The essence of land registration and certification is that farmland property rights are formally recognised by law, which legally guarantees the land rights. When farmland is not confirmed, the stability of land rights is insufficient, farmers lack awareness of farmland protection and tend to perform predatory management, leading to low adoption of conservation tillage techniques (Wang et al., 2020). The farmland ownership confirmation can stimulate farmers’ investment in farmland and increase the amount of organic fertiliser they use (Huang and Luo, 2020). Agricultural land titling has been widely promoted in many countries and regions, such as Ethiopia, the UK, Russia, and Peru. Scholars have found that when farmers perceive their farmland property rights as unstable, they reduce agricultural investments, such as land levelling, to improve the quality of cultivated land (Ma et al., 2015). Others have argued that the effect of increased security of land rights on farmers’ productive agricultural investments, such as soil improvement, is uncertain. However, some countries and regions have not achieved the expected results of land registration and certification (Jiao, 2018). For instance, a study of random villages in Zambia, where land registration and certification were carried out, found that the process did not lead to increased investment and conservation of land (Huntington and Shenoy, 2021). However, more studies have shown that land registration and certification can reduce the instability of land rights and enhance farmers’ protection of the ecological quality of agricultural land (Qian et al., 2021), help farmers focus on soil quality improvement (Navarro-Castañeda et al., 2021), motivate them to apply more organic fertilisers (Zhou et al., 2022), reduce the application of chemical fertilisers and pesticides (Zheng et al., 2023) and improve the selection of new agricultural technologies (Wen et al., 2020).

Institutional economics suggest that land registration and certification give farmers legal recognition of their land property rights, which can fully protect their land rights and interests and stimulate farmers to change their predatory production methods and choose ones that cause less damage to cultivated land (Qian et al., 2019). Clear and stable land property rights prevent farmers from concerns about losing their land and motivate them to protect it. Therefore, once the land rights are confirmed, farmers feel confident in investing in and protecting the land and improving the quality of farmland (Gao X. W. et al., 2021). However, from the perspective of individual farmers, farmland titling increased their land tenure stability expectations. Studies have demonstrated that farmland titling significantly increases farmers’ land tenure stability expectations (Qin et al., 2020) and promotes farmers’ arable land quality conservation behaviour.

Based on this, we proposed the following hypothesis:

In actual farming operations, the impact of land registration and certification on farmers’ adoption of conservation tillage techniques can differ, usually due to differences in farmers’ personal factors and family conditions (Wang et al., 2021). Farmers’ adoption of conservation tillage techniques is often affected by factors, including farmers’ own education level, the degree of fragmentation of the cultivated land they own, and the scale of crop planting (Ruzzante et al., 2021; Theis et al., 2018). Therefore, the adoption of conservation tillage technologies after land registration and certification is likely to differ between farmers.

More educated farmers have a better understanding of property rights, can confidently invest in their farmland after registration and certification, and understand the need for ecological protection of their farmland. Farmers tend to adopt new technologies considering profit maximisation. Finely fragmented farmland could negatively impact output (Sun et al., 2018), increasing labour time and costs, while also not being conducive to farmers’ adoption of conservation tillage technologies after land registration and certification. Regarding arable land area, large-scale farmers already have a high adoption rate of conservation tillage technology for gaining the scale effect before farmland titling (Feng et al., 2018). Small-scale farmers take more care of their farmland and are willing to adopt quality conservation measures and increase adoption of conservation tillage technology based on long-term awareness after land registration and certification, as their rights and interests are guaranteed and operational risks are reduced.

Therefore, we proposed the following hypothesis:

This study constructed a theoretical analysis framework for the adoption of conservation tillage techniques by farmers on agricultural land titling (see Figure 1).

This study used data from a questionnaire survey conducted by the School of Economics and Management of the Northwest Agriculture and Forestry University in August 2020. The survey covered the Yellow River Basin in the six provinces of Qinghai, Ningxia, Shanxi, Shaanxi, and Inner Mongolia (Figure 2). Using a combination of random sampling and stratified sampling, the survey randomly selected 13 counties based on economic development status and arable land, with three townships (towns) per county randomly selected for good, medium, and poor economic development, and three administrative villages per township randomly selected for good, medium, and poor economic development. The main contents of the questionnaire included: basic information on farmer’ households, household assets, household income and expenditure, agricultural production and operation, and village rules. The survey data were collected via questionnaires by investigators who interacted one-on-one with farmers. A total of 2,362 questionnaires were obtained. As this study focused on the straw-handling methods of cereal and oil crop growers, we screened samples that did not grow straw crops, samples with inconsistent information, and samples with missing key information. Finally, 830 valid samples were obtained after screening and sorting.

The basic characteristics of the interviewed farmers are shown in Table 1. Most participants were older, with 71.54% of them being over 50 years old. Moreover, education level was generally low, with 85.36% of the farmers having a junior high school or lower education. The overall health condition was high, with 72.77% of the farmers in good health. The majority (59.88%) of the family farming area was concentrated in less than 15 mu. The total household income was less than 60,000 yuan for 65.3% of the participants. The basic characteristics of the sample farmers were in line with the real situation of China’s rural areas at this stage and could be considered representative.

Explained variables. We selected straw retention to represent conservation tillage technology. Straw retention was defined as the behaviour of farmers who use the straw return technology when handling straw during the grain growth process.

To investigate the impact of the new round of land registration and certification on farmers’ adoption of conservation farming techniques, we simultaneously compared the impact of titling on the adoption of those techniques by the same farmers. As farmland can only be in a confirmed or unconfirmed state, data from both states cannot be observed at the same time, which creates a problem of missing data. Therefore, this study used the Propensity Score Matching (PSM) method based on a counterfactual framework to estimate the factors influencing farmers’ adoption of conservation tillage technologies and compared the impact of titled versus un-titled farmland and the new round of land registration and certification on farmers’ adoption of conservation tillage technologies by using the average treatment effect in the propensity score matching method. Specifically, we used appropriate control variables to match a characteristic-balanced unconfirmed household for each confirmed household to ensure that the two sample households were as balanced as possible. The propensity score value of each sample farmer was estimated, and the likelihood of land registration and certification of farm households was estimated using a binary logit regression, as shown in Equation 1. (1) P Xi = P D i = 1 | X i

Then, we used the propensity score matching method to analyse the data before matching each farmland-accredited household with balanced characteristics according to the propensity score and conducting a balance test between the two groups to ensure accuracy. The average treatment effect of farmland titling on the adoption of conservation farming technologies was obtained by calculating the difference in the adoption of conservation farming technologies between the treatment and control groups, as shown in Equation 2. (2) A T T = E y 1 i | D i = 1 − E y 0 i | D i = 1

Where D i = {0, 1} denotes whether the farmland is confirmed, y i represents whether farmers adopt conservation tillage techniques, y 1 i denotes whether farmers who have confirmed farmland adopt conservation tillage techniques, and y 0 i denotes whether farmers who have not confirmed farmland adopt conservation tillage techniques. To ensure the validity of the sample and the robustness of the matching results, multiple matching methods were used. Kernel (using the default kernel function with bandwidth), k-nearest neighbour (where k = 4), and radius matching (where ℇ = 0.01) were selected to estimate the average treatment effect.

Although the new round of land confirmation, registration, and certification is a policy arrangement led by the state, land registration and certification are not completely random (Chen et al., 2018). Promoting land registration and certification usually begins with the normal farmers’ families and testing it out on a pilot basis before gradually extending the scope. This means that the farmland ownership confirmation is not random. The results of the logit equation estimation showed that the propensity score of land registration and certification, age of the household head, education level, and degree of cultivated land fragmentation had significant effects on land registration and certification. The age variable was significant at the 1% level for land registration and certification, suggesting that older farmers were more likely to focus on security issues and, therefore, valued the security that comes with land property rights (Sun et al., 2023). Literacy was significant at the 1% level for land registration and certification, most likely because farmers with higher educational levels had a clearer understanding of property rights. The degree of cultivated land fragmentation was significant at the 1% level. Under the premise of voluntary farmer participation, certification could consolidate scattered plots and reduce the degree of cultivated land fragmentation; therefore, farmers with a high degree of cultivated land fragmentation were more willing to have their farmland confirmed to facilitate agricultural production and operation (Table 3).

To ensure accuracy in the propensity score matching, samples outside the range of values common to the treatment and control groups were excluded. However, excluding a large number of samples was considered a match failure. Thus, it was important to choose the appropriate matching method and cover as much of the sample as possible within the range of the common support domain. To examine the common support domain of treatment and control group farmers, probability density plots of propensity scores (Treated and Untreated, respectively) before and after matching are presented separately (Figure 3). The kernel density functions of the samples in the treatment and control groups converged after matching, indicating that a better matching effect was achieved.

To improve the reliability of the matching results, Table 4 presents the variance changes in the control variables after the balance test. The standard bias for most variables was substantially reduced after matching. The standard bias of the control variables after matching was <10%. In particular, the variables of age of the household head, education level, and the degree of cultivated land fragmentation were significant before matching, and the t-values of all control variables were less than the critical value of 1.96 after matching. Therefore, the covariates selected for matching were suitable, and the differences in the characteristic variables between the two groups were eliminated. Therefore, the sample matching passed the balance test.

As shown in Table 5, the mean treatment effects of the impact of land registration and certification on farmers’ adoption of conservation tillage techniques were obtained after applying kernel, k-nearest neighbour (k = 4), radius calliper, local linear regression, and martingale matching. These effects were significant at the 5% level for all five matching methods, four of which had significant treatment effects at the 1% level. Assuming that farmers did not participate in land registration and certification, the likelihood of adopting conservation tillage techniques ranged from 27.9 to 33.9%. Farmers in the treatment group were 49.1–49.3% more likely to adopt conservation tillage techniques, with a 15.2–21.4% increase in ATT values and an average treatment effect mean of 18.58%. Land registration and certification had a significant positive effect on the adoption of conservation tillage techniques by farmers (Zhang and Zhang, 2023). One possible reason could be that clear and stable land ownership stimulated farmers to take better care of their farmland and be willing to invest in it, thus influencing them to adopt conservation tillage techniques.

Therefore, H1 was confirmed.

The PSM method can only deal with the sample selection bias caused by observable variables; unobservable factors pose the problem of hidden bias. To test the robustness of the research results, this study adopted a double robust model to analyse the effect of land registration and certification on farmers’ adoption of conservation tillage techniques. We used regression-adjusted, inverse probability weighting, as well as inverse probability weighting regression-adjusted empirical methods for robustness testing.

Table 6 illustrates the ATT based on RA, IPW, and IPWRA. The estimation results of the three regression adjustment methods, inverse probability weighting, and inverse probability weighting regression adjustment were relatively similar. The new round of land registration and certification significantly increased farmers’ adoption of conservation tillage techniques. Regarding ATT values, the mean treatment effects of the three methods were 17.6, 23.6, and 19.6%, respectively, with a mean treatment effect of 20.3%; all were significant at the 1% level. The PSM estimation results were robust; thus, the findings of this study were valid.

This study demonstrated that the new round of land registration and certification had a positive impact on farmers’ adoption of conservation tillage techniques. However, the pathway of how land registration and certification specifically affected the adoption of those techniques requires further investigation. The above analysis argued that land registration and certification affect farmers’ adoption of conservation tillage technologies mainly through resource allocation. Drawing on previous research (Wen et al., 2022), the mesomeric effect was tested using the test path.

A bias-corrected non-parametric percentile bootstrap approach was adopted to perform robustness tests on the mesomeric effect of resource allocation. If the confidence interval did not contain 0, the mesomeric effect was present and significant. As shown in Table 7, the benchmark test results were robust. In particular, the 95% confidence interval of resource allocation did not contain 0, indicating that the mesomeric effect of resource allocation was significant. In addition, the ratio of the mesomeric effect to the total effect was 17.61%, implying that approximately 17.61% of land registration and certification were achieved through the mesomeric effect of resource allocation on technology adoption. This demonstrated the presence of a mesomeric path of land registration and certification-resource allocation to conservation tillage technology adoption.

Therefore, H2 was confirmed.

The PSM method was used to measure the effect of land registration and certification on farmers’ adoption of conservation tillage techniques. Nevertheless, this impact only reflected the mean value of the change in ATT after land registration and certification. Differences in conservation tillage technology adoption among farmers with different capital endowments were not evident. As the effects of urbanisation and industrialisation gradually spread to the countryside, the differentiation of farm household groups widened. Following previous studies reporting divisions in farmers’ education levels (Song et al., 2021; Yue et al., 2021), we distinguished the educational level of the household head into two categories: elementary school and below and junior high school and above. We distinguished the arable land area into two categories, as well: less than 10 mu and 10 mu and above. The degree of cultivated land fragmentation was grouped by mean value as the cut-off line. In addition, to further analyse the effect of land registration and certification on the adoption of conservation tillage techniques by farmers with different endowments, five matching methods were used to estimate the average treatment effect: kernel (using the default kernel function with bandwidth), k-nearest neighbour (where k = 4), radius (where ℇ = 0.01), local linear regression, and martingale matching. The differences in the average treatment effects estimated by these five matching methods were small, indicating that the results of the propensity score matching were robust (Table 8).

The results of the subgroup estimation based on the education level of the household head showed that in the lower education level subgroup, farmers’ participation in land registration and certification had a positive effect on ATT; however, the results were not significant. Moreover, land registration and certification had a significant positive effect on the adoption of conservation tillage techniques among the subgroups with higher education of the household head (Avemegah et al., 2024). The farmers’ probability of adopting conservation tillage techniques increased by 20.74% on average after land registration and certification. Thus, H3 was confirmed.