Stablecoins: Survivorship, Transactions Costs and Exchange Microstructure

Download as pdf or txt
Download as pdf or txt
You are on page 1of 43

Stablecoins: Survivorship, Transactions Costs and

Exchange Microstructure
∗1
Bruce Mizrach
1
Department of Economics, Rutgers University, New Brunswick, NJ USA
arXiv:2201.01392v1 [q-fin.TR] 5 Jan 2022

First Draft: April 27, 2021


Revised: December 27, 2021

Abstract
Seven of the ten largest stablecoins are backed by fiat assets. The 2016 and 2017
vintages of stablecoins have failure rates of 100% and 50% respectively. More than
one-third of stablecoins have failed. Tether has a 39% share of 1.77 trillion USD in
2021Q2 transactions, and USD Coin 28%. The top three stablecoins have an average
velocity of 28.3. Tether transacted between 3.8 million unique addresses, 63% of the
ERC-20 token network. Six of the top ten tokens have unconcentrated Herfindahl
indices, but Gemini, Pax and Huobi have single holders with more than 50% of the
supply. The median Tether transaction fee is similar to the cost of an ATM transaction,
but they are three to four times more for Dai and USDC. Fees, which are proportional
to the price of Ethereum, are rising though. Median fees for Tether rose 3,628% over
the last year, and 1,897% for USD Coin. 24 hour exchange turnover in Tether is nearly
$120 billion. This is comparable to the daily volume at the NYSE and almost 15 times
the daily flow in money market mutual funds. Narrow bid-ask spreads and depth have
attracted active HFT participation.

Keywords: Stablecoins; transactions; fee; hazard function; market microstructure;


cryptocurrency.

JEL Codes: G12; G23.


Correspondence: Department of Economics, Rutgers University, 75 Hamilton Street, New Brunswick,
NJ 08901 USA. email: [email protected], (908) 913-0253 (voice) and (425) 795-9942 (fax).
1 Introduction
The financial industry has continued to develop new forms of payment. Credit and debit
cards and automated clearinghouse (ACH) transactions were the key developments of the
late 20th century. Venmo and PayPal came along in the early 21st.
Credit and debit cards still dominate, making up 62% of purchases.1 Transaction costs
are still substantial. Credit card transactions are nearly instantaneous, but the merchants
pay fees of 1 to 3%2 . Debit cards fees are closer to 1%,3 yet consumers can face stiff payments
for overdrawing their account. Venmo is very convenient, but it also charges 3% for credit
card transactions.4 Paypal’s recent growth has largely been driven by opening up to credit
card companies.
There were 7.3 billion ACH transactions with a total value of $18.4 trillion in the second
quarter of 20215 with a median cost of $0.29.6 ACH transactions are quite slow though, and
can take up to five days. The G7 Working Group (2019) notes that there are still many
gaps in the payment system, particularly for cross-border transactions. With respect to fees,
speed and network size, there are substantial opportunities for new financial technologies.
Stablecoins try to harness the distributed ledger technology of Bitcoin and other digital
assets, while maintaining price stability of fiat assets and other stores of value. Regnard-
Weinrabe et al. (2019) note that there are three major types of stablecoins: (1) Fiat/commodity
collateralized; (2) Crypto-collateralized; (3) Non-collateralized. Each type differs on the col-
lateral backing the tokens. Among the top tokens I study, the coins are either fiat or
crypto-collateralized.
Collateralized stable coin companies are expected to actually hold the assets against
which their coin is pegged (e.g., US dollar or gold). They issue new units as they expand
their underlying assets.7
The valuation of crypto-collateralized coins is maintained through over-collateralization
and stability mechanisms. In the case of Dai, smart contracts called Collateralized Debt
Positions bring Dai into circulation. You can only retrieve your collateral by paying back
the debt.8
Non-collateralized stable coins use algorithms to dynamically expand and contract the
supply of tokens to maintain a predetermined peg. There are a number of prominent fail-
ures: SagaCoin (SAGA) which replicated the IMF SDR;9 Havven (HAV), now rebranded as
1
Aaron Black, ”Meet the New Payment Champions Same as the Old Ones, Wall Street Journal, January
11, 2019.
2
https://paymentdepot.com/blog/average-credit-card-processing-fees/
3
https://www.cardfellow.com/blog/debit-card-transaction-fees/
4
https://venmo.com/about/fees
5
https://www.nacha.org/news/strong-growth-continues-ach-network-volume-climbs-nearly-10-second-quarter-2021
6
Association for Financial Professionals, https://www.afponline.org/docs/default-source/
default-document-library/pub/2015-payments-cost-benchmarking-report
7
https://academy.binance.com/en/glossary/stablecoin
8
Why is Dai Stable?, https://medium.com/icovo/why-is-dai-stable-9a9fa84feca7
9
https://www.sogur.com/

1
Synthetix;10 and Basis,11 call into question the viability of the mechanism.
Table 1 contains the top ten Ethereum network stablecoins, pegged to the U.S. dollar,
ranked by market cap as of the end of the second quarter of 2021.12

Table 1: Top Ten Stablecoins by Market Capitalization

Name Symbol First Transaction Collateral Market Cap $bn.


Tether USDT 2017-11-28 Asset backed $30.91
USD Coin USDC 2018-09-10 Asset backed $24.33
Binance USD BUSD 2019-09-10 Asset backed $10.06
Dai DAI 2019-11-13 Defi $5.15
TrueUSD TUSD 2019-01-04 Asset backed $1.18
Paxos Standard PAX 2018-09-10 Asset backed $0.84
Huobi USD HUSD 2019-07-20 Asset backed $0.57
Terra USD UST 2020-12-03 Defi, algo stabilized $0.34
Gemini Dollar GUSD 2018-09-09 Asset backed $0.23
Synth USD sUSD 2020-05-11 Defi, algo stabilized $0.15
The market capitalization is based on the circulating supply on the Ethereum network at the
end of 2021Q2.

Ethereum technology has come to dominate the world of smart contracts, and this has
led most of the leading stable coins to adopt the ERC-20 (Ethereum Request for Comments-
20) standard, originally proposed by Fabian Vogelsteller in November 2015.13 The standard
makes tokens interoperable, and it requires that the total supply of tokens and the balance
of any address be transparent to the network.
Etherscan reports14 over 904 tokens based on the ERC-20 standard. It also lists more
than 199,000 verified smart contracts.15
The Ethereum network is more transparent than the interbank or equity market networks.
Any participant in the network can see all of the transactions in and out of Ethereum
stablecoins at the level of a blockchain hashtag.

1.1 Tether
Tether was created by iFinex, the same company that runs the Bitfinex exchange and was
originally called “RealCoin.” It was launched on the Bitcoin blockchain using the Omni Layer
Protocol on October 6, 2014.16 . The token I study here is an ERC-20 token that was issued
10
https://blog.havven.io/
11
https://www.coindesk.com/basis-stablecoin-confirms-shutdown-blaming-regulatory-constraints
12
Several tokens including Tether are issued on more than one network. I am showing the supply on the
Ethereum blockchain.
13
https://ethereum.org/en/developers/docs/standards/tokens/erc-20/.
14
https://etherscan.io/tokens Retrieved on December 11, 2021.
15
https://etherscan.io/chart/verified-contracts Retrieved on December 10, 2021
16
Blockdata (2018)

2
after a $31 million hack in late 2017.17 I first detect the token on the Mainnet on November
28, 2017. As of October 2021, 49.6% of Tether is on the Tron network, 46.7% is on Ether,
1.8% is still on Ommi, 1.6% on Solana, and 0.3% is on EOS, Algorand, Liquid, and SLP.18
The company originally represented Tether as something like a currency board: holding
“one U.S. dollar (“USD”) in reserve ‘backing’ the tether... Every tether is always backed
1-to-1, by traditional currency held in our reserves. So 1 USDT is always equivalent to 1
USD.”
In March 2019, Tether stepped back from this claim instead reporting that19 “Every
tether is always 100 percent backed by our reserves, which include traditional currency and
cash equivalents and, from time to time, may include other assets and receivables from loans
made by Tether to third parties, which may include affiliated entities...”
The New York Attorney (NYAG) pursued a fraud case20 against both Bitfinex and Tether,
documenting a series of questionable bank transfers. The NYAG has banned trading with
New York persons and entities and fined the company $18.5 million. The Commodity Futures
Trading Commission also fined iFinex an additional $41 million for misrepresentation of their
collateral.21 Griffin and Shams (2020) have shown that Tether’s price was manipulated to
support Bitcoin.
In accordance with the ruling by the NYAG, Tether released a breakdown of their reserves
on March 31, 2021.22 Tether holds 75.85% of its assets in cash and cash equivalents, 12.55%
in secured loans, 9.96% in corporate bonds. The statement was not backed by an auditor.
None of this has slowed the growth of the Tether stable coin. Between March 2019 and
June 2021, the market cap grew on the Ethereum network grew from $2.04 billion to $30.91
billion. Tether represents approximately 42% of stablecoin market capitalization and 39%
of all stablecoin transactions.

1.2 USD Coin


USD Coin is a stable coin developed by the fintech company Circle and the digital asset
exchange Coinbase. Coinbase originally claimed that each USD Coin is backed by U.S.
dollars in “a bank account.”23 Following an an auditor’s reportby Grant Thornton on July
16, 2021, Coinbase modified this claim to say that “Each USDC is backed by one dollar
or asset with equivalent fair value, which is held in accounts with US regulated financial
institutions.” Based on Grant Thornton’s report, USDC is 61% cash and cash equivalents
(which includes bank accounts, money market funds and other securities with less than 90
days maturity), 13% Yankee CDs, 12% U.S. Treasuries, 9% commercial paper, 5% corporate
bonds, and 0.2% municipal and U.S. agency bonds.
17
https://tether.to/2017/11/
18
These data were retrieved from wallet.tether.to/transparency on October 11, 2021.
19
https://investingnews.com/daily/tech-investing/blockchain-investing/tether-admits-not-fully-backed-us-dollar/
20
James (2021)
21
https://www.cftc.gov/PressRoom/PressReleases/8450-21
22
https://tether.to/wp-content/uploads/2021/05/tether-march-31-2021-reserves-breakdown.pdf
23
https://www.coinbase.com/usdc:“Each USDC is backed by one US dollar, which is held in a bank
account. Using the Internet time machine, I can verify this claim was on their website until July 20, 2021.

3
On August 22, 2021, Circle announced that “it will now hold the USDC reserve entirely in
cash and short duration US Treasuries. These changes are being implemented expeditiously
and will be reflected in future attestations by Grant Thornton.”
USD Coin, with 36% of transaction volume, along with Tether’s 39%, dominate the
stablecoin market.

1.3 Binance USD


Binance USD (BUSD) is a USD-denominated stable coin approved by the New York State
Department of Financial Services (NYDFS) launched in September 2019 in partnership with
Paxos and Binance.
The NYDFS maintains a “green list” of approved virtual currencies.24 At the moment,
the list includes thirteen currencies including Bitcoin, Ethereum and Litecoin. The two
leading stable coins on the list are Binance USD and Paxos (both standard and gold are
approved).
Binance USD is backed by U.S. dollar deposits which it claims are in FDIC insured U.S.
based institutions.25 Paxos is the custodian and issuer of the BUSD.

1.4 Dai
Bahachuk (2020) notes that Dai was conceptualized by a group called MakerDAO. It is
built on a decentralized Ethereum technology called the Maker Protocol. DAO stands for
Decentralized Autonomous Organization. The DAO holds four types of collateral, Ethereum,
Basic Attention Token (BAT), wrapped Bitcoin, and USD Coin. There are strong financial
incentives to keep the currency over-collateralized.
Kozhan and Viswanath-Natraj (2021) model the DAO stabilization mechanism both
theoretically and empirically. They find that while the Dai price covaries negatively with
risky collateral, the introduction of USDC as collateral has led to an increase in peg stability.

1.5 TrueUSD
TrueUSD provides real-time reports on the state of their collateral through the fintech firm
Armanino.26 At the time I first tried to retrieve the report in March 2021, the auditing firm
supplied a “ripcord” message, indicating that they had not been able to verify the collateral
for more than 72 hours. In private communication with Armanino, I received the following
reply: ”TrueUSD was sold to a 3rd-party and therefore the AT-C 205 reporting has ceased
until the new owners have completed the onboarding process. All other TrueCurrencies are
still owned by TrustToken and have live TrueCurrencies.”27
24
https://www.dfs.ny.gov/apps and licensing/virtual currency businesses/virtual currencies
25
https://www.binance.us/en/busd
26
https://www.armaninollp.com/software/trustexplorer/real-time-audit/
27
Communication between the author and Patrick Clancy, Senior Manager for Strategic Growth, March
8, 2021.

4
Since at least early June 2021, Armanino has resumed attestation reports of True USD.
The report of June 15, 2021 states that the new owner of is Techteryx, Ltd. based in
Shenzhen. As for collateral, the audit states: ”The USD balance, held in escrow accounts
in U.S.... and Hong Kong depository institutions, includes USD cash and cash equivalents
that include short-term, highly liquid investments of sufficient credit quality that are readily
convertible to known amounts of cash.”

1.6 Paxos
Paxos comes in two flavors, the standard and the gold backed version. I discuss the standard
Pax here. Pax maintains USD domiciled depository accounts at a 1:1 ratio with the number
of Paxos outstanding. The collateral includes both cash and U.S. Treasuries. Paxos Trust
Company has engaged Withum,28 a New Jersey based auditing firm, to provide monthly
statements about tokens outstanding and the collateral holdings. The technology is based
on the ERC-20 standard. On August 24, 2021, Paxos Standard was rebranded29 as Pax
Dollar (USDP).

1.7 Huobi
The Huobi USD is a stable coin pegged to the U.S. dollar and backed by USD in reserve ac-
counts.30 Attestation reports31 are provided by Eide Bailly, which reports that the company
is based in the British Virgin Islands. As of December 2021, going to the Huobi USD web-
site comes with this warning: “Currently, individual / institutional clients from Mainland
China, the United States, Iraq, Cuba, Iran, Sudan, Syria, Bangladesh, Ecuador, Tunisia,
Libya, Venezuela, etc. are not available to use this service.”

1.8 Terra USD


Terra USD is a USD pegged stablecoin, collateralized by a digital coin Luna. It is algorith-
mically stabilized, requiring destruction of one dollar of Luna for every newly minted Terra
USD token.32

1.9 Gemini Dollar


Gemini may be best known for its founders, the Winklevoss brothers. Gemini is on the green
list of the New York State Department of Financial Services, and guarantees a portion of
reserves through the FDIC. Gemini is audited monthly by BPM LLP, and the Ethereum
28
https://www.withum.com/our-locations/
29
https://www.paxos.com/the-digital-dollar-that-always-equals-a-dollar-paxos-standard-pax-is-now-pax-dollar-usdp/
30
https://www.huobi.com/en-us/usd-deposit/
31
https://www.stcoins.com/attestation/file?id=29
32
Do Kwon, “Announcing TerraUSD (UST)— the Interchain Stablecoin, https://medium.com/
terra-money/announcing-terrausd-ust-the-interchain-stablecoin-53eab0f8f0ac, September 21, 2020.

5
smart contract has been audited by Trail of Bits, a leading information security research and
development firm.33
Gemini has the most transparent disclosure34 of where it holds its’ assets:“The Gemini
dollar accounts are held and maintained by State Street Bank and Trust Company and
within a money market fund managed by Goldman Sachs Asset Management, invested only
in U.S. Treasury obligations.”

1.10 Synth sUSD


sUSD is a token made by the Synthetix team35 which tracks the price of USD. Users provide
collateral when creating sUSD. It can also be mined.36
Synthetix requires all synthetic tokens, or synths for short, be backed at a collateralization
ratio (C-Ratio) of 500%. Synthetix reported an aggregate collateralization ratio of 553% on
June 16, 2021.37
I was not able to determine the current holdings backing the synths. As of December
2021, Synth reports that the debt pool is in “Maintenance Mode. Data is out of date and a
new version is coming soon.”

Summarizing, the top three stablecoins hold USD, Treasuries and other cash equivalents.
There are varying degrees of transparency about where the assets are held. Only one of the
top ten (Gemini) names the bank where deposits are held. Two of the top three, USD Coin
and Binance, provide (at least partial) FDIC insurance. Dai, Terra USD and sUSD hold
digital assets as collateral.
Lyons and Viswanath-Natraj (2020) make the analogy between stablecoin projects and
currency pegs. They argue that even well collateralized stablecoins may face instability.
Grobysa et al. (2021) find that lagged Bitcoin volatility Granger-causess stablecoin volatility.
Hoang et al. (2021) show that Tether returns are highly correlated with BTC, even at the
daily frequency. For these reasons, I turn to survival analysis in the next section.

2 Survival Analysis
The analysis of digital assets has been influenced by the spectacular returns of a handful of
successful assets, most notably Bitcoin (BTC).
I construct a complete set of token transactions on the Mainnet. Between 2016 and the
frist quarter of 2021, there are more than 260,000 distinct tokens on the network.
To focus the analysis on the active tokens, I only include the 5,143 tokens with at least
10,000 transactions. I graph the growth in Figure 1.
33
https://www.gemini.com/dollar
34
https://assets.ctfassets.net/jg6lo9a2ukvr/2efRs3l0q3EKSv1WPwSvqe/871de898c778d2499d6a11acf697ae71/
Gemini Dollar Examination Report 01-29-21.pdf
35
https://synthetix.exchange/
36
https://defipulse.com/usd/SUSD
37
https://stats.synthetix.io/

6
Figure 1: Ethereum Active Digital Tokens Outstanding

Note: Issue dates are based on the token’s first transactions on the Mainnet. I only include
tokens with at least 10,000 cumulative transactions.

Gorton and Zhang (2021) compare these developments to the free banking era that
began in 1830s. While Gorton and Zhang claim that private currencies did not contribute
to banking panics, they do concede that ”varying discounts made actual transactions (and
legal contracting) very difficult...There was constant haggling and arguing over the value of
notes in transactions.” The National Bank Act of 1863, they note, effectively ended the era
of privately issued bank notes.
I also find that many tokens have failed, particularly those started in recent years. I plot
a hazard function in Figure 2.

7
Figure 2: Hazard Function for Ethereum Tokens

Note: I classify a project as dead if volume falls below 1% of the peak quarterly volume.

There were 1,953 active tokens outstanding at the end of the first quarter of 2021. The
peak year for issuance was 2018 when 1,995 tokens were issued.
With stable coins, the effects of survivorship are much more concerning because they are
not designed to provide any capital gains.
From CoinMarketCap, Nomics, and CoinCodex I identify 91 active and inactive stable-
coin projects. 65 of the 91 transact on the Ethereum Mainnet. These 65 are the most active
and include all the stablecoins in the top 5,000 in market capitalization. I graph the growth
of Ethereum stablecoins in Figure 3.

8
Figure 3: Ethereum Stablecoins Outstanding and New Issues

Note: The projects are from CoinCodex, CoinMarketCap and Nomics. The vintage date is
from the first transaction on the Mainnet.

I plot survivor probability in Figure 4 for the 65 ERC-20 stable coins on the Mainnet.
The vintage is the first year that the token transacts on the Mainnet.

9
Figure 4: Stablecoin Hazard Function

Note: The projects are from Blockdata, CoinCodex, CoinMarketCap and Nomics. The vintage
date is from the first transaction on the Mainnet. I classify a project as dead if volume falls
below 1% of the peak quarterly volume.

DigixDao (DGD) and Xaurum (XAU) are the failures from the 2016 vintage. Both
were backed by gold reserves. There were six tokens in the 2017 vintage. Three are still
active: Monolith (TKN); Noku (Noku); and Tether. There were 17 coins in the 2018 vintage
including Paxos, USD Coin, and Gemini from our top ten. There were also seven failures,
including Pecunio (PCO), and TrueVND (TVND).
2019 was the peak year of new tokens with 19. Binance USD, Dai, Huobi and True USD,
in our top ten, launched that year. Eight coins have failed. Issuance slowed in 2020, with
sixteen new stablecoins. sUSD and Terra USD are part of that vintage. Two tokens have
already failed from the 2020 vintage: Trust Token (TRU) and Xank (XANk).
Jalan et al. (2021) study the gold based stablecoins. The authors find that while gold was
stable during the COVID crisis, the gold-backed digital assets were as unstable as Bitcoin,
despite have no formal or statistical linkage to the leading digital asset.
The hazard functions understate the failure rates of stable coins since many projects never
move past the initial offering stage. Using data from Blockdata (2018), I can identify 103
projects from June 2019. At the time, 67 projects were active, with more than 30 projects
under development. One of those projects, still not live, is the Libra coin proposed by
Facebook. Libra has been rebranded as Diem and has plans to “... support single-currency

10
stablecoins (e.g., USD, EUR, and GBP) and a multi-currency coin (XDX).”38 Facebook has
launched a digital wallet called Novi that currently operates with Pax Dollar (USDP).39

3 Transaction Volume
I report daily estimates of transactions in each of the top ten stable coins for the second
quarter of 2021.

Table 2: Transactions Volumes for Top Ten Stablecoins

Token Name Transaction Volume ($bn) No. Transactions (mn)


Tether USD 692.94 15.3603
USD Coin 628.44 6.4196
Dai Stablecoin 198.95 1.8601
Binance USD 166.10 0.2121
Huobi USD 31.39 0.0424
Paxos Standard 16.93 0.0691
TrueUSD 14.57 0.1185
Synth sUSD 10.94 0.1002
Terra USD 6.43 0.0825
Gemini Dollar 3.21 0.0828
The totals are for Ethereum network transactions in 2021Q2.

There are $1.769 trillion in stablecoin transactions in the second quarter of 2021. Tether
has a 39.2% market share, USD Coin 35.5%, Dai 11.2%, and Binance 9.4%. All of the other
coins have less than a 1% market share.
I graph the growth of stablecoin transactions in Figure 5.
38
https://www.diem.com/en-us/vision/
39
https://www.novi.com/

11
Figure 5: Stablecoin Quarterly Transaction Volume

Note: The data are from the Ethereum Mainnet The legend is in order of 2021Q2 market share.

Over the last year, top stablecoin transaction volume has risen 1,123%, from $157.56
billion in 2020Q2 to $1,769.90 billion in 2021Q2. The fastest growth rate was for Synth
USD, 5,833%. Among the top four coins, USD Coin grew the fastest in the last year,
3,674%, to nearly equal the total for Tether.

3.1 Velocity
I take the transaction volume and divide by circulating supply over time. I have this infor-
mation from the Mainnet which I compute at the last block for the end of the quarter. I
plot velocity in Figure 6.

12
Figure 6: Stablecoin Velocity

Note: The data on supply is just for tokens circulating on the Ethereum Mainnet.

Dai has the highest velocity among the four leading stablecoins, 38.6. USD Coin is second
at 25.8, and Tether is third, at 22.4, and Binance USD is fourth at 16.5. The four token
(unweighted) average is 25.8. By comparison, the M1 velocity in 2019Q4 was 5.5, the last
quarter not impacted by Covid-19.

3.2 Network size


The Mainnet provides counterparty information from the transaction initiator and recipients.
I define a counterparty as a wallet or smart contract and include internal transfers in the
totals. I calculate the number of unique counterparties each quarter for each of the tokens.
The time series is plotted in Figure 7.

13
Figure 7: Stablecoin Network Size

Note: The data are from the Ethereum Mainnet and include wallets, smart contracts, and
internal transfers.

In the second quarter of 2021, Tether transactions involve 3.8 million unique counterpar-
ties. USD Coin has 1.6 million. Dai has 355,000. Binance USD has 77,900. The remaining
tokens have fewer than 50,000 counterparties.

3.3 Holders
The Mainnet allows us to compute the holdings of any ERC-20 token by wallet address.
Figure 8 shows the distribution of Tether holdings on March 15, 2021.

14
Figure 8: Tether Top Holders

Note: The graphic is from Blockchain.info for March 15, 2021.

Note that the top holder is just a network address node, 0xA929...490C. Other large hold-
ers include exchanges, Huobi, Binance (using at least three nodes) and Bittrex. There are
also three decentralized exchanges, Curve.fi40 , Sushi Swap41 and Uniswap.42 Fritsch (2021)
notes the two key aspects of decentralized exchanges (DEX): (1) they allow users to retain
custody of their tokens; and (2) they utilize automated market makers. He estimates $500
billion in trading volume for all decentralized exchanges in the first half of 2021. Park (2021)
discusses flaws in the automated market making mechanism on most DEX, and estimates
they may be raising transaction costs as much as 50 basis points on a substantial fraction of
trades.

To provide comparison across the top stablecoins, I report the largest holder and Herfind-
ahl (HHI) index for the ten largest stablecoins in Table 3.
Gemini, Paxos, and Huobi USD have the most concentrated holdings. For all three, the
largest holder has more than 50% of the issue. Gemini’s top holder is Curve.fi. Paxos
and Huobi’s largest holders are unnamed addresses, 0xe62...d4819b0322a for Paxos and
0xA9290...0D1e490C for Huobi.
40
https://curve.fi/rootfaq: “Curve is an exchange liquidity pool on Ethereum designed for (1) extremely
efficient stablecoin trading (2) low risk, supplemental fee income for liquidity providers, without an opportu-
nity cost. Curve allows users (and smart contracts like 1inch, Paraswap, Totle and Dex.ag) to trade between
DAI and USDC with a bespoke low slippage, low fee algorithm designed specifically for stablecoins and earn
fees. Behind the scenes, the liquidity pool is also supplied to the Compound protocol or yearn.finance where
it generates even more income for liquidity providers.”
41
https://docs.sushi.com/: “SushiSwap is an automated market-making (AMM) decentralized exchange
(DEX) currently on the Ethereum blockchain.”
42
https://uniswap.org/faq/: “Uniswap is a protocol for creating liquidity and trading ERC-20 tokens
on Ethereum. It eliminates trusted intermediaries and unnecessary forms of rent extraction, allowing for
fast, efficient trading. Where it makes tradeoffs decentralization, censorship resistance, and security are
prioritized. Uniswap is open-source software licensed under GPL.”

15
Table 3: Concentration of Stablecoin Hold-
ings

Stablecoin Max Share HHI


Gemini USD 76.12% 0.5837
Paxos Standard 69.15% 0.4806
Houbi USD 56.97% 0.4377
Binance USD 43.94% 0.2993
Synth sUSD 19.37% 0.0969
Terra USD 15.46% 0.0645
Dai Stablecoin 15.91% 0.0418
True USD 8.84% 0.0219
USD Coin 6.62% 0.0155
Tether 4.71% 0.0058
I retrieve from Etherscan all wallets with
more than 1% of the stablecoin on March 15,
2021.

I compare my stablecoin HHI estimates to those in the banking industry for context. At
the county level, Meyer (2018) found an average HHI of 0.3468. All of the tokens except
for Gemini, Paxos, and Huobi are below that level. Based on Department of Justice (DOJ)
merger guidelines, anything above 0.18 is “highly concentrated”. Mergers which raise the
HHI by 0.02 are also considered worth reviewing. By DOJ standards, six of the top ten
tokens are “not concentrated.”

4 Fees and Interest


Transferring ownership of a stablecoin requires intermediaries called miners to update the
blockchain network. The fees they earn are transparent to all network participants.

4.1 Aggregate transfer fees


A transfer begins with two counterparties, identified by addresses, on the blockchain, with
the owner requesting to process a transfer. I will use a real example to help understand the
nomenclature:
In the example in Table 4, wallet 0xb3eb794a375d802876f67f59d5494b2078f0bdd8 initi-
ates a transfer of 99.742115 Tether (USDT) to wallet 0x32034114ac386374d2f3e3057d61fdc3222c49ee.
Once the transaction begins, a hash number is assigned to the transfer. All of the information
in the example is public to the network, and I have retrieved it from the Mainnet.
The transfer request is publicized on the network and miners must make the effort to in-
sert the transaction into the blockchain. The miners compete and pricing varies considerably
with network congestion.

16
Table 4: Tether Transaction Example

Field Entry
Timestamp Mar-21-2021 12:00:40 AM +UTC
From 0xb3eb794a375d802876f67f59d5494b2078f0bdd8
To 0x32034114ac386374d2f3e3057d61fdc3222c49ee
Contract 0xdac17f958d2ee523a2206206994597c13d831ec7 (Tether)
Tokens Transferred 99.742115
Transaction Hash 0xd79cf3fea1ca4ddaac0f42f98c496159dbe8be1582c17ffbf146096573a373ef
Block 12078834
Gas Price (Ether Gwei) 140.4887
Miner 0xd224ca0c819e8e97ba0136b3b95ceff503b79f53 (UUPool)
Transaction Fee (Ether) 0.0078967293383
USD/ETH price $1,783.94 (at timestamp)
Fee in USD $14.08729134

The transfer was included in Block 12078834. The miner UUPool completed the block
in forty seconds. It charged a fee (priced in ether) using what is called the gas price. This
name reflects the fact that the miner must use computer time to incorporate the transfer
into the blockchain.
Gas prices are typically quoted in Gwei, billions of wei, the smallest divisible unit of
ether. The gas price for this transaction was 140.4887 Gwei, or 140.4887 ×10−9 in ether
terms. The transaction used 56,209 units of gas, making the ether cost of the transaction
7.90×10−3 ether.
I plot historical daily average ether gas prices in Figure 9.

17
Figure 9: Average Ether Gas Prices Jan. 1, 2018-June 30, 2021

Note: The data are from Etherscan, https://etherscan.io/chart/gasprice.

Average prices have ranged from 7.32 to 709.71 Gwei between January 2018 and June
2021. There is also substantial intra-daily fluctuation not captured in this figure. On March
14, 2021, for example, gas prices ranged from 70 to 600 Gwei.
The USD price of ether at the timestamp of the transaction completion in Table 4 was
$1,783.94, so the dollar cost of the transaction was $14.09.

4.1.1 Internal transactions


Internal transactions are transactions between smart contracts. ERC-20 tokens are them-
selves smart contracts, but a chain of token transfers may include numerous internal transfers
with additional intermediaries.
I discuss an example in Table 5. The transaction is two swaps, on the largest decentralized
exchange, Uniswap, that result in the exchange of Gemini Dollar for USD Coin.
In calculating the network size, I include the wallet that supplies Gemini USD to the
swap pools, the automated market maker smart contracts, and the wallet receiving the USD
Coin. I will average the gas fee across all of the steps of the transaction.
These assumptions have little impact on the fees for Tether and USDC. Only 0.2% of
Tether and 1.4% of USD Coin transactions involve multiple steps like in Table 5. For Dai,
while the number is higher, 10.8%, it has little effect on the distribution of transaction fees.

18
Table 5: Internal Transactions

Field Entry
Transaction Hash 0x00085bce0480ffae3a717aa1af9f72cecb7566b2ed8ce4cb50b4e8cc225eebe2
Date Jan-19-2021 07:54:01 AM
Transaction Action:
Swap 10,623.08 GUSD for 7.87795 Wrapped Ether On Uniswap V2
Swap 7.87795 Ether For 10,779.054879 USDC On Uniswap V2
Gas Price (Ether Gwei): 132.28
Transaction Fee (Ether): 0.06074
USD/ETH price: $1,367.65 (at timestamp)
Total Fee in USD: $83.07

I aggregate fees over the last three years for all the top ten stablecoins and plot them in
Figure 10

Figure 10: Quarterly Transaction Fees for Top Stablecoins

Note: I compute fees in Ether from the Mainnet and convert them to USD using daily 12:00
GMT Ethereum prices.

The total fees from 2021Q2 are $497.74 million and are dominated by three coins. USDT
generates $268 million in fees (53.9%), USDC $158 million (31.8%), and Dai, $52 million
(10.6%). The other seven coins only generate $18.7 million in fees, led by Synth sUSD at
$5.0 million.

19
4.2 Cross-section of fees
My examples show that transaction costs can vary substantially. Since block insertion time
is influenced by the transfer fee, both costs and time to completion can be hard to predict.
The cost of transfers has also been rising over time because of network congestion and the
upward trend in Ether prices.

4.2.1 Tether fees


I begin the discussion on fees for Tether transactions because it is by far the largest by
market capitalization.
To identify a large selection of representative transactions, I first gather the 19 network
addresses that hold 1% or more of Tether on March 15, 2021. I then search for all transfers
involving those counterparties on the Mainnet between October 1, 2018 and June 30, 2021.
I wound up with a sample of 11.78 million transactions. There are 1.34 million unique
senders and 2.30 million unique recipients. The most frequent sender is the Huobi exchange,
0x1062a747393198f70f71ec65a582423dba7e5ab3. This address initiates 1.68 million (14.2%)
of the transactions. The most frequent recipient is the Binance exchange, which receives
2.31 million (19.6%) of the transfers.
Summary statistics are in Table 6.

Table 6: Tether (USDT) Transactions

mean std. dev. 1% 25% 50% 75% 99%


Token Value 53,733 2,962,469 9 400 1,575 7,123 500,000
Gas Price (Gwei) 104.46 144.51 9.50 40.00 75.00 135.00 500.00
Gas Used 87,684 109,512 23,465 31,121 56,197 137,195 385,472
Transaction Fee (USD) $12.98 $47.60 $0.05 $0.62 $3.44 $12.62 $123.06
Transaction Fee (%) 0.58 0.97 0.00 0.03 0.20 1.13 67.14
# Obs. 11,783,050

The median fee is $3.44 with a median percentile fee of 0.20%. I plot a histogram of fees
in USD in Figure 11.

20
Figure 11: USDT Transaction Fee Histogram - USD

Note: The histogram describe transaction fees in USD for 11.8 million Tether (USDT) token
transfers from the largest holders of Tether between October 2018 and June 2021.

31% of fees are under $1.00, and 47.6% are less than the average $3.08 fee for out-of-
network ATM transactions.43 There is a long tail though: 13.7% of fees are over $25 and
1.5% have fees of over $100.
The fees in percentage terms are in Figure 12
43
https://www.bankrate.com/banking/checking/checking-account-survey/

21
Figure 12: USDT Transaction Fee Histogram - Pct.

Note: The histogram describes fees as a percent of the transferred amount for 11.8 million
token transfers from the largest holders of Tether between October 2018 and June 2021.

The median fee is 0.20%. Unfortunately, there are bad surprises though. More than
175,000 transactions (0.67%) have fees which exceed the transferred value.
There are economies of scale. Transactions over $10,000 have a median fee of 0.011%.
Transfers under that amount have a median fee of 0.38%.

4.2.2 USD Coin fees


I create a sample of 3.87 million USD Coin transactions between October 2018 and June
2021 from the largest holders of the token as I did with Tether. Descriptive statistics are in
Table 7
Table 7: USD Coin (USDC) Transactions from Largest Holders

mean std. dev. 1% 25% 50% 75% 99%


Token Value 94,155 2,356,790 3 350 1,826 9,969 1,192,906
Gas Price (Gwei) 111.10 210.90 2.00 40.00 80.00 138.01 580.00
Gas Used 212,782 234,922 41,588 133,328 175,654 206,165 1,266,310
Transaction Fee (USD) $30.04 $88.87 $0.05 $4.43 $13.03 $35.75 $237.21
Transaction Fee (%) 0.95 1.17 0.00 0.15 0.84 3.84 196.95
# Obs. 3,874,858

There are 454.2 thousand unique senders and 287.6 thousand unique recipients. The top
sender, with 1.27 million transfers (32.9%), is Uniswap: 0x55fe002aeff02f77364de339a1292923a15844b8.
The top recipient is again Uniswap with 1.19 million (30.8%) of the inbound transfers.

22
Transaction costs are higher for USD Coin than Tether. The median fee is $13.03 with a
median percentile fee of 0.84%. The gas used in the median transaction, 175,654, is nearly
three times the amount for Tether transactions.
I plot a histogram of fees in USD in Figure 11.

Figure 13: USDC Transaction Fee Histogram - USD

Note: The histogram describes transaction fees in USD for 3.8 million token transfers from the
largest holders of USD Coin between October 2018 and June 2021.

Only 17% of fees are below the ATM threshold of $3.00. More than a third of fees are
above $25.00, and 4.6% are above $100.
The fees in percentage terms are in Figure 14

23
Figure 14: USDC Transaction Fee Histogram - Pct.

Note: The histogram describes percentage transaction fees for 3.8 million token transfers from
the largest holders of USD Coin between October 2018 and June 2021.

4% of fees are larger than the transferred amount. Bigger trades (>$10,000) pay a median
fee of 0.044%, but smaller trades have a median of 1.65%.

4.2.3 Dai Stablecoin


I create a sample of 1.8 million Dai transactions between October 2018 and June 2021. This
is a sample from all holders of more than 1% of the stablecoin. There are 143,569 unique
senders, and 123,136 unique recipients. The most active sender and recipient is Uniswap
(0xa478c2975ab1ea89e8196811f51a7b7ade33eb11) which initiates 29.8% of the transfers and
receives 28.9%.
Descriptive statistics are in Table 8

Table 8: Dai Stablecoin Transactions

mean std. dev. 1% 25% 50% 75% 99%


Token Value 94,359 3,925,343 1.33026 313 1,951 13,132 872,241
Gas Price (Gwei) 112.61 335.50 1.01 41.00 78.62 130.00 600.00
Gas Used 274,600 329,336 84,633 128,491 174,296 268,470 1,766,001
Transaction Fee (USD) $33.55 $127.57 $0.08 $4.34 $11.68 $32.51 $333.46
Transaction Fee (%) 0.89 1.16 0.00 0.10 0.65 3.14 469.86
# Obs. 1,811,045

The average Dai transaction is similar in size to USD Coin, 94,350 tokens for Dai versus
94,155 for USDC. Fees are also similar, although the average fee for USDC is about $3.00

24
lower.
A histogram of fees in USD is in Figure 15.

Figure 15: USDC Transaction Fee Histogram - USD

Note: The histogram describes USD transaction fees for 1.8 million token transfers from the
largest holders of Dai stablecoin between October 2018 and June 2021.

17% of fees are below the $3.00 ATM threshold. More than 31% are above $25 and
almost 6% are above $100.
The fees in percentage terms are in Figure 16

25
Figure 16: Dai Transaction Fee Histogram - Pct.

Note: The histogram describes percentage transaction fees for 1.8 million token transfers from
the largest holders of Dai stablecoin between October 2018 and June 2021.

The median fee in percentage terms is 0.65% which is between Tether and USD Coin,
but closer to the latter. Bigger transfers (>$10,000) pay lower fees, 0.0038%, but these are
still higher than the fees for institutional size trades for the USDT and USDC. There are
534 transfers in which fees exceed the transferred amount.

4.3 Interest rates


A major trend in the digital asset space is DeFi, Decentralized Finance, which includes
the decentralized exchanges (DEX) I discussed earlier. The DeFi Pulse reports that, as of
December 20, 2021, over $95.7 billion has been committed to DeFi projects. I list the seven
largest projects with assets of over five billion USD in Table 9

26
Table 9: Largest DeFi Projects

Name Chain Category Locked (USD bn.)


Maker Ethereum Lending $17.88
Curve Finance Multichain DEXes $14.02
Aave Multichain Lending $11.00
InstaDApp Ethereum Lending $10.81
Convex Finance Ethereum Assets $9.99
Compound Ethereum Lending $9.74
Uniswap Ethereum DEXes $8.29
Assets from: https://defipulse.com/ Retrieved on December 20,
2021.

I focus in this section of lending projects in which owners of digital assets can earn interest
by allowing other traders to borrow their coins. A sample of current rates for stablecoins is
in Table 10

Table 10: Defi Lending Rates


Compound Aave dYdX Fulcrum BlockFi Nexo Celsius Bitfinex Poloniex Maker Coinbase
DAI 3.04% 2.77% 0.00% 5.18% 8.50% 8.00% 4.60% 9.13% 2.00%
USDC 2.70% 2.82% 0.00% 6.35% 7.50% 8.00% 8.50% 6.17% 4.00%
USDT 2.50% 2.69% 9.22% 7.50% 8.00% 8.50% 3.87%
Rates from https://defirate.com/lend/. Retrieved on December 20, 2021.

The interest rates are substantial but not guaranteed. A number of fees reduce the actual
returns. Most projects charge a fee on inbound fiat money. Conversion rates to stablecoins
are usually at prices away from the best prices on exchanges. Withdrawals are also limited.
The savings rates in the high single digits on BlockFi, Celsius, and Nexo have attracted a
great deal of attention from investors and concerns from regulators. SEC Chairman Gensler
wrote: “The world of crypto finance now has platforms where people can trade tokens
and other venues where people can lend tokens. The American public is buying, selling,
and lending crypto on these venues...I believe we need additional authorities to prevent
transactions, products, and platforms from falling between regulatory cracks.”44 BlockFi
which also offer credit cards has attracted attention from several state regulators, including
New Jersey which issued a cease and desist order45 on July 19, 2021, preventing them from
opening any new accounts. As of December 20, 2021, BlockFi continues to operate in New
Jersey.
44
https://www.warren.senate.gov/imo/media/doc/gensler response to warren - cryptocurrency
exchanges.pdf
45
https://cointelegraph.com/news/blockfi-faces-regulatory-heat-a-sign-of-possible-crypto-lending-regulations

27
4.4 Speed
The speed in which any transfer gets executed is a function of the price you are willing to
pay. In Figure 17, I show a screen shot from the popular Metamask46 Wallet App.

Figure 17: Price and Speed in the MetaMask Wallet App

Note: The example is from Chris Winfrey,“Gas Spectrum Transactions.,” Medium,


June 17, 2019. https://medium.com/authereum/gas-spectrum-transactions-bd34b65107b It

estimates network congestion and provides expected times for the transaction to complete
based on the amount you are willing to pay. The risk with setting the price too low is that
your transaction may never complete.
Etherscan provides47 a time series of the average transfer speed in the block chain. The
median over 2015-2021 is 14.13 seconds, with a maximum of 30.31 seconds. This data is
helpful, but it does not address the fact that some transactions can take days or weeks or
sometimes never complete. There is also no guaranteed time delivery, regardless of rate.

4.5 Trends over time


Ethereum prices rose dramatically in 2021. They averaged $1,536.97 in 2021Q1 and $2,590.91
in 2021Q2, up from an average of $494.15 in the last quarter of 2020. The increase over the
last year is 1,168%.
As Donmez and Karaivanov (2021) note, miners pay a flat fee for Ethereum transfers
between two wallets. They charge 10 to 20 times more though for smart contract interactions
which include our ERC-20 tokens. Gas fees rise with network congestion, particularly after
the network reaches 90% capacity. As gas fees and Ethereum prices rise, dollar transfer costs
46
https://metamask.io/
47
https://etherscan.io/chart/blocktime

28
rise with them. I graph the trend in transfer costs for the major stablecoins in Figure 18.
The median fee for Tether reached $15.47 in the second quarter of 2021.

Figure 18: Transaction Cost Trends in USD

Note: I use the sample from Section 4.1 for each of the three major tokens and show the rise
in median fees, measured in USD per transfer, over time.

Between the second quarter of 2020 and the second quarter of 2021, median USD fees
on Tether are up 3,628%. USDC median fees are $25.06, up 1,897%, and Dai at $27.21, up
1,724%. These increases reflect higher gas prices, 91% for USDT, 113% for USDC, and 99%
for Dai, as well as the increase in Ethereum prices.
The rise in median fees has made the median transfer cost in percentage terms much
higher. For USD Coin, it exceeded 1% in the 2020Q4 and 2021Q1.

29
Figure 19: Transaction Cost Trends in Pct. of Transfer Size

Note: I use the sample from Section 4.1 for each of the three major tokens and show the rise
in median fees, measured in percentage of the transferred amount, over time.

These rise in transfer fees has already spawned blockchain innovations. The first is that
more transactions are migrating to cheaper and faster blockchain networks. One example is
the the public open source Solana.48 The second is the upgrade of the Ethereum network
to version 2.0.49 The new standard includes a change in the fee structure using Ethereum
Improvement Proposal50 (EIP) 1559 which replaces the first price auction procedure in place
of a base fee. It is too early to tell how these changes will impact stablecoin fees. Catalini
et al. (2020) also raise concerns about the security of new network protocols.

5 Microstructure
The Mainnet had an average transaction time of 13.17 seconds per block51 in 2021Q2 which
is far too slow for low latency trading activity. This largely explains why the vast majority
of trading volume in stablecoins occurs off chain on centralized exchanges.52
I study the five active stablecoins with market capitalization over $500 million and with
$100 million in daily turnover.
48
https://docs.solana.com/introduction. Solana reports block formation time of 599 milliseconds on De-
cember 27, 2021 at 16:48:11 GMT.
49
https://ethereum.org/en/eth2/
50
https://notes.ethereum.org/@vbuterin/eip-1559-faq
51
https://etherscan.io/chart/blocktime
52
New, faster networks like Solana have achieved latencies under a second, and they are attracting a
growing share of the decentralized exchange activity.

30
5.1 Trading volume
I first provide trading volumes across all major currencies and digital asset cross rates in
Table 11. Tether has more than 11,000 market pairs in which there is a combined average
daily volume of nearly $120 USD billion. This is comparable to the average daily turnover
on the New York Stock Exchange.

Table 11: Most Active Stablecoins - All Currency Pairs

Name Symbol No. Market Pairs Avg. Daily Volume (USD bn.)
Tether USDT 11,008 $119.72
Binance USD BUSD 352 6.68
USD Coin USDC 773 $2.49
Dai DAI 379 0.52
Paxos Standard PAX 177 0.09
The table reports the average daily volume across all currency pairs and ex-
changes in the stablecoin for the second quarter of 2021. Source: CoinMarketCap.

I just want to clarify that these transactions are not part of the transfer volume studied
earlier. From the perspective of the blockchain, these trades are internal to a single wallet
and not reported on the Mainnet.
I study the five most active digital asset exchanges. These exchanges have more than
50% of the market share for the USD stable currencies I study. I report market shares in
Table 12.

Table 12: USD Stablecoin Exchange Market Shares

Binance Bitfinex Coinbase Huobi Kraken


Tether 61% < 1% < 1% 18% < 1%
USD Coin 64% < 1% 19% < 1% < 1%
Dai 71% < 1% 16% 19% < 1%
Binance USD 100% NA NA NA NA
Paxos Standard 89% 2% NA 2% NA
The table reports average daily market shares across all currency
pairs in the stablecoin on the exchange for the second quarter of
2021. Source: CryptoCompare API.

31
Figure 20: Trading Volume by Exchange of Major Stablecoins

(a) Tether (b) USD Coin

(c) Dai (d) Pax

Note: Volumes (in billions of USD) are from the CryptoCompare API for all currency pairs
in the stablecoin.

I report the daily trading volume for the four major currencies that trade across multiple
exchanges in Figure 20.
Binance is the most active exchange in Tether, with a daily average volume of $26 billion
in the second quarter of 2021. The only other exchange-currency pair with over one billion
in daily volume is Tether on Huobi ($7.4 billion daily average). USD Coin on Binance at
$512 million has the third largest average daily volume.
The exchanges do occasionally handle much larger volumes. Binance trades more than
$68 billion in Tether on May 19, 2021. All of the other stablecoins have major spikes on that
date across all exchanges.

5.2 Bid-ask spreads


The four pairs I study for Tether are BUSD/USDT on Binance, USDT/USD on Bitfinex,
USDT/HUSD on Huobi, and USDT/USD on Kraken. For USD Coin, I analyze three pairs:
USDC/USDT on Binance, BTC/USDC on Coinbase, and USDC/HUSD on Huobi. I analyze
four pairs of Dai, DAI/USD on Bitfinex, Coinbase, and Kraken and DAI/USDT on Huobi.
Finally, for PAX, I study PAX/USDT on Binance and PAX/USD on Bitfinex.
I next report the median bid-ask spreads for the four currencies by exchange-pair in

32
Figure 21: Median Bid-Ask Spreads (Pct.) by Exchange of Major Stablecoins

(a) Tether (b) USD Coin

(c) Dai (d) Pax

Note: The best-bid or offer is computed from direct API feeds for each exchange.

Figure 21.
All of the exchanges maintain generally stable bid-ask spreads of 0.01% or lower in Tether.
Bitfinex is most often the tightest. With Tether trading near $1.00, this is a cost of 0.0001$
for most transactions.
I consider USD Coin in two categories. On Binance, you exchange USD Coin for Tether.
On Huobi, you are exchanging for Houbi USD. On Coinbase, which sponsors USD Coin and
provides unlimited parity trades53 between the coin and the USD, you are exchanging for
Bitcoin. Binance bid-ask spreads are uniformly at the typical 0.01%. Huobi is frequently
double that at 0.02%. On Coinbase, since you are trading for a risky asset, the bid-ask
spread of 0.03% is similar to the BTC-USD rate.
Bid-ask spreads are much higher for Dai except on Coinbase which has a median average
of 0.017%. Huobi and Kraken are over 0.03%.
Paxos orders are at the 0.01% bid-ask spreads nearly every day on Binance, but it gen-
erally quotes at 0.02% spreads on Bitfinex.
On occasion, the spreads increase significantly though. I plot in Figure 22 the 95th
percentile of bid-ask spreads for each of the currency-exchange groups.
With Tether, even trading on the exchange with the widest spreads, Kraken, a trader
53
https://www.coinbase.com/usdc: “For customers with a US dollar bank account, one USDC can always
be redeemed for US$1.00.

33
Figure 22: 95th Percentile Bid-Ask Spreads by Exchange of Major Stablecoins

(a) Tether (b) USD Coin

(c) Dai (d) Pax

Note: The best-bid or offer computed from direct API feeds from each exchange.

34
would face a spread of 0.02% or more less than 5% of the time. This is quite important
because as Bianchi et al. (2020) show, deviations in Tether from parity impact the out-of-
sample returns on a broad index of digital currencies.
Huobi has occasional break downs in their usual tight spreads for USD Coin. 5% of the
time you would trade against a bid-ask spread of more than 3.31%.
With Dai, on Coinbase, bid-ask spreads occasionally exceed 1%. On Huobi, they some-
times exceed 2.0%. On Kraken and Bitfinex, the 95th percentile spreads never exceed 0.5%.
Pax is quite stable. Spreads reach 0.06% on the thinnest days at Binance and 0.3% on
Bitfinex.
I now turn to measure how much you can trade at the best bid or offer.

5.3 Depth
I calculate depth as

0.5 × [bid depth + ask depth],

in thousands of USD for the five currency-exchange groups. Figure 23 provides data on the
token-exchange pairs.
Tether is by far the deepest market. Binance displays a median depth that averages more
than 8.3 million Tether. The depth has also been increasing in the second quarter of 2021.
Huobi is the second deepest in Tether, but it is not even 20% as deep as Binance. The only
coin with comparable exchange depth is USD Coin on Binance, which has a median depth
which averages 545,000.
Paxos has a depth of 118,000 on Binance and and 97,000 on Bitfinex. Dai, on all four
exchanges I am streaming, Bitfinex, Coinbase, Huobi and Kraken, would see market orders
of more than 10,000 Dai likely breaking through the BBO; the largest depth on those four
exchanges is 9.38 thousand on Bitfinex.

5.4 High frequency trading


High-frequency trading (HFT) has become an important part of the digital asset trading
environment. Aleti and Mizrach (2021) found HFT market share at Coinbase of 15.5% for
Bitcoin. The authors did not examine stablecoins though, and they looked at a different set
of exchanges apart from Coinbase. Recent industry discussion suggests that HFT activity
has increased, with the arrival of equity HFT firms like Jump Trading and DRW. Industry
estimates suggest that 80 to 90% of the trading on Bitfinex is from HFT firms.54 .
One measure of HFT activity is the cancellation to execution ratio. I can compute these
estimates for only a subset of the exchanges, Binance, Bitfinex and Coinbase, where the API
provides accurate timestamps of the orders. These estimates are in Figure 24.
54
https://www.supercryptonews.com/over-80-of-trading-volume-on-bitfinex-is-generated-by-hft-firms/:“From
Jump to DRW, a lot of HFT firms] are diving in headfirst because of a lack of opportunities in the traditional
high frequency, low latency trading,”

35
Figure 23: Median Depth (USD thsd.) by Exchange of Major Stablecoins

(a) Tether (d) USD Coin

(c) Dai (d) Pax

Note: The depth (in thousands of USD) is one-half the sum of displayed depth at the best-bid
and offer computed from direct API feeds for each exchange.

36
Figure 24: Cancellation/Execution Ratios of Major Stablecoins

(a) Tether (d) USD Coin

(c) Dai (d) Pax

Note: Cancellations are computed from direct API feeds from each exchange.

37
Figure 25: HFT Pct. of Order Cancellations

(a) Tether (d) USD Coin

(c) Dai (d) Pax

Note: Cancellations are computed from direct API feeds from each exchange. Thresholds
are 500 milliseconds on Binance and Bitfinex and 50 millseconds on Coinbase.

Cancellation to execution (CE) ratios are under six on Binance for Tether, USD Coin,
and Pax. This is low compared to the typical US large cap equity. The SEC Midas website55
reports a 22-day moving average of 9.06 for the last month of 2020. Along with the narrow
spreads, this suggests liquidity providers face little risk of adverse selection.
Dai has higher CE ratios: nearly 12 on Coinbase and almost 300 on Bitfinex. Pax has a
CE ratio of 5.56 on Binance.56
A second measure is the percentage of cancellations under 500 milliseconds. These are
orders which are lifted (either through trades or cancellation) from the order book before
any non-HFT trader could react to them. I report estimates in Figure 25.
My highest share of HFT activity is 57.7% for Tether on Binance. Tether on Bitfinex
and USD Coin on Binance are in the 30% range. Coinbase has less HFT activity. 18.7% of
USD Coin cancellations are below the HFT threshold.
The HFT shares for Dai and Pax are lower, between 11 and 21%. The highest is at
Bitfinex, 22% for Dai, which is consistent with industry discussion.
There appear to be a variety of motivations for the HFT trading in stablecoins. The first
55
https://www.sec.gov/marketstructure/datavis/ma stocks canceltotrade.html
56
Trading activity is so low for Dai on Bitfinex, I exclude the pair from this figure. There are many days
in which the CE ratio is over 1,000 though.

38
is to arbitrage small price discrepancies on an exchange or across exchanges.57 The second
is as an on-exchange store of value. Getting assets into an exchange to exploit trading
opportunities can be quite slow. Coinbase, for example, requires 35 confirmations for a
token transfer, so it can easily take ten minutes to transfer stablecoins between exchanges.58 .
There is an open debate on whether the HFT activity is stabilizing for the digital asset
markets, see e.g. Alexander and Heck (2020), but it is clear that stablecoins will be a part
of that discussion.

6 Conclusion
I have analyzed the collateral structure of the ten leading stablecoins by market capitaliza-
tion. Seven of the ten coins are backed by fiat assets. Three hold digital assets, and two of
those are algorithmically stabilized.
While this paper has analyzed the successful tokens, failure rates of stablecoin projects
are almost as high for other digital assets. The 2016 and 2017 vintages of stablecoins have
failure rates of 100 and 50%. In total, more than 1/3 of the stablecoins that passed the
offering stage have failed.
I estimate $1.77 trillion in stablecoin transactions in the second quarter of 2021. Tether
leads stablecoin transactions, with a 39% market share. USD Coin is now a close second at
36%. Dai has the highest velocity at 38.6. USD Coin is second at 25.8, and Tether is third
at 22.4.
The Ethereum blockchain provides a great deal of transparency at the level of the hash
tag address. In the second quarter of 2021, Tether transactions involve 3.8 million unique
counterparties. USD Coin has 1.6 million. Dai has 355,000. The remaining tokens have
fewer than 50,000 counterparties.
The Mainnet provides the hashtag identity of the largest holders of any stablecoin. I
compute Herfindahl indices for the stablecoin network. Gemini, Paxos, and Huobi have the
most concentrated holdings. For all three, the largest holder has more than 50% of the
supply.
The total fees from 2021Q2 for the top ten stablecoins are are $497.74 million. USDT
generates $268 million in fees (53.9%), USDC $158 million (31.8%), and Dai, $52 million
(10.6%).
The median transaction fee for Tether is $3.44 or 0.2% of the transferred amount. 0.67%
of transfers have fees which exceed the transferred amount. USD Coin and Dai have higher
median transfer fees, $13.03 for USDC and $11.38 for Dai. For both of these stablecoins,
more than 25% of transactions are in excess of 3% of the amount transferred.
A major obstacle for the Mainnet is that network congestion is raising gas fees. As the
price of Ethereum has risen, so have transfer fees. Median fees for Tether rose 3,628% year-
over-year in the second quarter of 2021, They rose 1,897% for USD Coin, and 1,724% for
57
Ryan Adams, “How to make money trading stablecoins,” lhttps://newsletter.banklesshq.com/p/
how-to-make-money-trading-stablecoins
58
https://suredbits.com/lightning-201-high-frequency-trading/

39
Dai.
With the rapid growth of DeFi projects, reaching $96 billion at the end of 2021, there are
opportunities to earn attractive interest rates in stablecoins. I can only conjecture whether
these rates will stay high if non-stablecoin asset prices collapse again or if regulators will
allow this type of lending to continue.
The last stage of my analysis shows that the major digital asset exchanges provide low
cost, deep markets in the major stablecoins. 24 hour turnover in Tether is over $120 billion.
This is more than the daily turnover of all the FANG stocks (Facebook Amazon, Netflix,
and Google, now Alphabet). It is almost 15 times the daily flow in money market mutual
funds.59
Bid-ask spreads are nearly always $0.0001 on Tether, and the most liquid exchange,
Binance, typically displays more than eight million dollars in depth. USD Coin has similar
spreads, but much smaller depth. The spreads and depth are sufficient that there is active
high frequency trading participation, between 20 and 50%, in many of the liquid tokens.
The Basel Committee on Banking Supervision (2021) has started the discussion on capital
requirements for digital assets. The seven tokens backed by cash equivalents may qualify as
Group 1a assets requiring little or no reserves. The three tokens that hold digital assets would
likely fall into Group 1b with much higher capital set asides. The President’s Working Group
on Financial Markets60 has proposed that stable coin issuers become insured depository
institutions. There is an ongoing debate, see e.g. Malloy and Lowe (2021), about whether
central banks should offer competing digital currencies and how they should be funded. The
instability of the private offerings should impact both discussions.
59
The average outflow from the sector was $533.6 billion in 2020Q4,
https://fred.stlouisfed.org/series/MMMFTAQ027S. This averages to $8.08 billion per day, assuming
66 trading days.
60
President’s Working Group on Financial Markets, and Federal Deposit Insurance Corporation, and
Office of the Comptroller of the Currency (2021)

40
References
Aleti, S., & Mizrach, B. (2021). Bitcoin Spot and Futures Market Microstructure. Journal
of Futures Markets, 41, 194–225.
Alexander, C., & Heck, D. (2020). Price Discovery in Bitcoin: The Impact of Unregulated
Markets. Journal of Financial Stability, 50, Article Number 100776.
Bahachuk, V. (2020). What is the Stablecoin DAI? https://linen.app/articles/what-is-the-
stablecoin-dai
Basel Committee on Banking Supervision. (2021). Consultative Document, Prudential treat-
ment of cryptoasset exposures. https://www.bis.org/bcbs/publ/d519.pdf
Bianchi, D., Rossini, L., & Iacopini, M. (2020). Stablecoins and cryptocurrency returns:
What is the role of Tether? SSRN. https://dx.doi.org/10.2139/ssrn.3605451
Blockdata. (2018). Stablecoins: An overview of the current state of stablecoins. https : / /
blockdata.tech/stablecoin-report
Catalini, C., Jagadeesan, R., & Kominers, S. D. (2020). Markets for Crypto Tokens, and
Security under Proof of Stake. SSRN. https://dx.doi.org/10.2139/ssrn.3740654
Donmez, A., & Karaivanov, A. (2021). Transaction Fee Economics in the Ethereum Blockchain.
Economic Inquiry. DOI:%2010.1111/ecin.13025
Fritsch, R. (2021). Concentrated Liquidity in Automated Market Makers. Archiv.org. https:
//arxiv.org/abs/2110.01368v1
G7 Working Group, T. (2019). Investigating the impact of global stablecoins. https://www.
bis.org/cpmi/publ/d187.pdf
Gorton, G. B., & Zhang, J. Y. (2021). Taming wildcat stablecoins. SSRN. https://dx.doi.
org/10.2139/ssrn.3888752
Griffin, J. M., & Shams, A. (2020). Is Bitcoin Really Untethered? Journal of Finance, 75,
913–64.
Grobysa, K., Junttilab, J.-P., Kolaric, J. W., & Sapkotad, N. (2021). On the stability of
stablecoins. SSRN. https://dx.doi.org/10.2139/ssrn.3764457
Hoang, L. T., Klaus, & Baur, D. G. (2021). How stable are stablecoins? European Journal
of Finance. https://dx.doi.org/10.2139/ssrn.3519225
Jalan, A., Matkovskyy, R., & Yarovaya, L. (2021). Shiny” crypto assets: A systemic look
at gold-backed cryptocurrencies during the COVID-19 pandemic. SSRN. https : / /
papers.ssrn.com/sol3/papers.cfm?abstract id=3796837
James, L. (2021). In the Matter of Investigation by Letitia James, Attorney General of the
State of New York, of iFinex Inc., BFXNA Inc., BFXWW Inc., Tether Holdings Ltd.,
Tether Operations Ltd., Tether Ltd., Tether Intl. Ltd. Respondents. https://ag.ny.

41
gov/sites/default/files/2021.02.17 - settlement agreement - execution version.b-
t signed-c2 oag signed.pdf
Kozhan, R., & Viswanath-Natraj, G. (2021). Decentralized Stablecoins and Collateral Risk.
SSRN. https://dx.doi.org/10.2139/ssrn.3866975
Lyons, R. K., & Viswanath-Natraj, G. (2020). What Keeps Stablecoins Stable? NBER Work-
ing Papers, 27136.
Malloy, M., & Lowe, D. (2021). Global Stablecoins: Monetary Policy Implementation Con-
siderations from the U.S. Perspective. Finance and Economics Discussion Series,
Federal Reserve Board, 2021-20.
Meyer, A. (2018). Market Concentration and Its Impact on Community Banks. Regional
Economist.
Park, A. (2021). The Conceptual Flaws of Constant Product Automated Market Making.
SSRN. https://dx.doi.org/10.2139/ssrn.3805750
President’s Working Group on Financial Markets, and Federal Deposit Insurance Corpora-
tion, and Office of the Comptroller of the Currency. (2021). Report on Stablecoins.
https://home.treasury.gov/system/files/136/StableCoinReport Nov1 508.pdf
Regnard-Weinrabe, B., Vasu, H., & Al Nakib, H. D. (2019). Stablecoins. https://corpgov.
law.harvard.edu/2019/02/10/stablecoins/

42

You might also like