📈 economics

The Optimal Intensity of Performance Evaluation

This paper transforms Goodhart's and Campbell's Laws into a parameterized threshold model to demonstrate that university performance evaluation follows an inverted-U relationship with organizational effectiveness, identifying a critical tipping point where excessive intensity triggers dysfunction and proposing three qualitative levers to mitigate these effects across diverse institutional contexts.

Original authors: Zhongjun Wang

Published 2026-09-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zhongjun Wang

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the modern university, a quiet tension has grown between the desire to measure success and the reality of what actually creates it. For decades, administrators have relied on numbers—counts of published papers, citation tallies, and grant totals—to judge the quality of research and teaching. This approach rests on a simple premise: if you can measure it, you can manage it. However, a long-standing observation in economics and social science warns that this logic has a breaking point. When a measurement becomes the primary target, people begin to optimize for the number itself rather than the underlying value it was meant to represent. This phenomenon, often called Goodhart's Law, suggests that a metric stops being a good measure once it becomes a goal. In the academic world, this has led to concerns that faculty members are spending less time on deep, difficult scholarship and more time on activities that simply boost their scores, a process known as goal displacement. The question facing universities today is not whether to use metrics, but how much is too much before the system begins to work against itself.

A new study by Zhongjun Wang transforms this abstract warning into a concrete, measurable framework. Rather than treating the failure of performance evaluation as a vague possibility, the research builds a mathematical model to pinpoint exactly where the balance tips. The core finding is that the relationship between how intensely an institution evaluates its staff and the actual quality of its output follows a specific shape: an inverted U. Initially, adding more rigorous evaluation helps. It clarifies expectations, aligns efforts, and improves coordination. But this benefit does not last forever. The model identifies a critical threshold, a specific point of intensity beyond which the system collapses into dysfunction. Once universities cross this line, the costs of strategic publishing or ignoring teaching grow so rapidly that they overwhelm any remaining benefits, leaving the institution worse off than if it had used no metrics at all.

The research does not stop at identifying the danger zone; it offers a way to navigate it by breaking down the solution into three distinct tools. The study argues that simply swapping "quantitative" numbers for "qualitative" opinions is too blunt an instrument. Instead, it separates the role of human judgment into three specific levers: peer review, value alignment, and procedural fairness. Peer review, where experts judge the work directly, acts as a shield that pushes the dangerous threshold further out, allowing institutions to use more metrics safely. Aligning metrics with the university's specific mission—whether that is teaching, research, or service—raises the ceiling of what can be achieved. Finally, ensuring that the evaluation process feels fair to the faculty reduces the baseline resentment and resistance that drain energy from the system. These three elements are not interchangeable; each solves a different part of the problem, and using the wrong one will not prevent the collapse.

To understand how this breakdown happens over time, the author employs a dynamic model that tracks how faculty behavior shifts. The simulation shows that once the intensity of evaluation crosses the critical threshold, a subtle but powerful change occurs in the academic culture. Even as the number of publications continues to rise, the actual quality of the work begins to drift downward. This happens because the pressure to perform on the metrics becomes so overwhelming that it changes the social norms of the workplace. What was once considered a minor shortcut becomes a standard practice, and the collective effort of the university shifts away from creating genuine value toward optimizing the score. The model demonstrates that this divergence is not immediate; it accelerates as more people join the game, creating a feedback loop where the system rewards the wrong behaviors more and more heavily.

The study also clarifies why some universities seem to survive high-pressure environments while others crumble. The threshold is not a fixed number for everyone; it depends on the institution's design. Universities with strong peer-review systems and fair procedures can sustain a higher intensity of measurement before hitting the wall. In contrast, systems that rely heavily on raw numbers without these safeguards reach the point of dysfunction much sooner. The research suggests that the current trend of increasing metrics, driven by global rankings and competition, is pushing many institutions dangerously close to or past this limit. The model implies that simply reducing the number of metrics is not enough; the solution requires building the capacity for high-quality human judgment to buffer the system.

Ultimately, this work provides a diagnostic tool for university leaders and policymakers. It moves the debate away from the binary choice of "metrics versus no metrics" and toward a more nuanced question of design. The study suggests that there is an optimal intensity for evaluation, a sweet spot where accountability improves performance without triggering the counterproductive behaviors that Goodhart's Law warns against. By treating the threshold as a variable that can be managed through better peer review and fairer processes, the research offers a path forward. It indicates that the goal is not to eliminate measurement, but to ensure that the intensity of that measurement remains below the point where it begins to destroy the very value it seeks to measure. The findings are presented as a set of testable propositions, inviting further data collection to refine the exact location of these thresholds for different types of institutions, but the central message is clear: there is a limit to how much we can measure before we start measuring the wrong things.

Technical Summary: The Optimal Intensity of Performance Evaluation

Problem Statement
Performance evaluation has become a central governance instrument in higher education, yet the field lacks a formal understanding of the point at which evaluation intensity ceases to be beneficial and begins to undermine organizational goals. While Goodhart's Law (1975) and Campbell's Law (1979) are widely cited as rhetorical warnings that "a measure becomes a target and ceases to be a good measure," they remain interpretive metaphors rather than quantifiable models. Existing literature documents metric overload and goal displacement qualitatively but fails to provide a formal model of the non-monotonic relationship between evaluation intensity and outcomes, particularly one that incorporates institutional thresholds and endogenous mitigation mechanisms. Consequently, policymakers cannot determine when evaluation helps or harms, nor can they distinguish between different types of qualitative review.

Methodology
The paper develops a formal threshold model integrating organizational economics, motivation psychology, and higher-education policy analysis. The methodology proceeds through three distinct formulations:

Static Univariate Model: The model defines organizational effectiveness $U(k)$ as the net result of an alignment benefit $U_1(k)$ and a dysfunction cost $U_2(k)$ , where $k$ represents evaluation intensity.
- Alignment Benefit: Modeled as a concave exponential function $A(1 - e^{-\alpha k})$ , reflecting diminishing marginal returns to coordination and feedback.
- Dysfunction Cost: Modeled as a piecewise function. Below a critical threshold $k_0$ , costs grow linearly. Above $k_0$ , costs accelerate exponentially due to two reinforcing feedback mechanisms: a social-norm effect (where gaming becomes normalized) and an organizational-amplification effect (where management intensifies metrics in response to short-term gains).
- Result: This structure yields an inverted-U relationship between intensity and effectiveness, with a "break-even point" where excessive evaluation renders the organization worse off than having no evaluation at all.
Bivariate Extension (Qualitative Judgment): The model introduces a weight $w$ for qualitative judgment, decomposing it into three non-interchangeable levers:
- Peer Review ( $w_p$ ): Primarily raises the threshold $k_0$ by reducing information asymmetry.
- Value Orientation ( $w_v$ ): Primarily raises the benefit ceiling $A$ by aligning metrics with institutional missions.
- Procedural Justice ( $w_j$ ): Primarily reduces the baseline dysfunction cost $B$ by mitigating resentment.
- This formulation replaces the undifferentiated "quantitative vs. qualitative" debate with a parameterized design space where specific levers shift specific parameters.
Dynamic Drift–Diffusion Layer: To capture the temporal evolution of behavior, the paper employs a Fokker–Planck equation.
- Mechanism: Faculty effort is modeled as a density distribution $e(x,t)$ across a quality spectrum $x \in [0,1]$ , where $x=0$ is pure metric optimization and $x=1$ is true value production.
- Drift: Above the threshold $k_0$ , a drift velocity term activates, pushing the distribution of effort toward low-quality metric optimization.
- Outcome: This layer demonstrates that even as metric scores ( $M(t)$ ) rise, true value ( $V(t)$ ) declines monotonically, providing a dynamic micro-foundation for Goodhart's Law.

Key Results

Inverted-U Relationship: The model confirms that organizational effectiveness follows an inverted-U curve relative to evaluation intensity. There exists an optimal intensity $k^{**}$ and a break-even point $k_{be}$ beyond which net utility becomes negative.
Threshold Existence: A critical institutional threshold $k_0$ exists. Below this point, dysfunction grows linearly; above it, dysfunction accelerates exponentially due to social-norm tipping points.
Differential Levers: Qualitative judgment is not a monolithic solution. Peer review is the primary mechanism for raising the sustainable threshold; value orientation lifts the maximum potential benefit; and procedural justice lowers the baseline cost of administration.
Dynamic Divergence: In regimes where $k > k_0$ , the model predicts a progressive divergence where metric scores increase while true organizational value decreases, driven by a shift in faculty effort distribution toward metric gaming.

Six Falsifiable Propositions
The paper distills the model into six testable propositions with specific identification strategies for empirical research:

Non-monotonicity: Evaluation intensity and effectiveness follow an inverted-U shape (testable via institutional panel data).
Threshold Effect: Dysfunctions emerge rapidly beyond a critical threshold (testable via difference-in-differences using reforms like China's "Break Five Onlys").
Moderation: Qualitative judgment raises the threshold $k_0$ (testable via mediation models).
Differential Effects: Peer review raises $k_0$ more strongly than procedural justice (testable via interaction terms in hierarchical regression).
Dynamic Adaptation: The divergence between metric scores and true value accelerates over time when $k > k_0$ (testable via event studies).
Heterogeneity: The threshold $k_0$ varies by career stage (lower for early-career researchers) and discipline (higher in quantifiable STEM fields).

Significance and Claims
The paper claims to make three primary contributions to higher-education scholarship:

Formalization of Metaphor: It converts Goodhart's and Campbell's Laws from interpretive metaphors into a parameterized, falsifiable threshold model. Unlike previous economic models (e.g., Rablen, 2010) that rely on individual reference points, this model identifies an institutional threshold where dysfunction accelerates.
Decomposition of Qualitative Review: It moves beyond the binary "quantitative vs. qualitative" debate by identifying three distinct, non-interchangeable levers (peer review, value orientation, procedural justice) that operate through specific mechanisms. This allows for targeted policy design rather than generic calls for "more qualitative assessment."
Dynamic Mechanism: By integrating a Fokker–Planck drift-diffusion layer, the model explains how and why metric scores and true value diverge over time, offering a behavioral micro-foundation for the exponential cost of dysfunction.

The paper positions itself as a diagnostic tool rather than a prescriptive verdict. It argues that the "optimal" intensity is not a universal constant but a function of institutional design parameters. The model provides a framework to diagnose whether a specific evaluation system (e.g., China's "Break Five Onlys," the UK REF, or US performance-based funding) is operating within a sustainable regime or has crossed a threshold into dysfunction. The author emphasizes that the quantitative values (e.g., $k_0 = 0.5$ ) are illustrative and require empirical calibration, but the qualitative predictions regarding the existence of a threshold and the direction of drift are robust across parameter variations.

Technical Summary: The Optimal Intensity of Performance Evaluation

More like this