Does the median transaction track purchasing power? Five chains

The Bitcoin page found that the average on-chain transaction got 150 times larger in real terms, which breaks the whitepaper's gauge. Averages are dominated by exchanges and whales. This page repeats the test on every chain with free daily data, using the median transaction wherever it is available.

Generated , commit . Sources: . Monthly medians of daily values. Elasticity is the slope of log(y) on log(x): minus one means coins per transaction fall exactly as fast as the price rises, and zero for the USD version means the real size of a transaction did not change.

1. Price up, coins per median transaction down

Each panel indexes the price and the median transaction size (in the coin's own units) to the first month, on one axis. If the median transaction tracks purchasing power, the two lines are mirror images and the elasticity in the title is close to minus one.

Price, indexedMedian transaction in coin units, indexed (chain colour)

2. What the median transaction was worth

The direct test. A flat line means people moved about the same real value per transaction regardless of the coin's price.

Table view: yearly median transaction in USD

3. Elasticities side by side

Median coins per transaction vs price

-1 = tracks purchasing power exactly; 0 = no response

Median USD per transaction vs price

0 = real size unchanged; +1 = real size moves with price

Table view: elasticities by chain and era

4. Slice of supply per median transaction vs users

The coarse fixed-supply model predicts a slope of minus one on log-log axes. This corrects for coins minted along the way by dividing by circulating supply.

5. Reading

6. Caveats

Two sources. Litecoin, Dogecoin, Bitcoin Cash and Ethereum use Blockchair's daily aggregates; for the UTXO chains the transfer value is the total output of each transaction including the change a wallet sends back to itself, so levels overstate true transfers, the median far less than the mean. Bitcoin uses BitInfoCharts' daily median and mean transaction value in USD, converted to BTC at the CoinMetrics daily price, because Blockchair rate-limits bulk pulls. The two sources agree on Litecoin's yearly medians to within about 20%.
Ethereum: only plain ETH transfers of at least 0.001 ETH are counted, which drops the zero-value contract calls that would otherwise make the median zero. ERC-20 and stablecoin transfers, which carry most of Ethereum's payment activity today, are not counted anywhere here.
Dust and inscriptions: periods of spam with near-zero outputs (Dogecoin 2024, Bitcoin Cash 2019) drag the daily median toward zero. Those months show as dips in the USD chart and weaken the fits.
Round-number habits: Dogecoin's median sat at exactly 10,000 DOGE for years because of tipping conventions. A convention, not a price signal.
Active addresses count addresses, not people; exchanges and batching make it a loose proxy for users.

7. Reproduce

git clone https://github.com/pastacoin/pastacoin && cd pastacoin
python -m venv .venv && .venv/Scripts/pip install -e ".[dev,report]"
.venv/Scripts/python -m pasta.analysis.chains_report --figures figures         # committed data/chains/*.csv
.venv/Scripts/python -m pasta.analysis.chains_report --fetch ltc doge --figures figures   # refresh (Blockchair rate limits apply)