# Demand Under a Mechanical Price: Elasticity and Welfare in Ethereum's Data-Availability Market

Crypto.world

> Ethereum sells data-availability capacity to rollups in units called blobs, and the price is not set by any auction—it is set by a formula that moves the fee exponentially with cumulative usage above…

**arXiv**: 2610.16588

**Published**: 2026-10-12

**Authors**: Farbod Ghasemlu

## AI summary

## Overview

**Research area:** Blockchain economics and market microstructure—specifically the empirical industrial organisation of Ethereum's blob (data-availability) fee market, combining protocol-level measurement with applied econometrics and welfare analysis.

**Technical level:** Intermediate. The paper is readable without a protocol-engineering background, but the identification strategy (Poisson pseudo-maximum likelihood, placebo-date inference, wild score bootstrap, specification curves) assumes familiarity with applied microeconometrics.

**Scope (one sentence):** Using the Pectra upgrade of 7 May 2025 as a scheduled supply shock, the paper bounds the price elasticity of rollups' demand for blob capacity and uses that bound to decompose Ethereum's mechanical blob fee into transfers versus allocative deadweight loss.

## What This Paper Is About

Ethereum sells data-availability capacity to rollups in units called blobs, and the price is not set by any auction—it is set by a formula that moves the fee exponentially with cumulative usage above or below a target, with a floor of 1 wei. The paper asks what this mechanical price costs: whether a mis-set fee mainly changes *who gets capacity and how much is used* (an allocative cost, which requires elastic demand) or mainly *who pays* (a transfer, which is what inelastic demand implies). To answer this, the author estimates the price elasticity of blob demand from the one event in the sample that moved the price materially.

## Key Contributions

1. **A bound on the demand elasticity for blob capacity.** When the blob price fell from about 0.82 mETH per blob (about $1.40) to the 1-wei floor at Pectra, rollups' blob purchases rose 24%, with a reported 95% interval of 8% to 40%. The implied elasticity at the pre-shock price is -0.19, and no value below -0.31 is consistent with the interval.
2. **A welfare decomposition separating transfers from allocative loss.** Rollups paid 4,704 ETH ($14.8 million) in blob fees under the first (Dencun) regime, 67% of it in 2024-Q4 and 2025-Q1 when usage averaged 97% of target; in slack periods the deadweight loss of the EIP-7918 reserve price is at most 2.6 ETH ($8,039) over 296 days.
3. **Evidence that the floor regime reflects slack capacity, not queuing.** Matching 5.4 million blob transactions to their first mempool sighting shows that transactions first seen at the floor miss 1.03 blocks on average in Dencun—the fewest of any state—and 0.08% of the blocks they miss price them out.
4. **Two documented identification failures and a protocol audit.** The later capacity increases (BPO1 and BPO2) have no first stage because the reserve, not usage, set the price; the execution base fee fails the exclusion restriction because every blob rides in a transaction paying execution gas at that same base fee. Separately, the paper recomputes the excess-blob-gas state of all 6,650,443 blocks from their parents and confirms each regime's own rule explains every block.

## Main Findings

- **Demand is inelastic.** Removing a price of about $1.40 per blob raised purchases by 24%, and by no more than 40%; the elastic bound is -0.31. No interval reported in the paper contains an elasticity above 1 in absolute value.
- **The rise came through L2 activity, not intensity.** Blobs bought per L2 transaction changed by -2.8% (interval -19.9% to +11.6%), so rollups bought more blobs because they carried more L2 transactions, not because they bought more capacity per transaction.
- **The fee mostly acted as a transfer.** Under Dencun, 4,704.4 ETH ($14,813,911) was burned. In 2024-Q4 and 2025-Q1 capacity was full and usage averaged 97% of target, so those fees were scarcity rents on fixed capacity that any market-clearing mechanism would have collected in comparable size; 2024-Q3 fees came to 0.000004 ETH.
- **Deadweight loss under the reserve is small.** Over 296 days from Fusaka, when usage averaged 32% of target, the loss is 0.399 ETH ($1,241) at the primary point estimate, 0.644 ETH ($2,005) at the elastic end of the reported interval, and 2.582 ETH ($8,039) at the 5th percentile of the specification curve—9.8% of fees collected under the reserve.
- **The floor is slack capacity, not a queue.** In Dencun, transactions first seen at the floor missed 1.03 blocks versus 2.25 when the previous block was full; in Pectra, 0.89 versus 1.26. In congested Dencun states, 27% of missed blocks priced the transaction out, against 0.08% at the floor.
- **Pectra's first stage is large and clean.** The mean blob fee fell by 0.90 mETH per blob (Newey-West SE 0.33), and the Pectra regime recorded 7.8 ETH in total fees against Dencun's 4,704.4 ETH.
- **The specification curve is centred on inelasticity.** Across 450 specifications, the median elasticity at the pre-shock price is -0.22 with a 5th-to-95th percentile range of -0.39 to 0.72; 13% have the wrong sign and all of those use the 14-day window.
- **Heterogeneity across rollups.** Estimated one rollup at a time, 11 of 14 rollups increased purchases; the equal-weighted mean is 0.245 log points and Base's is 0.295. Base carries 78% of the variance of the score in the primary specification and 95% in the intensity specification.
- **A calibration supports at least this much inelasticity.** Given a fee standard deviation of 1.26 mETH per blob during congested quarters and detrended L2 activity with a standard deviation of 8.9%, the implied semi-elasticity is 0.071 per mETH per blob—less elastic than the primary estimate. The elastic tail of -1.36 would require demand shifts with a standard deviation of 171%.
- **Post hoc pass-through evidence (exploratory).** Comparing 13 blob-buying rollups with 7 L2s and sidechains that do not buy blobs over the same window (N = 1,120 chain-days), blob users' median L2 fees changed by -31% (SE 0.20 log points, placebo p = 0.35) and their L2 transactions rose 32% (SE 0.11, placebo p = 0.10).
- **The protocol schedule verifies exactly.** Each regime's own update rule explains 0.0000% unexplained blocks, while the EIP-4844 rule mispredicts 29.5% of Fusaka blocks, 76.9% of BPO1 blocks and 87.0% of BPO2 blocks.

## Methodology in Plain English

The central obstacle is that the blob fee is the same for every buyer in a block, so comparing rollups to each other reveals nothing about how they respond to price—and within a regime the fee is itself an outcome of demand. For long stretches the fee sits at a floor of 1 wei, where a log-price elasticity is not economically meaningful. The author therefore exploits a supply shock whose date was fixed weeks in advance by client releases: Pectra doubled the blob target from 3 to 6 per block on 7 May 2025, while the market was congested, and the fee collapsed to the floor within a day.

The estimator is a Poisson pseudo-maximum likelihood regression on rollup-day data over 28 days either side of the fork, with rollup fixed effects, rollup-specific linear trends and day-of-week effects. The headline outcome is blobs purchased by the 15 rollups with daily L2 transaction counts. The specification was committed in writing before estimation (commit 811ab9b) with blobs per L2 transaction as the primary outcome; when the data showed L2 activity itself responded to the fork, the unnormalised version was promoted to primary and the revision was recorded with its reason.

Because there is only one treatment date, standard errors clustered by rollup do not capture the risk of a common shock on the fork date. Inference therefore comes from applying the identical estimator at 516 placebo break dates whose windows exclude Pectra and reading off the quantiles of that distribution; a restricted wild score bootstrap and clustered intervals are reported alongside. A specification curve over 450 variants summarises sensitivity. The welfare calculation models demand as exponential in the fee, calibrated each day to the observed price and quantity, and prices slack capacity at zero in the counterfactual. A separate mempool analysis matches blob transactions to their first sighting in the public mempool to measure delay by market state.

## Why This Matters

**Impact on research.** The paper supplies the first design-based elasticity estimate for blob capacity and shows that the answer determines what a mechanical price costs—shifting the design question from allocative efficiency to revenue, volatility and capacity. It also reports two identification strategies that fail for mechanical reasons (BPO2's absent first stage, and the base fee's failure of the exclusion restriction, where blob purchases rise 0.060 log points per log point of base fee with SE 0.020 across 268,200 rollup-hours), which saves other researchers the attempt, and it verifies the blob schedule from the protocol's own update rule.

**Real-world applications:**

- **Blob-parameter planning.** Because demand is inelastic, raising the target is primarily a way to end a transfer rather than to correct a large misallocation; Pectra cut fees from 16.9 ETH per day ($29,182) to effectively zero while purchases rose 24%.
- **Evaluating reserve prices.** The EIP-7918 reserve can be assessed against a quantified deadweight loss (at most 2.6 ETH over 296 days) rather than by intuition.
- **Rollup fee-setting and pass-through.** The post hoc evidence that blob costs on OP Stack chains were of the same order as the entire median L2 fee before Pectra matters for how rollups price to their own users.
- **Mempool and inclusion monitoring.** The delay statistics give an empirical baseline for what normal, uncongested inclusion looks like—about one block, 12 seconds—against which congestion can be measured.

**Industry relevance.** The findings bear on protocol governance debates over blob capacity and reserve pricing, on rollup operators deciding how to pass data-availability costs through, and on the builder/relay layer, since prior work cited in the paper finds that 29.5% of blob-carrying blocks are built in a way that earns the builder less than an available alternative.

## Future Directions

- **A second supply shock in a congested market.** The author states explicitly that a pre-specified test on the next supply shock arriving in a congested market would settle whether the measured rise in L2 activity came from pass-through or from the ETH rally that began the day after Pectra.
- **Demand at higher fees.** The bound holds only for fees between zero and 4.45 mETH per blob; the 9 days above that range carried 27% of the fees paid in the first two regimes, and demand could be more elastic there.
- **The payload margin.** The outcome measures capacity bought, not how fully blobs are filled; rollups can respond to price by packing blobs more densely, and for small rollups the choice is coarse because a blob is a fixed 128 KiB.
- **Rollup-level cross-layer migration.** Movement between data-availability layers is not observable in this dataset—the Celestia indexer blocked collection—so the paper relies on aggregate cross-layer series that cannot show whether a given rollup moved its data.

## Target Audience

Protocol economists and mechanism-design researchers working on blockchain fee markets; applied econometricians interested in event studies with a single treatment date and few clusters; Ethereum core developers and governance participants weighing blob-parameter changes and reserve prices; and rollup operators or analysts who want a quantitative baseline for how blob costs transmit to L2 fees and activity.

## Authors’ abstract

Ethereum sells data-availability capacity to rollups through "blobs" priced by a controller: the blob fee rises and falls exponentially with cumulative usage above or below a target, and no auction sets it. We measure what this mechanical price costs. The Pectra upgrade of 7 May 2025 doubled blob capacity on a date fixed weeks in advance, while the market was congested, and the blob fee fell from 0.82 mETH per blob (about USD 1.40) to its 1-wei floor. Rollups' blob purchases…

[The full abstract on arXiv](https://arxiv.org/abs/2610.16588)

[Read the original paper](https://arxiv.org/abs/2610.16588)

---

Canonical: https://crypto.world/paper/demand-under-a-mechanical-price-elasticity-and-welfare-in-ethereum-s-2610.16588
