Principal investigator: Omar Barroso Khodr
Authors: Bel, Espaillat and Esteve (2025)
Original title: The performance of performance-based contracting in public outsourcing: a meta-regression analysis
Location of the Intervention: United States and United Kingdom.
Sample Size: 740 observations
Primary Variable of Interest:
Type of Intervention: Optimization in public policies
Methodology: meta-analysis and meta-regression.
Summary
Performance-Based Contracting (PBC) is presented as a model capable of improving results, raising quality, reducing costs, and increasing accountability. It has become a standard in government contracting worldwide and an essential component of the latest pay -for-success models . However, academic evaluations on the subject are still scattered and often lack a solid theoretical foundation.
This study conducts a meta-analysis of 740 observations drawn from 38 studies, covering ten service areas, with the aim of assessing the actual performance of public procurement. Based on principal-agent theory, incomplete contracts theory, and goal setting, we examine the factors that influence the performance of this type of contracting. By integrating these perspectives into a single empirical framework, the article offers a systematic test of fundamental assumptions related to incentive alignment, contractual incompleteness, and goal setting in public procurement.
The results indicate that contracts focused on final outcomes are more successful than those focused on intermediate processes or products ( outputs ). This finding offers quantitative evidence that corroborates the thesis of incomplete contracts theory, according to which performance depends on the contractability of results. Furthermore, it expands the principal-agent logic by demonstrating when and why incentives fail in multidimensional public contexts.
Although context is relevant, factors such as residual control, collaboration, and shared goals prove central to the success of PBC. Thus, the study contributes theoretically by connecting economic and behavioral theories of contracting and by empirically grounding its predictions with evidence from the public sector.
- Policy Problem
According to the authors, Performance-Based Contracting (PBC) proliferated in the public sector as a response to fiscal constraints and the demand for proof of value in the use of public resources, but lacked a holistic view of its determinants of success. This study fills this gap through meta-regression, situated at the intersection of principal-agent, incomplete contract, and goal-setting theories, and advances the theoretical debate by examining incentive alignment, empirically assessing contractability, and incorporating a behavioral perspective into contract theory.
The results indicate that the actual effect of PBC on performance is positive, but small and not statistically significant, which is common in policy interventions subject to external factors and contextual differences. Analyses show that metrics focused on process and product perform worse than those based on outcomes and impact, while human services offer weaker support for the hypothesis of lower effectiveness, suggesting that governments and providers are overcoming measurement and monitoring challenges to achieve more significant results. Other characteristics of PBC did not show statistical significance.
In this way, the authors emphasize that there was no previous systematic empirical effort to synthesize evidence on the success of public-private partnerships (PPPs) in the public sector, as existing reviews are sector-agnostic, rooted in a business perspective, and based mostly on conceptual and qualitative studies, with the exception of the health sector, whose meta-analyses do not distinguish between public and private schemes. Public procurement differs fundamentally from private procurement in market structure, complexity, composition (with common triadic relationships), and subjection to political demands and scrutiny, which justifies a specific analysis that considers its institutional particularities and additional challenges of accountability and transparency.
- Policy Implementation Context
According to the authors, PBC (Pay-as-you-go) is based on aligning incentives with measurable results, assuming that this alignment improves performance. Unlike traditional contracting, which focuses on inputs and activities, PBC emphasizes products and results, granting providers greater flexibility to innovate and linking incentives to public policy objectives. By linking payment to performance, it strengthens accountability, stimulates efficiency, and seeks a better cost-benefit ratio.
In this context, PBC is guided by three theoretical pillars. Through principal-agent theory, it addresses informational asymmetry and moral hazard by linking financial rewards to the achievement of clear goals, aligning the efforts of service providers with the objectives of the client.
The authors emphasize that, according to the theory of incomplete contracts, contracts cannot anticipate all contingencies, making residual control rights, which remain with the asset owner, essential. This implies that private ownership is associated with stronger cost-reduction behaviors than quality improvement behaviors, a challenge to which PBC responds by balancing the trade-off between cost and quality. According to goal-setting theory, clear, challenging objectives reinforced by feedback improve performance.
Thus, the implementation of PBC faces challenges in defining and measuring performance. Process and product metrics are easier to measure, but focus on short-term activities and neglect transformational results, creating a "performance paradox." Results-focused measures allow for innovation, but are difficult to define, monitor, and control, with the risk of unfair penalties due to factors beyond the influence of service providers.
However, human services present unique challenges. They involve dynamic and multifaceted needs that are difficult to capture with conventional measures. The literature argues that standardized systems often rely on simplistic metrics that fail to consider organizational complexity, professional judgment, equity, and intergovernmental fragmentation. Intangible outcomes, long-term impacts, and context dependence compromise the validity of PBC in these sectors.
Thus, market competition acts as an implicit incentive. In more competitive environments, managers respond better to incentives, and competition reduces information asymmetry. Contract renewal decisions are shaped by past performance. However, ideal levels of competition are rarely observed in public markets. In essential sectors, competition can destabilize services, while in markets with high asset specificity, reduced competition can improve performance through trust and long-term engagement.
Taken together, principal-agent and incomplete contract theories explain that PBC promotes performance when results are observable and verifiable, but fails when quality is multidimensional, ambiguous, or politically contested, risking encouraging a focus only on the measurable at the expense of the relevant.
In summary, the effectiveness of PBC is contingent on the type of measure adopted (process/product versus outcome/impact), the sector of activity (human services versus technical services), the level of competition, and the ability to balance extrinsic and intrinsic incentives, especially in mission-oriented sectors, where excessive financial incentives can undermine the intrinsic motivation of professionals.
- Evaluation Details
According to the authors, the construction of the meta-sample began with a review of leading journals in the areas of public administration, economics, transportation, and social services. Keywords were extracted from these studies and applied to multiple databases, including Web of Science, Scopus, EBSCO, JSTOR, PROQUEST, ETHOS (for doctoral theses), and Google Scholar.
In this way, studies published up to 2024 were included, with final verification in January 2025. The search also included references from selected articles, relevant reviews on PBC, and websites of government institutions and think tanks , such as the Urban Institute, Harvard Government Performance Lab, Government Outcomes Lab, the UK government, and the US OBM. The process followed PRISMA guidelines and best practices recommended for meta-analyses.
Next, four criteria were established for inclusion in the sample: (1) empirical nature, with multivariate regression methods; (2) dependent variable aligned with explicitly incentivized performance measures; (3) presence of a counterfactual, evaluating the effect of PBC compared to an alternative model without performance incentives; and (4) public context, with PBC predominantly funded with public resources and contractually managed by public entities.
According to the authors, Social Impact Bonds (SIBs) were excluded because they involve multiple partners and innovative financial mechanisms that confound the effects of PBC, in addition to the scarcity of peer-reviewed empirical studies. Public-public relations were also discarded, as they represent intergovernmental cooperation, not outsourcing contracts with private suppliers.
The authors emphasize that individual initiatives, with direct incentives to employees or clients, were excluded because they operate under motivational dynamics distinct from organizational ones. The education and hospital health sectors were removed: the former because it adopts performance-based financing models focused on individuals or public-to-public transfers; the latter because a specific meta-analysis already exists in the literature. Studies with distinct empirical relationships, such as gaming , selection bias, or equity, were also excluded because they analyze unintended consequences, not the direct effect of performance-based compensation on contracted performance.
Finally, the final sample comprised 38 studies, of which seven were not peer-reviewed, totaling 740 observations for the overall effects analysis. One observation was excluded from the meta-regression due to lack of information. The researchers contacted authors of six studies to obtain supplementary data or unreported information. Sensitivity analyses and additional regression models were performed for all studies.
- Method
The meta-analysis integrated 38 empirical studies, totaling 740 observations. Rigorous exclusion criteria were applied: Social Impact Bonds (SIBs), public-public relations, individual CBP initiatives, and the education and hospital health sectors were discarded – the latter because it already has a specific meta-analysis. However, services from long-term care facilities for the elderly were retained, as they are part of social welfare and had not been previously reviewed.
The final sample included seven non-peer-reviewed studies. Additional sensitivity analyses and regressions were performed, and authors of six studies were contacted for supplementary data. One observation was excluded from the meta-regression due to lack of information.
The effect was measured by the partial correlation coefficient, calculated from the t-statistics of the original studies, with conversion of odds ratios to Pearson correlations. All effects were transformed into Fisher's Z-scores to correct for biases. In one study, the lower bound method was used to estimate the effect.
Moderators for quality control and design were included: journal impact factor, sample size, presence of endogeneity strategies (such as difference-in-differences, matching, and fixed effects), and type of statistical model (continuous vs. dichotomous). Country variables were excluded due to multicollinearity.
The meta-regression model regressed the Z-score against type of measure, nature of service, competition, delivery model, impact factor, sample size, endogeneity control, and statistical model.
To assess publication bias, grey literature was included. The funnel plot showed symmetry, and Egger's (p = 0,132) and Begg's (p = 0,153) tests confirmed the absence of bias, including in the subsample of peer-reviewed studies.
- Main results
According to the results, the meta-analysis revealed that Performance-Based Hiring (PBC), considered in isolation, does not have a significant effect on performance (coefficient = 0,008; p = 0,303). This result indicates that the mere adoption of the model does not guarantee better results, and that an analysis of contextual and methodological moderators is necessary to understand its conditions of effectiveness.
According to the authors, when these factors are controlled, the average effect becomes positive and statistically significant (coefficient = 0,119; p < 0,01), showing that the success of PBC is contingent on specific variables. Among the moderators, the type of metric used stands out: process-oriented contracts show a negative and significant association with performance when compared to metrics based on final results ( outcomes ) (coefficient = -0,038; p < 0,05). This finding corroborates the thesis that a focus on results allows for greater innovation and autonomy for service providers.
In this context, regarding human services, a negative effect with marginal significance was observed in the GEE model (coefficient = -0,098; p < 0,10), suggesting specific difficulties in measuring and attributing performance in this sector. However, the statistical fragility recommends caution in interpretation.
The other variables analyzed — market competition, organizational structure (public, private, or mixed), use of endogeneity strategies, journal impact factor, sample size, and type of statistical model (continuous vs. dichotomous) — did not produce significant effects on the performance of the PBC.
In summary, the results indicate that the effectiveness of PBC is not automatic, but fundamentally depends on the contractual design, especially the choice of metrics oriented towards end results, while factors such as the methodological quality of the studies and organizational characteristics did not prove to be decisive.
- Lessons in Public Policy
According to the authors, the meta-analysis demonstrates that Performance-Based Hiring (PBC), in isolation, does not produce a significant effect on performance (coefficient = 0,008; p = 0,303). Governments should not adopt it as a generic reform without evaluating the contextual conditions.
Thus, the success of PBC depends on the design of the metrics: contracts oriented towards final results ( outcomes ) outperform those based on processes or intermediate products ( outputs ) (coefficient = -0,038; p < 0,05). It is recommended to prioritize indicators of effective social value, invest in results monitoring, and involve providers and beneficiaries in defining the metrics.
In this context, human services (social assistance, childcare, rehabilitation) present greater implementation difficulties (coefficient = -0,098; p < 0,10), requiring hybrid models that combine financial incentives with relational mechanisms and specific monitoring capabilities.
It is worth noting that market competition (p = 0,169) and the provider's organizational structure (p = 0,531) do not significantly influence performance. Factors such as incentive design, control over performance variables, contractual flexibility, and mutual trust are more relevant. In this sense, incomplete contracts do not harm performance when there are clear and shared goals, and can foster innovation and adaptation. It is recommended to balance specification with flexibility and invest in collaborative governance.
Therefore, PBC can lead to manipulation of metrics ( gaming ) and selection of beneficiaries ( cream-skimming ) when incentives are strong and monitoring is weak. Robust verification systems, incentives that discourage opportunism, and clear penalty mechanisms are essential. Studies controlling for endogeneity suggest that less rigorous research overestimates the impact of PBC. Robust counterfactual assessments and independent evaluations are needed before expanding programs.
Ultimately, PBC is not inherently superior to other forms of contracting. Its success depends on incentive design, choice of metrics, institutional capacity, service complexity, and relational governance. Evidence-based policies require careful design, continuous monitoring, rigorous evaluation, and adaptability. PBC is a valuable tool, but not a panacea.
