📈 economics

Inequality in the Age of Pseudonymity

This paper demonstrates that standard inequality measures, including the Gini coefficient, fail in pseudonymous digital environments due to Sybil attacks, necessitating the development of new, relaxed metrics that sacrifice fine-grained precision to ensure robustness against fake identities.

Original authors: Aviv Yaish, Nir Chemaya, Dahlia Malkhi, Lin William Cong

Published 2026-03-24

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Aviv Yaish, Nir Chemaya, Dahlia Malkhi, Lin William Cong

Original paper licensed under CC BY 4.0 (http://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

The Big Picture: The "Ghost Party" Problem

Imagine you are a party host trying to figure out how fair the distribution of cake is among your guests. You want to know: Is one person eating 90% of the cake while everyone else gets crumbs?

In the real world, you can usually see who is who. But in the digital world (like Bitcoin, crypto, or online games), people can wear masks. They can create fake identities (called "Sybils") to hide their true selves.

This paper asks a tough question: If people can wear infinite masks and split their cake among them, can we ever accurately measure who is actually rich and who is poor?

The authors' answer is a resounding "No." They prove that if people can create fake identities, the standard math we use to measure inequality (like the famous Gini Coefficient) breaks completely.

The Core Problem: The "Splitting" Trick

Let's use an analogy to understand why this is so tricky.

The Scenario:
Imagine two people, Alice and Bob.

Alice has 100 dollars.
Bob has 100 dollars.
Inequality: Zero. It's perfectly fair.

The Trick:
Bob decides to be sneaky. He creates 99 fake accounts (Sybils). He splits his 100 dollars among them so that:

One account has 99 dollars.
99 accounts have 1 dollar each.

Now, if you look at the list of "people" (identities), it looks like this:

Alice: 100
Fake Account 1: 99
Fake Accounts 2–100: 1 dollar each.

The Paradox:
If you use a standard inequality calculator (like the Gini Coefficient), the result changes!

Before: Perfect equality (0).
After: The calculator sees one person with 100 and many people with almost nothing. It screams, "This is highly unequal!"

But wait! Bob didn't actually get richer. He just split his money. The real inequality hasn't changed, but the measured inequality has skyrocketed.

The paper shows that this isn't just a small error. It's a fundamental flaw. If people can create fake names, you cannot trust any standard inequality number.

The "Impossible" Rules

The authors looked at the "Golden Rules" that economists usually demand from an inequality measure. They found that these rules fight each other when fake identities are involved.

Here are the rules they tested:

The "Rich to Poor" Rule (Transfer Principle): If you take money from a rich person and give it to a poor person, inequality should go down.
The "Name Doesn't Matter" Rule (Symmetry): It shouldn't matter if we swap names around; the math should be the same.
The "Scale Doesn't Matter" Rule: Whether we measure in dollars or cents, the result should be the same.
The "Fake-Proof" Rule (Sybil-Proofness): If someone splits their money into fake accounts, the inequality score shouldn't change.

The Big Discovery:
The authors proved that you cannot have all these rules at once.

If you want your measure to be Fake-Proof (Rule 4), you have to break the Rich-to-Poor Rule (Rule 1).
Basically, to stop people from cheating the system with fake names, you have to stop the math from caring about who is rich and who is poor in the way we usually understand it.

The Only Solution: The "Total Sum"

So, if we can't use the fancy rules, what can we do?

The authors found the only type of math that survives the "Ghost Party" is incredibly boring. It's called Sum-Dependence.

The Analogy:
Imagine a judge who refuses to look at the guests at all. Instead, the judge only looks at the total amount of cake on the table.

If there is 100kg of cake, the judge says, "Inequality is X."
If there is 200kg of cake, the judge says, "Inequality is Y."
It does not matter if one person has all the cake or if it's split among 1,000 people. The judge only cares about the total weight.

Why is this the only way?
Because if you look at how the cake is distributed, a cheater can always rearrange the slices (by creating fake names) to trick you. The only thing a cheater cannot hide is the total amount of cake.

However, this solution is useless for our original goal. If the measure only cares about the total sum, it can't tell you if society is fair or unfair. It's like a thermometer that only tells you the temperature of the room but can't tell you if the heater is broken.

What This Means for Bitcoin and Crypto

You might have heard headlines like: "Bitcoin is more unequal than North Korea!" based on a Gini Coefficient of 0.88.

This paper says: Ignore that number.

The authors show that the Gini Coefficient (and almost all other popular measures) is not Sybil-proof.

In Bitcoin, a wealthy person can easily split their coins into thousands of different wallets to make it look like the wealth is spread out, or conversely, hide their true concentration.
Because we can't distinguish between "one rich person with 1,000 wallets" and "1,000 poor people with 1 wallet each," the inequality number is meaningless.

The Takeaway

Digital Masks Break Math: In a world where people can easily create fake identities, the standard tools we use to measure inequality (like the Gini Coefficient) are broken. They can be easily manipulated.
No Perfect Fix: There is no mathematical formula that can both respect the "fairness rules" we love AND ignore fake identities. You have to choose one or the other.
The Only "Safe" Measure is Boring: The only math that can't be cheated is one that ignores distribution entirely and only looks at the total sum. But that doesn't help us understand inequality.
Be Skeptical: When you see reports claiming that a digital platform is "extremely unequal" based on wallet data, remember that the data might just be a magician's trick with fake names. We likely cannot know the true level of inequality in these pseudonymous worlds.

In short: You can't measure the fairness of a game if the players can change their names whenever they want. The scoreboard is lying to you.

1. Problem Statement

The paper addresses a fundamental flaw in applying classical economic inequality measures (such as the Gini Coefficient) to pseudonymous digital platforms (e.g., Bitcoin, DeFi, DAOs).

The Core Issue: In traditional economies, an individual's wealth is linked to a single, verified identity. In pseudonymous digital economies, actors can create multiple fictitious identities, known as Sybils, to split their wealth across multiple accounts.
The Consequence: Standard inequality metrics assume that the number of observed identities equals the number of real economic actors. When actors use Sybils, the observable wealth distribution ( $\rho$ ) diverges from the hidden true distribution ( $w$ ).
The Question: Can we reliably measure inequality in an environment where actors can manipulate the number of identities to hide their true wealth concentration? The authors argue that current metrics often produce misleading results (e.g., claiming Bitcoin is "more unequal than North Korea" based on a Gini of 0.88), potentially driving incorrect policy decisions.

2. Methodology and Framework

The authors employ an axiomatic approach, a standard method in welfare economics, to rigorously define the properties of inequality measures and test their robustness against Sybil attacks.

2.1 Formal Model

Hidden State: A set of $n$ real actors with a hidden wealth distribution $w = (w_1, \dots, w_n)$ .
Observable State: A set of $m$ identities with an observable wealth distribution $\rho = (\rho_1, \dots, \rho_m)$ .
Manipulation: A report matrix $R$ maps hidden wealth to observable identities. Actors can split wealth (Sybil attacks) or pool wealth (collusion).
Sybil-Proofness (Definition 6): A measure $I$ is Sybil-proof if $I(w) = I(\rho)$ for any wealth-conserving manipulation. The measured inequality must remain invariant regardless of how actors split their wealth among identities.

2.2 Canonical Axioms

The paper tests Sybil-proofness against the standard "desiderata" of inequality measures:

Scale Independence: Invariance to uniform scaling of wealth (e.g., changing currency units).
Population Insensitivity: Invariance to duplicating the population (e.g., $(5, 10)$ has the same inequality as $(5, 10, 5, 10)$ ).
Symmetry: Invariance to permutations of identities.
Transfer Principle (Pigou-Dalton): A progressive transfer of wealth from a richer to a poorer individual (rank-preserving) must strictly reduce inequality.
Egalitarian Zero Principle: Inequality is zero if and only if wealth is perfectly equal.

3. Key Contributions and Results

3.1 Impossibility Theorems

The paper establishes that no inequality measure can simultaneously satisfy Sybil-proofness and the standard Transfer Principle.

Theorem 1: No Sybil-proof measure satisfies the Transfer Principle.
- Logic: If a measure satisfies the Transfer Principle, splitting wealth (a Sybil attack) can be mathematically constructed to look like a "progressive transfer" between distinct entities, artificially lowering the measured inequality. Conversely, merging wealth can look like a regressive transfer. This creates a contradiction where the same underlying wealth distribution yields different inequality scores depending on the number of Sybils.
Proposition 2: Even relaxing the Transfer Principle to the Egalitarian Zero Principle (requiring only that perfect equality yields zero inequality) results in an impossibility. A non-equal distribution can be manipulated via Sybils to appear perfectly equal, forcing the measure to output 0, which contradicts the requirement that non-equal distributions have positive inequality.

3.2 Characterization of Sybil-Proof Measures

The authors fully characterize the class of measures that are Sybil-proof.

Theorem 2 (Aggregation Invariance): A measure is Sybil-proof if and only if it is Aggregation Invariant. This means the measure's value depends only on the sum of the wealth, not the distribution of that sum across identities.
Corollary 3 (Sum Dependence): A measure is Sybil-proof iff it is Sum-Dependent.
- Mathematically, $I(x) = I(y)$ if and only if $\sum x_i = \sum y_j$ .
- Implication: The measure treats the entire economy as a single "grand coalition" where all actors have perfectly colluded. It cannot distinguish between one rich person and one rich person with 1,000 fake accounts.
Corollary 2 (Constant Measures): If a Sybil-proof measure also satisfies Scale Independence or Population Insensitivity, it must be a constant function (i.e., it outputs the same value regardless of the wealth distribution).
- Result: The only measures that are both Sybil-proof and scale/population invariant are trivial constant functions (e.g., always outputting 0 or always outputting 1). These are useless for comparing inequality across different economies.

3.3 Application to Popular Metrics

The paper proves that all widely used inequality metrics fail Sybil-proofness:

Gini Coefficient (GC): Not Sybil-proof. It satisfies the Egalitarian Zero Principle but is not a constant function, violating the necessary conditions derived in Corollary 4.
Generalized Entropy (GE) Family: Includes Theil indices, Atkinson measure, and Coefficient of Variation. None are Sybil-proof because they are scale-independent and non-constant.

4. Significance and Implications

4.1 Theoretical Impact

Fundamental Limitation: The paper proves a "no-go" theorem for inequality measurement in pseudonymous settings. The very properties that give inequality measures their economic meaning (sensitivity to distribution, scale invariance) are incompatible with robustness to identity manipulation.
Sybil-Proofness implies Collusion-Resistance: The only Sybil-proof measures treat all identities as a single entity. Therefore, they cannot distinguish between a single wealthy actor and a perfectly colluding group of actors.

4.2 Practical Impact

Invalidation of Current Metrics: Claims about the extreme inequality of Bitcoin or DeFi based on Gini coefficients are theoretically unsound because they ignore the Sybil problem. The high Gini values may reflect the fragmentation of wealth across many addresses by a few users, not necessarily extreme concentration of power among distinct individuals.
Policy Warning: Policymakers and platform designers cannot rely on standard metrics to assess decentralization or fairness in digital economies.
Future Directions: The authors suggest that future work must either:
1. Relax the axioms significantly (accepting constant or sum-dependent measures).
2. Develop approximate guarantees (e.g., robustness to a bounded number of Sybils).
3. Rely on off-chain data or identity verification (KYC) to bridge the gap between observable and hidden actors, though this conflicts with privacy goals.

5. Conclusion

The paper concludes that in an age of pseudonymity, reliable inequality measurement is impossible without sacrificing the core properties that make such measures useful. The tension between robustness to pseudonymity and sensitivity to distribution is irreconcilable. Consequently, any attempt to measure inequality in systems like Bitcoin using standard metrics (like the Gini coefficient) is fundamentally flawed and likely to produce misleading conclusions about the true state of economic equality.