Goodhart’s Law

Narrow Street
No Space for 2 Targets

A couple of weeks back, I stumbled upon the concept of Goodhart’s Law and I can’t help wondering if the same is true of corporate performance indicators. Perhaps the case is weaker for corporate performance indicators but the idea may still hold some truth.

The Law based on Goodhart’s formulation in 1975 is “any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes

It is initially more or less based on the conduct of monetary policy and has much to do with statistics. But in the corporate setting, tying CEO’s financial rewards to share prices has somewhat the same sort of effect. Without the coupling of the 2 variables – share prices and CEO financial compensation, the share prices would ordinarily reflect the performance of the company, which is a proxy for the outcome of the management of the firm under the CEO (although people might argue that it is inaccurate, but in business, outcome is still the most important). When they are linked, CEOs might become obsessed with raising share prices of the firm and neglecting the core management of the firm.

The same applies for lower level sort of work. For example, if the waiting time at the government clinic is used as a measure for performance then doctors and nurses might quickly try to go through the patients and speed up consultation to hit their performance target rather than provide quality care and service. Likewise, if too much emphasis is on delivering good food at a local restaurant, service might be compromised, which explains why the boss of the pizzeria down the street has real bad attitude. Perhaps this is just part of human nature, the narrow focus of our minds.


  1. To add to the discussion: what are the conditions under which the law applies, and how can we use its implications for meaningful policy-making?

    In Information economics, we study how information affects decision-making at different hierarchies, and how information asymmetry (the idea that agents from different parties hold different sets of information) creates the need for information signalling (from the informed to the uninformed) or information screening (by the uninformed on the informed).

    Goodhart’s Law arises from the disparity in information between parties – and its solution lies in aligning the basket of signals/indicators used for screening with the correct motivation for the agents being assessed. Your examples used only singular variables for complex analysis, and hence the indicators are likely to fail on their own.

    Suppose (in your clinic example), there is a feedback process (surveillance of patient’s experience through feedback forms, reputation systems – when patients are regular and recurrent, and can choose the doctor who is tending to them), your doctors might consider it more economic to develop long-term relationships with their patients (eg. family doctors?). Will your “waiting time” indicator lose it’s value? Not necessarily, since it still serves it’s purpose of patient turnover rate in the overall function of how much service the doctor can provide to how many people.

    In monetary policy, the system (economy) you are trying to evaluate is far more complex than that in a corporate organisation – but I agree with you – Goodhart’s Law should still apply to corporate performance indicators if they (the indicators used) do not form a robust mapping (place a disproportionate amount of stress on some areas than others) of the information being screened (employee’s performance).

  2. Yup, using solely any particular indicator as a performance target would eventually put too much pressure on it such that the relationship breaks down as predicted by Goodhart’s Law, which is why corporations will need to evaluate performance with more complex systems that match the realities of the job the people are performing.

    Of course, measuring the performance of an economy is another sort of reality…

Leave a Comment

Your email address will not be published. Required fields are marked *