Author: Max Resnick, Chief Economist at Anza
Compiled by: Jiahuan, ChainCatcher
I am publishing this article as SIMD 550 (doubling the rate of inflation reduction) and SIMD 553 (resource costs) are about to enter the voting stage. This piece is not a specific commentary on these two proposals; I have already left my comments on the respective GitHub proposals. Here, I primarily discuss how the problems these proposals aim to address relate to the valuation of L1s.
Part 1: The Establishment of Asset Pricing Theory
Before asset pricing theory was formally established, investors were not short of methods to value companies. Some focused on hard assets and liquidation value, others emphasized profits, dividends, growth, management quality, or market psychology. There were many metrics available; what was truly missing was a rigorous framework to explain exactly which metrics determine value and how these metrics should be weighed against each other.
By the late 1920s, on the eve of the Great Depression, this ambiguity had become dangerous. Investors could cite facts like profit growth, market expansion, new technologies, and improved corporate governance, but these facts were often used to justify market prices rather than to derive asset value.
Graham and Dodd (1934) later described that era as one where analysis gave way to "potential and prophecy." Even when data was presented, it became "pseudo-analysis to support the illusions of the moment."
Anyone who frequently indulges in Crypto Twitter should find this scene familiar. Today's discussions around L1 tokens are similarly filled with the potential, prophecy, and pseudo-analysis that characterized the stock market of the late 1920s.
The number of people participating in blockchain development is at a record high. Transaction activity is at an all-time peak. Tokens will become currency, collateral, digital oil, or a call option betting on the future financial system.
Some of these claims may be true and could even imply upside for the underlying network. But if they cannot explain how these factors translate into token holder surplus, they do not constitute a coherent valuation framework.
John Burr Williams was the first to push asset pricing towards rigor. In "The Theory of Investment Value," Williams (1938) argued that value is "the present worth of future dividends, or of future coupons and principal in the case of a bond."
Gordon (1959) later expressed the same idea more succinctly: "A stock, like any other asset, is worth to a buyer only the benefits he expects to receive from it in the future."
Equity has value not because a company is impressive, active, important, or technologically irreplaceable. It has value because equity grants shareholders a claim on future income.
For L1 tokens, value can accrue in two ways. The first is fee burning, which economically resembles a buyback. The second is distributing fees to stakers, which economically resembles a dividend.
Staking rewards funded by issuance are different. They are neither income created by the network nor a cost borne by the network. The protocol creates new tokens and distributes them to stakers while diluting the holdings of non-stakers.
This mechanism may be necessary to secure the network and may determine who gradually comes to own the network over time. But from the perspective of all token holders, it neither creates nor destroys value.
A blockchain can process millions of transactions while creating almost no value for token holders because the surplus may be captured by users, applications, validators, or other intermediaries. Conversely, a chain with less activity may be more valuable if it converts a larger proportion of economic activity into token holder value.
But not all fees are created equal. The 'R' in ARR stands for recurring, meaning sustainable and repeatable income. Dichev, Graham, Harvey, and Rajgopal (2013) noted that high-quality earnings should be "sustainable and repeatable." The same standard applies to L1 fees.
One dollar of fees generated by long-term financial activity is not the same as one dollar generated by airdrops, meme coin mania, cascading liquidations, or temporary network congestion.
Some fees come from users' ongoing demand for scarce block space. Others are merely exhaust from speculative cycles. Once incentives disappear, volatility subsides, or users run out of money, these activities also vanish.
Fee quality depends on persistence and defensibility.
Are users paying because the chain provides long-term economic utility, or because a short-term activity happens to be on this chain? Can the protocol continue to charge these fees without driving users, applications, or order flow elsewhere? Can the token continue to capture these profits, or will this value eventually be competed away by validators, applications, searchers, block builders, users, or other chains?
Historically, crypto investors have often made two opposite errors regarding earnings quality simultaneously.
On one hand, they overestimated earnings quality because much crypto activity is speculative, reflexive, and episodic.
On the other hand, they underestimated earnings quality because they did not fully appreciate the strength of L1 network effects. Liquidity, applications, wallets, infrastructure, users, developers, assets, and order flow reinforce each other.
These network effects may make certain fees harder for competitors to capture than they initially appear. They also suggest that dominant blockchains like Solana and Ethereum may have stronger pricing power than the market generally believes, potentially benefiting from raising fees.
Part 2: Accounting Standards for Revenue, Inflation, and Total Supply
The next step is to specify a minimal fundamental model for L1s that can lead to valuation multiples.
It may be premature to call it the "standard model" now. There is currently no universally accepted standard model for L1 valuation. But the categorization below is the form I believe a standard model should take. It intentionally mirrors the methods stock analysts use to value companies.
The reason this needs to be explicitly written down is that there isn't even consensus on the most basic accounting objects.
I've discussed this framework with some of the smartest people I know, and they often disagree on fundamental questions. Are validator rewards funded by inflation considered a cost? Should foundation expenditures be treated as operating expenses? Should the unspent portion of foundation tokens be included in the token supply? Should MEV paid to validators count as protocol revenue, validator revenue, or neither?
Some of the confusion may stem from the fact that there can be more than one correct way to write a valuation model. Accounting classifications are not unique. As long as the corresponding offsetting items are also adjusted, you can move an item from one side of the ledger to the other while keeping the model correct.
But this flexibility is also dangerous. Many models are internally consistent, but many are not. Having multiple correct methods does not mean there are fewer incorrect ones.
Inflation rewards are the simplest example.
Staking rewards funded by issuance are essentially rewards paid by token holders to stakers via dilution. From the perspective of all token holders combined, the two offset each other. The protocol does not earn revenue when it mints new tokens, nor does it incur a real external cost when distributing these tokens to stakers.
You can create a correct model that treats inflation rewards as a cost, but only if you simultaneously count newly issued tokens as a source of value to offset that cost. Otherwise, the model leads to absurd conclusions, such as Solana being unprofitable because it pays large staking rewards.
The categorization below is my proposed baseline. It separates three objects: revenue, costs, and total supply.
I believe these definitions most closely resemble the models stock analysts are already familiar with, making them easier to understand and reason about.
Other classifications may be correct, but deviations from this one must have clear justification. Proprietary models come with two costs.
First, they are harder for others to understand. Second, it is easy to miss dependencies between items.
For example, if foundation expenditures are classified as a cost, then the unspent portion of foundation tokens cannot also be included in the total supply, or it would be double-counted. If inflation rewards are classified as a cost, newly issued tokens must be treated symmetrically.
Part 3: Supply, Demand, and Price
Recently, several blockchains have explicitly raised fees with the goal of increasing revenue. However, revenue equals price multiplied by quantity.
Raising fees increases the revenue contributed by transactions that remain, but it also causes some transactions to disappear. Therefore, the net effect of a fee increase on revenue is uncertain; it hinges on the price elasticity of demand for transactions.
To understand the reasoning behind these adjustments, I spoke with some of the decision-makers involved. They believed the fees on these chains were originally too low, so demand was relatively inelastic within that price range.
This argument might apply to specific cases but is not universally true.
I have done some past research utilizing randomness in EIP-1559 pricing. The analysis showed that the price elasticity of gas demand is roughly between 0.6 and 0.8. That is, a 10% price increase leads to a 6% to 8% decrease in quantity demanded.
Moreover, this data only reflects short-term price fluctuations and does not capture the aggregate effects of applications migrating off-chain or optimizing their programs.
Therefore, a flat fee is a rather blunt instrument for generating revenue. Different types of on-chain transactions have different willingness to pay.
A small wallet transfer, a large stablecoin transfer, and a liquidation may consume the same block space, but they generate different total surplus and have different willingness to pay.
Note: Blue dots represent fee payers who initiated more than 250 transactions in the period, suggesting they are more likely to be bots and thus have higher price elasticity. Bots typically have thin profit margins, so they tend to reduce resource consumption significantly when prices rise.
Protocols want to charge higher fees for transactions with a higher willingness to pay. Computation-based fees are a step in this direction but not thorough enough.
For financial activity, the nominal transaction amount is often a better indicator of willingness to pay than computational load. This is why exchanges typically charge fees based on basis points.
Token programs can provide a way to charge fees proportional to transaction value. By modifying token programs, a small percentage-based fee can be levied on token transfers. This way, even if a high-value transfer uses similar computational resources as a low-value one, the high-value transfer still pays more.
