Do staff capacity and performance-based budgeting improve organisational performance? Empirical evidence from Chinese public universities

The study addresses how Chinese public universities can enhance effectiveness and efficiency amid declining funding and pandemic-induced pressures. It focuses on performance-based budgeting (PBB) as a budgeting approach that links resources to results and examines the roles of staff capacity and top management support (TMS). Motivated by gaps in prior literature—limited empirical work on PBB outside the U.S., underexplored staff capacity in higher education, and lack of evidence on TMS as a moderator—the study integrates contingency theory and resource dependency theory to investigate: (1) whether staff capacity affects PBB implementation, (2) whether staff capacity affects organisational (university) performance, (3) whether PBB mediates the staff capacity–performance link, and (4) whether TMS moderates these relationships. The research provides empirical evidence from Chinese public universities, which operate within centralized, bureaucratic structures and lifetime employment systems that may constrain innovation adoption.

Conceptual clarification: Staff capacity is defined as the sufficiency in number and qualifications of accounting staff, including necessary professional knowledge, skills, and training. PBB is defined as integrating performance measurement into budgeting to link funding with outcomes. TMS is the provision of resources, clear direction, commitment, and coordination by top leaders to facilitate project success and innovation. Organisational (university) performance includes financial and nonfinancial (teaching and research) dimensions. Theoretical background: Contingency theory posits that organisational practices should fit contextual factors, implying that budgeting systems like PBB must align with organisational characteristics and environments. Resource dependency theory emphasizes power relations and control of resources; in Chinese universities, top management controls resources, influencing budgeting systems and performance outcomes. Hypothesis development: Prior research suggests qualified staff are crucial for adopting innovations and PBB, though findings are sometimes contradictory. The study formulates five hypotheses: H1: Staff capacity positively relates to PBB; H2: Staff capacity positively relates to university performance; H3: PBB positively relates to university performance; H4: PBB mediates the staff capacity–performance relationship; H5: TMS moderates relationships among staff capacity, PBB, and performance, specified as H5a (TMS moderates staff capacity→PBB), H5b (TMS moderates staff capacity→performance), and H5c (TMS moderates PBB→performance).

Design: Quantitative survey using an online questionnaire targeting accounting staff in Chinese public universities’ finance departments and relevant directors (e.g., audit) who understand PBB. Ethics approval was obtained from the University Malaya Research Ethics Committee. Sampling and data collection: Purposive sampling; pilot tested with five academic scholars and five potential respondents; revised for clarity. The final questionnaire (5-point Likert scale) was administered via Wenjuanxing (wjx.cn) for four weeks in May 2021, supplemented by WeChat and QQ groups for university accounting staff. Participation was anonymous and voluntary. A total of 271 valid responses were obtained (no missing data), exceeding the G*Power a priori minimum (~159) for small effects (f2=0.05, alpha=0.05, power=0.80). Sample profile: 132 male and 139 female respondents; ages: 47.6% over 45, 27.3% 26–35, 25.1% 36–45. Education: 53.1% bachelor’s, 36.2% master’s, 9.2% doctoral. Majors: 66.1% accounting; remainder related fields. Measures: Staff capacity (5 items; Amirkhani et al., 2019), two facets (number of employees, professional knowledge); Cronbach’s alpha=0.911. PBB (10 items adapted from Pratolo et al., 2020); alpha=0.881. TMS (6 items adapted from Islam et al., 2009), used as moderator; alpha=0.935. University performance (adapted from Bobe & Kober, 2018); alpha=0.899. All items measured on 1–5 Likert scales. Common method bias: Harman’s single-factor test showed the largest single factor accounted for 38.93% (<50%). An unmeasured latent marker construct approach in PLS showed average substantive variance=0.709 vs method variance=0.060 (~12:1), indicating minimal CMB. Analytic strategy: Employed SPSS for descriptives; PLS-SEM (SmartPLS) for measurement and structural model evaluation; PROCESS macro (v3.5, model 59) for moderated mediation with 5000 bootstrap samples and Johnson–Neyman analysis; NCA in R (NCA package). Measurement model assessed via reliability (Cronbach’s alpha, composite reliability), convergent validity (loadings, AVE), and discriminant validity (HTMT). Structural evaluation included R², f², Q², PLSpredict for out-of-sample predictive power. Measurement model results: Cronbach’s alpha ranged 0.881–0.935; composite reliability 0.910–0.949; loadings 0.707–0.930; AVE 0.629–0.755. HTMT values and confidence interval upper bounds were <0.85, supporting discriminant validity. Structural model results: R² for PBB=0.599 (adj. 0.596); R² for university performance=0.536 (adj. 0.531). Effect sizes: staff capacity→PBB f²=0.156; staff capacity→performance f²=0.003; PBB→performance f²=0.406. Predictive relevance Q²: PBB=0.362; performance=0.376. PLSpredict showed medium out-of-sample predictive power (most indicators with lower RMSE/MAE than linear model). Hypothesis tests: Direct effects: staff capacity→PBB β=0.712, t=20.502, p<0.001 (H1 supported); PBB→performance β=0.694, t=10.943, p<0.001 (H3 supported); staff capacity→performance β=0.059, t=0.833, p=0.405 (H2 not supported). Mediation: indirect-only mediation with staff capacity→PBB→performance β=0.494, t=10.543, p<0.001 (H4 supported). Moderated mediation (PROCESS model 59): TMS×staff capacity→PBB β=0.159, t=5.532, p<0.001; TMS×staff capacity→performance β=0.112, t=2.903, p<0.01; TMS×PBB→performance β=-0.226, t=-3.956, p<0.001; PBB main effect β=1.425, t=6.998, p<0.001; TMS main effect β=0.752, t=5.121, p<0.001. Johnson–Neyman points for TMS at 1.291 and 2.983; moderation significant below 1.291 and above 2.983. Robustness checks: CTA-PLS supported reflective specifications. Ramsey RESET indicated no nonlinearity (p=0.108 for R² change). Quadratic effects were non-significant. NCA: Both staff capacity and PBB were necessary conditions for university performance with effect sizes d=0.140 (staff capacity) and d=0.324 (PBB), p<0.001. Bottleneck table indicated for 50% performance, minimum staff capacity 2.2% and PBB 29.8%; for 100% performance, staff capacity 25.5% and PBB 71.7%.

• Staff capacity strongly and positively predicts PBB implementation (β=0.712, p<0.001), but does not have a significant direct effect on university performance (β=0.059, p=0.405).
• PBB has a strong positive effect on university performance (β=0.694, p<0.001).
• Mediation: PBB mediates the relationship between staff capacity and performance with an indirect-only mediation (β=0.494, p<0.001).
• Moderated mediation: Top management support (TMS) significantly moderates key paths—enhancing staff capacity→PBB (β=0.159, p<0.001) and staff capacity→performance (β=0.112, p<0.01), but weakening PBB→performance (β=-0.226, p<0.001). Johnson–Neyman analysis identified regions of significance for TMS below 1.291 and above 2.983.
• PLS-SEM model fit and prediction: R²=0.599 for PBB and 0.536 for performance; f² shows meaningful effects (PBB→performance f²=0.406; staff capacity→PBB f²=0.156); Q² indicates medium predictive relevance; PLSpredict shows medium out-of-sample predictive power.
• NCA: Both staff capacity (d=0.140) and PBB (d=0.324) are significant necessary conditions for performance. Bottleneck thresholds: to reach 50% performance, staff capacity ≥2.2% and PBB ≥29.8%; to reach 100% performance, staff capacity ≥25.5% and PBB ≥71.7%.
• Measurement validity and CMB checks support reliability and validity of constructs; CMB unlikely (largest factor 38.93%; ULMC substantive variance ~0.709 vs method 0.060).

The findings show that enhancing staff capacity is crucial for adopting PBB, which in turn improves university performance. However, staff capacity alone does not directly translate into performance gains; its impact operates primarily through PBB as a management control mechanism, aligning with contingency theory’s emphasis on fit and RDT’s focus on resource control. The moderated mediation results indicate that top management support shapes the strength and direction of these relationships: with strong support, the translation of staff capacity into PBB and performance is amplified, while the marginal benefit of PBB on performance may diminish at very high levels of support, reflecting power and resource allocation dynamics in highly centralized Chinese universities. NCA complements these insights by identifying minimum thresholds of staff capacity and PBB necessary for achieving targeted performance levels, offering actionable guidance for resource allocation and capacity building. Together, these results address the research questions, demonstrate the central role of PBB as a conduit between human capital and outcomes, and underscore leadership’s pivotal role in enabling reforms.

The study contributes by integrating PLS-SEM and NCA to examine how staff capacity and PBB influence performance in Chinese public universities under the moderating influence of top management support. Empirically, staff capacity drives PBB adoption, PBB improves performance, and PBB mediates the staff capacity–performance link; TMS moderates these pathways. Theoretically, the work extends contingency theory and resource dependency theory to a higher education budgeting context and introduces conditional process analysis with Johnson–Neyman probing to delineate moderation ranges. Practically, the results guide universities to invest in staff capability and PBB implementation, and to secure active top management backing to realize performance gains. Future research should test the model across different national and institutional contexts, explore additional contextual moderators, and employ longitudinal or mixed-method designs to strengthen causal inference.

• Single-country context (China) may limit generalisability due to unique governance and bureaucratic structures; cross-cultural validation is needed.
• Cross-sectional survey limits causal inference; longitudinal designs could better capture temporal dynamics.
• Focus on accounting and finance-related staff may not reflect perspectives across all university units.
• Self-reported measures may be susceptible to bias despite applied remedies.