Adoption intensity of whole-cycle reproductive management practices and economic returns in smallholder beef breeding-cow systems in China

1 Introduction

With the increasing demand for high-quality animal protein, beef cattle production has become an important component of livestock systems. In cow–calf production, breeding cows provide the biological foundation for calf output, herd renewal, and long-term production stability. Their reproductive and productive performance therefore has direct implications for both biological efficiency and the economic sustainability of beef production systems (Diskin and Kenny, 2014; Lancaster and Larson, 2022). In China, however, smallholder beef breeding-cow farming continues to face relatively low economic returns, extensive management practices, and uneven management capacity across households, which constrain the performance and sustainability of smallholder cow–calf systems (Pulina et al., 2021; Gao et al., 2023). Compared with more efficient beef production systems in developed countries, reproductive performance in Chinese breeding-cow systems remains relatively weak, suggesting that management continuity, decision-making quality, and systematic reproductive management remain important areas for improvement (Li et al., 2018).

Beef breeding-cow production involves a sequence of biologically connected stages, including pre-breeding preparation, gestation management, calving and postpartum recovery, and remating or culling decisions. Because these stages form a continuous reproductive cycle, weak management at one stage may reduce the value of management efforts at later stages (Short et al., 1990; Diskin and Kenny, 2016; Bogado Pascottini et al., 2025). For example, inadequate pre-breeding preparation may affect conception, insufficient gestation-stage management may influence calving outcomes and postpartum recovery, and delayed remating or inappropriate culling may reduce herd renewal efficiency. Farm-level returns therefore depend not only on whether individual practices are adopted, but also on the adoption intensity, continuity, and coordination of reproductive management practices across the whole cycle. Economic performance in breeding-cow systems is closely associated with reproductive efficiency, body condition, calf survival, herd health, and market fluctuations (Damiran et al., 2018; Greenwood, 2021; Kappes et al., 2023). Among these factors, management practices are more readily adjusted at the farm level than genetic endowment or external market conditions (Moorey and Biase, 2020; Jiang et al., 2023; Nsabiyeze et al., 2025). For smallholder farmers, achieving sufficient adoption intensity of whole-cycle reproductive management may require not only technical knowledge but also the capacity to sustain continuous management practices under resource constraints (Cooke et al., 2021; Nkadimeng et al., 2024). Accordingly, improving whole-cycle reproductive management may provide a practical pathway for enhancing farm performance without requiring heavy reliance on capital-intensive technologies or facility investment (Funston et al., 2012; Pessoa et al., 2018; de França e Melo et al., 2024).

Whole-cycle reproductive management can be understood as knowledge- and management-intensive practices that require farmers to adjust decisions according to the physiological status, reproductive stage, and production objective of breeding cows. These practices cover key reproductive processes, including breeding preparation, gestation management, postpartum recovery, remating, and culling decisions (Hentzen and Holm, 2025; Bernabucci et al., 2025). In principle, such practices may support body condition maintenance, reproductive recovery, production continuity, and management efficiency, and may contribute to the economic sustainability of beef breeding-cow farming (Hess et al., 2005; Lovarelli et al., 2020). Unlike embodied technologies such as automated equipment, facilities, or digital monitoring systems, whole-cycle reproductive management relies more heavily on farmers’ knowledge, observation, decision-making, and coordination across reproductive stages (Hörner et al., 2022). Traditionally, agricultural technology adoption studies have focused on whether farmers adopt specific technologies or management practices and have examined the role of farmer characteristics, resource endowments, and institutional factors in adoption decisions (Doss, 2006; Dimara and Skuras, 2003). Recent studies have increasingly emphasized adoption intensity and the combined adoption of complementary technologies, suggesting that the extent and coordination of adoption may shape economic outcomes (Palma-Molina et al., 2023). However, for lifecycle-dependent management systems such as beef breeding-cow reproduction, economic returns may depend not only on individual practice adoption but also on the intensity, continuity, and coordination of multiple interrelated practices across stages (Moser and Barrett, 2006; Kassie et al., 2013; Takahashi et al., 2020). Therefore, an adoption-intensity perspective is needed to capture the economic relevance of coordinated whole-cycle reproductive management.

This feature suggests that the association between whole- cycle reproductive management adoption intensity and economic returns may be nonlinear rather than simply linear. At low levels of adoption intensity, farmers may implement only selected practices without establishing a coordinated management system across the reproductive cycle. During this stage, farmers may face learning costs, adjustment costs, and additional managerial efforts, while the potential economic benefits of improved management remain limited because complementary practices across different reproductive stages are not fully developed. As adoption intensity increases, however, management practices implemented throughout the reproductive cycle may become increasingly coordinated and complementary. Improvements in breeding preparation, gestation management, postpartum recovery, remating, and culling decisions may accumulate over time and become increasingly associated with improved economic returns. Therefore, the economic implications of whole-cycle reproductive management may depend not only on whether individual practices are adopted, but also on the intensity and coordination with which multiple related practices are implemented across the reproductive cycle. Based on this reasoning, we propose the following hypothesis: Hypothesis 1 (H1): Whole-cycle reproductive management adoption intensity is associated with economic returns in a nonlinear manner.

However, the realization of potential economic benefits from whole-cycle reproductive management adoption intensity may depend on farmers’ ability to access and organize complementary resources. According to Resource Orchestration Theory, resources do not automatically create value; rather, value creation depends on actors’ ability to acquire, integrate, and utilize resources effectively (Sirmon et al., 2007, 2011). For smallholder farmers, financial constraints and limited risk-bearing capacity may hinder the sustained implementation of coordinated reproductive management practices. Previous studies further suggest that external support can improve farmers’ access to complementary resources and enhance their capacity to adopt and maintain agricultural technologies requiring sustained investment and management adjustment (Yigezu et al., 2018). Therefore, subsidy and credit support may serve as important boundary conditions shaping the strength of the nonlinear association between whole-cycle reproductive management adoption intensity and economic returns. Based on this reasoning, we propose the following hypothesis: Hypothesis 2 (H2): External support conditions strengthen the nonlinear association between whole-cycle reproductive management adoption intensity and economic returns.

Against this background, this study examines the association between whole-cycle reproductive management adoption intensity and economic returns in smallholder beef breeding-cow systems in China. Using survey data from 384 beef breeding-cow farmers in Jilin, Heilongjiang, and Shaanxi Provinces, we develop an adoption-intensity index covering four reproductive management stages: pre-breeding, gestation, postpartum management, and remating/culling. We then estimate the nonlinear association between adoption intensity and net income per breeding cow and examine whether external support conditions shape this association. This study contributes to the literature in three ways. First, it extends research on livestock management and agricultural technology adoption by conceptualizing reproductive management as a whole-cycle management system rather than a collection of isolated practices. Second, it moves beyond binary adoption measures by developing a continuous adoption-intensity index that captures the systematic implementation, continuity, and coordination of multiple lifecycle-related management practices. Third, it provides empirical evidence on the nonlinear association between management-intensive practices and economic returns and highlights the boundary role of external support conditions in shaping this association among smallholder farmers. The theoretical framework of the study is presented in Figure 1.

2 Materials and methods

2.1 Data source and sample description

This study uses household survey data collected by the research team in 2024 and 2025 in major beef-producing areas in China. The two survey waves followed the same questionnaire design and sampling framework but interviewed different households. Therefore, this study uses pooled repeated cross-sectional data. To ensure regional representativeness, the survey covered Taonan City in Jilin Province, Longjiang County in Heilongjiang Province, and Linyou County in Shaanxi Province, involving 3 counties, 9 townships and 24 administrative villages. These three areas are important local beef cattle breeding regions and are broadly representative of the production and management characteristics of breeding-cow farmers across different ecological zones in northeastern and northwestern China.

The respondents were households primarily engaged in beef breeding-cow production. The questionnaire covered household head characteristics, household endowments, production and management conditions, reproductive management, adoption of management practices, and production costs and returns. The survey was conducted mainly through face-to-face household interviews and was supplemented by centralized interviews organized through village committees. Before the formal survey, the questionnaire was revised on the basis of a pilot survey, and all enumerators received standardized training. Key indicators were verified through a combination of on-site checks and post-survey review to ensure data reliability. During data cleaning, observations with substantial missing information, abnormal values in key variables, or logical inconsistencies were excluded. Ultimately, 384 valid observations were retained, including 256 households surveyed in 2024 and 128 households surveyed in 2025. Because the sampled households were not repeatedly tracked across years, the final dataset was treated as pooled repeated cross-sectional data in the subsequent analysis.

2.2 Variable definition and measurement

2.2.1 Adoption-intensity index of whole-cycle reproductive management practices

In this study, the adoption intensity of whole-cycle reproductive management practices is defined as the extent to which farmers systematically implement coordinated reproductive management practices across the entire reproductive cycle of beef breeding cows. Unlike individual technology adoption, whole-cycle reproductive management represents a management system composed of multiple biologically connected practices implemented across different production stages. Accordingly, we constructed an indicator system covering four biologically connected reproductive management stages: (1) pre-breeding management, (2) gestation-stage management, (3) postpartum management, and (4) remating and culling management. The detailed indicators and corresponding measurement items are presented in Table 1.

Table 1. Indicator system and measurement framework for whole-cycle reproductive management adoption intensity.

(I)–(IV) denote the four key reproductive management stages across the full reproductive cycle of breeding cows. The listed practices represent stage-specific management measures used to construct the adoption-intensity index for whole-cycle reproductive management practices.

The adoption-intensity index consists of 19 specific management practices derived from farmer survey responses. These indicators include both binary and continuous measures reflecting different aspects of reproductive management implementation. For binary practices, adoption was coded as 1 and non-adoption as 0. For continuous indicators, scores were assigned according to the degree of conformity between farmers’ practices and recommended reproductive management standards. All indicators were subsequently standardized into values ranging from 0 to 1, with higher values indicating stronger implementation intensity of whole-cycle reproductive management practices. The detailed survey questions, measurement types, and coding rules for all indicators are reported in Supplementary Table S1.

Because whole-cycle reproductive management represents a knowledge-intensive and process-oriented management system, different reproductive stages and specific management practices may have different functional roles and theoretical importance throughout the production cycle. Such differences cannot be fully identified based solely on variation within the sample. Therefore, this study adopts an expert-weighting approach to construct the baseline adoption-intensity index by incorporating expert knowledge regarding the production functions and relative importance of different management stages and practices. Specifically, expert evaluations were conducted to determine the relative importance of the 19 individual management practices and the four lifecycle management stages. The stage-level weights were subsequently used to aggregate the four stage-specific adoption-intensity scores into the overall whole-cycle index. The expert panel consisted of seven specialists with extensive knowledge and practical experience in beef cattle production and reproductive management, including researchers and technical extension experts from the National Beef Cattle and Yak Industry Technology System. These experts were involved in beef cattle breeding, production management, and technology extension. The expert-weighted whole-cycle reproductive management adoption-intensity index was calculated as follows:

PFMTi=∑k=14WkStageki

where Stagekirepresents the standardized adoption-intensity score of the k-th reproductive management stage for farmer i, and Wkrepresents the corresponding expert-derived weight of each stage. The detailed expert weights of the 19 individual management practices and the four reproductive management stages are reported in Supplementary Tables S2, S3, respectively.

To examine whether the empirical findings are sensitive to the expert-based weighting approach, we further constructed an alternative adoption-intensity index using the entropy-weighting method. Unlike expert weighting, the entropy method is entirely data-driven and determines indicator weights according to the relative information contribution reflected by variation within the sample. The entropy-weighting procedure includes Min–Max normalization, information entropy calculation, redundancy coefficient estimation, and entropy weight determination. The detailed entropy weights of the 19 management practices and the four reproductive management stages are reported in Supplementary Tables S4, S5, respectively.

Importantly, the entropy-weighted index is not intended to validate or replace the expert-weighted index. Instead, the two approaches capture different dimensions of information. The expert-weighted index reflects the theoretical importance and biological relevance of reproductive management practices based on domain knowledge, whereas the entropy-weighted index captures the relative information contribution of indicators within the observed sample. Consistent empirical results across these two alternative weighting approaches suggest that the main findings are robust to different weighting schemes.

2.2.2 Economic returns of beef breeding-cow farming

This study measures economic returns in beef breeding-cow farming from both per-cow and household-level perspectives. The main dependent variable is net income per breeding cow, measured as 10,000 Chinese yuan per breeding cow (10,000 CNY/head). It is calculated as household net income from breeding-cow production during the survey period divided by the number of breeding cows. By expressing returns on a per-cow basis, this measure reduces the influence of herd size differences and provides a more comparable indicator of economic returns across households.

The variable enters the regression models in its original economic unit without logarithmic transformation, standardization, or mean-centering. The possibility of negative values reflects genuine production losses experienced by some sampled households rather than any data transformation. For robustness analysis, an alternative outcome variable, total net income from breeding-cow farming (10,000 CNY/household), was constructed to capture household-level operating returns.

2.2.3 Control variables

To reduce potential omitted-variable bias, control variables were selected based on previous studies on agricultural technology adoption, household production decisions, and livestock production performance. Considering that whole-cycle reproductive management represents a knowledge-intensive and process-oriented management innovation, its economic implications may depend not only on the extent of adoption but also on farmers’ implementation capacity, production conditions, and access to external resources. Therefore, control variables were selected from three dimensions: household capability and resource endowments, production and market conditions, and institutional support and external resource environments.

(1) Household capability and resource endowments

Household capability and resource endowment variables were included to control for differences in farmers’ ability to acquire information, learn management knowledge, and continuously implement complex management practices. Previous studies on agricultural technology adoption suggest that farmers’ human capital, production experience, and resource endowments influence technology perception, adoption decisions, and subsequent implementation performance (Doss, 2006). Accordingly, this study includes household head age, education level, cattle-raising experience, and labor input in beef cattle production as control variables.

(2) Production and market conditions

Production and market condition variables were included to control for differences in production foundations, resource allocation patterns, and market environments that may simultaneously influence technology adoption and economic returns (Pan et al., 2021; Eyasmin and Ghosh, 2024). These variables include breeding scale, roughage share, feeding system, breeding method, and calf price. Breeding scale captures differences in production scale, while roughage structure, feeding systems, and breeding methods reflect heterogeneity in production organization, resource allocation, and cost structures. Calf price is included to control for variations in market conditions affecting economic returns.

(3) Institutional support and external resource environments

Institutional support and external resource environment variables were included to capture differences in farmers’ access to financial resources, technical services, and organizational support, which may influence the sustained implementation and value realization of management-intensive technologies (Manda et al., 2016). Specifically, this study includes organizational participation, beef cattle loan access, current technical training, technology subsidy, and government training as control variables.

Finally, to examine whether the nonlinear association is sensitive to different control variable specifications, control variables were introduced progressively according to farmers’ decision-making processes and beef production management logic. Model 1 includes only PFMT adoption intensity and its quadratic term to identify the baseline nonlinear association between whole-cycle reproductive management adoption intensity and economic returns. Model 2 further incorporates household capability and resource endowment variables to control for differences in farmers’ learning capacity and implementation ability. Model 3 additionally introduces production and market condition variables to account for differences in production foundations and market factors. Model 4 further incorporates institutional support and external resource environment variables, while including county fixed effects and year fixed effects to control for time-invariant regional heterogeneity and common temporal shocks.

2.3 Model specification and estimation strategy

2.3.1 Baseline nonlinear model and estimation strategy

To examine the nonlinear association between whole-cycle reproductive management adoption intensity and economic returns, this study specifies a quadratic regression model that includes the PFMT adoption-intensity index and its squared term, following the nonlinear relationship literature (Haans et al., 2016). Because the dependent variable is continuous and the dataset consists of pooled repeated cross-sectional observations from two independent survey waves, the model is estimated using ordinary least squares (OLS) (Wooldridge, 2010). County fixed effects and year fixed effects are included to control for time-invariant regional heterogeneity and common temporal shocks. In addition, standard errors are clustered at the village level to account for potential within-village correlation arising from the sampling structure.

The baseline nonlinear model is specified as follows:

Yirt=α0+α1PFMTirt+α2PFMTirt2+βXirt+λt+μr+ϵirt

where Yirt denotes the net income per breeding cow of household i in county r and year t; PFMTirtrepresents the adoption-intensity index of whole-cycle reproductive management practices; PFMTirt2is the squared term of adoption intensity; Xirtrepresents a vector of control variables; λtand μr denote year fixed effects and county fixed effects, respectively; and ϵirtis the error term.

2.3.2 Robustness check strategy

To examine the robustness of the baseline findings, this study conducts three additional checks. First, the dependent variable is replaced with an alternative measure of total net income from breeding-cow farming (Y3) to assess whether the main conclusion remains unchanged under a different outcome definition. Second, the core explanatory variable is replaced with the entropy-weighted index to examine whether the result is robust to alternative weighting schemes. Third, all continuous variables are winsorized at the 1% level to reduce the potential influence of extreme observations on the estimated curvature and turning point of the quadratic model. These tests assess whether the observed nonlinear association satisfies formal statistical requirements and remains stable under repeated sampling. All robustness analyses are conducted within the same baseline regression framework, including county fixed effects, year fixed effects, and village-clustered standard errors.

2.3.3 Heterogeneity and moderation analyses

Because external resource conditions may influence farmers’ ability to sustain management-intensive innovations, this study further examines whether the nonlinear association between whole-cycle reproductive management adoption intensity and economic returns differs under different support environments (Omotilewa et al., 2019; Regassa et al., 2023; Addai et al., 2023). Specifically, heterogeneity and moderation analyses are conducted from two dimensions: subsidy support and credit support.

First, subgroup regressions are conducted according to whether households receive subsidy support and whether they have access to credit support. The purpose of the heterogeneity analysis is to explore whether the nonlinear association between PFMT adoption intensity and economic returns differs across households facing different external support conditions. Specifically, we compare the estimated coefficients of the linear term, quadratic term, and turning points of PFMT adoption intensity across different groups.

Second, based on the heterogeneity results, moderation analysis is further employed to formally examine whether external support conditions systematically alter the strength of the nonlinear association between PFMT adoption intensity and economic returns. Specifically, subsidy support, credit support, and their interaction terms with both PFMT and its squared term are introduced into the nonlinear regression model to assess the moderating role of external support on the linear and nonlinear components of the adoption–returns relationship. The moderation model is specified as follows:

Yirt=β0+β1PFMTirt+β2PFMTirt2+β3Mirt+β4(PFMTirt×Mirt)+β5(PFMTirt2×Mirt)+γXirt+λt+μr+ϵirt

Where Mirtrepresents the moderating variable, including subsidy support and credit support. The interaction terms PFMTirt×Mirt and PFMTirt2×Mirtare introduced to examine whether external support conditions alter the strength of the linear and nonlinear associations between adoption intensity and economic returns, respectively. Xirt represents the vector of control variables, while λt and μr denote year fixed effects and county fixed effects, respectively. Other variables are defined consistently with the baseline model.

2.3.4 Additional analysis accounting for potential endogeneity

Although the baseline models control for a wide range of household characteristics, production conditions, market factors, and institutional support variables, PFMT adoption intensity may still be correlated with unobservable factors that simultaneously influence adoption decisions and economic returns. Therefore, this study further employs a nonlinear control function (CF) approach as a supplementary analysis to examine whether the nonlinear association remains robust after accounting for potential endogeneity concerns (Wooldridge, 2015).

Because PFMT adoption intensity enters the outcome equation in both linear and quadratic forms, PFMT and PFMT² are treated as two potentially endogenous components. To identify these components, this study employs two leave-out peer adoption measures constructed at different spatial scales, together with their squared terms, as excluded instruments. Specifically, Z2represents the village-level leave-out average adoption rate, capturing local peer learning and information diffusion, whereas Z3represents the cross-village leave-out average adoption rate within the same township, capturing broader diffusion environments beyond the focal village.

The nonlinear control function approach includes two first-stage equations:

PFMTirt=π0+π1Z2irt+π2Z3irt+π3Z2irt2+π4Z3irt2+γXirt+μr+λt+v1,irt

PFMTirt2=θ0+θ1Z2irt+θ2Z3irt+θ3Z2irt2+θ4Z3irt2+γXirt+μr+λt+v2,irt

where PFMTirt denotes the adoption intensity of whole-cycle reproductive management practices for household i in county r and year t; Z2irtand Z3irt represent the village-level and cross-village leave-out average adoption rates, respectively; Xirt denotes the vector of control variables; μr and λt denote county and year fixed effects; and v1,irt and v2,irt capture unobservable components associated with PFMT and PFMT², respectively.

The second-stage control function specification is:

Yirt=β1PFMTirt+β2PFMTirt2+ρ1v^1,irt+ρ2v^2,irt+γXirt+μr+λt+ϵirt

where v^1,irt and v^2,irt are the estimated residuals from the two first-stage equations. Including these residual terms helps account for potential unobservable factors correlated with both PFMT adoption decisions and economic returns. Unlike conventional IV-2SLS approaches that replace endogenous variables with predicted values, the control function approach retains the original endogenous variables and incorporates the estimated residual components into the outcome equation.

The instrumental variables are constructed using leave-out peer adoption measures. Both Z2and Z3exclude the focal household’s own adoption behavior (and Z3further excludes the focal village) to reduce mechanical correlation and reflection concerns. Detailed construction procedures and formulas are provided in the Supplementary Materials.

As an additional robustness check, the nonlinear control function model is further estimated with township fixed effects to account for potential township-level common factors.

3 Results

3.1 Descriptive statistics of the main variables

Table 2 presents the descriptive statistics of the main variables for the 384 sampled households. The expert-weighted PFMT adoption-intensity index, which serves as the core explanatory variable, has a mean of 0.49, a standard deviation of 0.16, and a range from 0.15 to 0.92, indicating substantial variation in adoption intensity across households.

Table 2. Descriptive statistics of the main variables.

Compiled by the authors based on the survey questionnaire data.

Net income per breeding cow, measured in 10,000 CNY/head, is the main dependent variable. It has a mean of −0.15, a standard deviation of 0.70, and ranges from −6.00 to 3.71. The negative average value reflects genuine economic losses experienced by some households during the survey period rather than any data transformation, as the variable enters the regression models in its original scale without logarithmic transformation or standardization. The substantial variation in economic returns indicates considerable heterogeneity in the profitability of smallholder beef breeding-cow production. The raw distributions of the two outcome variables are presented in Figure 2.

Among the control variables, the household head is, on average, 50.19 years old, has 7.74 years of education, and has 10.32 years of cattle-raising experience. In addition, breeding cow scale, the share of purchased roughage, and calf price all show considerable variation, supporting the use of the sample for subsequent regression analysis.

3.2 Baseline regression results

Table 3 reports the baseline nonlinear regression results for the association between whole-cycle reproductive management adoption intensity and net income per breeding cow. Across the four specifications, the coefficient of PFMT adoption intensity is negative, whereas the coefficient of PFMT² is positive. The quadratic term remains statistically significant at conventional levels across models, indicating a nonlinear association between adoption intensity and economic returns.

Table 3. Baseline nonlinear regression results.

*, , and * indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

In Model 1, the coefficients of PFMT and PFMT² are −1.959 and 2.073, respectively, yielding an estimated turning point of 0.472. This turning point falls within the observed range of PFMT adoption intensity (0.155–0.923), suggesting that the nonlinear pattern is empirically relevant. The estimated turning points remain stable after sequentially adding household capability variables (Model 2), production and market condition variables (Model 3), and institutional support variables (Model 4), ranging from 0.461 to 0.478. Overall, the baseline results indicate that the association between whole-cycle reproductive management adoption intensity and economic returns is characterized by a U-shaped pattern. These findings provide initial support for Hypothesis 1, which posits that the relationship between whole-cycle reproductive management adoption intensity and economic returns is nonlinear rather than monotonically increasing.

Figure 3 presents the estimated marginal association between whole-cycle reproductive management adoption intensity and net income per breeding cow based on the baseline quadratic regression. The marginal association increases continuously as the adoption-intensity index rises. At relatively low adoption intensity levels, the marginal association is negative or close to zero, whereas it becomes positive after the estimated turning point (approximately 0.46) and continues to increase at higher adoption levels. This pattern visually illustrates the nonlinear association identified in the baseline regression results.

3.3 Formal tests and stability analysis of the nonlinear association

Although the baseline regression results reveal a significant quadratic relationship between PFMT adoption intensity and economic returns, a statistically significant quadratic term alone does not necessarily establish a strict U-shaped relationship. Therefore, this study further applies the Lind–Mehlum strict U-test to formally examine whether the estimated nonlinear association satisfies the required conditions for a U-shaped pattern. Table 4 reports the results of the strict U-test. Across the four model specifications, the estimated turning points range from 0.461 to 0.478, all of which fall within the observed range of PFMT adoption intensity (0.155–0.923). In addition, the slopes at the lower and upper bounds are consistently negative and positive, respectively. The overall U-test p-values range from 0.029 to 0.052 across specifications, providing formal statistical support for the presence of a U-shaped association between PFMT adoption intensity and economic returns.

Table 4. Lind–Mehlum strict U-test results for the nonlinear association.

The Lind–Mehlum strict U-test examines whether the slope is negative at the lower bound and positive at the upper bound within the observed PFMT adoption-intensity range. The turning point is calculated as −β1/(2β2).

To further assess the stability of the nonlinear association and the reliability of the estimated threshold, a Bootstrap procedure with 1,000 replications was conducted. As shown in Table 5, the Bootstrap mean turning points remain close to the original estimates across alternative model specifications, ranging from 0.457 to 0.482. The percentile-based 95% Bootstrap confidence intervals indicate that the estimated turning points remain within reasonable ranges around the original values, suggesting that the identified threshold is not highly sensitive to sampling variation. Moreover, the quadratic coefficient remains positive in 97.3%–99.1% of Bootstrap replications, providing additional evidence that the upward curvature of the nonlinear association is stable across repeated samples.

Table 5. Bootstrap stability test of the PFMT nonlinear turning point and curvature.

The Bootstrap stability analysis is based on 1,000 replications. In each replication, the nonlinear regression model is re-estimated and the turning point is recalculated. The reported confidence intervals are percentile-based 95% Bootstrap confidence intervals for the estimated turning points. The proportion of replications with β₂>0 indicates the stability of the positive curvature of the nonlinear association. The turning point is calculated as −β₁/(2β₂).

3.4 Results of the robustness checks

Table 6 presents the robustness checks. The results remain consistent across alternative outcome measures, alternative PFMT adoption-intensity indices, and winsorized samples. In all specifications, the linear term remains negative and the quadratic term remains positive, with estimated turning points ranging from 0.431 to 0.492. Replacing the outcome variable with household-level net operating income (Y3), using an entropy-weighted PFMT index, and winsorizing continuous variables at the 1% level all yield similar nonlinear patterns. These results indicate that the nonlinear association between PFMT adoption intensity and economic returns is robust to alternative measurements and sample treatments.

Table 6. Summary of robustness checks.

All models include the same control variables, county fixed effects, year fixed effects, and village-clustered standard errors. The turning point is calculated as −β1/(2β2). * and ** indicate statistical significance at the 10% and 5% levels, respectively.

3.5 Boundary conditions of the nonlinear association

3.5.1 Exploratory heterogeneity analysis by external support conditions

Figure 4 illustrates heterogeneous nonlinear associations between PFMT adoption intensity and economic returns under different external support conditions. The subgroup analysis shows that the nonlinear association is mainly observed among farmers with access to external support. Specifically, among farmers without subsidy support (N = 169) or credit support (N = 281), the linear and quadratic terms of PFMT adoption intensity are not statistically significant. In contrast, farmers receiving subsidies (N = 215) and credit support (N = 103) exhibit a significant U-shaped association, with estimated turning points of approximately 0.53 and 0.50, respectively.

These patterns suggest that external support conditions may influence the extent to which farmers can translate higher-intensity PFMT adoption into economic returns. Without sufficient external resources, the association between adoption intensity and returns remains relatively weak, whereas supportive resource environments are associated with a more pronounced nonlinear pattern. These subgroup results provide descriptive evidence of heterogeneous associations across different resource conditions and motivate the subsequent moderation analysis of external support as a boundary condition.

3.5.2 Moderation analysis of external support

While the subgroup analysis identifies differences in nonlinear associations across support conditions, it does not formally test whether external support changes the strength of the PFMT–economic returns relationship. Therefore, we further estimate moderation models by introducing interaction terms between external support indicators and both the linear and quadratic terms of PFMT adoption intensity.

Table 7 reports the moderation results. The interaction terms between external support and PFMT adoption intensity are statistically significant for both subsidy and credit support. Specifically, the interaction between PFMT and subsidy support is negative and significant (β=−7.446,p<0.01), while the interaction between PFMT² and subsidy support is positive and significant (β=6.750,p<0.01). Similar patterns are observed for credit support, with significant interaction effects for both PFMT × Credit (β=−13.863,p<0.01) and PFMT² × Credit (β=13.319,p<0.01).

Table 7. Moderation analysis results for subsidy and credit support.

The dependent variable is net income per breeding cow. External support refers to subsidy support in column (1) and credit support in column (2). All models include the same set of control variables, year fixed effects, and county fixed effects. Village-clustered standard errors are reported in parentheses. *** indicates statistical significance at the 1% level.

These results indicate that external support modifies the shape of the nonlinear association rather than merely shifting income levels. Specifically, subsidies and credit support strengthen the curvature of the PFMT–economic returns relationship, making the transition from low-intensity adoption to higher economic returns more pronounced. Therefore, the results provide support for H2 that external support functions as an important boundary condition influencing the extent to which economic returns are associated with intensive PFMT adoption.

3.6 Results of the nonlinear control function analysis

Although the baseline models control for a wide range of observable characteristics, PFMT adoption intensity may still be correlated with unobservable factors that simultaneously influence adoption decisions and economic returns. Therefore, we further employ a nonlinear control function approach as supplementary evidence to examine whether the nonlinear association remains robust after accounting for potential endogeneity concerns.

Because PFMT enters the outcome equation in both linear and quadratic forms, PFMT and PFMT² are treated as two potentially endogenous components. Table 8 reports the results of the nonlinear control function approach. The first-stage results indicate that the excluded instruments jointly explain significant variation in both PFMT adoption intensity and its quadratic component. Specifically, the village-level and cross-village leave-out adoption measures and their squared terms are jointly significant in the PFMT and PFMT² first-stage equations (F = 8.65 and F = 8.67, respectively; p<0.001).

Table 8. Results of the control function approach.

The dependent variable is net income per breeding cow. Z2 and Z3 denote village-level and cross-village leave-out average adoption rates, respectively. Joint F-tests report the significance of excluded instruments in the corresponding first-stage equations. All models include control variables, year fixed effects, and county fixed effects. Standard errors are clustered at the village level. v^1 and v^2 denote the estimated residuals from the first-stage equations for PFMT and PFMT², respectively. *, , and * indicate significance at the 10%, 5%, and 1% levels, respectively.

In the second-stage control function estimation, the coefficient of PFMT remains negative, while the coefficient of PFMT² remains positive, consistent with the baseline nonlinear specification. The estimated turning point is approximately 0.526, which is close to the baseline estimate and remains within the observed range of adoption intensity. These results suggest that the nonlinear association between PFMT adoption intensity and economic returns remains qualitatively similar after accounting for potential endogeneity concerns.

To further address concerns that spatial peer instruments may capture township-level common factors, we re-estimate the nonlinear control function model using township fixed effects instead of county fixed effects. The results remain highly consistent, with the turning point estimated at 0.534 (Supplementary Table S6), suggesting that the estimated nonlinear association remains stable after controlling for additional township-level heterogeneity.

4 Discussion

This study shows that whole-cycle reproductive management adoption intensity is significantly associated with economic returns in smallholder beef breeding-cow systems, but this association is neither immediate nor linear. Instead, the estimated relationship follows a U-shaped pattern. At relatively low levels of adoption intensity, economic returns remain limited, whereas the positive association with net income per breeding cow becomes more evident once adoption intensity exceeds the estimated turning point. This finding is broadly consistent with previous studies showing that improved reproductive management and management-related innovations can enhance farm performance in breeding females (Inchaisri et al., 2010; Taylor et al., 2017; Tse et al., 2018; Wicaksono et al., 2024). At the same time, the present study extends this literature by showing that the economic value of reproductive management depends not only on whether farmers adopt specific practices, but also on the intensity, continuity, and coordination with which these practices are implemented across the reproductive cycle.

A possible explanation for this nonlinear pattern is that whole-cycle reproductive management does not operate as a single input with an immediate economic payoff. Rather, it represents a package of knowledge- and management-intensive practices that require continuous adjustment across multiple stages of the breeding-cow cycle, including pre-breeding preparation, gestation, calving, postpartum recovery, remating, and culling. When adoption intensity is low, farmers may implement only part of this management package. Under such conditions, they may incur learning costs, adjustment costs, record-keeping costs, and additional managerial effort without yet achieving sufficient coordination across reproductive stages. As a result, short-term economic gains may remain weak or even negative. As adoption intensity increases, however, management becomes more systematic and better aligned with the physiological and reproductive requirements of breeding cows. Improved coordination across breeding timing, body condition management, postpartum recovery, remating decisions, and culling strategies may therefore be increasingly associated with improved economic returns.

This interpretation highlights the importance of a whole-cycle perspective when evaluating reproductive management in beef breeding-cow systems. Existing studies often focus on specific performance indicators, such as reproductive rate, feed efficiency, calf survival, or production costs (Consentini et al., 2021; Umaña Sedó et al., 2023; Ojo et al., 2024). These indicators are important, but the present findings suggest that the economic value of reproductive management cannot be fully understood by examining isolated practices or individual stages alone. The reproductive cycle of breeding cows is sequential and interdependent. Weak implementation at one stage may reduce the value of improvements made at another stage. For example, gestation-stage feeding may be less effective if pre-breeding preparation is inadequate, while postpartum recovery management may have limited value if it is not linked with timely remating and rational culling decisions. Therefore, whole-cycle reproductive management should be evaluated not only through individual management outcomes, but also through its capacity to coordinate practices across the reproductive cycle and support the overall functioning of cow–calf production systems.

From the perspective of animal production management, these findings suggest that whole-cycle reproductive management should be understood as a management-intensive innovation rather than as a set of isolated practices. This distinction is particularly important for smallholder beef production systems. Compared with capital-intensive technologies, such as automated equipment, advanced facilities, or digital monitoring systems, management-intensive practices rely more on farmers’ knowledge, observation, timing, record keeping, decision-making, and sustained implementation. For small- and medium-scale farms with limited capital, such practices may represent a relatively accessible management pathway associated with improved economic sustainability. However, their benefits are unlikely to emerge automatically. Farmers need to implement these practices with sufficient intensity, continuity, and coordination before management improvement can be translated into higher and more stable returns.

The heterogeneity and moderation results further indicate that the association between management adoption and economic returns depends partly on external support conditions. Among farmers receiving subsidy support or having access to credit, the U-shaped relationship between adoption intensity and net income per breeding cow is more pronounced. This pattern suggests that external support may be related to farmers’ capacity to sustain management-intensive practices by alleviating financial constraints and supporting continuous implementation. Subsidies may reduce resource constraints associated with management adoption, while credit access may improve farmers’ ability to maintain production and management adjustments over time.

From a broader perspective, this study contributes to the literature on livestock management innovation and smallholder beef production systems. Previous studies suggest that precision livestock farming and related management innovations are associated with improvements in productivity, sustainability, and management efficiency, particularly when they are integrated into production decision-making rather than treated as isolated tools (Aquilani et al., 2022; Carillo and Abeni, 2020). Building on this literature, the present study reframes reproductive management in beef breeding-cow systems as a whole-cycle management strategy. This perspective is relevant because management decisions are distributed across a sequence of biologically linked stages, including pre-breeding, gestation, postpartum recovery, and remating or culling. The findings also show that management-based practices should not be evaluated solely through a binary adopted/not adopted distinction. Instead, greater attention should be given to adoption intensity, implementation depth, continuity, and the extent to which reproductive management practices are coordinated across stages. In this sense, the study extends existing research by emphasizing the cumulative and threshold-type nature of management benefits in smallholder beef breeding-cow systems.

This study has several limitations. First, because the analysis is based on pooled repeated cross-sectional data, it cannot fully identify the dynamic process through which farmers learn, adjust, and coordinate reproductive management practices over time. Second, although the results are consistent with the argument that learning costs and coordination constraints shape the nonlinear association, these mechanisms are inferred rather than directly measured. Third, net income per breeding cow is an important indicator of economic performance, but future research could combine economic outcomes with biological and reproductive indicators such as calving rate, conception rate, calf survival, body condition score, cow longevity, and herd renewal efficiency. Future studies could benefit from panel data, follow-up surveys, or more detailed process-level indicators to identify more explicitly the dynamic pathways through which whole-cycle reproductive management adoption improves both biological performance and economic sustainability in beef breeding-cow systems.

5 Conclusions

Using survey data from 384 beef breeding-cow farmers in Jilin, Heilongjiang, and Shaanxi Provinces, China, this study developed an adoption-intensity index for whole-cycle reproductive management practices and examined its association with net income per breeding cow. The results reveal a significant U-shaped association between adoption intensity and economic returns. At low levels of adoption intensity, economic benefits remain limited, likely because management practices are implemented only partially and coordination across reproductive stages is insufficient. Beyond the estimated turning point, its positive association with net income per breeding cow becomes increasingly evident. Further analyses show that subsidy and credit support are important boundary conditions associated with a stronger nonlinear relationship between whole-cycle reproductive management adoption intensity and economic returns.

These findings suggest that the economic value of whole-cycle reproductive management depends not only on whether farmers adopt individual practices, but also on the intensity, continuity, and coordination with which these practices are implemented across the reproductive cycle. Whole-cycle reproductive management should therefore be understood as a management-intensive innovation rather than as a set of isolated technical measures. For smallholder beef breeding-cow systems, policies that improve farmers’ ability to implement coordinated reproductive management, together with appropriate financial and technical support, may help create more favorable conditions for realizing the potential economic value of management-intensive innovations. Future research could further examine the biological and behavioral mechanisms linking whole-cycle reproductive management adoption to farm performance by using panel data and more detailed indicators of calving rate, reproductive efficiency, calf survival, body condition, herd renewal, and long-term economic outcomes.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fanim.2026.1878033/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Rondineli Pavezzi Barbero, Federal Rural University of Rio de Janeiro, Brazil

Reviewed by: Ana Carolina Espasandin, Universidad de la República, Uruguay

Jing Zhang, Fujian Academy of Agricultural Sciences, China