The Target That Moved Itself - How Goodhart's Law Rewrites Financial Metrics

Every number used to govern money eventually risks drifting away from what it was designed to measure. Not because anyone breaks the rules, but because once a rule has a number attached to it, people naturally learn to satisfy the number rather than the intention behind it.

This isn’t necessarily cynicism or misconduct. It is closer to a structural feature of incentives. Attach consequences to a measurement, and the measurement starts adapting to the pressure around it - often quietly, legally, and with everyone acting rationally. Economists call this Goodhart’s Law. And few areas show it more clearly - or more expensively - than corporate debt covenants and executive compensation KPIs.

This piece starts from first principles: what Goodhart’s Law actually means, how it appears in debt and compensation contracts, why even intelligent and well-intentioned people can fall into the trap, and what can be done to keep a metric meaningful when everyone has an incentive to improve it.

Part 1: A Rule Too Simple to Ignore

In 1975, British economist Charles Goodhart was examining why the Bank of England’s monetary targets kept losing their effectiveness. Whenever policymakers selected a measure of money supply and began managing the economy toward it, the relationship between that measure and the broader economic objective started to weaken.

The observation, later popularized by anthropologist Marilyn Strathern, is usually summarized as:

“When a measure becomes a target, it ceases to be a good measure.”

The mechanism is straightforward. A metric is initially chosen because it correlates with something we actually care about - a healthy company, a well-managed fund, or a creditworthy borrower. That relationship generally holds as long as nobody is specifically optimizing for the metric.

Once real consequences - a bonus, covenant breach, or credit downgrade - are tied to that number, the relationship can begin to break. There are now two ways to improve the score: improve the underlying business, or improve the number. The second is often faster and cheaper, so the incentive to take that route grows.

Finance is an especially good laboratory for this because so much of it is built around contracts tied to specific numbers. Two of the clearest examples are debt covenants and executive compensation. To understand how Goodhart’s Law operates in practice, it helps to build each one up from first principles.

Part 2: Covenant EBITDA - When the Target Redefines the Company

Start with the basics. When a company borrows money, the lender doesn’t simply hand over cash and hope for the best. The loan agreement includes covenants - financial tests the borrower must continue to meet throughout the life of the loan.

A common example is a leverage covenant, typically measured as Debt/EBITDA. A company might agree to keep Debt/EBITDA below 4.0x. Breach that limit, and the lender may be able to declare a default, demand repayment, or renegotiate the loan.

EBITDA was originally a useful shorthand for operating performance, stripping out financing structure, taxes, depreciation, amortization, and other non-cash items to make companies easier to compare. But EBITDA is a constructed measure, not a single fixed number. Loan agreements often give borrowers flexibility over what can be included or excluded for covenant purposes.

Line Item	Amount ($mm)
Reported EBITDA	58.0
+ Restructuring & severance	4.5
+ Pro-forma run-rate synergies	6.0
+ Non-cash stock compensation	3.0
+ Transaction & advisory fees	2.5
Covenant (“adjusted”) EBITDA	74.0

That is where “covenant EBITDA” can begin to diverge from the EBITDA an independent analyst would calculate.

The add-backs total $16.0mm - a 28% uplift over reported EBITDA. Against $280mm of net debt, that swings leverage from 4.83x on a reported basis to 3.78x on a covenant basis: the difference between a business that is close to breach and one comfortably inside its terms.

Leverage(reported) = 280 / 58.0 = 4.83x vs Leverage(covenant) = 280 / 74.0 = 3.78x

None of this is necessarily fraudulent. Each add-back can usually be justified on its own, and lawyers on both sides negotiate the definitions carefully before the loan closes. But when enough individually defensible add-backs are stacked together, the covenant ratio can look comfortably healthy even as the company’s underlying cash generation deteriorates.

The result is that the lender’s early-warning system - the very reason the covenant exists - can quietly stop working just when the warning matters most.

Part 3: The Denominator Game - EPS and ROE

The same pattern appears in executive compensation, often in an even more mechanical way. Two common metrics used in executive bonuses and long-term incentive plans are earnings per share (EPS) growth and return on equity (ROE). Both are ratios, which means they can improve in two ways: increase the numerator or reduce the denominator.

1. EPS accretion from buybacks alone

EPS = Net Income / Shares Outstanding

Suppose net income is $120mm and the company has 100mm shares. EPS is $1.20. If it buys back 10mm shares using existing cash or new debt, EPS rises to $1.33 ($120mm / 90mm) - an 11% increase with no improvement in operating performance.

If the buyback is debt-funded, higher interest expense will reduce net income. But if borrowing costs are low enough, the benefit from the lower share count can still outweigh the added interest, creating EPS accretion even when enterprise value is unchanged or has deteriorated.

2. ROE decomposition (DuPont)

ROE = (Net Income / Revenue) × (Revenue / Assets) × (Assets / Equity)

These three components are net margin, asset turnover, and leverage. Management can improve ROE through genuine operational improvements in the first two, or simply increase leverage and reduce the equity base.

For example, increasing the equity multiplier from 8x to 10x raises ROE by 25%, even if profitability and asset efficiency remain unchanged. That is why credit analysts use DuPont analysis before treating higher ROE as evidence of stronger business quality - the same ROE can reflect either a more profitable business or a more leveraged and fragile one.

In both cases, the metric rises and the bonus may vest, but the underlying question - has the business actually become more valuable or resilient? - remains unanswered. The headline number can improve while the underlying business moves in the opposite direction.

Part 4: Why This Happens Without Misconduct

It’s tempting to see this as a story about managers deliberately gaming the system. A more useful explanation is psychological - and it doesn’t require anyone to be dishonest.

Once a target becomes important and tied to rewards, attention naturally shifts toward whatever moves the measured number. Steven Kerr described this decades ago as “the folly of rewarding A while hoping for B”: organizations often reward one behavior while actually wanting another, then act surprised when people optimize for what is on the scoreboard.

Negotiated definitions make this effect even stronger. When people debate what counts as a one-time cost or a reasonable leverage add-back, it is natural for each side to argue for the interpretation that benefits them. No one needs to be acting in bad faith.

The result is gradual, collectively reasonable drift: each adjustment may be defensible on its own, yet over time, the final number can move far from what it was originally meant to represent.

Part 5: Keeping the Measure Honest

This is where the focus should shift from diagnosis to design. Metrics can never be made completely ungameable, but they can be structured so that gaming them requires either genuine operational change or enough effort to become visible. Five mechanisms do most of the work.

1. Cap and Sunset the Add-Back Schedule

Instead of allowing an open-ended list of adjustments, credit agreements can cap total add-backs at a fixed percentage of reported EBITDA - often 15–20% - and require a “non-recurring” item to lose that treatment if it appears again within the following four quarters.

This turns a subjective question - “Is this really one-time?”- into a simple, auditable test: did it occur in the previous four quarters? No fresh negotiation is needed at every measurement date.

2. Force Multi-Metric Consistency

A single ratio can be gamed in isolation. Two or three related metrics that should move together are much harder to manipulate without genuine improvement.

For example, pair EPS growth with free cash flow per share, or ROIC with WACC. Pair ROE with a minimum equity or capital-adequacy floor. If EPS rises mainly because of buybacks while net income and FCF remain flat, the divergence becomes the warning signal.

This is similar to how auditors compare covenant EBITDA with statutory EBITDA over time: the gap between related measures can reveal more than either number alone.

3. Replace Cliff-Edge Tests with Trends and Relative Benchmarks

A binary threshold - breach 4.0x and trigger a default, or miss the EPS target and lose the bonus - creates a strong incentive to manage the number around the measurement date.

Two alternatives help. First, use trailing-twelve-month or multi-quarter averages instead of a single-period snapshot, reducing the impact of one-off actions. Second, benchmark performance against peers or the wider sector, such as relative TSR or ROIC, rather than relying only on absolute targets. This makes it harder to manage the metric without considering the broader market.

4. Track the Gap, Not Just the Number

Lenders and auditors can independently calculate a metric using a stricter definition - for example, statutory EBITDA instead of covenant EBITDA, or unlevered ROE instead of reported ROE - and track the gap between the two over time.

A small, stable gap is usually unremarkable. But a gap that keeps widening quarter after quarter can signal weakening underlying performance, often before the headline ratio breaches the covenant. That is precisely the early-warning function the covenant was supposed to provide.

5. Use Equity Cures and Clawbacks as Pressure Valves, Not Loopholes

Well-designed agreements can combine strict definitions with controlled ways to deal with breaches. Equity cure rights can allow sponsors to inject capital to fix a covenant breach, but only a limited number of times over the life of the loan. Clawback provisions can allow boards to recover bonuses if the underlying results are later restated.

Both reduce the incentive to manipulate the metric. A genuine breach can be fixed properly rather than hidden, while an inflated bonus can be recovered later. Gaming the number becomes less attractive because it is no longer a one-way win.

Part 6: Closing

Goodhart’s Law isn’t a flaw in financial contracts. It is an unavoidable consequence of reducing something as complex as “Is this business healthy?” to one negotiable number. That number will naturally be influenced by whoever has the most at stake in how it looks, and that influence often won’t appear as misconduct while it is happening.

The answer isn’t to find a metric that can never be gamed - no such metric exists. It is to design contracts with that reality in mind: limit adjustments, make multiple metrics corroborate each other, measure trends rather than snapshots, and track the gap between the reported number and a stricter version underneath it.

Done consistently, a covenant or compensation plan stops being a number management simply needs to hit and returns to its original purpose:

an early-warning system, not a target.