A brief analysis of ETH selling pressure after Shanghai upgrade: What is the impact of partial withdrawals?

The exact date for Ethereum’s Shanghai upgrade has not yet been determined, but it is very close.

The discussion around selling pressure is rife again, but it seems no one has explored this topic in depth from a data perspective. My earlier model exploring this topic is outdated but still being referenced, and this post will be an update.

Reading this article requires you to have a basic understanding of the mechanism behind Ethereum withdrawals. If you are not familiar with the two types of credentials discussed below, they will be explained in detail later in the content.

The main changes compared to my earlier models are:

• The control process for switching between the two types of withdrawal vouchers has been clarified.

  All withdrawals require 0x01 credentials, which are not set by early validators;

  The operation rate of adding 0x01 withdrawal vouchers is limited to 16 times per block;

• All 0x01 credential validators will automatically perform partial withdrawals, with a maximum of 16 partial withdrawals processed per block.

• The process for all withdrawals remains the same, but 0x00 validators must set their 0x01 credentials before processing a withdrawal.

For more information on this process, I recommend the following resources:

• ETH Withdrawal FAQ

• Christine Kim's detailed record

Which validators set which credentials?

Most analysts assume that the two types of validators (0x00, 0x01) have consistent characteristics, but this is a big mistake. When we look at the beacon chain deposits in the figure below, we see that all genesis validators were created without setting up 0x01 credentials, which severely distorts the data.

As of today, approximately 20% of new validators do not have 0x01 credentials set up, likely individuals using outdated guidelines and institutions that have not updated their procedures.

The most prominent adopter of the 0x01 certification system is Lido; 88% of liquidity staking providers have set up withdrawal credentials and control over 60% of validators that are set up to receive automatic withdrawals. A naive observer might worry about the centralization issues this will cause, but in this case, Lido's mass adoption of the standard is a positive for the network.

Since Lido will recycle their rewards back to new validators to optimize their staking derivative yields, mass adoption will clog some withdrawal queues and reduce the initial volatility of withdrawals.

Aside from heavy adopters of 0x01 tokens, most of Lido’s validators were activated early on, with the largest spike occurring in the spring of 2022 when their staking derivatives began trading at a discount. As a result, the average Lido 0x01 validator is older than the average 0x01 validator and holds a large amount of automatically withdrawn rewards.

Dividing the accumulated rewards by the number of validators, we get the result in the figure below. Validators with Lido 0x01 certificates have accumulated an average of 1.23 ETH in rewards, while non-Lido 0x01 validators have accumulated an average of only 1.06 ETH.

Validators with 0x00 credentials are heavily skewed by genesis validators who have accumulated rewards for over two years. As of February 4, 2023, such validators have an average of 2.47 ETH in rewards (which can be withdrawn via partial withdrawals), while the best performing validators receive a base reward of 4.9 ETH.

After Shanghai's upgrade

Once withdrawals are enabled, Ethereum will begin processing some withdrawals; scanning validators and distributing rewards to validators in the 0x01 pool. At the same time, 0x00 pool validators who requested to switch their credentials to 0x01 will begin filling the 0x01 pool. Both operations occur at a rate of 16 validators per block.

The 0x01 pool will start to heavily favor Lido validators. If the Shanghai fork occurs on January 29, 2023, the expected distribution of validators in the first block will be:

• Initially 16 validators set up 0x01 credentials

  9.95 Lido: A total of 13.19 ETH was withdrawn.

  6.05 Non-Lido: A total of 6.89 ETH was withdrawn.

• There is no authenticator that initially sets the 0x00 credential.

By block 1,000 (approximately 3 hours and 20 minutes), the expectation is that:

• 14.83 validators who initially set 0x01 credentials

  9.22 Lido: 12.22 ETH withdrawn

  5.61 non-Lido: 6.39 ETH withdrawn

• 1.17 validators who initially set up 0x00 credentials: 3.10 ETH withdrawn

By block 10,000 (approximately 1 day and 9 hours), the expectation is that this will be:

• 7.50 validators who initially set up 0x01 credentials

  4.66 Lido: Total withdrawal 6.17 ETH

  2.84 Non-Lido: 3.23 ETH withdrawn

• 8.50 validators who initially set up 0x00 credentials: 22.57 ETH withdrawn

The location of the peak withdrawal rate (amount of withdrawals per block) depends on what percentage of 0x00 validators switch their credentials immediately, but in the most extreme case where all 0x00 validators request a change, the peak would be reached after 70 hours, with approximately 37.34 ETH withdrawn per block - 3.01 of which would come from Lido validators.

Anyone who suggests the peak is imminent hasn’t looked deeply enough into the data.

The animation below shows the evolution of daily fractional withdrawals after a different share of 0x00 validators choose to switch their credentials immediately. I think the most likely scenario is between 25% and 50%, suggesting that the peak rate will be reached early the next day and last for about two days.

In the most extreme case, the first day should see partial withdrawals of around 110,000 ETH (excluding Lido 0x01 validators). This number is in stark contrast to naive models that predict that the first day could see partial withdrawals of over 400,000 ETH.

Considering that non-Lido validators include most centralized institutions and liquid staking derivatives, it is reasonable to expect that these validators will only sell part of their rewards. In this light, it seems unlikely that autonomous partial withdrawals will lead to catastrophic events in the market. The decision to automatically withdraw partial withdrawals may reduce the initial selling pressure by clogging the queue with validators who do not intend to sell.

Before Shanghai Upgrade

The data is not definitive, so a natural question to ask is: how should we expect the profile to change before enabling withdrawals? My intuition is as follows:

All types of validators will continue to receive rewards, and new validators will continue to emerge.

New 0x00 validators will be very scarce compared to the existing 310,000+ validators, so their average balance will continue to climb, reaching 2.5 to 2.6 ETH per validator.

The number of Lido 0x01 validators will continue to climb, and the average balance per validator will increase to 1.3 to 1.4 ETH.

The number of non-Lido 0x01 validators will accelerate, putting downward pressure on average balances, which I expect to be in the 1.0 to 1.1 ETH range per validator.

The net result of these increases should be a 10% increase in the maximum withdrawal rate over the estimates in this article.

Adding validators (particularly 0x01) before withdrawals are enabled will flatten the curve; extending the duration of sell pressure but slightly reducing the peak of sell pressure.

Full withdrawal

The process for full withdrawals remains the same. You can track the churn limit in real time here, but the churn limit quotient of your balance at the start of the withdrawal is probably 8, which would indicate that 57,600 ETH can be withdrawn per day via full withdrawals.

A queue will emerge for new validators to activate and validators to exit. The double-ended queue will lock the churn limit and active validator count at a value until one side of the queue dies down. I suspect the community will see the Shanghai upgrade as a sufficiently de-risking event that new validator deposits will surge, but there will also be many validators who decide that staking is not the right choice.

Over time the number of validators may increase, but in the short term anything can happen.