inputs into the model. The results indicate that a simultaneous reduction in livestock feed trade barriers and an introduction of forage-growing subsidies lifts net household incomes by approximately 10%. Different scenarios have different impacts on incomes, land allocation decisions and grain purchases, with heterogeneity also found among household income responses. Livestock-development policies have the potential to lift household incomes however there are tradeoffs between income and grain self-sufficiency. Ó 2011 Elsevier Ltd. All rights reserved.

Introduction Rural income growth has become a fundamental goal for Chinese policy makers especially given China’s longstanding commitment to reducing rural poverty (Park et al., 2002). Many of China’s poorest rural households are situated in low rainfall areas on the margin of cropping and pastoral regions where population densities are high but agro-ecological potential is limited. Qingyang Prefecture in Gansu Province in western China is one such area. Average rural per capita income in Qingyang Prefecture in 2007 was only half of the national rural average. The low level of income in areas such as Qingyang has captured the attention of high-level government officials with Premier Wen Jiabao calling on Gansu to increase farmers’ incomes to reach the 2007 national average by 2012 (Brown et al., 2009). In response, local governments have pursued various interventions, primarily agricultural-based, in an attempt to improve these incomes. In the broader context, numerous development options have been pursued by the government at different administrative levels to improve rural household livelihoods in western China. These encompass measures focusing on creating local employment opportunities (Goodman, 2004) and developing transportation infrastructure (Fan and Chan-Kang, 2008). Agricultural-based measures are also being canvassed and gradually implemented, for example, improving crop total factor productivity (Shen et al., 2009), providing financial payments for lucerne production ⇑ Corresponding author. Tel.: +61 2 4570 1475; fax: +61 2 4570 1750. E-mail address: [email protected] (A.M. Komarek). 0306-9192/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodpol.2011.10.001

(Uchida et al., 2005) and advancing livestock husbandry techniques (Squires et al., 2010).1 In this paper, two policy-relevant interventions are analysed namely: (1) the role of subsidies in promoting livestock development; and (2) reducing barriers to livestock feed trade. The objective of this paper is to examine the influence of these two policies on farm household enterprise choices and incomes using the case of two townships in Qingyang Prefecture of Gansu Province. A heterogeneous-agent model with the capacity to examine the specific options is developed to determine the local consequences of these two policies. Policy makers perceive livestock development as a useful means of raising incomes and promoting development in western China. This follows international debates about the role of livestock in development, with some consensus emerging that farm households can experience modest income and production efficiency gains when appropriate livestock policies are applied (Herrero et al., 2010; McDermott et al., 2010; Tarawali et al., 2011). In the case of Qingyang, livestock development strategies have become visible with specific lucerne acreage and small ruminant inventory targets being set (Brown et al., 2009). Achieving agricultural production targets has been seen as an intermediate objective to the main goal of lifting net incomes. By assessing the divergent net income effects of the two interventions on the heterogeneous households that characterise these areas, this paper sheds insights into the suitability of the policy interventions. 1 Lucerne (or alternatively, alfalfa) is a deep-rooted perennial forage crop that is used to feed livestock.

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

Previous policy-relevant studies in this region have focused on examining tradeoffs between production, food security, income and environmental outcomes associated with different policy directions (Lu and van Ittersum, 2004; Lu et al., 2004; Komarek et al., 2010). This paper complements these more bio-physical focused studies by providing empirical evidence on the whole-farm effects of alternative policies. The model used shares common features with previous heterogeneous-agent models (Berger, 2001; Johnson et al., 2006; Komarek and Ahmadi-Esfahani, 2011) as it captures the responsiveness of multiple households to policy shocks. The model also incorporates equilibrium trade conditions for livestock feed, labour and land. Although spatial equilibrium models have previously been used to examine trade and border effects (McCarl and Spreen, 1997), the model applied in this paper uses detailed household and agronomic data to investigate these effects; a level of detail often missing from broader scale agricultural sector models.

Study site and livestock-development scenarios Livestock are an important component of many household enterprise mixes in the developing world and are especially important in parts of western China (Brown et al., 2008, Chapter 6). To address the study objectives, three villages within Quzi township and three villages within Shishe township of Qingyang Prefecture in Gansu Province are chosen as case studies.2 The townships are approximately 100 km apart, and within in each township the villages are within a 5 km radius of each other. Despite their relatively close proximity, the two townships experience different agroclimatic conditions with Quzi having lower rainfall, and hence lower crop yields, than Shishe. Shishe farmers use more intensive production practices and their population density is greater. In addition, Shishe is located much closer to the prefectural capital and so households are closer to agricultural markets and non-agricultural offfarm employment opportunities. As mentioned, the development of local livestock industries is being used in an attempt to grow household incomes to help meet government goals of raising incomes to the national rural average. One specific livestock target involves increasing the number of improved-breed goats in Qingyang Prefecture from 2.4 million (2008) to 3 million (2012) (Brown et al., 2009). To meet these targets, various bundles of specific interventions are being used. One specific bundle of interventions examined in this paper are subsidies to grow lucerne, to buy breeder goats and to build warm sheds for goat housing (collectively known as livestock-development subsidies). Apart from developing feed sources to support the push into ruminant livestock, lucerne subsidies are also closely linked with other government goals of reducing environmental externalities (Liu et al., 2008). In Qingyang, this program involves a cash payment of 2400 RMB/ha/annum for households to plant and maintain lucerne on sloping land.3 Another set of interventions examined in this paper relate to reducing forage trade barriers. Forages refer to maize stover, wheat stover and lucerne.4 There are no official physical restrictions on forage trade. However, there are numerous economic barriers to trading forages across regions. Farmers trade stover locally within their township using motorised bicycles or horse-drawn carts based on word-of-mouth information flows. Despite this intra-township trade, fieldwork revealed that stover trade with neighbouring townships does not occur. Key drivers of this lack of inter-township trade 2 Quzi is part of Huan County and Shishe is part of Xifeng district. Both Huan County and Xifeng district belong to Qingyang Prefecture. 3 At the time of research in 2009, US$1  6.3 Chinese Renminbi (RMB). 4 Stover is the part of the crop that is left in the field after grain harvest and includes leaves and stalks.

13

are stover’s low value to volume ratio, poor infrastructure leading to high transportation costs, incomplete information, a lack of capital for purchasing suitable trucks and limited competition among traders.5 Widespread grazing restrictions aimed at reducing grassland degradation (Dong et al., 2007) have also seen systems gradually evolve from a reliance on native pasture for livestock feed to include expanded lucerne acreage and a more efficient usage of stover. This has increased stover’s relative value, and hence its trade. Nevertheless, stover markets are still rudimentary and households do not possess the transportation means necessary to access other township markets. The trading situation for lucerne differs to that of stover. Lucerne area declined in the 1980s as both food demand increased and as a rising usage of fertiliser reduced the relevance of lucerne in providing soil nitrogen (Li and Huang, 2008). This trend has reversed in the past decade with lucerne acreage expanding across western China. This is partly the result of a paradigm shift towards integrated crop-livestock systems (Jia et al., 2009), positive linkages between income growth and meat consumption and government incentives to plant lucerne for environmental benefits (Liu et al., 2008). An expanding agribusiness sector focused on intensive ruminant feeding has further strengthened lucerne demand. Local level data indicate a similar pattern with Brown et al. (2009) reporting that Qingyang lucerne acreage doubled between 2000 and 2007. Within this context, inter-township lucerne trading in Qingyang has been observed as lucerne has a high value to volume ratio, and local agribusiness development is strengthening its demand. Despite this, solid trade networks with neighbouring prefectures have yet to emerge as current transportation and marketing costs remain relatively high for traders. Potential exists for longer distance trade to occur if market integration and trading costs are addressed. In essence, the lucerne market is more developed than the stover market. It, therefore, appears useful to examine the possibility of stover trade between different townships and lucerne trade between different prefectures becoming feasible. These trade options may eventuate naturally as the scarcity value of forages is rising. However, to render trade feasible, trading costs need to be reduced along with improved market integration. Based on these observations, the two livestock-related interventions and study objectives, four specific scenarios are canvassed in this paper: 1. A base-case scenario where there are no livestock-development subsidies available. In this scenario, intra-township trade of stover and inter-township trade of lucerne are possible. This trading situation is based on existing trade patterns. 2. Livestock-development subsidies are available and intra-township trade of stover and inter-township trade of lucerne are possible. This implies that no changes to external trading conditions have occurred, but that there is government support to provide financial incentives to households to intensify lucerne and livestock production. 3. No livestock-development subsidies operate but the intratownship trade of stover and inter-township trade of lucerne restrictions are relaxed. Inter-township trade of stover and inter-prefecture trade of lucerne are possible in this scenario. This scenario reflects attention being directed at improving trade flows for stover and lucerne rather than providing direct financial assistance for lucerne and livestock production. 4. Livestock-development subsidies are available and the intratownship trade of stover and inter-township trade of lucerne restrictions are relaxed. Inter-township trade of stover and

5

For instance, maize stover has 6% crude protein compared with 20% for lucerne.

14

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

inter-prefecture trade of lucerne are possible in this scenario. This scenario entails a concurrent improvement in forage trading as well as the provision of livestock-development policies.

Model Western China comprises a wide range of agro-ecological zones and there is a corresponding diversity among rural households. A typology commonly used in China to reflect this diversity distinguishes between pastoral, semi-pastoral and agricultural households. The model developed in this study focuses on agricultural households, especially former semi-pastoral households that have been subject to grazing bans to combat over-grazing and are now encouraged to raise livestock using more intensive systems (Brown et al., 2008; Han et al., 2008). These agricultural households produce crops, raise goats and have also become more involved in buying and selling livestock feed. They are also involved in off-farm migration. The key decision variables in the model relate to allocating land, labour and cash to different crop and livestock activities, buying and selling livestock feed stocks and allocating labour to working off-farm. Rather than employing a representative farm approach, the model draws on information and explicitly models the case of 96 farm households as discussed below. The model’s objective is to maximise the combined net present value of all 96 household net incomes from 2004 to 2008.6 The model is solved as a multi-year, mixed integer programming problem with township-level and district-level constraints existing for specific activities. The aggregate objective function means that the model output for any one particular farm may not necessarily correspond to the optimal solution for the individual farm. Nonetheless, the model can provide insights into the situation at a village level and so enable the investigation of the township-level and districtlevel constraints which is a purpose of the paper. An algebraic specification of the model appears in Appendix A. Heterogeneity among households within a township arises from differences in their area of arable land, available starting cash and family size. As elaborated below, the average surveyed household has access to 1.1 ha of arable land. Distinct differences arise between the two townships, with Quzi households cultivating on average 0.5 ha more than their Shishe counterparts. Shishe households have on average 13,729 RMB of available cash each year, exceeding Quzi households by approximately 2000 RMB. Heterogeneity is evident among individual households in both locations, with the coefficient of variation of area cropped and available cash being 0.70 and 0.76 respectively. Family size is generally uniform between the two townships. A principal driver of heterogeneity is inter-township rainfall differences (530 mm in Shishe vs. 403 mm in Quzi). Although inter-township crop yields vary in the model, households within each township share identical crop yields and face the same market prices. Household diversity implies that individual households could respond differently to alternative policies, thus households are classified into different groups to provide insights into whether divergent policy effects exist. The length of the planning horizon was set at 5 years (2004– 2008) and follows the principles outlined in McCarl and Spreen (1997). This time horizon broadly corresponds to the period that decisions on lucerne replanting have to be made as well as the period needed to optimise flock structures. The value of any change in livestock numbers and grain inventories was included in the objective function to ensure that the change in net wealth across the 5 years was captured. 6 The model is written and solved in the General Algebraic Modelling System (GAMS). The complete GAMS code and data files are available from the corresponding author on request.

The model accounts for changes in livestock inventories, feed and cash supplies through time. Decisions regarding the buying, selling and retaining of goats in 1 year impact on future years, and so a dynamic model is used to capture the flock dynamics. Lucerne is a perennial crop with yields varying depending on age. Thus, a dynamic model incorporating different lucerne crop ages was also employed. The main activities in the model are as follows:  Livestock activities – Three breeds of goats are modelled: Liaoning, Inner Mongolian and native goats. Within each breed there are different classes of goats, including starting breeders, purchased breeders, culled breeders and kids retained for breeding or sale. Each breed and class produces different outputs (cashmere and meat) with different energy requirements, gross margins, kidding rates and mortality rates. Goats can be housed in a warm shed or in a common shed. The revenue obtained from a breeder goat varies between each year (Table 3), but is the same across households.  Crop activities – Maize, wheat and lucerne can be grown, sold, stored and consumed. Crop production is simulated from 2004 to 2008 using APSIM (Keating et al., 2003).7 Maize (a summer crop) and wheat (a winter crop) are grown in a rotation and these crops can be cultivated using either machinery or draught power. There are three categories of lucerne sales: intratownship; inter-township; and inter-prefecture. There are two categories for stover sales: intra-township; and inter-township. Each category has a different transportation cost and labour requirement.  Subsidies – In the scenarios involving subsidies, lucerne planting is subsidised at a rate of 2400 RMB/ha/annum in fieldwork sites. The government also subsidises 50% of the cost to purchase breeder goats and to build a warm shed.  Labour – Households allocate their time to four activities: agricultural activities on their own farm; agricultural activities on other households’ farms (hired out labour); full year migration to an urban centre; and 6 months migration over winter to an urban centre. In each month, agricultural labour can be hired in and hired out. Daily wages for agricultural work are 25 RMB in the harvest season and 20 RMB in the slack season.  Land – Land can be rented in and rented out at a rate of 1800 RMB/ha/annum subject to the constraints outlined below. The main constraints in the model for each household are:  Land – The area of arable land available for a household is equal to the amount observed in the 2009 survey.  Livestock feed – Goats must obtain a minimum amount of energy each month. Goat energy demands were calculated using NRC (1981), Paten (2008) and local field observations. Livestock feed can be sourced from home-grown crops or market purchases. Livestock feed can be stockpiled and carried forward to future months.  Labour – Activities have monthly labour requirements that must be balanced with available household labour and labour hired in. The amount of labour available to conduct agricultural activities is limited by the number of people living on the farm in any month, and the amount of cash available to hire in labour. Technical input–output coefficients for seasonal labour requirements per unit of activity are fixed across households and are based on survey data. Family labour can be supplemented with hired labour. Hired labour is constrained at the township level 7 APSIM is a widely used simulation model that simulates crop, forage and soilrelated processes and the influence of climate and management activities on these processes using local climate and soil data.

15

A.M. Komarek et al. / Food Policy 37 (2012) 12–20 Table 1 Base-case trade conditions (based on field work observations). Commodity

Inter-town trade

Inter- prefecture trade

Maize and wheat stover Lucerne Maize and wheat grain Agricultural labour Arable land

No Yes Yes No No

No No Yes No No

Notes: Intra-town trade was observed and is assumed for all commodities. When the forage trade barrier is relaxed, inter-town wheat and maize stover trade can occur as can inter-prefecture lucerne trade.

with labour used in each township equal to labour available in each township. As the study sites are approximately 100 km apart, it is not feasible to travel between the townships on a daily basis to conduct wage-based agricultural activities.  Cash – Based on fieldwork observations, the amount of cash that can be spent on agricultural activities in the initial year of the model is set equal to the observed net agricultural income plus any remittances available for farm spending.8 In following years, the total year’s spending on agricultural activities is limited to the previous year’s net income (after accounting for non-agricultural costs). Surplus cash in December is carried forward to the following January.  Crop equilibrium conditions – For lucerne, maize and wheat, the amount produced plus stored from the previous year equals the amount fed to animals, consumed by humans, sold and stored for future usage.  Livestock equilibrium conditions – The number of goats at the start of a year plus the number purchased equals the amount sold plus those kids retained for breeding plus the adult breeder goats retained. Based on a 5% mortality rate, the number of breeder goats at the start of a year is equal to 95% of the previous year’s ending stock plus kids retained. Breeder goats are culled after 6 years. As mentioned, a feature of the model is the incorporation of trade constraints at different levels of aggregation as shown in Table 1. Township and district level constraints are as follows:  Within each year, the sum of all household wheat and maize stover purchases must not exceed sales within each individual township. This constraint ensures that stover consumed in a town is sourced from within the town.  Within each year, the sum of lucerne traded in Shishe and Quzi equals the sum of all purchases in Quzi and Shishe. This implies that lucerne can be traded between townships within Qingyang but not with townships outside Qingyang.  Each year aggregate grain supply, and carryover from previous seasons, for all 96 households must be enough to cover home consumption needs of 220 kg of grain/capita/annum. Grain demand is met through own production and purchases.  The sum of all labour time incurred across the households within a township in wage earning agricultural activities equals the time that all households hire in labour for agricultural activities in that township.  The sum of the total area cropped across all households within a township is less than the sum of the available land for all households. Within each township there is zero excess demand for rental land as area rented in must equal area rented out. Forage trade barriers are imposed by excluding activities where households can buy or sell stover from outside their own township, 8 Remittances are the portion of off-farm earnings that are received by family members living on the farm.

and by excluding activities where households can buy or sell lucerne for outside the case-study Prefecture. This was achieved in the model by incorporating these activities into the model at prohibitive per unit trading costs. The scenarios with more open forage trade were introduced by reducing prohibitive trading costs to levels that reflected better transportation networks and improved market integration. For example, the transfer cost for inter-prefecture lucerne trade is approximately 200 RMB/tonne for 300 km, based on cost data obtained from provinces with more developed road networks. Data Data to calibrate the model are from three sources: a household survey conducted in 2009; APSIM crop simulations; and data from research trials conducted by the Huan County Animal Husbandry Bureau. The 2009 surveys provide data on land usage, socio-demographic characteristics, agricultural production and consumption, income and migration for a random sample of households. The target population for this study consists of farm households that live below the national rural average income in the two townships of Qingyang. Three villages were selected in each township, with 48 households being interviewed in each township. Different household types were developed based on Principal Component Analysis and subsequent Cluster Analysis. The major distinguishing feature between the two derived household clusters was the land to labour ratio (LLR). The LLR ratio is the number of hectares of land per working family member living on the farm (aged 15–65 years). In Shishe, there are 46 low LLR households and two high LLR households, whilst in Quzi there are 23 low LLR households and 25 high LLR households. Small ruminant numbers per household are higher on the low LLR households. Although low LLR households have less available land per capita, the relative importance of grain and lucerne to farm acreage is similar across the clusters. The high LLR household generally sells forages to the low LLR household. Remittances comprise a large portion of household income and the two household groups show a similar level of dependency on remittances (Table 2). Despite these similar dependencies, income responses to policy shocks are expected to differ among households as rainfall and land and labour assets vary between the groups. Fieldwork revealed that households displayed a tendency to use remittances for non-agricultural purposes like education. Thus, remittances are not expected to greatly impact on household responses to policy measures. Goat revenues hinge on cashmere and meat price changes, as well as productivity parameters and flock dynamics (Table 3). Breeder gross margins include the sale of cashmere and the sale of cull goats, which are sold on a liveweight basis. Revenues from kids are derived from sales made on a liveweight basis. Native goats are small-framed but attract a premium of 2 RMB/kg liveweight because of consumer preferences for their meat. They have Table 2 Characteristics of household groups. Variable

Low land to labour ratio (n = 69)

High land to labour ratio (n = 27)

Area cropped (ha) Household size (all people) Land to labour ratio Income from crops (%) Income from livestock (%) Income from remittances (%)

0.99 (0.62) 4.9 (1.05) 0.4 (0.18) 15 10 75

1.53 (0.63) 4.4 (1.28) 0.63 (0.18) 19 19 62

Notes: Mean values are reported with standard deviation in parentheses. Data are from the authors’ survey. Areas cropped and land to labour ratios differ between the groups (P = 0.00). Household size also differs between the groups, but at a lower level of significance (P = 0.06).

16

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

Table 3 Goat revenues (RMB/year) by type of goat – 2004–2008. Year

Breeder

2004 2005 2006 2007 2008

Kid

Native

Liaoning

Inner Mongolian

Native

Liaoning

Inner Mongolian

79 88 95 106 48

228 254 273 301 139

215 239 257 283 130

398 403 506 591 422

513 526 660 774 518

426 438 547 640 424

Notes: Data are from authors’ survey data and Huan County Animal Husbandry Bureau. The revenue for a breeder does not account for the value of kids produced. The model has a separate category for retained kids.

Table 4 Distribution of crop prices (RMB/kg) and yields (t/ha). Year

Crop prices Wheat grain

2004 2005 2006 2007 2008

1.24 1.24 1.26 1.40 1.66

Crop yields Wheat stover

0.24 0.30 0.30 0.34 0.40

Maize grain

1.26 1.24 1.30 1.44 1.52

Maize stover

0.10 0.12 0.13 0.14 0.15

Lucerne

0.9 0.9 0.8 0.5 0.8

Wheat grain yield

Maize grain yield

lucerne biomass yield

S

Q

S

Q

S

Q

2.9 1.9 2.3 2.7 2.9

2.4 1.6 1.9 2.4 3.1

5.6 5.4 5.5 5.6 6.6

2.4 1.8 1.6 2.4 3.1

6.8 5.4 5.4 5.8 6.8

4.6 3.7 2.5 0.1 2.5

Notes: Data are from authors’ survey data and authors’ APSIM simulation results. S = Shishe and Q = Quzi. Grain comprised approximately 40% of wheat and maize total biomass.

a low cashmere yield and the price of their cashmere is discounted because of its dark colour. The cashmere of the Inner Mongolian breed is finer (15 lm) than Liaoning cashmere (16 lm). However, China’s undeveloped cashmere marketing system means that traders do not offer households a price premium for finer cashmere (Waldron et al., 2011). Liaoning goats have a high cashmere yield and heavier frames, and in the absence of price-grade differentials for their coarser cashmere, generate the highest revenues of the three breeds (Table 3). APSIM crop yields in Shishe generally exceed those in Quzi (Table 4). Wheat and maize grain prices are similar and increased over the study period (Table 4). Results Although the model explicitly incorporates the 96 households, the results are reported for the two different land to labour ratio household types so as to simplify the presentation. Results of the base-case model are reported in Table 5 and are for the average household in the average year. The household effects of introducing livestock-development subsidies and reducing forage trade barriers are reported in Table 6. Income distributions among households for different scenarios are also presented (Fig. 1). Base-case results Agricultural sales comprised approximately one third of predicted total incomes with the remaining income derived from off-farm employment (Table 5). Fourteen households live below the Qingyang-specific poverty line of 1196 RMB/capita/annum (CSP, 2006), 12 of these households are low LLR households and two are high LLR households. Households choose to raise Liaoning goats in a warm shed. The average household has four Liaoning goats, with low LLR households having five Liaoning goats and high LLR households having two Liaoning goats (Table 5). The average household spent approximately 18 person-months working away from the farm each year. Grain acreage exceeded lucerne acreage. There was a net transfer of 33 tonnes of stover and 26 tonnes of lucerne in the base case from high LLR to low LLR households. These volumes represent

the sums of sales minus purchases made by the different household types within each township. For example, low LLR households purchased 144 tonnes and sold 111 tonnes of stover in an average year. In an average year, 7 tonnes of grain were purchased for human consumption from other districts outside Qingyang Prefecture, and this accounts for approximately 7% of consumption demand.9 Within the low LLR category, 2215 man days were hired in throughout the average year and 2209 man days were spent working on other households’ farms. To compare the predictive performance of the model with reality, the results from the scenario that includes livestock-development subsidies and intra-township trade of stover and intertownship trade of lucerne (current situation in Qingyang) are compared to survey data. In the model, there were on average 430 goats compared with 409 goats in the survey. The model slightly over predicted average annual per capita net income relative to observed data (RMB 5524 vs. RMB 4751). Scenario results When forage trade barriers are relaxed and no livestock-development subsidies are used, the area of lucerne fell by 35% and 20% for low LLR and high LLR households respectively (Table 6). Total cropped area remained constant as grain crop area increased. This resulted in a decline in grain purchases for human consumption. Livestock numbers rose by 24% in this scenario. Across the sample, lucerne sales fell from 100 tonnes to 41 tonnes and lucerne purchases rose from 100 tonnes to 136 tonnes. When livestock-development subsidies are introduced without a reduction in forage trade barriers, total income gains exceeded those achieved when forage trade barriers are reduced and no livestockdevelopment subsidies are used, namely a 6.2% income gain versus 3.3% (Table 6). Four less households live below the poverty line in this scenario, compared with no change in poverty levels in the scenario when barriers to trading in forages are reduced and no livestockdevelopment subsidies are used. When livestock-development 9 In this context, the average year refers to the average of the 5 years that are modelled.

17

A.M. Komarek et al. / Food Policy 37 (2012) 12–20 Table 5 Comparison of average of total sample with average for the two land-to-labour-ratio household groups for base case scenario. Variable

Units for base model

Total

Low land to labour (LLR) ratio

High land to labour (LLR) ratio

Net present value of total per capita net income Net present value of per capita net agricultural income Households living below the poverty line Off farm employment Total goat numbers Total area lucerne: grain Total net stover traded Total net lucerne traded Total grain purchased for home consumption

RMB (000) RMB (000) Number Years Number Ha Tonnes Tonnes Tonnes

21.9 4.4 14 1.5 365 44:60 0 0 7

20.9 4.2 12 1.5 309 28:39 33 26 6

24.6 4.7 2 1.5 56 16:21 33 26 1

Notes: Activity level results are for the average year and average household within the two groups. All goats are Liaoning goats raised in a warm shed. Total net stover, total net grain and total net lucerne traded = sales  purchases. No land renting occurs in the base case. The Qingyang-specific poverty line is 1196 RMB/per capita/annum (CSP, 2006). There are 69 low land to land households and 27 high land to labour ratio households in the sample.

Table 6 Comparison of average of total sample with average for the two land-to-labour-ratio household groups: percentage changes from the base case for the three experiments (poverty figures are the number of households living below the poverty line). Variable

Net present value of total per capita net income Net present value of per capita net agricultural income Households living below the poverty line Total goat numbers Total area lucerne: grain Total net stover traded Total net lucerne traded Total grain purchased for home consumption Notes:

⁄⁄

Livestock-development subsidies and forage trade barriers

No livestock-development subsidies and reduced forage trade barriers

Livestock-development subsidies and reduced forage trade barriers

Total

Low LLR

High LLR

Total

Low LLR

High LLR

Total

Low LLR

High LLR

6.1 29.1 10 17.9 32:23 0 0 57

6.7 30.9 9 19.4 14:10 33.6 115 23

5.0 25.4 1 9.1 65:48 33.6 115 269

3.3 17.9 14 24.3 30:21 0

4.4 23.7 11 24.4 35:25 23.2 294 25

0.9 4.7 3 23.8 20:15 23.2 70 37

10.4 60.4 7 67.5 11:8 0

9.0 51.0 6 48.1 7:5 121.8 283 6

13.3 81.7 1 173.8 47:34 121.8 71 306

⁄⁄

26

⁄⁄

46

In these scenarios, the base case value is zero so the percentage increase is undefined. No land is traded in any of the scenarios.

100 90

Cumulave per cent

80 70 60 50 40 30 20 10 0 0

3000

6000

9000

12000

15000

Net present value of per capita net agricultural income (RMB) No livestock-development subsidies and forage trade barriers (base case)

Livestock-development subsidies and reduced forage trade barriers

Livestock-development subsidies and forage trade barriers

No livestock-development subsidies and reduced forage trade barriers

Fig. 1. Shifts in the distribution of the net present value of per capita net agricultural income for the base case and three scenarios.

subsidies are introduced without a reduction in forage trade barriers, livestock numbers rose by 17.9%. The subsidies raise household incomes and alter crop-livestock enterprise mixes. In this scenario, lucerne area increased at the expense of grain area. The volume of lucerne traded increased and the volume of stover traded declined. Grain purchased for home consumption rose from 7.1 to 11.5 tonnes.

The simultaneous addition of livestock-development subsidies and the reduction of forage trade barriers increased total household incomes by 10.4%. This is because of the positive supply effects of lucerne-growing subsidies and the increased ability to source forages from other districts to support larger goat flocks (Table 6). The high LLR household group enjoyed an income gain of 13%, compared to the 9% gain realised among low LLR households. Total goat numbers owned by all households rose from 365 to 611. Goat numbers increased as more incentives are provided to grow lucerne and buy breeders. In addition, stover can be purchased and sold between the two townships and lucerne can be exchanged with districts outside Qingyang Prefecture. For example, purchases of lucerne rose by 35 tonnes, although sales fell by 59 tonnes. In this scenario, lucerne area increased by 11% and grain area fell by 8%. The subsidy increased lucerne area and the relaxation of forage trade barriers increased grain areas. When the two interventions are combined, the subsidy effect outweighs the forage trade barrier effect resulting in lucerne areas increasing and stover trade declining.

Distributional effects of scenarios Household assets and location resulted in varying scenario returns (Fig. 1). In the base case, the income coefficient of variation across the 5 years for low and high LLR households was 1.05 and 0.62 respectively. The higher income variability faced by the low LLR households is partly due to the more variable rainfall these households face. Reducing trade barriers induced a fall in the income coefficient of variation to 0.81 and 0.55 for the low and high LLR households respectively. An even sharper decline in the income coefficient of variation arose when subsidies are introduced in conjunction with the removal of trade barriers, namely 0.55 and 0.37 for the low and high LLR households respectively. The

18

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

policies implicit in the scenarios may be able to reduce income variability among the households.

Concluding remarks The analysis above highlights the impacts of livestock and forage measures designed to raise farm household incomes in Qingyang. Although specific to Qingyang, the measures broadly reflect policy directions in agricultural and semi-agricultural areas in several other parts of western China where poverty and concern over household incomes are greatest. Apart from the focus on livestock and forage measures, the analysis also highlights the extent to which improving forage trade flows may enhance any of the income generating effects of these policy measures. Better developed transportation systems and lower forage-trader entry costs may help unlock the benefits of more integrated forage-livestock systems. The use of subsidies alone appears to have larger income effects for households than the use of instruments that reduce forage trade barriers, although subsidies can introduce distortions that may be unsustainable in the long run. The greatest income growth and poverty reduction occurs when subsidies are introduced in conjunction with relaxing forage trade barriers. The measures analysed in this paper represent only part of the full suite of rural development measures and programs relevant in western China, which now fall largely under the ‘‘Building a New Socialist Countryside’’ blueprint (Su, 2009). Brown et al. (2009) highlight a range of rural development measures applied in Qingyang Prefecture including agricultural, labour migration, structural adjustment, science and technology, market and especially infrastructure programs. With regard to the latter, roads are a major investment item, and include local (inter-village) roads constructed by the Poverty Alleviation Bureau and main road improvements funded by the Development and Reform Commission. China’s large-scale investment in roads and other infrastructure are viewed as public goods used to stimulate broad-based investment across the services, industrial and primary (including agricultural) sectors. Thus, it is challenging to accurately attribute the costs of these infrastructure projects to a particular (small) agricultural activity (stover and lucerne trading). It is possible to more directly attribute costs to livestock-development policies. However, these policies also deliver environmental services that are not explicitly measured in this study. Moreover, households are not the sole beneficiaries of the environmental effects – there are positive environmental externalities for society (Liu et al., 2008). Given the inherent complexity involved in attributing costs to programs, the paper focuses on net income effects at the household level, which is often the over-arching objective of development programs. Despite local governments in western China generally have limited funding sources there has been a flood of transfers and investments in recent years from different government and donor agencies (Su, 2009). Regardless of funding levels, any agency will be conscious of a program’s economic efficiency, thus a comparative analysis of the development options in terms of their cost effectiveness and social efficiency is warranted. Nevertheless, given the reality that a mix of development measures will be pursued and especially in the context of regions with scarce resources, an analysis of how interventions impact on household incomes is crucial to better understanding alternative development pathways. This is particularly true given the incremental and trial-and-error approach that characterises policy making in rural China (Gulati et al., 2007). Livestock price and production risks undoubtedly influence enterprise adoption choices. In that the model covers multiple years and draws on information over a period (2004–2008) when

there was significant price and yield variability, some of the commodity price and production fluctuations have been captured (Tables 3 and 4). However, a systematic risk modelling approach has not been undertaken partly due to the focus on the dynamic elements of the model crucial in modelling the livestock and lucerne relationships. Accommodating risk more directly into the analysis (see, for instance, Hardaker et al., 2004) may be useful especially if risk management measures such as agricultural insurance are being considered. Acknowledgements The Australian Centre for International Agricultural Research funded this research. The authors thank Yuman Liu, Lihua Li, Chongyue Gao, Qin Jiang and Helan Bai for their diligent assistance during the data collection stage of this research. Valuable comments from Michael Robertson and two anonymous referees and a journal editor have benefited this article. Appendix A. Summary of main equations in the heterogeneousagent model The objective of the model is to maximise the net present value of total net household income across all households:

Z¼

H Y X X

I X

h¼1 y¼2004 i¼wheat ðmachine producedÞ



xh;i;y GMh;i;y þ FYM h;y FW h;y þ WMh;y WW h;y þ V h;2008 ð1 þ dÞr

 ð1Þ

where h = 1, . . . , H (index of households, with H = 96); i = wheat (machine produced), . . . , I (index of agricultural activities); y = 2004, . . . , Y (index of years, with Y = 2008); Z = total aggregate net income (2008 RMB); xh,i,y = level of agricultural activity i conducted by household h in year y; GMh,i,y = the gross margin of agricultural activity i conducted by household h in year y; FYMh,y = an integer variable representing whether or not a family member migrates away from household h in year y; FWh,y = the wage received for 1 years off-farm employment for a family member of household h in year y; WMh,y = an integer variable representing whether or not a family member migrates away from household h for a 6 month period over winter in year y; WWh,y = the wage received over winter in off-farm employment for a family member of household h in year y; d = the discount rate, set at 6%; and Vh,2008 is the value of livestock at the end of 2008. To discount the objective function, r = 0 in 2004, and increases sequential to r = 4 in 2008. Households 1–48 are in Shishe and 49–96 are in Quzi. Eq. (1) is maximised subject to constraints A1–A7, where A1–A6 are household-level constraints and A7 are township-level and district-level constraints: A.1. Land

Y X y¼2004

2 4

I X

i¼wheat ðmachine producedÞ

ROy 6 FS  0:25; RIy 6 FS  0:25;

8y 8y

3 xcr;y  ROy 5 6 FS þ

Y X

RIy

cr # i ð2Þ

y¼2004

ð3Þ ð4Þ

where cr = area of wheat (machine produced), area of maize (machine produced), area of wheat (draught power), area of maize (draught power) and area of lucerne; ROy = area of land rented out in year y; RIy = area of land rented in year y; and FS = total observed land area available under cultivation in survey year.

19

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

A.2. Livestock energy

X

Y X

2 4

f 2GCP;MJ y¼2004

6

M X m¼1

Y X

X

f 2GCP;MJ y¼2004

#

A.6. Livestock breeder equilibrium

2

3

I X

4

3

I B X X

½xi;y Df ;m;i;y  þ T f ;m;y 5 þ TDf ;y 5

i¼wheat ðmachine producedÞ

2 2 M X 4 4

I X

m¼1

i¼wheat ðmachine producedÞ

i¼eb b¼CCS y¼2004

½xi;y Sf ;m;i;y  þ T f ;m1;y þ Bf ;m;i;y 5

ð5Þ

where m = 1, . . . , M (index of monthly periods, with M = 12); where f = grams of crude protein (GCP), megajoules (MJ); Df,m,i,y = livestock activity i demand for f in month m in year y; Sf,m,i,y = supply of f by cropping activity i in month m in year y; Tf,m,y = net transfers of f from month m to m + 1; TDf,y = net transfers of f from December in year y to January in year y + 1; Bm,f,i,y = supply of f by buying crop i in month m of year y. A.3. Labour

m¼1 y¼2004

6

3

I X

4

2008 X

½1:05

i¼eb b¼CCS y¼2004

 xb;sb;yþ1  eb; rk; sb # i

ð9Þ

A.7. Township and district equilibrium constraints The amount of lucerne sold across all 96 farms equals the amount purchased. This indicates that lucerne can be traded between the two different townships.

½xh;sell

lucerne;y

 xh;buy

lucerne;y 

¼0

ð10Þ

h¼1 y¼2004

½xi;y LDm;i;y   HIm;y 5

48;96 X

i¼wheat ðmachine produced

M Y X X

I B X X

where eb = end breeder, rk = retain kid, sb = starting breeder, b = Inner Mongolian goats common shed (CCS), Inner Mongolian goats warm shed, native goats common shed, native goats warm shed, Liaoning goats common shed, Liaoning goats warm shed; xb,eb,y = number of ending breeders in class b in year y; xb,rk,y = number of kids retained in class b in year y; and xb,sb,y+1 = number of starting breeders in class b in year y + 1.

H Y X X

M Y X X

½xb;eb;y þ xb;rk;y  ¼

3

þ TDf ;y1

2

2008 X

Y I X X ½xh;ss;y  xh;bs;y  ¼ 0 bs; ss # i

ð11Þ

h¼1;49 y¼2004 i¼bs

½HLm;y  FYM m;y  WM m;y  HOm;y 

ð6Þ

m¼1 y¼2004

where LDm,i,y = technical input–output coefficients for activity i’s labour requirements in month m of year y, expressed in days; HIm,y and HOm,y = the days of labour hired in and out in month m of year y, respectively; HLm,y = the days of family labour available in month m of year y (based on total household size); FYMm,y = the days of family labour used in full year migration in month m of year y (an integer variable); and WMm,y = the days of family labour used in winter migration in month m of year y (winter migration lasts 6 months). A.4. Cash

where bs = maize stover purchased, wheat stover purchased and ss = maize stover sold, wheat stover sold. Within each township stover purchases cannot exceed sales. 48;96 X

Y X

2 4

h¼1;49 y¼2004

6

48;96 X

3

CR X

xh;cr;y þ ROh;y þ RIh;y 5

cr¼wheat ðmachine producedÞ Y X

FSh;y

ð12Þ

h¼1;49 y¼2004

The area of all land cropped with a township cannot exceed the amount of land available in that township. 48:96 X

M Y X X

½HIh;m;y  HOh;m;y  ¼ 0

ð13Þ

h¼1;49 m¼1 y¼2004 I X

Y X

xi;y ci;y 6

i¼wheat ðmachine producedÞ y¼2005

Y X

Within a township, the amount of labour hired in for local agricultural employment equals the amount supplied.

½NAIy1 þ ARy1 þ COy1 

y¼2005

ð7Þ where ci,y = the cost in RMB of activity i in year y; NAIy1 is net agricultural income in year y  1; ARy1 is remittances available to spend on agricultural activities in year y  1; and COy1 is unused cash in year y  1.

H Y I X X X h¼1 y¼2004 i¼1

xh;i;y Ph;i;y P

H Y I X X X

Dhh;i;y

ð14Þ

h¼1 y¼2004 i¼1

The sum of total household wheat and maize demand must not exceed the total amount of maize and wheat produced.

A.5. Crop material balance References Y X

I X

½xcr;y Pcr;y þ St cr;y1 

y¼2004 i¼wheat ðmachine producedÞ

¼

Y X

I X

½Dlcr;y þ Secr;y þ Stcr;y

y¼2004 i¼wheat ðmachine producedÞ

þ Dhcr;y  cr # i

ð8Þ

where Pcr,y = the yield of cr in year y; Stcr,y1 and Stcr,y = amount of cr retained in years y  1 and y, respectively; Dlcr,y = the amount of cr fed to livestock in year y; Secr,y = amount of cr sold in year y; and Dhcr,y = the amount of cr consumed by humans in year y.

Berger, T., 2001. Agent based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agricultural Economics 25, 245–260. Brown, C., Waldron, S., Longworth, J., 2008. Sustainable Development in Western China: Managing People, Livestock and Grasslands in Pastoral Areas. Edward Elgar Pub, Cheltenham. Brown, C., Waldron, S., Liu, Y., Longworth, J., 2009. Forage-livestock policies designed to improve livelihoods in western China: a critical review. China Agricultural Economic Review 1 (4), 367–381. China Statistics Press (CSP), 2006. Qingyang Yearbook. China Statistics Press Drafting Committee, Qingyang City. Dong, S., Gao, H., Xu, G., Hou, X., Long, R., Kang, M., Lassoie, J., 2007. Farmer and professional attitudes to the large-scale ban on livestock grazing of grasslands in China. Environmental Conservation 34 (3), 246–254. Fan, S., Chan-Kang, C., 2008. Regional road development, rural and urban poverty: evidence from China. Transport Policy 15 (5), 305–314.

20

A.M. Komarek et al. / Food Policy 37 (2012) 12–20

Goodman, D.S.G., 2004. The Campaign to ‘‘Open Up the West’’: national, provinciallevel and local perspectives. The China Quarterly 178, 317–334. Gulati, A., Fan, S., Delafi, S., 2007. The Dragon and the Elephant: A Comparative Study of Agricultural and Rural Reforms in China and India. International Food Policy Research Institute, Washington, DC. Han, J., Zhang, Y., Wang, C., Bai, W., Wang, Y., Han, G., Li, L., 2008. Rangeland degradation and restoration management in China. The Rangeland Journal 30 (2), 233–239. Hardaker, J.B., Huirne, R.B.M., Anderson, J.R., Lein, G., 2004. Coping with Risk in Agriculture. CABI, Wallingford. Herrero, M., Thornton, P., Notenbaert, A., Wood, S., Msangi, S., Freeman, H., Bossio, D., Dixon, J., Peters, M., Van de Steeg, J., 2010. Smart investments in sustainable food production: revisiting mixed crop-livestock systems. Science 327, 822– 825. Jia, Y., Li, F.-M., Zhang, Z.-h., Wang, X.-L., Guo, R., Siddique, K.H.M., 2009. Productivity and water use of alfalfa and subsequent crops in the semiarid Loess Plateau with different stand ages of alfalfa and crop sequences. Field Crops Research 114 (1), 58–65. Johnson, M., Masters, W., Preckel, P., 2006. Diffusion and spillover of new technology: a heterogeneous-agent model for cassava in West Africa. Agricultural Economics 35 (2), 119–129. Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J., Holzworth, D., Huth, N.I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L., Asseng, S., Chapman, S., McCown, R.L., Freebairn, D.M., Smith, C.J., 2003. An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18 (3), 267–288. Komarek, A.M., Ahmadi-Esfahani, F.Z., 2011. Impacts of price and productivity changes on banana-growing households in Uganda. Agricultural Economics 42 (suppl.), 141–151. Komarek, A.M., McDonald, C.K., Bell, L.W., Whish, J.P.W., Robertson, M.J., MacLeod, N.D., Bellotti, W.D., 2010. Whole-farm effects of livestock intensification in smallholder systems in Gansu, China. In: Proceedings of the CAER-IFPRI Annual International Conference on Agriculture and the Wealth of Nations; 16–17 October 2010, China Agricultural University. Li, Y., Huang, M., 2008. Pasture yield and soil water depletion of continuous growing alfalfa in the Loess Plateau of China. Agriculture, Ecosystems & Environment 124 (1), 24–32. Liu, J., Li, S., Ouyang, Z., Tam, C., Chen, X., 2008. Ecological and socioeconomic effects of China’s policies for ecosystem services. Proceedings of the National Academy of Sciences 105, 9477–9482.

Lu, C.H., van Ittersum, M.K., 2004. A trade-off analysis of policy objectives for Ansai, the Loess Plateau of China. Agriculture, Ecosystems & Environment 102 (2), 235–246. Lu, C.H., van Ittersum, M.K., Rabbinge, R., 2004. A scenario exploration of strategic land use options for the Loess Plateau in northern China. Agricultural Systems 79 (2), 145–170. McCarl, B.A., Spreen, T.H., 1997. Applied Mathematical Programming Using Algebraic Systems. http://agecon2.tamu.edu/people/faculty/mccarl-bruce/ books.htm (accessed 28.04.11). McDermott, J., Staal, S., Freeman, H., Herrero, M., Van de Steeg, J., 2010. Sustaining intensification of smallholder livestock systems in the tropics. Livestock Science 130, 95–109. National Research Council (NRC), 1981. Nutrient Requirements of Goats: Angora, Dairy and Meat Goats in Temperate and Tropical Countries. National Academy Press, Washington, DC. Park, A., Wang, S., Wu, G., 2002. Regional poverty targeting in China. Journal of Public Economics 86 (1), 123–153. Paten, A.M., 2008. Management of Livestock Nutrition under Subsistence Farming Conditions in Eastern Gansu Province, China. Bachelor of Science (Animal Science), School of Agriculture, Food and Wine, The University of Adelaide, Adelaide. Shen, Y., Li, L., Chen, W., Robertson, M., Unkovich, M., Bellotti, W., Probert, M., 2009. Soil water, soil nitrogen and productivity of lucerne-wheat sequences on deep silt loams in a summer dominant rainfall environment. Field Crops Research 111, 97–108. Squires, V., Hua, L., Zhang, D., Li, G. (Eds.), 2010. Towards sustainable use of rangelands in northwest China. Springer, Dordrecht. Su, M., 2009. China’s Rural Development Policy: Exploring the ‘‘New Socialist Countryside’’. First Forum Press, Boulder. Tarawali, S., Herrero, M., Descheemaeker, K., Grings, E., Blümmel, M., 2011. Pathways for sustainable development of mixed crop livestock systems: taking a livestock and pro-poor approach. Livestock Science 139 (1), 11–21. Uchida, E., Xu, J., Rozelle, S., 2005. Grain for green: cost-effectiveness and sustainability of China’s conservation set-aside program. Land Economics 81 (2), 247–264. Waldron, S., Brown, C., Komarek, A.M., 2011. China, the reconfiguration of GCVs and cashmere. In: Proceedings of the European Conference on Agricultural and Rural Development in China; 8–10 February 2011, Aarhus University.