Peasants are not only the basic unit of rural society, but are also the main body of cropland management. Since the implementation of household contract responsibility system, peasants have had autonomy in the operation of production as the basic unit of agricultural production decisions. With China’s market economic development, the peasant household economic behaviors show a transition from survival rationality to economic rationality (Li and Fu, 2008). Moreover, the overall behavior of peasant households is more consistent with the principle of rationality, which provides a theoretical basis for the study of peasant household behaviors.

Classical economics emphasizes the fact that the division of labor is the source of economic growth, and the production specialization and division of labor lead to more rapid economic growth. In inframarginal economic theory, the division of labor is further divided into personal specialization, professional diversification, production circuitry and economic organization. In terms of the internal division of peasant households, the individual specialization of family members is consistent with the profession diversification of overall families, leading to indirect economic development, which promotes the expansion of individual peasant land management scale and the improvement in the efficiency of small-scale peasant economy, so that the multi-occupation becomes the long-term balanced organization form (Xiang and Han, 2007). In neoclassical economics, which focuses on the allocation of resources, it is considered that peasant multi-occupation hinders the improvement of agricultural labor productivity, thus reducing the land output and utilization rate without an ideal state of scale economy, which is not the most effective method of resource allocation. However, based on the cognitions of the socio-economic development stage, land expectation and social security functions, peasants are wary of land circulation, which uses surplus agricultural labor force resources to increase the income of peasants for seeking a reasonable configuration method of agricultural production elements, which is in line with the “economic rationality” decision criteria. In accordance with the “economic rationality” decision, peasants will strive consider a variety of environmental and resource conditions to configure elements, in order to obtain the maximum benefit and allocation efficiency of production factors (Li et al., 2012; Schultz, 2010). Peasants are direct land users, and their rational allocation of production factors determines the intensive utilization degree of cropland.

Based on relevant theoretical analysis, this paper argues that, even in the traditional agricultural production, peasants also seek the configuration mode of agricultural production elements to maximize their interests. Due to this element configuration concept, under the condition of specific productivity, the cropland input behaviors of peasants as rational economic persons are subject to benefit maximization principle, i.e. the essence of cropland intensive utilization is the benefit-maximum-based element configuration concept of peasants. The behavior results of cropland input and use are proportional to the peasants’ expectations regarding cropland earnings.

2.2.1 Evaluation index and intensive utilization degree index

According to the scientific connotation of cropland intensive utilization, the literature review and expert consultation methods are used to select the evaluation index system of cropland intensive utilization degree (Table 1). Then, based on provincial panel data from 1996 to 2008, the factor analysis and other methods are adopted to analyze the cropland intensive utilization evaluation index. The specific steps are shown as follows: (1) use the maximum difference normalization method for standardized processing of the original data; (2) carry out the use condition judgment of the factor analysis method; (3) calculate the eigenvalue and eigenvector of the correlation matrix, and extract the public factor; and (4) calculate the factor loading matrix. On this basis, the cropland intensive utilization degree index (CII) is established as follows:

where F_i is the contribution rate of the i-th common factor, and V_i is the variance contribution rate of the i-th common factor.

2.2.2 Driving factor selection and data processing

From the perspective of the influencing operators’ decision, natural resource factors are selected, including cropland quality (X₁), cropland area proportion (X₂), and per labor cropland area (X₃); social and economic factors selected include per capita net income of peasant farming (X₄), cropland production comparative benefit (X₅), and labor productivity (X₆); and bearing capacity factors selected include cropland non-agricultural comparative advantage (X₇), peasant agricultural productive fixed assets (X₈), and grain yield per unit area (X₉). The above are the leading driving factors of cropland intensive utilization.

Cropland quality is a natural basic condition to influence cropland intensive utilization. In general, in the case of better cropland quality, the yield will be higher under the same input conditions. Cropland quality reflects the background characteristics of regional cropland resources, measured by the cropland standard coefficient. The cropland area proportion characterizes the abundance of regional cropland, measured by the proportion of cropland area in the total regional land area. The per labor cropland area is used to reflect the scarcity of peasant agricultural production object. The per capita net income of peasant planting industry reflects the expectations of peasants in agricultural production, which is also an important representation of input ability in agricultural production. Cropland production comparative benefit affects the enthusiasm of peasant household production, and is measured by the ratio between per capita net income of the peasant planting industry and net income of the non-planting industry. Labor productivity is an important symbol of comprehensive changes in agricultural output and labor. The cropland non-agriculturalized comparative advantage is measured by the added values of the secondary and tertiary industries per unit area of residential areas and mining areas as well as the added values of cropland agriculture per unit area, which affects the land use conversion. A portion of cropland non-agricultural conversion leads to irreversible cropland change. The peasant agricultural productive fixed asset is expressed by peasant purchase data of productive fixed assets, such as draught animals, commodity animals, large and medium-sized farming tools, agriculture, forestry, animal husbandry and fishery mechanics, as defined by the National Bureau of Statistics. The calculation method of cropland standard coefficient is shown below:

where CLSI is the cropland standard coefficient, MCI is the multiple crop index, Y is the grain yield per unit area, and i, j and c distribution respectively represent the region, year and nation.

2.2.3 Driving force model for cropland intensive utilization

In order to explore the driving force of cropland intensive utilization, in this paper the multivariate regression model is used to judge the relationship between cropland intensive utilization degree and various variables. The modeling steps are as follows: (1) select the index based on correlation analysis method; (2) use the multivariate statistical factor analysis method to determine the contribution rate of each variable and solve the main factor; and (3) carry out multicollinearity diagnosis on the independent variables of the main factor according to the correlation analysis results. If there are no multicollinearity between driving factors, the multivariate linear regression model is used with the following calculation equation:

where Y is the estimated value calculated according to the independent variable X, and b_i (i=1, 2, 3…n) corresponds to the partial regression coefficient of x_j (j=1, 2, 3…n). If a multicollinearity exists, then the following step is carried out. (4) By means of the ridge trace to calculate the value of parameter k, the ridge regression model is established.

During the research process, the calculated basic index data mainly include more than 10 terms, such as cropland area, fertilizer usage, and effective irrigation area. The 1996-2008 provincial (excluding Hong Kong, Macao and Taiwan regions) data include the following: cropland area, rural labor force, pesticide usage and effective irrigation area, sourced from the China Rural Statistical Yearbook (1997-2009), China Population and Employment Statistical Yearbook (1997-2009) and China Agricultural Statistics (1997-2009); total planting area, added values of agriculture, secondary and tertiary industries, mechanical power and fertilizer usage data, sourced from the China Statistical Yearbook (1997-2009) and Six-decade New China Statistical Data Collection; and peasant per capita agricultural income and agricultural expenditure, sourced from the China Rural Household Survey Yearbook (1992-2009) and National Agricultural Product Cost-benefit Data Collection.

3.1.1 Evaluation factor identification of cropland intensive utilization

The 1996-2008 evaluation index of cropland intensive utilization is taken as a sample to run the related module of SPSS software for calculation. The results show that the KMO statistic is 0.60, and the p value of the Bartlett sphericity test is 0.000, indicating that the research sample has passed the applicability test of the factor analysis. According to the principles that the contribution rate of the eigenvalue is more than 85% and the absolute values of the factor load are quite different, five common factors are extracted (Σλ_i = 7.8), and the cumulative variance contribution rate is 87.0% (Table 2). Among these, the first common factor has a higher load in V₂, V₃ and V₆, and the second common factor has a higher load in V₄ and V₉, which can be collectively known as conservative input factors. The third common factor has a higher load in V₁ and V₅, which can respectively be referred to as labor and capital input factors. The fourth common factor has a higher load in V₇, which reflects the intensive utilization degree of cropland use intensity, and can be referred to as the use intensity factor. The fifth common factor has a higher load in the variable of cropland change rate V₈, which can be referred to as the cropland conversion factor.

Based on the above analysis, it may be concluded that the conservative labor input factor effectively reflects the cropland intensive utilization degree, of which the variance contribution rate is 24.9%; this is followed by the labor and capital input factor, use intensity factor and cropland conversion factor, which reflect the fact that the variance contribution rates of cropland intensive utilization degree are 12.0% and 11.7%, respectively. These four factors can explain 89.04% of the changes in cropland intensive utilization degree, and can thus fully reflect the cropland intensive utilization degree.

3.1.2 Spatial differentiation characteristics of cropland intensive utilization

According to Equation (1), the comprehensive score value of provincial cropland intensive utilization is calculated. With the support of ArcGIS, the natural clustering method is used to divide the cropland intensive utilization degree into four types, i.e. extensive use area, low intensive area, moderate intensive area and high intensive area, so as to determine the spatial heterogeneity of provincial cropland intensive utilization.

The results show that China’s provincial cropland intensive utilization degrees are significantly different (Figure 1). (1) In 1996, there were more extensive use areas with large area proportions, mainly distributed in the western region and three eastern provinces, i.e. Beijing, Shanghai and Guangdong; the low intensive areas have the second largest quantity, distributed in Hebei, Hubei, Guangxi, Hainan and the three northeast provinces (Heilongjiang, Jilin, Liaoning); and the moderate and high intensive areas are concentrated in the central and southern North China Plain and the middle-lower reaches of the Yangtze River Plain. In addition, Xinjiang is classified as a high intensive area of cropland use. (2) Compared with 1996, the provincial cropland intensive utilization degree was greatly raised in 2008, and the provinces in extensive use areas became fewer in number with significant concentrated distribution characteristics; the moderately intensive areas are concentrated in Hubei, Anhui, Jiangsu, Beijing and Tianjin; and the highly intensive areas are significantly increased, distributed in the Huang-Huai-Hai Plain, the middle and lower reaches of the Yangtze River Plain, Xinjiang, and Heilongjiang. Provincial cropland use in 2008 basically showed a spatial differentiation pattern of “first ladder: higher intensive utilization degree and second and third ladders: lower intensive utilization degree”.

In order to reveal the dynamic change characteristics of provincial cropland intensive utilization degree, 31 provinces are further divided into three types, i.e. low increase speed, moderate increase speed, and high increase speed. It can be seen that the provincial cropland intensive utilization degrees were improved in 1996-2008 (Figure 2). The low increase areas are concentrated in the western region; moderate increase areas are mainly distributed in eastern China, central China, and Gansu, Ningxia, Chongqing and Guizhou provinces in the western region; and the high increase areas are concentrated in the northeast Heilongjiang-Jilin, Beijing-Tianjin, and Shanxi-Shaanxi-Ningxia-Inner Mongolia resource areas.

According to the element gravity model, the gravities of cropland intensive utilization degree in 1996, 2000, 2004 and 2008 are calculated. It can be seen from Figure 2 that the gravity of provincial cropland intensive utilization degree experienced a rapid transition from the central-western regions to the central-eastern regions in 1996-2000, after which it moved toward the southeast in 2000-2004, then quickly moved northward in 2004-2008. It is shown that in 1996-2008, the cropland intensive utilization degree in the western regions was ascending slowly, while in the central-eastern regions, especially in the northern regions, the cropland intensive utilization degree was rising greatly, and was significantly higher than that in the south.

3.3.1 Main impact factors of cropland intensive utilization

At first, a normal distribution test is carried out on the data, and then the related analysis method is used to judge the relationship between cropland intensive utilization degree and each variable (Table 3). The results show that X₁, X₄, X₅, X₆, X₇ and X₉ have significant correlations with cropland intensive utilization degree (Y) (P < 0.001), and X₂, X₃ and X₈ have low correlations with Y in an undetermined state. Therefore, the impacts of cropland area proportion (X₂), per labor cropland area (X₃) and peasant agricultural productive fixed assets (X₈) on cropland intensive utilization degree can be excluded.

Through KMO and a Bartlett sphericity test, the five driving factors are shown to be in accordance with the requirements of the factor analysis. The calculation results show that the cumulative variance contribution rate of the extracted first four common factors reaches 94.63%, indicating that these four common factors can fully explain the changes in cropland intensive utilization degree. Based on this, the common factors are determined, and varimax rotation is carried out to solve the principal factor solution. The results are shown in Table 4.

It can be seen from Table 4 that the first common factor has a higher correlation with cropland quality (X₁) and grain yield per unit area (X₉), reflecting the cropland productivity conditions and representing the influence of cropland natural background conditions on the cropland intensive utilization degree. The second common factor has a correlation with the per capita net income of peasants in the planting industry (X₄) and cropland production comparative benefit (X₅), representing the influence of cropland economic benefits, i.e. peasant household production expectations for cropland intensive utilization degree. The third common factor has a higher correlation with labor productivity (X₆), representing the influence of peasant ability access to consumer goods on the cropland intensive utilization degree. The fourth common factor has a higher correlation with cropland non-agriculturalization comparative benefit (X₇), representing the influence of cropland use conversion.

3.3.2 Regression equation of driving force

It can be seen from Table 3 that a stronger correlation exists between partial variables, e.g. the correlation coefficients between X₁ and X₉, X₅ and X₃, and X₅ and X₄ respectively reach 0.76, 0.69 and 0.63, indicating significant multicollinearity. Therefore, the ridge regression method is adopted in this paper. Through mapping the ridge trace, the k value is calculated to establish the regression model, so as to reduce the influence of multicollinearity on the regression results (Figure 3).

It can be seen from Figure 3 that when k < 0.8, the ridge regression line fluctuates greatly, and the regression coefficients of various driving factors are not stable. When k > 0.8, the ridge regression line gradually tends to be gentle. Therefore, in this paper the k value is set as 0.8, and the following regression model is obtained as follows:

t =12.16 8.30 1.14 6.11 -3.10 8.66

( =0.767; R²=0.781; DF=56.01; P=0.0000)

The determination coefficient of the model can reach 0.781, which indicates that driving factors can explain 78.1% of the changes in the cropland intensive utilization degree (Y). The regression model results show that only X₇ has a negative correlation with Y, suggesting that the higher the regional cropland non-agriculturalized efficiency is, the higher the expectations of cropland use conversion will be, thus leading to non-strong cropland protection enthusiasm, reduced farming input, and lower cropland intensive utilization degree. Cropland quality, per capita net income of peasants in the planting industry, labor productivity, cropland production comparative benefit and grain yield per unit area have a significantly positive influence on cropland intensive utilization, showing that the better the cropland natural foundation conditions are, the higher the cropland production yield will be, and the more tending to intensive utilization the cropland will be; otherwise, extensive use will result. Therefore, the research hypothesis proposed in this paper is validated herein.

In the provincial scale, according to provincial cropland intensive utilization degree from 1996 to 2008 and the change characteristics in 1996-2002, 2002-2008 and 1996-2008, the SPSS clustering method is used for classification. The results show that the study area can be divided into four types (Figure 4): (1) Region I is comprised of six provinces in eastern China, three provinces in central China, as well as Tianjin, Hebei and Xinjiang, of which the cropland intensive utilization degree reaches the maximum. The main reason for this is that the agricultural production mechanization degree is high in Xinjiang, where the labor productivity and cropland production comparative benefit reach a higher level, thus promoting cropland intensive utilization. This type of province is concentrated in the North China Plain and the middle and lower reaches of the Yangtze River Plain, characterized by strong basic farming conditions, abundant precipitation, and abundant agricultural labor force, but at the same time large conservative labor input exists. In addition, agricultural production activities pose serious environmental problems. (2) Region II consists of the three northeast provinces and Sichuan Province, which are characterized by very superior cropland productivity and natural conditions, as these provinces form China’s important production base of agricultural products with higher agricultural labor productivity and cropland intensive utilization degree. (3) Region III is composed of three resource-enriched provinces, i.e. Shanxi, Shaanxi and Inner Mongolia, which are characterized by fragile ecological environment, poorer cropland quality, farming-pastoral intersection and serious cropland degradation. With the progress of agricultural technology, although the cropland intensive utilization degree has increased by a certain extent, the overall level is still low. (4) Region IV covers Gansu, Ningxia, Qinghai, Tibet, Yunnan, Guizhou, Chongqing and Guangxi in the western regions, and Beijing, Shanghai, Guangdong and Hainan on the east coast, characterized by poorer cropland productivity and natural production conditions, as well as lower cropland production yield, especially in the western provinces, leading to the cropland intensive utilization degree always being at a lower level.

(1) Based on “rational peasants” and informarginal economics theory, the mechanism of cropland intensive utilization is explained. In a certain divided organization structure, the peasants are in accordance with the “economic rationality” decisions to consider all types of environment and resource conditions and configure production factors as much as possible, in order to obtain the greatest benefit. As a result, the cropland production configuration method of peasants determines the cropland intensive utilization degree.

(2) The conservative input factors, such as fertilizers, machinery and pesticides, are the greatest reflection of cropland intensive utilization degree, followed by the labor and capital input factor, utilization intensity factor, and cropland conversion factor. These four factors can explain 89.04% of the changes in the cropland intensive utilization degree.

(3) In the spatial and temporal scales, compared with 1996, the provincial cropland intensive utilization degree was greatly raised in 2008 with significant regional differences, initially showing a spatial differentiation pattern of “the first ladder: higher intensive degree, and the second and third ladders: lower intensive degree”. Among these, the natural background conditions of croplands are the first positive driving factors influencing the cropland intensive utilization degree. In addition, the higher the cropland production comparative benefit and labor productivity are, the more tending to intensive utilization the cropland will be. However, the higher the regional cropland non-agriculturalized efficiency is, the more unfavorable to intensive utilization the cropland will be.

The above research conclusions show the following beneficial policy implications: (1) Financial support can be increased, and agricultural production conditions, such as comprehensive improvement of field, water, road and forest, can be improved to construct high standard farmland and improve the middle- and low-yielding fields, so as to improve the overall cropland production capacity and promote more intensive utilization of cropland. (2) As viewed from the perspective of the positive influence of labor productivity on cropland intensive utilization, it is recommended to accelerate the formulation and perfection of the cropland circulation system, cultivate new agricultural management object and new-type peasants, integrate production factors, and improve labor productivity. This will provide policy guidance for regions with poor basic cropland conditions or higher cropland non-agriculturalized efficiency, to effectively solve cropland abandonment, low use efficiency, and other problems. (3) The inputs of fertilizers, pesticides and other non-renewable resources in agricultural production are important characterizations of a rise in cropland intensive utilization. However, based on the cognition of their impacts on ecological environment, this should be considered dialectically. It is urgently required to strengthen policy guidance for scientific and reasonable use. (4) Tasks which should be performed are improve agricultural pricing, strengthen the research and implementation of direct subsidy policies for grain producers and agricultural material subsidy policies, perfect the compensation mechanism, optimize farming system and the subsidy systems, raise subsidy standards, reduce agricultural production costs, and improve the comparative benefits of agricultural production and peasant enthusiasm in farming production. In addition, it is necessary to be wary of excessive cropland non-agriculturalization and non-food phenomena resulting from the lower cropland production comparative benefits.