冀县卿 刘守英 等：中国集体所有制下土地经营权的保障对提高农业生产效率是否重要？

本文刊于《China & World Economy》2021年第1期，原题为《Does Security of Land Operational Rights Matter for the Improvement of Agricultural Production Effciency under the Collective Ownership in China?》，未经授权禁止转载，经授权转载请注明来源！

文 | 冀县卿 刘守英 颜嘉楠 李友艺

Abstract

Under the “separation of three rights” policy, the impact of security of land operational rights on agricultural production efficiency has attracted much attention in recent years. Data envelopment analysis and mediation effect analysis were applied to 888 family farms run by new-type agricultural operators from Songjiang to identify the mechanism of the effect of land operational rights security on agricultural production efficiency through long-term investment. The results show that greater security of land operational rights generally increased agricultural production efficiency. Approximately 37.94 percent of the impact could be explained by long-term investment. The results also indicate that significant heterogeneity exists in the effect of land operational rights security on agricultural production efficiency at various levels of the family farms’ efficiency distributions. It is suggested that government should legalize land operational rights and give them a status equal to those of households’ contractual rights and land ownership rights in China’s future land tenure reform.

Key words: agricultural production efficiency, new-type agricultural operator, security of land operational rights

1

Introduction

China’s agricultural system is based on small-scale farms. This is a result of its household responsibility system (HRS), which was introduced between 1978 and 1984 (Che, 2012). The average farm size in China fell from 0.7 hectares in 1985 to 0.55 hectares in 2000 (Huang et al., 2012). The operation of small-scale farms led to lower agricultural production efficiency (Jia and Petrick, 2014; Chen et al., 2015). In 2014, the central government released a new rural land tenure reform called the “separation of three rights,” which aimed to encourage moderate-scale agricultural operations to improve agricultural production efficiency (Shen, 2015; Wang and Hu, 2016). The ratio of scale operation farmland area to actual cultivated farmland area in rural China had already reached 28.6 percent at the time of the third agricultural census in 2016. Liu and Long (2020) found that more farmland had been transferred to operate than in earlier years, and the ratio of farmland area transferred by family farms to the total transferred farmland was 58.38 percent in 2016. The “separation of three rights” policy may enable those new operators to transfer land, which will have a profound impact on agricultural production efficiency in China (Ma et al., 2017; Wang and Zhang, 2017; Zhou et al., 2018).

Empirical studies from Africa and other countries in Asia showed that improved security of land tenure may increase agricultural production efficiency by encouraging long-term investment (Barrows and Roth, 1990; Place and Otsuka, 2001; Michler and Shively, 2015), although the study by Migot-Adholla et al. (1991) stated that no relationship was found between security of land tenure and farmers’ investment behavior. For studies in China, with its collective ownership regime of rural land, Zhang et al. (2011) and Zhou et al. (2018) indicated that land tenure security increased agricultural production efficiency, but Ma et al. (2017) found that increasing land tenure security by providing land certificates to households had a negative impact on agricultural production efficiency. Gao et al. (2017) argued that farmers who operated on transferred-in land (with land operational rights) had less incentive to increase long- term investment than those who operated land for which they were responsible (with household contractual operation rights). Qiu et al. (2017) also found that transferred- in land tended to be less efficient than land for which the farmers were responsible. We found, on the one hand, that most existing studies concerning China’s rural land tenure only explored the effects of the security of whole contractual operation rights on agricultural production efficiency; on the other hand, studies examining the impact of land operational rights on agricultural production efficiency also had ambiguous results. The impact of the security of land operational rights on agricultural production efficiency therefore needs to be re-scrutinized against the background of the “separation of three rights.”

This study attempts to investigate empirically the relationship between the security of land operational rights owned by new-type agricultural operators under China’s unique land tenure system and agricultural production efficiency. To achieve this goal, we set three objectives. First, we examined the effect of security of land operational rights on agricultural production efficiency using instrumental variables (IV) to address the endogeneity of land operational rights. Second, we illustrated how and to what extent security of land operational rights affects agricultural production efficiency by considering the ratio of long-term investment to total investment. Finally, we highlighted the heterogeneous association between security of land operational rights and agricultural production efficiency using conditional quantile regression. The marginal contribution of this study is to shed light on the mechanism of the effect of security of new land operational rights on agricultural production efficiency using data from 888 family farms after the implementation of the “separation of three rights” policy, and to help policymakers formulate land tenure policies to improve China’s agricultural production efficiency.

The outline of the paper is as follows. Section II introduces the separation of rural land tenure under the collective ownership regime in China. Section III describes sample and data collection. Section IV describes the econometric models. Section V reports and discusses our results. Section VI concludes and considers policy implications.

2

Separation of China’s rural land tenure

under the collective ownership regime

Since the founding of the People’s Republic of China in 1949, the Chinese government has mandated a series of rural institutional reforms. Collective ownership was established as a form of socialist public ownership in China’s countryside. Since then, rural land tenure experienced three stages of change.

The first stage is that the land tenure was separated from the people’s communes and was transferred into the system of “three-level ownership with production teams as basic accounting units” in 1962, and this entitled production teams to more autonomous land-farming rights. However, incomplete ownership by production teams provided less incentive for the farmers to produce and led to agricultural production inefficiency.

The second stage is that land tenure was separated from production teams and transferred into the HRS in the late 1970s. The individual farmers who were members of the collectives were entitled to contractual operational rights relating to rural farmland. Under collective rural land ownership regimes, households’ contractual operation rights have gradually become stronger. Liu (2009) pointed out that the Chinese central government gave rural households the contractual operation rights to use, transfer, and receive proceeds from farmland to boost agricultural production efficiency. Specifically, the strengthening of land contractual operation rights could be seen as improving the right to transfer land to address the problem of low agricultural production efficiency. An official document named Notifications about Work Arrangement in Rural Areas in 1984 issued by the CPC Central Committee permitted the land contracted by rural households to be subcontracted on a voluntary basis. The land transfer principle is “voluntary, lawful, and compensatory,” and was codified into The Rural Land Contract Law of the People’s Republic of China, implemented in 2003.

The third stage is that the land tenure was subdivided into households’ contractual rights and land operational rights. These policies aimed to encourage land consolidation by land transfer. China’s land transfer rate reached 36.98 percent in 2017 (Liu, 2009). At the same time, China’s rapid urbanization and industrialization made farmers seek off-farm jobs and migrate to urban areas. Zhao et al. (2018) showed that the new- generation migrants, which are better educated and skilled, tend to migrate to urban areas with their families. Most of the new-generation migrants will never go back to rural areas to participate in agriculture production (Liu and Long, 2020). Hence, this led to the separation of households’ contractual rights and land operational rights, without changing the collective ownership nature of rural land. Given this situation, the “separation of three rights” policy was launched in 2014, calling for the implementation of land collective ownership rights, stabilizing households’ contractual rights, and releasing land operational rights.

Land operational rights were given to agricultural operators with farming rights. Agricultural operators have the operational rights to occupy, cultivate, and obtain proceeds from transferred lands with stable expectations for a contracted period, and to use land operational rights as mortgage to obtain finance. According to the provisions of the policies, the rights being transferred are land operational rights, while households’ contractual rights cannot be transferred. Wang and Zhang (2017) pointed out that the purpose of policymakers was to further provide agricultural operators with stable expectations regarding land use and to encourage them to increase agricultural production efficiency by releasing land operational rights.

China’s rural land tenure system has changed dramatically since 2014. Besides the “separation of three rights” reform, the Chinese central government also indicated that it aimed to build and foster “new-type agricultural operators” to improve agricultural production efficiency in a 2013 official document, Proposals on Speeding up the Development of Modern Agriculture and Further Enhancing the Vitality of the Rural Area Development (Lin and Wang, 2014; Gong et al., 2019). In 2020, The Ministry of Agriculture and Rural Affairs issued The Plan for High Quality Development of New- type Agricultural Operators and Service Entities (2020–2022) to further promote development of new-type agricultural operators, which comprised family farms, large specialized farms, agribusinesses, farmers’ professional cooperatives, and others, engaging in large-scale, intensive, and commercialized agricultural operations. Among those new-type agricultural operators, Graeub et al. (2016) pointed out that family farming is the predominant form of agriculture based on global agricultural census data. According to survey data from the Ministry of Agriculture and Rural Affairs of China, there were approximately 600,000 family farms recorded by the end of 2018, which was a fourfold increase over the number of family farms in 2013. The family farms occupied 1.07 million hectares, of which 71.7 percent had been transferred from smallholders or village collectives to family farms in 2018. The rapid development of new-type agricultural operators in China’s agricultural system has led to a need to motivate them to enhance their production efficiency. We will therefore consider the land tenure practices associated with the “separation of three rights” in Songjiang and conduct further detailed empirical study.

3

Sample description

and data collection

1.Sample description

The data used in this study were obtained by a cross sectional survey in Songjiang district in Shanghai, China. Songjiang is located on the upper reaches of the Huangpu River, southwest of Shanghai and eastern coast of China, with a total area of 604.64 km2. It is located at the bottom of the dish-shaped lowland in the Taihu Basin, with a subtropical and humid climate. Songjiang features mostly plains, which are extremely flat compared with the rolling lowlands. At the same time, the effective irrigation rate of the cultivated land has reached 100 percent. By the end of 2017, the sown area of grain was 10,533.33 hectares in Songjiang. This region has two main crops grown annually – rice and wheat – which occupied 10,200 hectares and 333.33 hectares, respectively in 2017. In the same year, the family farms of Songjiang operated 9,266.67 hectares of farm, accounting for 95 percent of the region’s total sown grain area.

To improve agricultural production efficiency, since 2007, the Songjiang district government has made a special arrangement for land transfer to develop family farms. First, the contracted farmland was transferred from the collective farmers with land operation rights to the village collectives. After transferring out the land, the collective farmers received an annual rent equivalent to 250 kg paddy per mu (1 mu = 0.0667 hectares), and the elderly people whose land contractual operation rights were entirely transferred to village committees (males aged above 60, females aged above 55) received a monthly allowance of 150 yuan. Second, the village committees transferred the consolidated and moderate-scale farmland (about 100–150 mu) to family farms with land operational rights in the form of official land operational contracts. Since 2007, the Songjiang district government has tried its best to encourage family farms as new-type agricultural operators to improve grain production efficiency.

Village committees determine the purpose of land use and farm size; sign land contracts with collective farmers; select, administer, and dismiss new-type agricultural operators; and assess the performance of family farms. Farmers, as members of a collective, have rights to decide whether to transfer their land operational rights to obtain transferring fees while keeping land contractual rights, or to withdraw their contractual operation rights in exchange for pension insurance payment. Family farms as new-type agricultural operators have the ultimate rights over the decision of what and how to produce. Besides agricultural income from the sale of their products, family farms also receive subsidies for grain, seeds, machinery, and land transfer from central, municipal, and district governments, respectively.

This paper uses a unique data set from a family farm-level survey in Songjiang to investigate the impact of land operational rights security on the production efficiency of new-type agricultural operators for the following reasons: first, family farms’ land operational rights have been separated from households’ contractual operation rights for 11 years in Songjiang. Their practices can provide valuable evidence for the “separation of three rights” policy. Second, agricultural production efficiency is associated not only with land operational rights but also with local social and economic environment, regional climate, natural resources endowment, crop varieties, and cropping technologies, these are almost the same production conditions faced by Songjiang family farms. This is, therefore, suitable for us to identify the effect of the security of operational trights on agricultural production efficiency.

2.Survey and data collection

This paper draws on a unique data set from our interviews in 2017. In August 2017, we collected all of the input information and partial output information related to the family farms’ production. In March 2018, we collected the remaining output information regarding those family farms’ production. All of the data used in the paper were collected by the authors in collaboration with colleagues at Nanjing Audit University, Renmin University of China, and Nanjing Agricultural University. The survey covered all 11 towns in Songjiang in 2017, including Chedun, Dongjing, Gongyequ, Maogang, Sheshan, Shihudang, Xiaokunshan, Xinbang, Xinqiao, Yexie, and Yongfeng. There was a total of 945 family farms in Songjiang by August 2017. Fifty-five family farms were involved in poultry-raising activities, which were different from crop family farms. There were two invalid questionnaires from the family farms. The sample for this paper therefore included 888 crop family farms.

The questions for each family farm covered three essential inputs to the farm to generate grain production: land, labor, and capital (Kumbhakar and Lovell, 2003). The input indicators are included in the present paper as follows: (i) land is measured as the operational farm size of the family farms; (ii) labor is defined as total number of family farms’ family workers and hired workers working on the farms in 2017; (iii) capital is calculated by the total expenditure on fertilizer, organic fertilizer, pesticide, machinery services, green manure, depreciation of agricultural machinery, and other inputs necessary for grain production (e.g. expenses for drainage-irrigation, fix machinery, and diesel for machinery). The interviewed owners of family farms were also asked for information about the outputs (e.g. sales revenue) of rice and wheat. The output value is defined as the total value of the family farm’s income and subsidies, and is used as an output indicator to estimate the family farms’ efficiency.

4

Econometric models

1.Evaluating the efficiency of family farms: Data envelopment analysis

Data envelopment analysis (DEA) with the assumption of constant returns to scale (CRS) was first introduced by Charnes et al. (1978), and has been widely used to evaluate the relative efficiency of a set of decision-making units (DMUs) with multiple inputs and outputs in the agricultural sector. Data envelopment analysis is a non-parametric method with the advantage that it does not require assumptions about the fundamental data distribution and the assumption of a functional relationship between the inputs and outputs (Choo et al., 2018). Because of the specificity of the agriculture sector, which relies on limited inputs, an input-oriented model, which measures the ability of a DMU to minimize inputs while maintaining the same level of outputs, is more appropriate (Javed et al., 2010; Toma et al., 2015). However, it is impossible for each DMU in practical production to meet the CRS assumption that every DMU is set as CRS. Banker et al. (1984) developed the DEA by taking the assumption of variable returns to scale (VRS) into account, which is in line with real production practice. In our study, each family farm is applied as a DMU, which may be in different stages of production (increasing, constant, or decreasing to scale). The input-oriented DEA model, with the assumption of VRS, was therefore proposed to estimate the relative production efficiency of family farms in our paper. The constrained optimization model for DMUk (k represents a family farm) is specified in Equation (1):

where θ represents the production efficiency score, generally between 0 and 1, which is set as the upper limit, and λ represents a coefficient vector of weights, which defines the linear combination of the peers of the jth family farms (Javed et al., 2010); x and y represent the input vector and output vector of the family farms, respectively. In this paper, i is the number of inputs of family farms (i = 1, 2, 3), x1 denotes land, x2 denotes labor, x3 denotes capital. The j represents jth family farms in this paper (j = 1, 2, …, n). The kth family farm represents the family farm that is evaluated. The optimum solution of Equation (1) is θ*, which represents the optimal efficiency value of the kth family farm. If θ* = 1, it indicates that the kth family farm is relatively efficient and is on the grain production frontier (Coelli et al., 1998), and indicates that the current input level cannot be reduced (Pradhan and Kamble, 2015). If θ* < 1, it means the kth family farm is relatively inefficient, hence there is a possibility of reducing current input level for the same output (Pradhan and Kamble, 2015). The smaller θ* is, the lower the efficiency of the kth family farm will be, which implies that there is more possibility for reducing the current input level to obtain the same level of output. Other things being equal, 1 – θ* represents the maximum probability for reducing the kth family farm’s input level without changing the output level. Table 1 presents the summary statistics of the major input and output variables used in our DEA model.

2.Benchmark model

To investigate the relationship between security of land operational rights and agricultural production efficiency, we set up the following benchmark model:

where Efficiency represents the agricultural production efficiency of the jth family farms, and RS represents the security of land operational rights measured by the duration of the land operational contract. Longer contract duration provides greater land tenure security (Zhou et al., 2018). We also employ family-farm level data to create control variables (Controls). Four control variables are used to measure characteristics of the family farm owners, including Age (age of the family farm owner, measured in year), Education (education of the family farm owner, measured in year), Experience (time of the family farm owner engaged in agriculture production, measured in year), Drive_License (whether or not the family farm owner has a driving license for agricultural machinery, Yes = 1, No = 0). Three control variables are used to measure family farms’ characteristics, including Farm_Size (hectare), Family_Labor (the number of family laborers measured by person), and Going_Concern (whether or not the family farm operates continuously? Yes = 1, No = 0). One control variable Government_Policy is used to measure government supporting policy (ratio of government subsidies received by family farms to their incomes).

When land operational rights security is an exogenous variable, β0 in Equation (2) can be used to examine the impact of land operational rights security on agricultural production efficiency. However, the security of land operational rights is affected by some unobserved factors, which might also influence the level of agricultural production efficiency, implying the potential existence of endogeneity-correlated omitted variables. There may also be a reverse causality relationship between agricultural production efficiency and security of land operational rights. Without considering the endogeneity, the estimation results of the effect of land operational rights security on agricultural production efficiency are biased. We therefore use IV estimation to solve potential endogeneity problems (Greene, 2008; Wooldridge, 2018). Generally, characteristics of geography, such as geographical distance, could be used as potential valid instruments (Easterly and Levine, 2003; Redding and Venables, 2004). Thus, we use Distance, which is the distance between a village and the location of Songjiang district government, as an IV.

Distance is used as a valid instrument for the duration of land operational contract for two reasons (Wooldridge, 2018). First, geographical location may affect contract duration through its influence on the cost of economic behaviors (Redding and Venables, 2004). The Songjiang district government provides information about land transfer (including the policies, procedures, contract terms, and farm size), examines family farm operators’ qualifications, handles transfer disputes involving land operational rights, supervises family farms’ production, and evaluates family farms’ performance (Wang and Zhang, 2017). Family farms in a certain village closer to the location of Songjiang district government may obtain more specific information on land operational rights and policy support, and may have lower risks and transaction costs in land operational rights transfer, and are more likely to sign long-term contracts (Gao et al., 2019). Second, the geography variable represents the relative geographical distance, which cannot directly affect economic behaviors; it is reasonably regarded as an exogenous variable (Liu et al., 2007; Li et al., 2011). Hence, Distance is an exogenous variable that should not affect the production efficiency of family farms directly.

3.Mechanism analysis of mediation effect

To understand better how the mechanism of land operational rights security affects agricultural production efficiency, we use the ratio of long-term investment to total investment to test the mediation effect. The mediation effect is applied to account for the hypothesis of the mediating variable, when the independent variable and dependent variable are related (Edwards and Lambert, 2007). A widely used test method for the mediation effect is the causal steps approach introduced by Baron and Kenny (1986). The basic causal chain involved in the conceptual mediation effect mechanism in this paper is shown in Figure 1. The following three conditions must be satisfied for the mediation effect (Baron and Kenny, 1986; Kim et al., 2014). Condition 1: the independent variable should be significantly related to the dependent variable (path β in Figure 1). Condition 2: the independent variable should strongly account for mediating variable (path ρ in Figure 1). Condition 3: the mediating variable should be associated significantly with the dependent variable (path γ in Figure 1).

Assuming that the three conditions are satisfied, the meditation effect is significant. It can be categorized into two types of meditation effect from a theoretical perspective. One is a partial meditation effect showing a significantly decreased degree of relationship between the independent variable and the dependent variable when considering the mediator (mediating variable), while the independent variable is still significantly associated with dependent variable. The other is the full mediation effect showing the major influence of independent variable on dependent variable is through mediator. In this paper, based on the benchmark model Equation (2), our mediation effect mechanism can be expressed as Equations (3) and (4). Investment will be regressed on RS, and Efficiency will be regressed on both RS and Investment.

where Investment is the ratio of the long-term investment to the total investment of grain production undertaken by family farms. Family farm operators usually have two types of investments to support the production of both rice and wheat short-term and long-term investment. The inputs of grain production, including the fertilizer, pesticide, machine services, and other inputs, are defined as short-term investments, while the costs including deprivation charges of machinery (fixed) assets, green mature,and organic fertilizer are defined as long-term investment (Wang et al., 2017). Table 2 presents summary statistics of the variables used in Equations (3) and (4).

5

Empirical results

1.Agricultural production efficiency of family farms in Songjiang

The first task of this study is to determine the family farms’ production efficiency based on the data from 888 family farms from Songjiang using the DEA technique. The mean of overall family farm production efficiency is estimated, which is 0.37 with minimum level of 0.18 and maximum level of 1 (Table 3). This suggests that if the sample family farms in Songjiang practiced at full operational level, they could increase efficiency by 63 percent (1 – 0.37) with available technology.

To observe how well family farms in Songjiang have operated their farmland we analyze the frequency of different efficiency levels for farmers’ operation in Table 3. Table 3 shows that the production efficiency levels of the majority of all sample family farms range from 0.20 to 0.40. It reveals that approximately 659 of 888 family farms have efficiency scores between 0.20 and 0.40. Specifically, of all 888 sample family farms, 85.69 percent of the family farms had production efficiency scores less than 0.50, 6.53 percent from 0.50 to 0.60, 3.72 percent from 0.60 to 0.70, 1.58 percent from 0.70 to 0.80, 1.35 percent from 0.80 to 0.90, and only 1.13 percent more than 0.90. These results indicate that a very large proportion of family farms are far from operational efficiency, implying there is much room for them to improve production efficiency with available technology.

2.Land operational rights security and agricultural production efficiency

We use different methods to estimate the impact of security of land operational rights on agricultural production efficiency. The effect of security of land operational rights on production efficiency is presented in Table 4. First, OLS is used to estimate Equation (2). The result shows that the RS coefficient is statistically significant at the 1 percent level, indicating that land operational rights security significantly and positively influenced production efficiency (row 1, column (1), Table 4). If all the assumptions of the classical linear regression model do not hold, application of OLS regression in such a scenario may give an incomplete picture and yield biased estimates (Verbeek, 2008). Then, quantile regression (QR) is developed to estimate the entire conditional distribution of the dependent variable based on minimization of weighted absolute deviations (Koenker and Bessttt, 1978; Koenker and Hallock, 2001). Median regression using symmetric weights is preferred to mean regression to reduce susceptibility to outliers (Cameron and Trivedi, 2010). We find that RS has a significantly positive impact on Efficiency from QR estimator with a quantile of 0.5 (row 1, column (2), Table 4), as well as from the IV estimations using both IV–2SLS and IV–QR estimators with a quantile of 0.5 (row 1, columns (3) and (4), Table 4). The condition 1 in Section IV.3, which is that the independent variable (RS), should be significantly related to the dependent variable (Efficiency), is therefore satisfied.

To check the validity of our instrument variable, in the first place, a Durbin–Wu– Hausman (DWH) test of endogeneity was applied to test whether the land operational rights security variable was indeed endogenous. We find that the endogeneity test p-value is 0.00 (Table 4), implying land operational rights security is indeed an endogeneity variable. Under these circumstances, the IV estimation might be unbiased. We use the Cragg-Donald Wald F statistic to test whether weak identification (ID) occurs in the instrument variable. The result of the Cragg–Donald Wald F statistic is 20.55, which is larger than the result of Stock–Yogo weak identification test critical value of 16.38 (Table 4), suggesting that the IV estimation has no weak instrument problem.

3.Mechanism analysis

In this subsection we use Equations (3) and (4) to test whether there is a mediating effect (Baron and Kenny, 1986; Wen et al., 2004); that is, whether security of land operational rights affects production efficiency through investment. We run Equation (3) using IV– 2SLS and IV–QR estimators with a quantile of 0.5. The results show that the coefficient of RS is statistically positive at the 5 and 1 significance levels in the mediation function (row 1, columns (2) and (4), Table 5). The results from the IV test indicate that our instrument variable is validity. Condition 2 in Section IV.3, which is that the independent variable (i.e. RS) must affect the mediator (i.e. Investment) is therefore satisfied. The results indicate that the greater security of land operational rights enhanced family farm operators’ stable expectations and ability to maintain long-term use over their land, therefore, family farm operators will have greater incentive to undertake long-term investment.

Then, we use IV methods to estimate Equation (4) named the overall equation, which inserted the mediator (investment) into the benchmark Equation (2). The results show that Investment is significantly correlated with Efficiency when RS is controlled (row 3, Table 6). Evidence from the endogeneity test (the endogeneity test p-value is 0.01) shows that the hypothesis that the variable is exogenous is rejected at 5 percent level of significance. The result of Cragg–Donald Wald F statistic of weak identification is 14.89, which is larger than the result of the Stock–Yogo weak ID test critical value 8.96 (column (2), Table 6), suggesting that the IV estimation has no weak instrument problem (Andrews and Stock, 2005; Xu et al., 2020). Condition 3 in Section IV.3, which is that the mediator (investment) must affect the dependent variable (Efficiency), is satisfied. After the three conditions were satisfied, RS had a significantly positive effect on Efficiency at 5 percent significance (row 1, column (4), Table 6), implying that partial mediation effect was inferred when the investment was controlled.

4.Robustness check

As mentioned above, according to the causal steps approach introduced by Baron and Kenny (1986), there is a partial mediating effect in our study; that is, RS affects Efficiency indirectly through Investment. Here we use another approach developed by Wen and Ye (2014) to test whether there is a partial mediating effect. The results are consistent with the test results using the casual steps approach, confirming the stability of the partial mediating effect findings of our study. Using this kind of approach, the percentage of mediating effect to total effect should be reported if a partial mediating effect exists. Specifically, for the median (percentile = 0.5) regression, based on Tables 4–6, we calculate the partial mediation effect of Investment on Efficiency (Table 7). The effect equals the impact of RS on Investment (0.0343) multiplied by the impact of Investment on Efficiency (0.5221), and then divided by the effect of RS on Efficiency (0.0472). The size of the effect of security of land operational rights on agricultural production efficiency through long-term investment is approximately 37.94 percent (column 4, Table 7).

5.Heterogeneity of the correlation between land operational rights security and agricultural production efficiency

The analysis of heterogeneity can help to identify which subgroups of family farms could increase or decrease their production efficiency when security of land operational rights improves or worsens. This paper highlights the heterogeneous relationship between security of operational land rights and production efficiency using the IV–QR method by estimating eight various percentiles (20th, 30th, 40th, 50th, 60th, 70th, 80th, and 90th percentiles) of efficiency. The results indicate that significant heterogeneity exists in security of land operational rights at various levels of the distribution of agricultural production efficiency. The results show that the effect of security of land operational rights on agricultural production efficiency is positive and tends to increase along the various efficiency percentiles ranging from the 20th to 90th percentiles, without including the mediating variable of Investment (column (1), Table 8). After controlling for Investment, the results show that there is no significant relationship between security of land operational rights and agricultural production efficiency at the 90th percentile (column (2), Table 8). The results also show that, although the magnitude of the security of land operational rights estimation decreases, the significance still remains; this also implies that Investment played a significant role when land operational rights security affected agricultural production efficiency.

6

Conclusions and policy implications

Using the DEA technique, this paper revealed that the production efficiency of the sample family farms from Songjiang could be improved by 63 percent with available technology. After using Distance as a valid instrument to address the potential endogeneity of RS, we found that greater security of land operational rights generally increased agricultural production efficiency. Approximately 37.94 percent of the impact of security of land operational rights on agricultural production efficiency could be explained by the family farms’ long-term investment. The results also showed that there was significant heterogeneity for the effect of land operational rights security on agricultural production efficiency across various efficiency quantiles.

Our study has far-reaching policy implications as far as agricultural production efficiency is concerned. China’s future agricultural production efficiency will depend mainly on agricultural operators, especially the new-type agricultural operators. The security of land operational rights under the new rural land tenure system should therefore be considered as a higher priority in China’s future agricultural development. The central government should legalize land operational rights and give equal status to households’ contractual rights and land ownership rights to further stabilize the investment expectation of agricultural operators. Only when land operational rights are clearly stated in a legal form can the security of land operational rights be guaranteed. The local government should also make target policies according to the different production situations experienced by the new-type agricultural operators to both secure the land operational rights and improve agricultural production efficiency.

（参考文献与注释略）

编辑：程果  熊雪锋