Research on incentive mechanisms for veterans’ participation in grassroots governance: An analysis based on behavioral expectations Shuqiong Wang, Zhao Zhang ABSTRACT Under the modernization of grassroots governance, veterans constitute a high-quality human resource for enhancing grassroots social governance capacity. Drawing on questionnaire survey data from 525 veterans in Guizhou Province, China, this study identifies three distinct veteran groups using latent class analysis (LCA): Utilitarian (51.3%), Mission-Oriented (23.8%), and Development-Focused (24.9%). Empirical results reveal clear heterogeneity in policy responsiveness and behavioral mechanisms across groups. First, the Utilitarian group responds most strongly to Subsidies, with participation behavior predominantly driven by economic rationality. Second, the Mission-Oriented group demonstrates honor-driven motivations, with symbolic recognition playing a central role in stimulating governance engagement. Third, the Development-Focused group prioritizes career advancement and exhibits significant responsiveness to promotion-related incentives. Based on these findings, this study constructs a mediation model of “Policy Instruments (PI)—Behavioral Expectations (BE)—Participation Behavior (PB)” and proposes an intervention pathway centered on “expectation diagnosis” and “instrument matching.” By revealing the internal mechanism through which policy instruments (PI) influence participation behavior (PB) via Behavioral expectations (BE), this research provides empirical evidence and theoretical insights for shifting incentive strategies from “uniform provision” to “precision adaptation.”. 1. Introduction Grassroots governance serves as the cornerstone of modern governance systems and plays a decisive role in translating public policies into tangible social outcomes. In many countries, increasingly complex social structures, demographic shifts, and governance fragmentation have intensified the demand for diverse governance actors beyond traditional administrative frameworks [ 1 – 3 ]. In this context, veterans—with their organizational discipline, crisis response capabilities, and collective orientation cultivated during military service—have emerged as a distinctive group capable of injecting vitality into grassroots governance, garnering growing attention from both academia and policymakers [ 4 – 6 ]. International research on veterans’ post-service participation has primarily focused on three interrelated themes. First, studies on veteran reintegration highlight the complex psychological and social challenges associated with the transition from military to civilian life [ 7 – 9 ]. Recent research emphasizes that successful reintegration requires not only institutional support but also strong community engagement and identity reconstruction [ 10 – 12 ]. These studies demonstrate that veterans’ post-service roles are shaped by both social networks and institutional opportunities. Second, a growing body of literature examines civic and governance participation. Research on public participation indicates that institutional arrangements and policy feedback mechanisms can significantly influence citizens’ engagement in governance activities [ 2 , 13 , 14 ], while broader political and cultural contexts may also shape patterns of civic participation [ 15 ].Within veteran populations, military socialization and shared service experiences may enhance collective identity and civic commitment, thereby facilitating participation in community governance [ 4 , 6 ]. Third, policy design research increasingly highlights the importance of policy instruments in shaping behavioral responses. Scholars emphasize that different types of instruments—such as economic incentives, symbolic recognition, and capacity-building measures—operate through distinct behavioral mechanisms [ 16 – 18 ]. However, the effectiveness of these instruments depends largely on how target groups perceive and interpret policy signals [ 19 , 20 ]. Within the Chinese context, research on veterans’ participation in grassroots governance has developed rapidly in recent years, driven by the institutionalization of veterans’ affairs and the enactment of the Veterans Protection Law. Existing studies have examined veterans’ engagement from perspectives such as policy support, organizational embeddedness, and social role transformation [ 21 ]. In particular, research on Policy Instruments (PI) has drawn on public policy design theory to categorize incentive tools into material, symbolic, and developmental types [ 16 – 17 ]. Meanwhile, psychological studies have underscored the lasting influence of military identity, honor, and service ethos on veterans’ post-service behavior [ 22 – 23 ]. Despite these advances, two significant research gaps remain inadequately addressed. The first gap concerns the heterogeneity within the veteran population. Most existing studies implicitly treat veterans as a relatively homogeneous group, overlooking substantial variations in their post-service expectations, career trajectories, and governance motivations. Although recent research has begun to apply Latent Class Analysis (LCA) to identify subgroups of veterans with different risk profiles or behavioral tendencies [ 24 ], such approaches have not yet been systematically integrated into governance participation research, particularly in relation to incentive design. The second gap lies in the mismatch between Policy Instruments (PI) and Behavioral Expectations (BE). While various policy tools aim to encourage veterans’ participation in grassroots governance, their effectiveness is often hindered by a neglect of veterans’ subjective expectations. Existing studies typically assess either Policy Instruments (PI) or Behavioral Expectations (BE) in isolation, failing to elucidate how specific Policy Instruments (PI) activate—or fail to activate—the psychological mechanisms that translate into Participation Behavior (PB). Addressing these gaps requires moving beyond purely structural or institutional explanations of incentives toward a mechanism-oriented analytical framework. This study conceptualizes veterans’ participation as a process mediated by Behavioral Expectations (BE)—that is, individuals’ subjective anticipations regarding material returns, social recognition, and personal development. To clarify the theoretical positioning of Behavioral Expectations (BE), this study distinguishes the concept from several related constructs in behavioral research. Behavioral expectations refer to individuals’ anticipatory evaluations of the outcomes associated with a potential action. In contrast to behavioral intention in the Theory of Planned Behavior [ 25 ], which reflects an individual’s readiness to perform a behavior, BE focus on how individuals assess the potential returns of action prior to forming such intentions. BE also differ from motivation and attitudes: motivation refers to the internal forces that drive behavior, while attitudes represent relatively stable evaluative orientations. By comparison, behavioral expectations are situational and policy-sensitive assessments regarding expected outcomes. This understanding aligns with expectancy theory [ 26 ], which emphasizes that expectations about outcomes shape behavioral choices. Conceptualizing BE in this way provides a psychological mechanism linking policy instruments to participation behavior. Unlike approaches that assume a universal model of economic rationality, this method recognizes that incentive effects depend on how different groups interpret policy signals. It should be explicitly noted that this study does not claim causal inference in a strict experimental sense. Based on cross-sectional survey data, the proposed analytical framework focuses on identifying empirically consistent mediation pathways between Policy Instruments (PI), Behavioral Expectations (BE), and Participation Behavior (PB). Although alternative causal directions remain theoretically plausible, examining these mediation patterns provides valuable insights into how different veteran groups psychologically process Policy Instruments (PI), thereby offering an evidence-based foundation for more precise policy design. Using questionnaire data from 525 veterans in Guizhou Province, this study integrates Latent Class Analysis (LCA) and mediation modeling to achieve three research objectives: first, to identify distinct subgroups of veterans based on their Behavioral Expectations (BE); second, to analyze the sensitivity of different groups to specific Policy Instruments (PI) and their combinations (Combined-PI); and third, to test whether Behavioral Expectations (BE) function as a mediating mechanism linking Policy Instruments (PI) to Participation Behavior (PB). Through this research approach, the study contributes to public governance research in two ways: empirically, by revealing the structural heterogeneity in veterans’ incentive responses; and theoretically, by refining—rather than replacing—existing explanations of economic rationality through an expectation-mediated perspective applicable to governance participation. For clarity and conciseness, the following abbreviations are used throughout the remainder of this manuscript: Policy Instruments (PI), Behavioral Expectations (BE), and Participation Behavior (PB). Unless otherwise specified, these abbreviations are used consistently in all subsequent sections. 2. Methods 2.1. Definition of core constructs The study examines three central concepts: Veterans refer to military personnel who have completed active service, including discharged soldiers, non-commissioned officers, and commissioned officers. This group shares distinctive attributes: rigorous military training that develops strong discipline and teamwork abilities, specialized skills in emergency response and resource coordination, and the challenging transition from military to civilian roles requiring social readjustment. Grassroots Governance encompasses public affairs management at basic administrative levels like townships, sub-districts, villages and communities. Its key features include multiple participating stakeholders (government bodies, social organizations, residents), direct handling of livelihood-related matters such as public safety and environmental maintenance, and governance approaches emphasizing consultation and co-creation. Incentive mechanisms denote institutional designs aimed at motivating participation, consisting of three key dimensions: material incentives (compensation packages, welfare benefits), recognition-based incentives (honors, social acknowledgment), and Development-focused incentives (career advancement opportunities, skills training programs). 2.2. Survey instrument and rationale for question design The questionnaire employed multiple measurement formats tailored to the characteristics of each construct. Behavioral Expectations (BE) were measured using a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree), capturing veterans’ expectations regarding material returns, social recognition, and personal development through items B1–B3. Policy Instrument (PI) perception was assessed using two complementary approaches: exposure frequency (never, occasional, frequent) and attractiveness evaluation on a 0–10 rating scale for specific instruments such as subsidies, honors, training, and promotion. Participation Behavior (PB) was measured using two indicators reflecting behavioral engagement: participation frequency in governance activities and the degree of decision-making involvement. This multi-format measurement design enabled the study to capture both subjective expectations and observed participation behaviors, providing an empirical basis for subsequent latent class analysis (LCA) and mediation modeling. The questionnaire is structured as follows in Table 1 10.1371/journal.pone.0351162.t001 Table 1 Veterans’ participation in grassroots governance survey. Instructions: This survey is solely for academic purposes and is anonymous. Completion time: ~ 12 minutes. Thank you for your support! A. Demographic Information A1 . Years since discharge: □ < 2 □ 2-5 □ 5-10 □ > 10 Rationale: A1(Critical timepoints affect social readjustment [ 19 ]). A2 . Highest military rank: □ Enlisted soldier( soldier ) □ Non-commissioned officer ( NCO ) □ Commissioned officer( officer ) Rationale: A2(Rank correlates with resource acquisition capacity [ 27 ]). A3 . Monthly income (CNY): □ < 3000 □ 3000-5000 □ 5000-8000 □ > 8000 Rationale: A3(Controls economic status’ effect on PI sensitivity (ref. EVS scale)). B. Behavioral Expectations (BE) (7-point Likert scale: 1 = Strongly disagree, 7 = Strongly agree) B1 . I hope that participation in grassroots governance can provide substantial income. □1 □2 □3 □4 □5 □6 □7 Rationale: B1(Measures instrumental BE [ 22 ]). B2 .I hope to improve the community’s respect for me through participation. □1 □2 □3 □4 □5 □6 □7 Rationale: B2(Assesses military identity [ 22 ]). B3 . I hope to improve management skills through participation □1 □2 □3 □4 □5 □6 □7 Rationale: B3(Evaluates developmental BE [ 22 ]). C. Policy Instruments (PI) C1 . Exposure frequency (past 2 years) PI Never Occasional Frequent C11 Subsidies □ □ □ C12 Honors □ □ □ C13 Training □ □ □ Rationale: C1(Measures PI exposure intensity [ 22 ]). C2 . Single PI attractiveness (0–10 scale) PI Rating C21 Subsidies □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 C22 Honors □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 C23 Training □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 C24 Promotion □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 Rationale: C21(Baseline economic incentive (Art.25, China’s Veterans Support Law)); C22(Military-specific symbolic needs [ 16 ]);C23(CLong-term human capital [ 19 ]); C24(Career development focus (ref. EVS)). C3 . Combined-PI attractiveness (0–10 scale) Combined-PI Rating C31 Subsidies+Honors □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 C32 Training+Promotion □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 C33 Subsidies+Training+Honors □ 1 □ 2 □ 3 □ 4 □ 5 □ 6 □ 7 □ 8 □ 9 □ 10 Rationale: C31(Material-symbolic synergy [ 20 ]);C32(Dual capability-position path [ 23 ]); C33(Comprehensive incentive mix [ 16 ]). D. Participation Behavior (PB) D1 .Governance participation frequency (past 6 months): □0 □1-3 □4-6 □ > 6 D2 .Decision-making involvement in participated activities: □ < 20% □20-50% □ > 50% Rationale: D2(Differentiates executive vs. decision-making PB [ 5 ]). To ensure the credibility of subsequent analyses, reliability tests were conducted for the core modules of the survey instrument, including Behavioral Expectations (BE), Policy Instrument perception (PI), and Participation Behavior (PB). Different reliability assessment methods were applied according to the measurement structure of each construct. Cronbach’s α was used to assess the internal consistency of the reflective Behavioral Expectations (BE) scale, while Composite Reliability (CR) was employed for Policy Instruments (PI), which represent a formative construct composed of different policy tools. For Participation Behavior (PB), which contains only two indicators, the Spearman–Brown coefficient was used to evaluate reliability. The results demonstrate acceptable reliability levels for all constructs and are presented in Table 2 . 10.1371/journal.pone.0351162.t002 Table 2 Reliability and validity analysis of the survey instrument. Module Measurement Items Test Method Result Interpretation BE B1-B2-B3 Cronbach’s α α=0.76 Acceptable internal consistency PI C21-C22-C23-C24 Composite Reliability (CR) CR = 0.71 Meets minimum reliability standard for formative constructs PB D1-D2 Spearman-Brown Coefficient SB = 0.66 Moderate reliability 2.3. Data compilation The survey was administered via a mobile application through veterans affairs bureaus across 10 counties in Guizhou Province, yielding 525 valid responses. The frequency distributions are presented in Table 3 . 10.1371/journal.pone.0351162.t003 Table 3 Survey data summary. Demographic Information A1 <2 years 2-5years 5-10 years >10 years 38 142 224 121 A2 soldier NCO Officer 386 104 35 A3 (CNY) <3000 3000-5000 5000-8000 >8000 255 167 75 28 BE Item 1 2 3 4 5 6 7 Mean B1 9 10 36 52 78 126 214 5.69 B2 21 46 56 100 109 104 89 4.71 B3 8 40 133 127 75 90 52 4.33 PI C1 Never Occasional Frequent C11 16 446 63 C12 6 421 98 C13 102 377 46 C2 1 2 3 4 5 6 7 8 9 10 Mean C21 1 6 8 7 20 32 79 80 157 135 8.22 C22 22 12 20 56 61 46 73 78 56 101 6.78 C23 33 28 20 103 125 67 55 41 11 32 5.18 C24 8 11 20 16 22 6 11 105 115 211 8.31 C3 1 2 3 4 5 6 7 8 9 10 Mean C31 0 8 4 11 5 17 20 55 106 299 9.00 C32 1 7 3 11 38 55 53 108 124 125 7.98 C33 0 0 0 2 9 13 13 26 100 362 9.10 PB D1 0 1-3 4-6 >6 19 271 147 88 D2 <20% 20-50% >50% 351 132 42 2.4. Data collection protocol The participant recruitment for this study was completed between February 15 and May 20, 2024. Throughout the research process, all data underwent rigorous anonymization procedures to ensure that researchers could not access any information that could identify individual participants. The processed dataset has been stripped of all sensitive personal information, fundamentally safeguarding participant privacy. Prior to the study implementation, standardized ethical procedures were followed to obtain verbal informed consent. All participants provided oral confirmation of their agreement to participate under conditions of full informed consent. 2.5. Identification of behavioral expectation types The veteran population exhibited significant heterogeneity. Using Latent Class Analysis (LCA) on behavioral expectation measures (B1-B2-B3) from the survey, we classified veterans into distinct subgroups. The analysis was conducted in R, with results shown in Table 4 . 10.1371/journal.pone.0351162.t004 Table 4 LCA classification results. Number of Classes AIC BIC LMR p-value Entropy 2 2002.3 2046.7 0.001 0.71 3 1987.6 2068.2 0.003 0.84 4 1975.4 2080.1 0.112 0.83 The selection of latent classes followed standard LCA procedures. With a significant LMR test (p < 0.05), the three-class model showed the lowest AIC and BIC values while maintaining an entropy above 0.8, indicating accurate classification. Based on posterior probabilities obtained from LCA analysis, the 525 samples were classified into their most probable categories. The characteristics of these three categories are presented in Table 5 . 10.1371/journal.pone.0351162.t005 Table 5 Characteristics of three behavioral expectation groups. Type Proportion B1 B2 B3 Demographic Profile Utilitarian 51.3% High (6.2) Low (3.8) Medium (3.9) 87% soldiers 68% discharged <5 years Mission-Oriented 23.8% Medium (4.5) High (5.5) Low (4.4) 59% officers 72% discharged >10 years Development- Focused 24.9% Low (4.7) Medium (4.9) High (5.2) 51% NCOs 81% discharged 2–10 years As shown in Table 5 , the Utilitarian group scored highest on B1 (mean = 6.2). This group mainly consists of recently discharged enlisted soldiers facing financial pressures, who prioritize economic returns and make decisions based on cost-benefit calculations, thus named “Utilitarian”. Similar reasoning applies to the naming of “Mission-oriented” and “Development-focused” groups. Table 5 further reveals systematic differences in the demographic profiles of the three groups. For example, 87% of enlisted soldiers were classified as the “Utilitarian” type, 59% of officers were categorized as the Mission-oriented type, and 51% of NCOs were grouped as the Development-focused type. Among those discharged within five years, 68% were classified as the “Utilitarian” type; among those discharged for more than ten years, 72% were categorized as the Mission-oriented type; and among those discharged between two and ten years, 81% were grouped as the Development-focused type. These characteristics collectively provide a direct and intuitive explanatory basis for understanding the differentiated motivations behind veterans’ participation in governance. 2.6. Policy instruments (PI) sensitivity analysis Based on the LCA classification (Utilitarian, Mission-oriented, Development-focused) established in Section 2.5 , this analysis examines the differential responses of the three groups (Utilitarian, Mission-oriented, and Development-focused) to both individual PI and Combined- PI . The evaluation encompasses four discrete policy tools (Subsidies C21, Honors C22, Training C23, Promotion C24) and three policy combinations (Subsidies+Honors C31, Training+Promotion C32, Subsidies+Training+Honors C33). This provides an empirical foundation for the subsequent development of targeted incentive strategies. Descriptive statistics were first conducted for individual PI, with results presented in Table 6 below: 10.1371/journal.pone.0351162.t006 Table 6 Mean attractiveness scores of single PIs by group. PI Utilitarian (n = 269) Mission-Oriented (n = 125) Development-Focused (n = 131) Full Sample (n = 525) C21 8.99 7.68 7.15 8.22 C22 6.12 7.64 7.31 6.78 C23 4.31 6.18 6.01 5.18 C24 8.15 8.07 8.87 8.31 Table 6 , through descriptive statistics, presents the mean attractiveness scores of four individual PI (C21-C24) across three veteran groups, clearly revealing their differentiated incentive preferences. The Utilitarian group rated Subsidies (C21) the highest (8.99), significantly exceeding their ratings for other instruments, reflecting a decision-making logic centered on economic returns and short-term gains. The Mission-oriented group scored notably high on Honors (C22) (7.64), while showing comparable responsiveness to Subsidies (C21) and Promotion (C24) (7.68 vs 8.07), indicating a multidimensional incentive structure with a distinct honor-driven characteristic. The Development-focused group demonstrated the strongest preference for Promotion (C24) (8.87), further validating their long-term development orientation focused on career growth. Overall, the preferences of the three groups align closely with their core group characteristics. Descriptive analysis was subsequently conducted for Combined-PI, with comparative results shown in Table 7 . 10.1371/journal.pone.0351162.t007 Table 7 Mean attractiveness scores of combined-PI. PI Combination Utilitarian (n = 269) Mission-Oriented (n = 125) Development-Focused (n = 131) Full Sample (n = 525) C31 9.16 8.66 8.99 9.00 C32 6.61 9.37 9.47 7.98 C33 9.27 8.82 9.02 9.10 Table 7 further illustrates the differences in attractiveness of three policy combinations (C31-C33) across different groups, revealing the synergistic effects of tool combinations and their group-specific adaptability. For the Utilitarian group, the “subsidies + recognition” combination (C31) received the highest score (9.16), indicating that adding symbolic recognition on top of material incentives yields optimal results. In contrast, the Development-focused group responded most positively to the “training + promotion” combination (C32) (9.47), demonstrating the significant advantage of this pathway in meeting both their skill enhancement and career development needs. Notably, although the comprehensive policy package (C33) scored the highest in the full sample (9.10), it provided limited additional benefit to the Utilitarian group (only a 3.1% increase compared to C31). This suggests that when designing incentive programs for this group, excessive tool stacking should be avoided, and focus should instead be placed on the most core and effective “economic-honor” combination. These results collectively indicate that the effectiveness of policy combinations is highly dependent on the specific needs of the target group, and precise matching is key to enhancing incentive efficiency. 2.7. Mediation analysis of behavioral expectations(BE) Section 2.5, through Latent Class Analysis (LCA), identifies three distinct groups among veterans: Utilitarian, Mission-oriented, and Development-focused, whose BE (B1–B3) exhibit fundamental differences. Section 2.6 further reveals that these three groups also show differentiated response patterns to various PI (C21–C24) and Combined-PI (C31–C33). However, the underlying logical chain connecting group classification and policy effectiveness remains unclear—how PI influence veterans’ psychological expectations and subsequently drive their PB remains a critical question to be verified. To address this, this section constructs and tests mediation effect models for the three groups, aiming to establish the theoretical pathway of “PI → BE→ PB” thereby validating the core hypothesis of this study: BE serve as the key psychological mediator through which PI influence PB. By selecting independent variables from the perspective of PI, mediating variables from the perspective of BE, and and dependent variables from the perspective of participation indicators for each group, and employing the Bootstrap sampling method to test the mediation effects, this section seeks to uncover the intrinsic psychological pathways through which different veteran groups respond to policies. This approach advances policy design from “differentiation based on group characteristics” to “targeted intervention based on psychological mechanisms” and provides direct empirical support for subsequent policy recommendations. Building upon the findings from Section 2.5 and Section 2.6 , we first specify the selection criteria for independent, mediating, and dependent variables in the mediation effect models for all three veteran groups, as detailed in Table 8 . 10.1371/journal.pone.0351162.t008 Table 8 Variable selection for mediation models. Utilitarian Group Model Variable Type Selected Variable Rationale Independent Variable Subsidies (C21) Highest score (M = 8.99) ( Table 6 ) Mediator Instrumental BE (B1) Highest score (M = 6.2; reflects cost-benefit logic)( Table 5 ) Dependent Variable Participation frequency (D1) Short-term economic drivers prioritize quantity over depth Mission-Oriented Group Model Independent Variable Honors (C22) Exceeded material incentives (M = 7.64,) ( Table 6 ) Mediator Identity-based BE (B2) Peak score (M = 5.5); 59% officers ( Table 5 ) Dependent Variable Decision-making involvement (D2) Long-service veterans value governance voice Development-Focused Group Model Independent Variable Promotion (C24) Exceptional response (M = 8.87) ( Table 6 ) Mediator Developmental BE (B3) Highest score (M = 5.2); 51% NCOs transitioning ( Table 5 ) Dependent Variable Participation depth (D1 × D2) Career-focused veterans emphasize quality The modeling approach utilized raw scores for B1-B2-B3 and C21-C22-C23-C24, while employing the following coding scheme for D1 and D2 as specified in Table 9 . 10.1371/journal.pone.0351162.t009 Table 9 Coding scheme for D1, D2, and D1 × D2. D1 Original Options 0 1-3 4-6 >6 Assigned Value 0 2 5 7 Rationale True zero point Median value Interval mean Upper-bound adjustment D2 Original Options <20% 20-50% >50% Assigned Value 1.0 1.5 2.0 Rationale Executive participation Symbolic participation Substantive participation D1 × D2 Example: D1 = 2, D2 = 1.5,then D1 × D2 = 3 Following the specifications in Table 8 and Table 9 , mediation analyses were conducted for all three veteran groups using 500 bootstrap samples, with results presented in Table 10 . 10.1371/journal.pone.0351162.t010 Table 10 Mediation analysis results across veteran groups. Utilitarian Group (n = 269) Path β p-value 95% CI Significance Subsidies→Instrumental BE (A) 0.51 <0.001 [0.28, 0.74] Significant Instrumental BE→Frequency (B) 0.29 0.010 [0.09, 0.49] Significant Direct effect (c’) 0.38 <0.001 [0.22, 0.54] Significant Mediation effect (a × b) 0.15 0.008 [0.07, 0.23] Significant Total effect (c)=0.53 Proportion mediated: 28.3% Mission-Oriented Group (n = 125) Honors→Identity BE (A) 0.63 <0.001 [0.34, 0.92] Significant Identity BE→Decision-making (B) 0.41 0.002 [0.20, 0.62] Significant Direct effect (c’) 0.17 0.190 [-0.08, 0.42] n.s. Mediation effect (a × b) 0.26 0.002 [0.14, 0.38] Significant Total effect (c): 0.43 Proportion mediated: 60.5% Development-Focused Group (n = 131) Promotion→Developmental BE (A) 0.58 <0.001 [0.33, 0.83] Significant Developmental BE→Depth (B) 0.33 0.005 [0.13, 0.53] Significant Direct effect (c’) 0.09 0.130 [-0.07, 0.25] n.s. Mediation effect (a × b) 0.19 0.003 [0.09, 0.29] Significant Total effect (c): 0.28 Proportion mediated: 67.9% Table 10 demonstrates that the mediation effects are consistently established across all groups. For all three veteran types, both path A (PI → BE) and path B (BE → PB) reached statistical significance (p < 0.05), with all mediation effects confirmed by 95% confidence intervals (excluding zero).The effect size gradient of mediation follows: Utilitarian (0.15)