Introduction

The grassland ecosystem is one of the most important terrestrial ecosystems globally, maintaining the livelihoods of more than 1.3 billion people (Herrero et al. 2013) and ensuring 30% of the global meat food supply (Hou et al. 2021). However, overgrazing of open pastures under public or collective property rights has led to grassland degradation worldwide and restricted the sustainability of grassland animal husbandry production (Pulido et al. 2018). Under the influence of the “tragedy of the commons” (Hardin 1968) and Coase’s property rights theory (Coase 1960), defining private property rights has become the mainstream means for many developing countries in Asia and Africa to solve the disorderly competition of open grassland (Su et al. 2021; Undargaa and McCarthy 2016). For example, China, Mongolia, Kenya, and Botswana gradually privatized grassland at the end of the 20th century (Banks 2003; Galaty 2016; Lesorogol 2008). The implementation of private property rights solves the problem of competition for public common pastures but leads to other unintended ecological and livelihood consequences (Li and Huntsinger 2011; Shi et al. 2023b). Grassland privatization is usually accompanied by herders’ settlement, pasture area reduction, and pasture fencing (Tan et al. 2018b), which leads to grassland fragmentation, further leads to the decline of traditional nomadic modes (Gongbuzeren et al. 2018), and breaks the integrity of the grassland ecosystem and the stability of the animal husbandry production system (Boone and Hobbs 2004; Hobbs et al. 2008; Shi et al. 2021).

In Ellis and Swift’s Non-equilibrium Ecosystem model, the nomadic movement strategy is the core means for herders to maintain the elasticity of the grassland grazing system under the dual pressure of livestock production growth and climate change, especially extreme drought (Ellis and Swift 1988). Settled grazing reduces the resilience of the grassland ecosystem and the livelihood of herders, which is considered an important inducement for grassland ecological degradation and the vulnerability of herders’ livelihood (Gongbuzeren et al. 2018; Li and Huntsinger 2011; Tan and Tan 2017). Therefore, in the context of grassland privatization, how to promote herders to optimize their grazing modes from unreasonable small-scale high-intensity grazing to more reasonable rotational grazing is a key issue related to grassland ecological restoration and the production sustainability of the grazing system. Still, it has received less research attention (Shi et al. 2021). Although some studies have proposed restoring the rotational grazing behavior through joint household grazing and grazing quota management (Gongbuzeren et al. 2018; Qi and Li 2021), it has not fundamentally solved the “second-order dilemma” of the supply of the above collective action institution (Ostrom 1990).

The grassland rental market is an important way to reallocate grassland resources in grassland privatization countries (Shi et al. 2022; Tan et al. 2018a), which provides opportunities to improve grazing modes (Feng et al. 2022). With urbanization, the increase of non-pastoral employment opportunities for herders, and the deepening of grassland privatization, the grassland rental market came into being in some developing countries (Su et al. 2021). For example, some recent micro-survey data from China show that grassland rental has been prevalent in main pastoral areas, and the incidence of grassland rental is as high as 30–70% (Huang et al. 2017; Su et al. 2021; Tan et al. 2018a). Under the market economy, grassland rental is the spontaneous reallocation of livestock production factors under the economic rationality of herders (Su et al. 2021). It has strong endogenous power (Hu et al. 2014) and avoids the problem of the institutional supply of collective action. Therefore, if grassland rental can promote the transformation of herders’ grazing mode, it will provide an effective solution for the sustainable development of grassland animal husbandry. However, in carrying capacity management under the discourse system of equilibrium ecosystem, overgrazing is considered the main human factor in grassland ecological degradation (Shi et al. 2023a). The existing research on grassland rental has focused on evaluating the impact of grassland rental on herders’ grazing intensity or overgrazing (Feng et al. 2022; Su et al. 2021), ignoring the impact on their grazing modes. Whether grassland rental can promote the optimization of herders’ grazing modes needs further research. China is the second-largest grassland country globally, accounting for 12% of the world’s grassland area (Hua and Squires 2015). It is also a typical example of grassland privatization by implementing the grassland household responsibility system (HRS) (Liu et al. 2020). Meanwhile, China’s grassland rental market has developed rapidly in recent years. Therefore, clarifying the impact of grassland rental on Chinese herders’ grazing modes has a strong reference significance for other countries with private grassland property rights.

To our best knowledge, only a few studies qualitatively describe the impact of grassland rental on herders’ grazing modes based on the case analysis of individual pastoral villages in China (Gongbuzeren et al. 2016; Qi and Li 2021). There is no systematic theoretical analysis and empirical test on the impact of grassland rental on herders’ rotational grazing behavior. Based on the survey data of large sample herders in pastoral areas of northern China, our study theoretically analyzes and empirically tests the impact, mechanism, and heterogeneity of grassland rental on herders’ rotational grazing. The main contributions of the study are as follows: First, in the context of HRS, from the perspective of grassland resource reallocation, it provides empirical evidence for grassland rental to promote the optimization of grazing mode and provides new empirical support for the development of grassland rental market in countries with grassland property right privatization. Second, theoretically analyze and empirically test the path of grassland rental promoting herders’ grazing mode from high-intensity continuous grazing to rotational grazing, deepening the knowledge of grassland rental’s ecological impact. Third, clarify the heterogeneity of grassland rental’s impact on optimizing herders’ grazing modes; that is, it reveals the context-dependence of grassland rental’s impact on herders’ rotational grazing behavior.

The remainder of this paper is organized as follows. Section 2 presents the theoretical analysis and conceptual framework. Section 3 introduces the materials and methods. Section 4 presents the results and discussion. Finally, Section 5 provides the conclusions and implications.

Theoretical Analysis and Conceptual Framework

Shock of Grassland Privatization on Grazing Modes in China

The grassland property rights privatization process has referred to some experiences of the HRS in farming areas (Li et al. 2021). As the reform of grassland property rights follows the principle of equal allocation, grasslands of different quality are divided, and each household gets at least one grassland plot at a different quality level, which leads to the fragmentation of the originally integrated grassland collectively ownedFootnote 1 (Su et al. 2021). Since then, the separation of herder households has led to more subdivision of grassland areas, further reducing the contracted grassland area of a single herder and aggravating grassland fragmentation (Tan et al. 2018a; Zhang et al. 2020b). The range of livestock activities is gradually reduced, and the traditional nomadic grazing is gradually replaced by small-scale high-intensity continuous grazing (Li and Huntsinger 2011). Since then, the fence construction to clearly define private property rights has solidified the fragmentation pattern of grassland and completely broken the integrity of the grassland ecosystem (Gongbuzeren et al. 2015). Fragmentation subverts the nomadic mode in a large space in the traditional animal husbandry system, intensifies the grazing intensity of grassland in a small range, inhibits the self-restoration ability of grassland ecology, and leads to serious degradation of grassland (Hobbs et al. 2008; Li and Huntsinger 2011). In addition, the small-scale, fragmented, and decentralized management pattern also limits the ability of livestock to obtain heterogeneous resources (such as water and different grass species) (Boone and Hobbs 2004), inhibits the mobility and flexibility of livestock, and reduces the ability of grazing system to avoid the risk of local forage fluctuation caused by natural disasters such as drought (Hobbs et al. 2008), which may lead to the low efficiency of animal husbandry production and aggravate the overuse of grassland. It is not conducive to the sustainability of animal husbandry management of herders.

Non-equilibrium Characteristics of Grassland Ecosystem in Northern China

Ellis and Swift (1988) proposed the non-equilibrium ecosystem theory based on long-term observation and research on African grasslands and herders. They believed that grazing intensity is not the only factor affecting grassland vegetation. The impact of abiotic factors such as precipitation and another climate on grassland is sometimes even more important. They pointed out that in the non-equilibrium ecosystem, traditional herders do not achieve grassland ecological restoration in local areas by controlling grazing intensity but maintain long-term survival by expanding moving range, reducing human demand, or increasing the supply of substitutes (Fernandez-Gimenez and Allen-Diaz 1999). The value of expanding the moving range is to distribute the grazing pressure more widely and evenly in the pastoral area or even obtain the grassland resources outside the local system. This theory has also been supported and applied by Fernandez-Gimenez and Allen-Diaz (1999), Sullivan and Rohde (2002), Zhang et al. (2013) and other scholars. Under the discourse system of non-equilibrium ecosystem theory, the grassland ecosystem is mainly driven by random abiotic factors (mainly precipitation reduction). Controlling grazing intensity cannot play a decisive role in grassland ecological restoration, and the mobility of livestock in a larger space is more critical to promoting grassland ecological restoration (Ellis and Swift 1988). Some scholars concluded that the grassland in northern China has similar non-equilibrium characteristics to African grassland based on the long period of climate data analysis (Li and Li 2012). Grassland in the arid and semi-arid areas has high climate variability, large interannual rainfall variation, extremely uneven temporal and spatial distribution, and extremely uncertain vegetation growth (Wang et al. 2014). So, it is more suitable to diagnose the grassland ecosystem in northern China using the non-equilibrium ecosystem theory or combining it with the equilibrium ecosystem theory (Li and Li 2012; Sullivan and Rohde 2002).

Reasonable Grazing Mode: Rotational Grazing

Reasonable grazing mode refers to the grazing mode adopted by herders in the process of grazing adapted to the growth law of grassland vegetation. According to the non-equilibrium ecosystem theory, the resource distribution of the grassland ecosystem has a high degree of temporal and spatial heterogeneity (such as precipitation, forage species, etc.) (Ellis and Swift 1988). In dry years, forage growth is poor, and even mild grazing will lead to vegetation degradation. Traditional herders avoid the adverse effects of climate disasters such as drought on forage growth through nomadic and transfer strategies to provide time and space buffer for vegetation restoration and adapt to the internal working mechanism of the grassland ecosystem (Ellis and Swift 1988; Oba et al. 2000). Therefore, seasonal rotational grazing can be considered a reasonable grazing mode. Based on the non-equilibrium theory or the local knowledge formed by herders’ long-term animal husbandry practice, many scholars have analyzed seasonal rotational grazing’s ecological and livelihood rationality (Fernandez-Gimenez 2000; Zhang et al. 2020a; Zhang et al. 2007; Zhuang et al. 2019). Compared to the large-scale nomadism, the high-intensity settled grazing on a small scale violates the internal working mechanism of vegetation growth and the high variability of the non-equilibrium ecosystem and intensifies grassland degradation (Wang et al. 2020). Therefore, continuous grazing with high intensity in a small range can be considered an unreasonable grazing mode.

Some pastoral areas have developed regional rotational grazing in a small range based on settled grazing (Shi et al. 2021; Teague et al. 2013). Regional rotational grazing can be regarded as a “condensed version” of traditional nomadism. Its operation mode has the concept of nomadism (Ren 2012). By dividing herders’ grassland into multiple independent blocks with fences, grazing plans are formulated. Livestock is grazed in turn in each independent block for a certain period according to a certain order to provide buffer time for vegetation restoration in each grazing block (Ren 2012; Wang et al. 2020; Zhou et al. 2015). Based on grazing experiments, many scholars globally have verified the ecological advantages of regional rotational grazing in promoting vegetation growth, soil restoration, livestock growth (Baronti et al. 2022; Cristina Vecchio et al. 2019; Lawrence et al. 2019; Li et al. 2020; Ravetto Enri et al. 2017). Therefore, regional rotational grazing can also be considered a reasonable grazing mode. In 2016, “the 13th Five-Year Plan for National Grassland Protection, Construction, and Utilization,” issued by the Ministry of Agriculture of PRC, took the regional rotational grazing of natural grasslands as a key indicator to reflect the level of scientific utilization of grasslands. It proposed that natural grasslands’ regional rotational grazing area should cover more than 28 million ha by 2020.

Influence of Grassland Rental on Herders’ Rotational Grazing

Combined with the existing theories and literature analysis, grassland rental may encourage herders to adopt rotational grazing behavior by saving fence costs and expanding grazing space.

First, grassland rental may increase the number of grassland plots with fences, save fence expenses, and promote herders to adopt regional rotational grazing. Regional rotational grazing needs to divide the grassland into multiple independent blocks with fences to graze livestock in each independent block for a certain time in a certain order (Briske et al. 2008). Fences construction and maintenance is the main cost of implementing regional rotational grazing (Wang et al. 2020; Zhou et al. 2015). The strong and durable fences have greatly increased the investment in fixed assets of herders and increased the cost burden (Wang 2009). When the collective grassland is contracted to households, it is divided into different quality grades and dispersed to each household, resulting in the decentralization and fragmentation of the herders’ grassland. Since then, most herders have enclosed their grassland plots with fences to clarify grassland boundaries and protect private property rights (Li and Huntsinger 2011). Some herders who are still engaged in animal husbandry but lack pastoral labor will rent out some small grassland plots with fences to reduce the difficulty of grazing. Herders who withdraw from animal husbandry production will rent out all the enclosed grasslands with fences. Therefore, nearly all grassland plots in the grassland market are plots with fences. For herders who continue to engage in animal husbandry production, renting in these fenced grassland plots can obtain grassland resources and greatly save the construction cost of fence subdivisions simultaneously, thus promoting their adoption of regional rotational grazing to a certain extent.

Second, grassland rental may promote herders to adopt regional rotational grazing or seasonal rotational grazing by expanding the grazing space (Feng et al. 2022). Although the grassland is divided into small plots for use when regional rotational grazing, it also has certain requirements for the area of grassland. Regional rotational grazing will not be carried out when the grassland area is lower than the scale threshold of the grazing block (Wang et al. 2020). Because too small a grassland area will increase the division difficulty of grazing blocks and reduce the grazing time of a single block, the moving space of livestock, and the buffer time for grassland restoration, thus reducing the positive ecological effect of rotational grazing (Zhou et al. 2015). Besides, the above problems will further increase the difficulty of rotational grazing and the labor input intensity of herders. Renting grassland can increase the total grassland area and the single block grazing area operated by herders simultaneously, reduce the difficulty of block division and unnecessary labor input, satisfy the spatiotemporal principle of regional rotational grazing, and promote herders to carry out regional rotational grazing. Seasonal rotational grazing has greater requirements for grassland areas due to the grazing for longer periods on a single grassland plot. Renting in large grassland plots and increasing the grazing durability of a single plot is conducive for herders to reasonably allocate the grazing time in the contracted grassland plots and rented grassland plots according to the number of livestock and seasonal changes to meet the requirements of seasonal rotational grazing. In addition, with the increase of grassland plots rented and the expansion of grassland rental transaction radius, the spatiotemporal distribution of grassland plots will be more discrete (Shi and Zhao 2022). It is helpful to increase the response-ability of grassland resources to the high spatiotemporal variability of precipitation (Wang et al. 2014) and promote herders to adopt seasonal rotational grazing.

To sum up, we build the following analysis framework (Fig. 1) and put forward the following research hypotheses:

Fig. 1
figure 1

Analysis framework of the impact of grassland rental on herders’ grazing mode under the grassland HRS

Full size image

Hypothesis 1: Grassland rental can promote herders to carry out rotational grazing (seasonal rotational grazing or regional rotational grazing);

Hypothesis 2: Grassland rental can promote rotational grazing by increasing the number of grassland plots fenced (grazing blocks) operated by herders, which reduces fence and labor costs;

Hypothesis 3: Grassland rental can promote rotational grazing by increasing the grazing area operated by herders, which expands the grazing space and the temporospatial variability of grassland resources.

Materials and Methods

Study Area and Data Source

Research data were obtained from our herder household surveys conducted in the pastoral areas of Inner Mongolia and Gansu in 2020. The data collected are mainly the information of herders in 2019. In addition, in terms of animal husbandry production, we also traced the data of 2018 to verify the data of 2019 through the continuity of animal husbandry production to avoid apparent deviation in the data of animal husbandry production in 2019. Data of 2019 was used in the study. Inner Mongolia is the biggest livestock production base and has the second-largest grassland area in China. The natural grassland grazing system here plays an irreplaceable role in ensuring the livelihood of millions of herders and the meat supply throughout the country (Liu et al. 2019). Natural grassland in Inner Mongolia is also an important ecological barrier to maintaining ecological security in Northern China. Inner Mongolia faces the dual pressure of the sound and rapid development of animal husbandry and grassland ecological restoration. Inner Mongolia is the first province to implement the grassland HRS, and its grassland rental market is active (Su et al. 2021). Gansu is the fifth largest livestock production base in China, and its natural grassland also plays an indispensable role in maintaining herders’ livelihood. Grassland in Gansu is mainly located in important ecological functional or ecologically fragile areas. It also faces the problem of coordination between livestock production and grassland restoration. Where the grassland rental market has also developed to a certain extent (Hu et al. 2014). Six pastoral counties (Fig. 2) were selected to conduct a household survey. The counties were selected because they have similar problems regarding the ecological environment and livestock production: animal husbandry as the primary industry and the main source of herders’ livelihoods, serious grassland degradation, fragile grassland ecology, severe overgrazing, settled grazing, and animal husbandry transformation, and active grassland rental markets. These areas cover three main grassland types: typical grassland, desertified grassland, alpine grassland, and counties with different economic development levels. They can reflect the decision-making of herders at different economic levels and grassland types. The above regions are typical and representative in studying grassland rental’s impact on herders’ grazing mode.

Fig. 2
figure 2

Study area and respondents’ distribution

Full size image

Stratified random sampling was adopted. First, all townships with animal husbandry as the main industry were selected from each county. Three to four townships were randomly selected from each county based on the distance between the township and county governments. Second, all pastoral villages were selected from each township, and three to five villages were randomly selected according to the distance between the village committee and the township government. Finally, we randomly selected 10–15 herders in each village and conducted face-to-face interviews with the main livestock laborers in each household. 857 questionnaires were collected, and 820 valid questionnaires (see Table S.1) were obtained (effective rate = 95.68%).

Sample Characteristics

Table S.2 shows the basic characteristics of the sample households and respondents. The respondents are mainly male (accounting for 74.51% of the total sample), ethnic minorities (accounting for 74.15%), and middle-aged and elderly (respondents over 45 years old account for 61.37%). The overall education level of the respondents is not high, and the respondents with junior high school education or below account for 77.56%. The characteristics of the respondents are consistent with the reality of the pastoral area and similar to the survey of other scholars (Su et al. 2021; Tan et al. 2018a). The family size ranges mainly from 3 to 6 persons (accounting for 82.31%), with an average population of 3.9 and a small family size. In 2019, herders with a net household income of less than 200 thousand CNY accounted for 63.42%, with an average net household income of 195 thousand yuan. Except that herders’ income is higher than the statistical data published by the government, other characteristics of the samples are similar to the government statistical yearbook. In general, the samples are representative.

Econometric Model

The Probit model is used to estimate the impact of grassland rental on herders’ grazing mode since the rotational grazing behavior is a discrete binary variable. The model is set as follows.

$$P\left( {Rotational\_grazing_i = 1} \right) = \Phi \left( {\alpha Grassland\_rental_i + {{{\mathbf{\beta X}}}}_i + \mu _i} \right)$$

Grassland_rental represents herders’ grassland rental behavior; X is a control variables vector affecting herders’ rotational grazing behavior, including multilevel variables at the individual, village, and region levels. μ is the random error term; β and α are the corresponding parameters matrix and parameter to be estimated, respectively. The coefficient α is of our primary interest.

Although control variables can help alleviate the endogeneity problem, there may still be some unobservable variables that may be omitted to affect herders’ grassland rental behavior and rotational grazing behavior simultaneously. Therefore, the Extended Probit Regression (EPR) model based on the instrumental variable method is used to test whether there is an endogenous problem. If the test results show endogeneity, EPR is used to obtain consistency estimation; Otherwise, the Probit model is used to obtain consistency estimation. This method can ensure the real causal effect of grassland rental on herders’ rotational grazing behavior. The key to the EPR method is to select appropriate instrumental variables. Further, the treatment effect of grassland rental on herders’ rotational grazing can be calculated under the counterfactual framework, including the average treatment effect of the treatment group (ATT) and sample average treatment effect (ATE).

Variables

Dependent Variable

Rotational grazing behavior. Ask herders, “do you carry out rotational grazing in a small range (including seasonal rotational grazing and regional rotational grazing)” to obtain the grazing mode of herders in the questionnaire survey. When respondents answer “yes,” the value is 1; Otherwise, the value is 0; And through the follow-up questions (such as the number of blocks, time for rotational grazing, etc.) to verify the authenticity of herders’ answers.

Core Independent Variable

Grassland rental. It is reflected by whether the herders have rented the grassland and rented area.

Control Variables

include contracted grassland endowment, family animal husbandry labor force characteristics, animal husbandry management characteristics, cognitive characteristics, climate factors, village characteristics, and regional characteristics (Hyland et al. 2018; Shi et al. 2021). The contracted grassland endowment includes the contracted grazing grassland area, the number of plots fenced, the square term of the number of plots fenced, and the contracted grassland quality. It is expected that the larger the contracted grassland area and the more plots fenced, the greater the possibility of rotational grazing, the better the grassland quality, and the less the possibility of rotational grazing. The family animal husbandry labor force characteristics include nation, animal husbandry experience, education, and the number of pure animal husbandry labor forces. It is expected that the more animal husbandry experience, higher education level, and the number of pure animal husbandry labor force, the greater the possibility of rotational grazing; compared to Han nationality, ethnic minorities are more likely to rotational grazing due to their nomadic tradition (Zhang et al. 2007). The animal husbandry management characteristics are mainly reflected by livestock stock at the beginning of the year. Cognitive characteristics include rotational grazing belief, attitude, and fence cognition. The higher the rotational grazing belief, the more positive the rotational grazing attitude, and the more positive the fence cognition, the greater the possibility of rotational grazing (Shi et al. 2021). The climate factors include temperature and precipitation. These are believed to affect grassland ecology, thus affecting herders’ rotational grazing behavior. Government township’s location was used to reflect village characteristics; County-level dummy variables reflect regional characteristics to control unobservable regional natural, economic, social, and policy differences.

Instrumental Variables

Village-level rental and grassland rental transaction costs. Village-level rental refers to the percentage of grassland rented at the village level after excluding the respondent and reflects the development of the village grassland rental market. The more mature the grassland rental market is, the more convenient it is for herders to rent grassland, and the more likely they are to rent-in grassland (Su et al. 2021). In addition, according to the peer effect theory, herders will imitate and learn from the grassland rental behavior of surrounding herders to make similar decisions (Manski 2000). Therefore, there is a correlation between the village-level rental and herders’ grassland rental behavior. Meanwhile, the development degree of the village rental market is not related to herders’ rotational grazing behavior, which meets the exogeneity of instrumental variables. Transaction cost refers to herders’ cognition of the transaction friction in the process of grassland rental (Shi and Zhao 2022). The higher the transaction cost, the lower the possibility of herders renting-in grassland (Deininger and Jin 2005), which meets the correlation between instrumental and explanatory variables. Besides, the transaction cost is not directly related to the rotational grazing behavior of herders, which meets the exogeneity of the instrumental variable.

Mechanism Analysis Variables

Grazing grassland area operated and the number of fenced plots operated. Grazing grassland area operated was obtained by summarizing the area of grazing grassland contracted and rented. The number of fenced plots operated was obtained by summarizing the number of fenced plots contracted and rented.

The definition and description statistics of the above variables are shown in Table 1. 65% of herders have conducted rotational grazing, and 35% still conducted continuous grazing in a small range. 43.4% of sample herders rented grassland. The average rented grassland area is 10.330 hundred mu per household. It can be found that the grassland rental market in the study area is active, and grassland rental has improved herders’ grassland endowment to a great extent. Under the background of private grassland property rights and the grassland rental market development, different grazing modes such as high-intensity continuous grazing and rotational grazing coexist.

Table 1 Variables Definition and Descriptive Statistics
Full size table

Empirical Results and Discussion

Effects of Grassland Rental on Herders’ Rotational Grazing Behavior

Table 2 shows the benchmark results of grassland rental’s impact on herders’ rotational grazing. Models 1–3 in Panel A gradually adds the core independent variable (Grassland rental), county dummy variables, and other control variables. The results show that whether other variables are controlled or not, the grassland rental can significantly increase the probability of herders’ rotational grazing, indicating that the role of grassland rental in increasing herders’ rotational grazing probability is relatively stable. It also shows ignoring regional differences may underestimate the impact of grassland rental on herders’ rotational grazing, and omitting other control variables will overestimate the impact of grassland rental on herders’ rotational grazing. Further, combined with the analysis of the results in model 3, it is found that the estimated coefficient of grassland rental is 0.2310 and significant at the 1% statistical level; that is, grassland rental can increase the average probability of herders’ rotational grazing by 23.10%. In model 4, the core independent variable is further replaced by the rented grassland area. The estimation coefficient of the rented grassland area is 0.0035 and significant at the 1% statistical level; that is, the larger the rented grassland area is, the higher the probability of herders’ rotational grazing is. The average marginal effect of every thousand mu increase in the rented grassland area is that the probability of herders’ rotational grazing is increased by 3.5%. In conclusion, grassland rental can increase the probability of herders’ rotational grazing, and the larger the rented area, the higher the probability of herders’ rotational grazing. Hypothesis 1 was preliminarily verified.

Table 2 Effects of Grassland Rental on Herders’ Grazing Modes
Full size table

Models 5 and 6 in Panel B further report grassland rental’s impact on herders’ rotational grazing after correcting endogenous problems. The estimation results show that the coefficient of ρ1 is −0.6710 and significant at the 1% level, the coefficient of ρ2 is −0.7650 and significant at the 1% level; that is, grassland rental-in and rented grassland area are endogenous variables. The estimation results of the first stage in models 5 and 6 show that the instrument variables (the transaction cost) have significant impacts on herders’ grassland rental-in and rented grassland area; that is, it is appropriate to correct endogeneity through the above variables. After correcting the endogenous problems, the coefficient of grassland rental is 1.8112 and significant at the 1% statistical level, and the coefficient of the rented grassland area is 0.0427 and significant at the 1% statistical level. In conclusion, after considering the endogenous problem, the grassland rental-in and rented grassland area can still significantly promote the rotational grazing of herders, which is consistent with the benchmark results. Based on model 5, we also calculated the treatment effect of grassland rental-in on herders’ rotational grazing under the counterfactual framework. Table 3 shows ATT is 0.5827 and significant at the 1% statistical level; that is, for herders who have rented-in grassland, the possibility of rotational grazing will be reduced by 58.27% if they do not rent-in grassland. ATE is 0.4719 and significant at the 1% statistical level, which means that grassland rental-in can significantly increase the probability of rotational grazing by 47.19% for all samples. Hypothesis 1 was further verified.

Table 3 Treatment Effect of Grassland Rental-in
Full size table

Based on model 5 (See Table 9 for complete results), we further explain the influence of control variables on herders’ rotational grazing. As expected, the contracted grassland area has a significant positive impact on rotational grazing. With the increase of the contracted grassland area, the possibility of herders’ rotational grazing increases. The influence of plots fenced on rotational grazing shows a significant inverted “U” shape, with an inflection point of about nine plots. The number of plots in our study is, at most, 13. Although it is located on the right side of the inflection point, only five households operate more than nine plots of grassland, less than 1% of the total sample; thus, overall, the increase in plots fenced will significantly increase the possibility of herders’ rotational grazing, but the promotion effect decreases with the increase in grassland plots fenced. Grassland quality significantly negatively impacts rotational grazing behavior, which means that herders with better grassland quality are less likely to conduct rotational grazing. The possible reason is that compared to herders in the same village, the better the quality of individual contracted grasslands, the higher their grassland production capacity, which can meet livestock’s grazing and feeding needs to a greater extent. Therefore, the motivation of herders to optimize grassland utilization through rotational grazing is lower. Besides, minority herders, herders with more livestock production experience, and herders with higher education levels are more likely to conduct rotational grazing. The possible reason is that compared with herders of Han nationality, minorities (such as Mongolian nationality and Zang nationality) have a historical tradition of nomadism (Zhang et al. 2007) and prefer this grazing mode. The more experienced herders are in animal husbandry and the higher their education level is, the better they understand the growth rules of livestock and grassland, so they conduct rotational grazing to adapt to these rules. Fence recognition and rotational grazing belief have a significant positive impact on herders’ rotational grazing behavior; that is, the more herders believe in the protective effect of rotational grazing on grassland ecology and the positive effect of fences for realizing regional rotational grazing, the more willing they are to conduct rotational grazing, which is consistent with (Shi et al. 2021). In addition, livestock stock has a significant negative impact on the rotational grazing behavior of herders. The possible reason is that the excessive number of livestock has greater requirements for the grassland area, the number, and quality of animal husbandry labor force, which increases the difficulty of rotational grazing. In general, the estimated results of control variables are consistent with expectations, which also shows the reliability of the research results.

Robustness Checks

Firstly, to avoid the incorrect setting of the model form, we use the nonparametric method of propensity score matching (PSM) to re-identify the impact of herders’ grassland rental-in on rotational grazing. PSM can overcome the problem of self-selection of grassland rental-in caused by observable factors. To ensure the stability of the estimation results, three different matching methods, nearest-neighbor matching, caliper matching, and kernel matching, are used to test the treatment effect of herders’ grassland rental-in behavior. The results are shown in Table 4. Under the three matching methods, the results of ATT and ATE were significantly positive at the level of 1%, and the same type of treatment results was close (0.2164–0.2371, 0.1939–0.2080). Taking the k-nearest neighbor matching results as an example, the ATT is 0.2230; that is, if herders with grassland rented did not rent grassland, the possibility of rotational grazing will be reduced by 22.3% (See Fig. S.1 and Fig. S.2 for more details). In conclusion, grassland rental-in can significantly increase the possibility of rotational grazing. Although the treatment effect is reduced, it is still positive, consistent with the results in Table 3.

Table 4 Robustness Check 1: Treatment Effect Calculated by PSM Method
Full size table

Secondly, we make three sub-sample estimates by dropping some samples. The specific strategies are as follows: (1) considering samples of Bordered Yellow Banner are relatively small and 50% of them rented grassland, to avoid the possible impact of this part of samples on the research conclusion, the model is re-estimated after dropping these samples; (2) Considering that a small number of herders have rented out the grassland, resulting in the change of grassland endowment, the model is re-estimated after dropping the samples rented out the grassland; (3) Considering that some of the herders rented out and rented in grassland simultaneously, this part of the sample is retained based on strategy (2), and the model is estimated again. The results in models 1–3 in Table 5 show that the estimated coefficient and significance level of grassland rental-in are basically consistent with the estimated results of model 5 in Table 2. Table 6 further reported the treatment effects of grassland rental-in. The results show that the coefficient and significant levels of ATT and ATE are nearly consistent with Table 3.

Table 5 Robustness Checks 2 and 3: Results of Subsamples and Changing Core Independent Variables
Full size table
Table 6 Treatment Effect Calculated by Subsample
Full size table

Thirdly, after changing the core explanatory variables, re-estimate the model. The specific strategy is to use the grassland rental rateFootnote 2 to represent the grassland rental. Model 4 in Table 5 shows that the estimated coefficient of grassland rental rate is positive and significant at the 1% statistical level. Increasing the grassland rental rate can encourage herders to carry out rotational grazing. It is consistent with the benchmark regression results.

Based on the above results, we can conclude that grassland rental can significantly and robustly increase the possibility of herders’ rotational grazing.

Impact Mechanism of Grassland Rental on Rotational Grazing Behavior

Model 1 in Table 7 shows that the grassland rental has a significant positive impact on the grassland area of herders; that is, the grassland rental-in can significantly improve herders’ grassland endowment; Model 2 further replaced the core independent variable with the rented area, and the results were similar. Model 5 shows that the grassland area has a significant positive impact on the rotational grazing behavior of herders; that is, the larger the grassland area operated by herders, the greater the possibility of rotational grazing. The marginal effect estimation result (Model 6) shows that the coefficient of the operated grassland area is 0.0018 and is significant at the 1% statistical level; that is, the average marginal effect of every thousand mu increase in the operated grassland area is to increase the probability of rotational grazing by 1.8%. To sum up, grassland rental can increase the area of grazing grassland operated by herders to match the grassland endowment with the large-scale livestock activity space required for rotational grazing and promote herders to conduct rotational grazing. Hypothesis 2 was verified.

Table 7 Influence Path of Grassland Rental on Herders’ Rotational Grazing Behavior
Full size table

Model 3 in Table 7 shows that grassland rental has a significant positive impact on the number of fences operated by herders; that is, transferring to grassland can significantly increase the herders’ operated plots fenced; Model 4 further replaced the core independent variable with the rented area, and the results were similar. Model 5 shows that the operated plots fenced and their square term significantly impact herders’ rotational grazing behavior; that is, there is a significant inverted “U” relationship between the rotational grazing behavior and the operated fenced inflection point is about 14 plots. Since only one household in the sample has more than 14 plots fenced, the number of plots fenced has a gradually decreasing positive effect on the rotational grazing. The marginal effect estimation results (Model 6) show that the coefficients of the plots fenced and their square terms are 0.1346 and −0.0047, respectively, which are significant at the 1% statistical level; that is, the average marginal effect of every increase in the number of plots fenced is increasing the probability of rotational grazing by 12.04%Footnote 3. In addition, compared to increasing the grazing grassland area, grassland rental-in promoting herders’ rotational grazing by increasing operated plots fenced is more obvious. The probability of rotational grazing can be increased by about 3.3%Footnote 4 by increasing grazing grassland area, and the probability of rotational grazing can be increased by about 16.20%Footnote 5 by increasing plots fenced, which is about 4.9 times that of the former. In conclusion, grassland rental can increase the plots fenced operated by herders, thus increasing the grazing blocks and saving the construction cost of fences to encourage herders to conduct rotational grazing. Hypothesis 3 was verified.

Further Discussion: Heterogeneous Effects of Grassland Rental on Rotational Grazing

Theoretical analysis and the previous empirical results show that the herders’ rotational grazing behavior depends significantly on the number of plots fenced and the scale of grazing grassland. Therefore, we further compare the heterogeneous effect of grassland rental-in on herders’ rotational grazing behavior through grouping regression according to the number of plots fenced operated by herders and the scale of the grazing grassland area.

Herders are divided into two groups according to whether the number of plots fenced operated is less than or equal to 3Footnote 6. The results in Table 8 show that for herders with few plots fenced, ATT and ATE were 0.5581 and 0.4700, and significant at the 1% statistical level, respectively; higher than herders with more plots fenced (ATT and ATE are 0.4683 and 0.3371, and significant at the 1% statistical level, respectively), which means the promotion effect of grassland rental-in for herders with fewer plots fenced is greater than that for herders with more plots fenced. The possible reason is that at least two clearly fenced blocks is needed to enable herders to carry out small-scale rotational grazing (Zhou et al. 2015). For herders with few plots fenced, the grassland rental-in can make their grazing blocks cross this threshold to change from non-rotational grazing to rotational grazing. In contrast, herders with many plots fenced have better rotational grazing conditions even if they do not rent grassland. For herders who had conducted rotational grazing before renting grassland, grassland rental-in only improves the rotational grazing conditions in quantitative change, not a qualitative leap. Therefore, the grassland rental-in has a higher role in promoting the rotational grazing of herders with fewer plots fenced. To sum up, the impact of grassland transfer behavior on herders’ rotational grazing behavior has heterogeneity in the number of fences. The promotion effect of grassland rental on rotational grazing for herders with fewer fences is greater than for herders with more fences.

Table 8 Heterogeneity of Grassland Rental’s Treatment Effect
Full size table

Herders are divided into “small-scale” and “large-scale” groups for estimation according to whether the grassland area operated by herders is less than or equal to the medianFootnote 7 by sample herders in the study area. The results in Table 8 show that for small-scale herders, ATT and ATE are 0.1880 and 0.1807 but not significant; lower than large-scale herders (ATT and ATE are 0.6782 and 0.5303, respectively, and significant at the 1% statistical level), which means for herders with large-scale grassland area who have been rented in grassland, not renting in grassland will reduce the possibility of rotational grazing by 58.66%; however, for herders with small-scale grassland area, grassland rental has no significant positive effect on herders’ rotational grazing behavior. The possible reason is that for small-scale herders, even if they have rented-in grassland, the operated scale of grazing grassland has not yet reached the threshold of grassland area expected by herders to conduct rotational grazing. Rotational grazing behavior is “laborious and thankless” for these herders. On the contrary, for large-scale herders, rent-in grassland makes the scale of their operated grazing grassland reach the lowest threshold or ideal scale of rotational grazing expected, making them think that rotational grazing is reasonable and then promotes herders to carry out rotational grazing. To sum up, the impact of grassland rental-in on herders’ rotational grazing is heterogeneous due to the operated grassland scale difference. It only plays a significant role in promoting large-scale herders.

Conclusions and Implications

Based on the survey data of herders in pastoral areas in northern China, this study theoretically analyzes and empirically tests the impact of grassland rental on herders’ rotational grazing. The main conclusions are as follows: (1) Grassland rental is conducive to promoting herders’ rotational grazing to improve the rationality of grazing modes. This result still exists after considering endogeneity and multiple robustness checks. Specifically, grassland rental increased the rotational grazing probability of herders by 58.27%. (2) By increasing the operated area of grazing grassland and the number of grassland plots fenced, grassland rental promotes the matching of herders’ grassland endowment with the large-scale livestock activity space and the number of grazing blocks required for rotational grazing, reduces the input cost and operation difficulty required for rotational grazing, and increase herders rotational grazing probability. (3) The impact of grassland rental on herder’s rotational grazing is heterogeneous, showing the dependence of the number of plots fenced and the scale of grazing grassland. It has a higher promotion effect on herders with more plots fenced; It cannot promote the generation of herders’ rotational grazing behavior when the rented grassland area fails to make the operation scale of the grassland reach the minimum threshold of rotational grazing.

The above conclusions have the following policy implications: First, further develop the grassland rental market, make use of the market mechanism to make up for the defects of grassland fragmentation under the privatization of grassland property rights, and promote the optimization of herders’ grazing modes. Grassland rental provides opportunities for herders to optimize their grazing modes. However, due to the vast and sparsely populated pastoral area and the lag of information construction, grassland rental still exits high transaction costs, which restricts the transfer of herders’ grassland (Shi and Zhao 2022; Tan et al. 2017). By improving the online and offline rental trading platforms and standardizing grassland rental contracts to reduce the transaction cost of grassland rental, herders can participate in the grassland rental market at a lower cost and promote the grassland rental market prosperity. The restrictions of rotation grazing to herders due to the lack of private grassland resources will be released through the spontaneous reallocation of grassland resources. Second, with the development of modern animal husbandry, fences have become the necessary infrastructure for herders to carry out regional rotational grazing and are no longer just an obstacle for herders to optimize their grazing modes. Grassland plots with fences have become popular in the rental market and favored by herders to maintain private property rights and optimize grazing modes. Therefore, policymakers should dialectically look at the role of fences in modern animal husbandry and seek advantages and avoid disadvantages. Third, moderate scale management is the premise for herders to optimize their grazing modes. The total grassland resources are still the core limiting factor restricting herders’ rotational grazing. The grassland rental still needs to encourage moderate scale to a certain extent. Too small a rental scale cannot achieve the purpose of promoting herders’ rotational grazing and lead to a new mismatch of grassland. Policymakers should consider incorporating the principle of optimizing herders’ grazing modes when intervening in the grassland rental market construction. Promote herders to realize the appropriate scale management of grassland guided by optimizing grazing modes.

Our results emphasize the importance of developing the grassland rental market to promote the optimization of grazing modes in grassland privatization countries. The results can be used as a reference for those pastoral areas where the traditional nomadic modes have declined due to grassland fragmentation in the countries with grassland privatization. However, since our sample data are from grasslands in arid and semi-arid areas in northern China, the above research conclusions should be extended to a limited range of areas carefully. In addition, because our data are cross-sectional data, we cannot analyze the changes in herders’ grazing modes over a longer period. Besides, there are natural deficiencies in cross-sectional data in causal identification. The follow-up research needs to use a longer micro survey panel data to conduct a more detailed analysis of this issue and carry out a more accurate causal identification of the grassland rental on herders’ rotational grazing.