What is the Correlation Penalty?
The correlation penalty is an additional slashing penalty in Ethereum's proof-of-stake system that increases based on how many other validators are slashed around the same time. When a single validator is slashed in isolation, they receive minimal additional penalties. When many validators are slashed together, suggesting either a coordinated attack or dangerous infrastructure centralization, penalties can scale up to the validator's entire stake.
This mechanism is specifically designed to make large-scale attacks prohibitively expensive while being relatively lenient toward isolated honest mistakes. The correlation penalty embodies Ethereum's security philosophy: individual errors should be forgiven, but systemic risks and attacks must be severely punished.
How it Works
When a validator is slashed, they immediately lose 1/32 (approximately 3.1%) of their effective balance as an initial penalty. They then enter a roughly 36-day withdrawal delay period (8,192 epochs). During this period, the protocol calculates the correlation penalty.
The correlation penalty formula examines how much total stake was slashed in the 18 days before and after the validator's slashing event. The additional penalty equals: (validatorbalance × 3 × totalslashedbalance) / totalactivebalance.
If only one validator is slashed, the correlation penalty is negligible since totalslashedbalance is tiny compared to totalactive_balance. But if 33% of validators are slashed together, the multiplier approaches 1, meaning each slashed validator loses their entire stake.
The formula creates strong incentives for validator diversity. Running validators on the same client software, same cloud provider, or same geographic region creates correlated failure modes. If a bug or outage causes mass slashing, all correlated validators face maximum penalties.
After the withdrawal delay, slashed validators receive whatever balance remains after the initial penalty, correlation penalty, and any other accumulated penalties. In severe cases, this can be zero or near-zero.
Practical Example
A bug in a minority Ethereum client causes 500 validators running that client to accidentally double-attest. All 500 are slashed within the same few epochs. With roughly 900,000 active validators total, these 500 represent about 0.055% of the network. Each slashed validator faces a correlation penalty of approximately 0.17% of their balance (3 times 0.055%). Combined with the initial 3.1% penalty, total losses are around 3.3% per validator. However, if the same bug affected a 33% client, penalties would be catastrophic.
Why it Matters
The correlation penalty is fundamental to Ethereum's decentralization incentives. It economically rewards diversity in client software, infrastructure providers, and geographic distribution. Validators who choose minority clients and independent infrastructure face much lower tail risk than those concentrated on popular choices. Understanding correlation penalties helps stakers make informed infrastructure decisions. Fensory analyzes validator distribution across clients and infrastructure providers to help you minimize correlation penalty risk in your staking setup.