The SEMs in Figure 3
were estimated using AMOS 17.0. Table 2
reports the estimated results for both the measurement model and structural model. The χ2
coefficient for the SEM is 338.131 (df = 204, p
= 0.000, PMSEA = 0.043, NFI = 0.903, CFI = 0.919), indicating that the SEM fits the data well. The graphical representation of Table 2
is shown in Figure 4
. Table 3
reports the standardized direct, indirect and total effects between latent variables.
3.1. The Influencing Factors of Land-Use Behavior
With regards to the influencing factors of farmers’ land-use behavior, the coefficients (shown in Table 2
) for urbanization and internal factors have better explaining power, which also proves that the PRI framework is acceptable. Urbanization and internal factors both have positive impacts on land-use behavior, significant at the level of 1% and 5%, respectively. The direct effects (represented by the standardized coefficient) of urbanization and internal factors on land-use behavior are 0.699 and 0.11, respectively. This indicates that, with other conditions unchanged, farmers’ land-use behavior would rise by 0.699 and 0.11 when urbanization and internal factors increase by one unit. As can be seen in Table 3
, the indirect and total effects for pathway ‘urbanization → internal factors → land-use behavior’ are 0.299 and 0.928. Consequently, compared to internal factors, the changes of external social-economic context stemming from urbanization are the major determinants of land-use behavior. Moreover, urbanization also has a positive impact on a household’s internal factors. Rapid urbanization facilitates the improvement of household endowment.
For the eight observable variables representing urbanization, VCD, APP, AST and TTN have significant and positive impacts on land-use behavior. With VCD, APP, AST, and TTN increased by one unit, land-use behavior would increase by 0.891, 0.573, 0.038 and 0.063, respectively. In other words, increasing the distance from the town centre, raising the price of agricultural products, increasing agricultural subsidy and increasing the frequency of technology training would make farmers’ land-use behavior more sustainable. Among these, regional location and price of agricultural products are the main factors.
Frequency of land adjustment, NFN, MPP and LN have significant and negative effects on land-use behavior, valued at −0.069, −0.771, −0.107 and −0.363, respectively. Therefore, land tenure insecurity, increasing number of off-farm labor force, rising price of farm-related inputs, and land fragmentation would all impede the sustainability of land-use behavior. Among them, the number of off-farm labor force and land fragmentation are the main factors.
The internal factors, including AGE, EDU, YEAR, ALN, HIT, and LRN, all have significant and positive impacts on land-use behavior. The standardized coefficients are 0.857, 0.468, 0.739, 0.128, 0.187 and 0.213, respectively. Age, education and farming experience of household head are the main factors, which indicate that the ability of the decision maker is of more importance than other internal factors.
3.2. The Impact of Land-Use Behavior on Land Quality
With regards to the influencing factors of land quality, the results shown in Table 2
indicate that farmers’ land-use behavior has a significant and positive effect, with the standardized coefficient valued at 0.773. In turn, land quality also imposes an effect on farmers’ land-use behavior, with the standardized coefficient valued at 0.247. In other words, while land quality could be largely improved by guiding farmers’ land-use behavior, it can also act back on famers’ land-use behavior. Moreover, as is shown in Table 3
, the indirect and total effects of ‘urbanization→land-use behavior→land quality’ are both 0.717, implying that, with an urbanization increase of one unit, land quality would rise by 0.717.
For the three observable variables representing land-use behavior, the standardized coefficients for GCC and LII are 0.823 and 0.803, respectively. The positive effects of land-use behavior imply that farmland on which cash crop was grown and more capital input per unit of land was received is associated with higher soil quality. This is because, in the study sites, the planting of cash crops (vegetables, fruits) requires the application of base fertilizer (manure) which is beneficial to soil. The higher capital input means more chemical fertilizer inputs, which is also a good supplement to soil nutrient balances when it is applied appropriately.
The multiple cropping index, however, has a negative effect on land quality, as expected. Land that grows multiple crops indicates more intensified use. The more crops grown, the more nutrients, such as N, P and K, are absorbed, and this leads to land degradation. In addition, the diversity of planted crops (high MCI) may result in land fragmentation, which would affect fertilization negatively [30
For the five observable variables representing land quality, AVN, AVP and AVK are mostly affected by farmers’ land-use behavior, with the standardized coefficient valued at 0.686, 0.887 and 0.756, respectively. The finding provides strong support in our conceptual framework that land-use behavior, especially inputs of N, P and K, may have the strongest and most direct effect on land quality. The influence of OM is relatively weak, though positive as well. This can be explained by the fact that, in our survey area, farmers paid little attention to input of organic fertilizer (such as manure), which leads to the excessive use of chemical fertilizers. As a result, the soil is acidized and the influence of pH value is negative.