BRL Working Paper Series No.

26

Liquidity shocks, token returns and market capitalization

in decentralized finance (DeFi) markets

Lennart Ante 1, *

1
Blockchain Research Lab gGmbH, Weidestraße 120b, 22083 Hamburg
*
[email protected]

Published: 06 Aug 2022

Abstract: This paper investigates the market reaction to large positive or negative
liquidity shocks on the value of tokens traded on decentralized exchanges (DEXes)
on the Ethereum blockchain. Automated market makers (AMMs) and constant product
markets provide transparent and decentralized ways to directly swap two blockchain
tokens for each other via the use of liquidity pools. Using trade-by-trade data of 2.77
million swaps of 14 different tokens traded on Uniswap v2, v3 and SushiSwap, we
find that the size of sell orders significantly correlates with negative future token
returns, while buy size positively correlates with future token returns. Using an event
study approach, we quantify the market reaction of unusually large sell and buy orders
(top 1% percentile) and identify that the market reaction outweighs the economic
value of the event by a factor of -7.4 for sell orders and +4.4 for buy orders over a
short-span trading window. In the case of sell orders, a high proportion of the
abnormal return is already realized before the event, which indicates informed trading
in the form of arbitrage or frontrunning via Miner Extractable Value (MEV). Looking
at individual crypto assets, we find a mean reassessment of token value following
short sales of up to -0.79% within just one follow-up trade (buy orders up to 0.50%).
The findings indicate that price shocks may have a signaling effect but also that
market capitalization may be an insufficient metric for assessing the liquidity and
valuation of (inefficient) crypto assets. The results suggest multiple challenges for
investor protection in decentralized finance (DeFi) markets.

Keywords: Market efficiency, Liquidity; Informational efficiency, Price discovery,

Asset pricing, Event study; Uniswap

1 Introduction

On January 17, 2022, tokens worth $11.3 million of the blockchain project Olympus DAO (OHM)
were sold on the decentralized exchange (DEX) SushiSwap in a single transaction (Etherscan
2022b). Up to that point, the OHM cryptocurrency had a market capitalization of $1.3 billion. As
a result, the market capitalization of OHM dropped to $900 million over a short period of time
(CoinGecko 2022). The fact that a single transaction with a value of ~0.86% of the implied value
of all OHM tokens caused a market cascade or overreaction of ~ -30% suggests that market
capitalization is not a meaningful metric for the actual liquidity or “value” of cryptocurrencies
and that decentralized finance (DeFi) markets are inefficient.
It should be noted that such large transactions on DEXes are by no means a rarity in DeFi
markets. In fact, there are already services that automatically track transparent markets on
blockchains and report particularly large swaps so that traders or market makers can incorporate
the information into their strategies. One example is the DeFi Sniper project, which has over
32,000 followers on Twitter (DeFi Sniper 2022). However, even if part of the market already
knows about the occurrence and timing of large swaps, the actual market impact or relevance
remains largely unclear.
The trading of tokens on DEXes has increased significantly in recent years, due in part to the
great success of the automated market making (AMM) protocol Uniswap. Uniswap is a system
of smart contracts, i.e., decentrally anchored computer code on blockchains (Ante 2021), that
functions as an automated liquidity protocol using a constant product market maker formula.
More precisely, two assets are anchored in a decentralized liquidity pool and can be swapped
directly with each other. Liquidity providers are incentivized via a transaction fee (Adams,
Zinsmeister, and Robinson 2020; Adams et al. 2021). Traders can directly interact with the
protocol by sending one token to the liquidity pool and receiving the other one in return. The
shifting ratio of the tokens in the pool then determines the new exchange rate between them. The
use of AMM protocols for token trading makes logins, central authentication or the use of
intermediaries superfluous or replaces them with smart contracts, i.e., decentralized trust. The
decentralized nature of DEXes such as Uniswap offers a unique degree of transparency and
certainty of settlement, as all trades and traders’ accounts can be monitored in real-time. This is
not the case for centralized exchanges that self-report trading volume and trades, which can lead
to trust issues due to, e.g., wash trading, as the information cannot be independently validated
(e.g., Le Pennec et al., 2021). Additionally, it is usually not possible to monitor individual
accounts on centralized exchanges.
AMM protocols are a rather recent phenomenon whose capital efficiency is constantly
developing. Older protocols, such as Uniswap v2 or SushiSwap, use the constant product formula
of x ∗ y = k, where x and y represent the reserves in token A and token B in a liquidity pool. If
a user wants to withdraw a number of tokens A from the pool, a proportional amount of token B
must be sent to the pool. Thus, the price of token A can be calculated as the reserve of token B
divided by the reserve of token A. Other protocols, most of which were released at a later date,
use more advanced logic to improve capital efficiency, such as concentrated liquidity (e.g.,
Uniswap v3 or Balancer) or multi-asset pools (e.g., Curve) (Adams et al. 2021; Curve.fi 2022;
Martinelli and Mushegian 2019). A key characteristic of AMMs—especially DEXes employing
a constant product formula—is that the underlying liquidity for the exchange of two tokens is
publicly visible at all times and that swaps are almost exclusively market orders. The fact that
each swap shifts the market price by changing the ratio of the two tokens means that large
transactions can have a significant impact on prices and liquidity. While the price effect of a swap
is immediately obvious, there are currently no precise findings on the extent to which the market

2
interprets (particularly large or significant) swaps as price signals that trigger overreactions, i.e.,
effects beyond the economic impact of the transaction itself.
There are various reasons why supply or price shocks in the form of exceptionally large
cryptocurrency sales or purchases on DEXes could or should cause an abnormal market reaction,
i.e., one that exceeds its actual economic impact. The withdrawal or addition of tokens shifts the
supply curve or the underlying liquidity of the token, thus making future trading more or less
efficient. Accordingly, and in line with Signaling Theory (Spence 1973), this suggests that an
increase in liquidity should represent a positive market signal for traders and a decrease a
negative signal. So if a very strong unforeseen price or liquidity shock occurs, traders will update
their expectations of the token price, which may in turn result in direct buying (more liquidity),
selling (less liquidity) or temporary absence from any activity in the market (uncertainty due to
possible adverse selection) (Beaver 1968; Chae 2005). A higher risk of encountering informed
counterparties reduces the willingness of uninformed traders, such as liquidity traders or market
makers, to participate in a market (Black 1986; Milgrom and Stokey 1982). However, in the case
of liquid assets that are besides DEXes also traded on other (central) trading venues,
extraordinary market reactions also provide an opportunity for arbitrageurs to offset these effects
and exploit market inefficiencies.
The academic literature has examined price shocks and similar unforeseen events in a wide
variety of settings and concluded that they can have both short- and long-term effects on financial
markets. This includes, for example, oil price shocks and stock returns (Hu et al. 2018; Kang,
Ratti, and Vespignani 2016; Cunado and Perez de Gracia 2014), short sales and stock returns
(Aitken et al. 1998) or Bitcoin volatility (Diaconaşu, Mehdian, and Stoica 2022). Other studies
focus on the relationship of on-chain (‘on-the-blockchain’) transactions and financial metrics.
Koutmos (2018) finds that the cumulative amount of transaction activity in the Bitcoin network
has an effect on the returns and trading volume of Bitcoin, and Wei (2018) identifies that
issuances of the stablecoin Tether do not affect returns on Bitcoin but raise its trading volume.
For other stablecoins, too, it has been found that issuances and transactions relate to the pricing,
return and volume of Bitcoin and other cryptocurrencies (Kristoufek 2021; Griffin and Shams
2020; Ante, Fiedler, and Strehle 2021b; 2021a; Saggu 2022). Ante (2020) identifies that large
Bitcoin transfers are part of Bitcoin’s microstructure, as informed traders adjust their
expectations (trading volume) based on the degree of information asymmetry. For large on-chain
transactions of Bitcoin, Ante and Fiedler (2021) further identify different price effects depending
on size and type (initiator/receiver) of transactions.
To our knowledge, there is currently no scientific research that addresses the question of how
positive or negative price shocks on DEXes affect the prices of crypto-assets. This study therefore
aims to investigate and quantify these effects. An accurate understanding of the factor liquidity
is a strikingly important aspect of understanding financial markets and an essential criterion for
the behavior of market participants such as liquidity traders. In traditional financial markets or
centralized cryptocurrency exchanges, order books with bids and asks serve a similar role as
liquidity pools in DeFi markets. The same cryptocurrencies that are traded in centralized order
books are also traded in DeFi markets. Thus, arbitrageurs, for example, need to understand
exactly how these markets interact and differ from each other. Academic research on order book
dynamics and its information content has a long history (e.g., Cao et al., 2009; Chen et al., 2019;

3
Cornelli and Goldreich, 2003), which is why assessing the liquidity and risk factors of assets and
prices is easier on traditional exchanges than it is for novel forms of supply and demand
matching. Accordingly, we see a significant research gap in studying the relationship between
liquidity and prices in DeFi markets. The analysis of price shocks can play a small part in filling
the research gap and contributes to research on DeFi in general, the role of information for
financial markets and cryptocurrency market efficiency. Furthermore, we include the extremely
important cryptocurrency market metric ‘market capitalization’ in our analysis (=market price
multiplied by circulating token supply) to analyze how high relative swaps and market reactions
turn out to be. If the market capitalization metric represents a proper proxy for the realizable
value of a crypto asset, then the relative economic impact of a large swap should be at most as
high as it is large relative to the market capitalization. However, we suspect that this is not the
case, especially for comparatively illiquid cryptocurrencies. It seems reasonable to assume that
the market capitalization metric sometimes puts the actual value of crypto assets much higher
than is realistically the case. Therefore, if the price impact of a large swap is higher than it would
rationally be using market capitalization as a fair proxy for value, we can interpret this as an
indication that our assumption is correct.
To assess the significance of price and liquidity shocks in DeFi markets, we use trade-by-trade
data from three major DEXes (Uniswap v2, Uniswap v3 and SushiSwap) on the largest
blockchain infrastructure for DeFi, Ethereum. These DEXes likely represent over 95% of the
respective DeFi market for the time frame considered. Based on a sample of 2.77 million
individual trades of 14 tokens (i.e., crypto assets), we determine the extent to which the swap
size of buy and sell orders relates to token returns at different intervals. We find that the size of
sell orders is significantly related to negative future token returns, while the size of buy orders
positively relates to future token returns. Using an event study approach, we identify and quantify
the market reaction to unusually large sell and buy orders and find that the resulting short-term
market reaction is 640% higher than the actual swap for sales and 340% higher for buys. For
sales, a high proportion of the abnormal return is realized before the event, suggesting informed
trading in the form of arbitrage via Miner Extractable Value (MEV) (Daian et al. 2020; Strehle
and Ante 2020). Looking at individual crypto assets, we show that token prices on average drop
by -0.79% after sales in the 1% percentile within just one subsequent trade (up to 0.50% for buy
orders). However, the relative size of the swaps is only 0.059% for sells and 0.043% for buys.
The results thus suggest that price shocks on decentralized exchanges may have a signaling
effect, but they also indicate that market capitalization may be an insufficient metric for assessing
the liquidity and value of (inefficient) crypto assets.
This article proceeds as following. Section 2 describes the data collection, the sample and the
empirical approach. Section 3 presents correlations between swap size and token returns as well
as the event study results. Section 4 reflects on the main results of the study, its limitations, future
research paths and implications. Section 5 concludes.

4
2 Data and methods

2.1 Data and sample description

There are thousands of different trading pools for DEXes (especially for Uniswap v2), making it
difficult to select specific pairs for swap data extraction and subsequent analysis. We adopt a
multi-step approach to selecting tokens for our analysis. The first step is the exploratory
identification of tokens with exceptionally large swap activity. For this purpose, we use a Python
script to extract all large swaps published by market intelligence bot DeFi Sniper between June
05, 2021, and January 23, 2022, from their Telegram channel (t.me/defisniper). This yields
29,654 swaps with an average size of $1.16 million (SD = $1.54 million) involving 511 unique
crypto assets.
We then exclude the major reference currencies from the sample. This includes Ether (ETH),
Wrapped Ether (WETH), and their interest bearing (e.g., Aave’s aETH or aWETH), staked (e.g.
Lido’s stETH or Ankr’s aETH) and leveraged (e.g., ETH 2x Flexible Leverage Index; ETH2x-
FLI) variations. Wrapped Bitcoin (WBTC), renBTC and HBTC are likewise dropped from the
sample, as are stablecoins, (e.g., USDT, USDC, MIM, EURT and UST), their interest-bearing
variations (e.g., Compound’s cUSDC), and mirrored stocks such as Mirrored Apple (mAAPL)
and Mirrored Tesla (mTSLA). We individually check the suitability of the remaining tokens for
swap activity analysis based on the following criteria: Is there a high number of unusually large
swaps? Does the token have a high average trade volume and token value locked (TVL), i.e. the
liquidity anchored in AMM pools, on DEXes? These checks leave a sample of 14 tokens, for
which we use subsequently collect data and conduct the analyses: Aave (AAVE), Alchemix
(ALCX), Compound (COMP), Ethereum Name Service (ENS), Chainlink (LINK), LooksRare
(LOOKS), Decentraland (MANA), Maker (MKR), Olympus DAO (OHM), Ribbon Finance
(RBN), The Sandbox (SAND), Shiba Inu (SHIB), Spell Token (SPELL) and Sushi (SUSHI). Of
course, this selection of tokens is subject to some extent arbitrary, but given the lack of research
or approaches to systematically select tokens on this topic, we consider our approach suitable for
such an explorative analysis.
We use Flipside Crypto (flipsidecrypto.xyz) to generate custom application programming
interfaces (APIs) for each of the 14 tokens, which provide us with a historic dataset of all swaps
made on the DEXes Uniswap v2, Uniswap v3 and Sushiswap from May 20, 2021, to March 20,
2022. The number of liquidity pools depends on the token—for example, the highest number of
different pools is 46 for LINK. However, 99% of the volume accrues to a few trading pairs with
ETH or stablecoins as the reference pairs (i.e., LINK/ETH or LINK/USDT). The swap data is
enriched with the circulating supply of each token, which is obtained via the CoinGecko API
(coingecko.com). This data is updated daily (while the swaps are specific to seconds or rather
Ethereum blocks) and is curated by CoinGecko. This means that vested or burned tokens are
excluded from the calculation, which allows a meaningful assessment of the actually circulating
supply of a token. This results in the justifiable limitation that intraday fluctuations or inflation
of the circulating token supply are not accounted for. Accordingly, the price of a token in USD
is multiplied by its daily circulating supply to calculate the market capitalization at the time of
the swap. Table 1 shows an overview of the 14 tokens, their average market capitalization, the
periods under consideration, their swap count and distribution across the DEXes.

5
Table 1. Overview of token samples.

Avg. Swaps Distribution across DEXes

Token market First swap
cap ($b) All Buys Sells Uniswap v2 Uniswap v3 Sushiswap
AAVE 3.751 May 2021 58,503 54% 46% 26% 38% 36%
ALCX 0.224 May 2021 61,443 43% 57% 0% 11% 89%
COMP 1.943 May 2021 59,492 45% 55% 26% 36% 38%
ENS 0.864 Nov 2021 145,356 33% 67% 8% 70% 22%
LINK 10.539 May 2021 186,072 49% 51% 35% 44% 21%
LOOKS 0.525 Jan 2022 317,837 52% 48% 44% 56% 0%
MANA 3.444 May 2021 95,625 52% 48% 36% 31% 33%
MKR 2.449 May 2021 52,022 48% 52% 29% 45% 25%
OHM 1.546 Dec 2021 107,084 55% 47% 1% 11% 88%
RBN 0.116 Oct 2021 36,022 56% 44% 24% 75% 1%
SAND 2.981 May 2021 222,653 59% 41% 72% 28% <1%
SHIB 10.917 May 2021 1,060,915 61% 39% 68% 31% 1%
SPELL 1.001 Jun 2021 174,319 57% 43% <1% 26% 74%
SUSHI 1.761 May 2021 192,334 47% 52% 15% 9% 76%
All 3.004 May 2021 2,769,677 55% 45% 44% 35% 21%

Blocks and thus transactions on the Ethereum blockchain are confirmed on average every 12-14
seconds (Etherscan 2022a). Miners bundle initiated transactions from the Mempool and confirm
them. They can independently determine the order of settlement, which has significant
implications for activity and trading on DEXes. Via MEV, users can bid to be placed in front of
and/or behind specific transactions, which, in conjunction with flash loans (loans that are
reimbursed within the same block), enables applications such as sandwich attacks, frontrunning
or arbitrage (Daian et al. 2020). The ordering of settled transactions therefore represents a direct
sequence and can be considered as a time series (with irregular intervals). So to quantify the
short-term impact of an individual token swap on the price of a token, the calculation must
include the immediate successor swap(s), even if they may have the exact same timestamp as the
actual swap event. We therefore define the sequence of swaps as a time series, with each swap
as a unit of time, and use this basis to calculate log returns. Table 2 shows descriptive statistics
for the log returns of each token.

The statistics clearly show that the log returns are not normally distributed, as the very large
kurtosis indicates thick tails. Ten of the tokens have a negative average return over the respective
periods, four have a positive average return. The fact that the minimum swap-to-swap returns per
token range between -14.38% (MKR) and -68.48% (RBN) illustrates how volatile the tokens are
and that there are significant downward outliers. Likewise, there are outliers on the upside, as
evidenced by the maximum swap-to-swap log returns ranging from 10.53% (LINK) to 38.19%
(ENS).

6
Table 2. Descriptive statistics for token log returns. For the number of observations, see Table 1.

Token Mean (%) SD (%) Min (%) Max (%) Skewness Kurtosis
AAVE -0.0020 0.51 -17.08 13.39 -1.23 199.07
ALCX -0.0042 0.49 -59.08 16.02 -27.37 3,455.01
COMP -0.0032 0.48 -23.06 12.79 -2.93 245.94
ENS -0.0002 0.29 -27.56 38.19 17.16 3,130.91
LINK -0.0003 0.26 -18.26 15.01 -3.31 653.97
LOOKS -0.0003 0.22 -63.05 21.56 -68.69 30,965.39
MANA 0.0005 0.43 -20.69 20.13 1.83 287.26
MKR -0.0019 0.47 -14.38 12.14 -0.29 154.25
OHM -0.0025 0.20 -29.10 12.09 -32.44 5,455.47
RBN -0.0038 0.73 -68.48 25.59 -19.66 2,332.22
SAND 0.0011 0.30 -15.47 17.84 5.80 562.03
SHIB 0.00003 0.14 -16.27 20.04 14.74 3,831.42
SPELL 0.0008 0.53 -36.94 37.73 4.53 1,750.85
SUSHI -0.0007 0.31 -21.82 14.88 -1.84 639.22

For each swap we collect basic metrics like timestamps, transaction ID, direction of the swap
(sell or buy order), pool name, the exchanged number of tokens, and the USD value of these
tokens at the time of the swap. The latter refers only to the side of the swap that contains the
respective token. For example, an exemplary swap of 1 SUSHI for 16.88 USDT only counts for
$16.88 USD, rather than the total volume of the swap ($37.66). From this absolute swap size, we
calculate the relative size by dividing the former by the current market capitalization of the token.
Table 3 shows descriptive statistics for absolute and relative swap sizes for each token sample
and for subsamples based on the direction of the swap. “in” indicates that the token in question
is sent to the pool and is thus being sold for another token, while “out” means that other tokens
are sent to the pool and the tokens under consideration are transferred out of the pool, i.e., they
are being purchased.
On average, the absolute size of the swaps ranges between $11,984 (MANA) and $29,093
(LINK). Half of the tokens considered feature an excess of sales (out transactions). The share of
sales is highest for ENS (67%), which is probably due to its distribution by airdrop (ENS 2022).
On average, sales of ENS were significantly larger than purchases (∆=$13,370***), which
indicates that an overselling of ENS has taken place—at least on the DEXes considered. The
other tokens with statistically significantly larger sells compared to buys show the opposite
picture. LOOKS, MANA, OHM, RBN, SAND, SHIB (the lowest share of sales, at 39%) and
SPELL all exhibit significantly higher average sales than purchases. This could indicate that
smaller (retail/individual) investors buy tokens on DEXes and larger/professional investors or
stakeholders use DEXes to sell tokens. Another picture emerges with tokens like AAVE, ALCX,
COMP or MKR which, have a larger overall share of buys than sells. In relation to the tokens’
market capitalization, ALCX has the largest swaps (average swap of 0.01% of the "total value"
of all tokens / market capitalization) and SHIB the smallest, at 0.00008%.

7
Table 3. Descriptive statistics for token swap size. ‘Swap $’ indicates the mean dollar value of the swaps of a
given token. ‘Swap %’ shows the mean swap size in relation to the crypto assets’ market capitalization in percent.
The ∆ columns show the difference between the sell and buy orders group and the significance level of a Mann—
Whitney U test. The asterisks *** indicate significance at the 1% level.

All Sell orders Buy orders

Swap $ Swap % Swap $ Swap % Swap $ Swap % ∆ Swap $ ∆ Swap %

N N N
(SD) (SD) (SD) (SD) (SD) (SD) (SE) (SE)

AAVE 58,503 26,995 0.0007 31,301 26,870 0.0007 27,202 27,140 0.0007 270*** 0.0001***
(288,650) (0.0071) (389,943) (0.0097) (70,964) (0.0017) (2,393) (0.0006)
ALCX 61,443 23,443 0.0100 35,283 21,122 0.0091 26,160 26,573 0.0112 5,451*** 0.0021***
(107,165) (0.0458) (112,478) (0.0485) (99,467) (0.0420) (874) (0.0004)
COMP 59,492 18,774 0.0010 32,463 17,454 0.0009 27,029 20,359 0.0011 2,905*** 0.0002***
(66,729) (0.0036) (82,907) (0.0044) (39,252) (0.0022) (549) (0.00003)
ENS 145,356 18,068 0.0022 97,535 13,670 0.0017 47,821 27,039 0.0033 13,370*** 0.0016***
(43,649) (0.0052) (38,616) (0.0045) (51,280) (0.0064) (241) (0.00003)
LINK 186,072 29,093 0.0002 95,527 28,931 0.0003 90,545 29,264 0.0003 333*** 0.00003***
(93,705) (0.0009) (92,435) (0.0008) (95,027) (0.0009) (435) (0.00004)
LOOKS 317,837 15,104 0.0041 153,057 15,577 0.0042 164,780 14,665 0.0039 -912*** -0.0003***
(48,264) (0.0143) (50,289) (0.0136) (46,301) (0.0148) (171) (0.00005)
MANA 95,625 11,984 0.0004 45,695 12,366 0.0004 49,930 11,635 0.0004 -731*** -0.00002***
(24,498) (0.0008) (25,025) (0.0008) (23,999) (0.0007) (159) (0.000002)
MKR 52,022 22,789 0.0009 27,117 22,501 0.0009 24,905 23,102 0.0010 601*** 0.00003***
(77,470) (0.0033) (100,283) (0.0043) (39,833) (0.0017) (680) (0.00003)
OHM 107,084 22,481 0.0020 47,826 25,519 0.0023 59,258 20,029 0.0018 -5,090*** -0.0005***
(75,306) (0.0065) (91,729) (0.0079) (58,682) (0.0052) (463) (0.00004)
RBN 36,022 35,320 0.0336 15,891 41,585 0.0395 20,131 30,375 0.0290 -11,209*** -0.0105***
(120,978) (0.1289) (150,171) (0.1593) (91,286) (0.0982) (1,282) (0.0014)
SAND 222,653 12,017 0.0009 92,039 13,241 0.0011 130,614 10,379 0.0007 -3,963*** -0.0003***
(32,710) (0.0023) (35,710) (0.0025) (30,342) (0.0021) (141) (0.000001)
SHIB 1,060,915 7,950 0.00008 410,006 10,193 0.0001 650,909 6,537 0.00007 -3,656*** -0.00004***
(34,533) (0.0004) (39,986) (0.0004) (30,520) (0.0003) (68) (0.0000008)
SPELL 174,319 23,949 0.0062 74,442 27,617 0.0073 99,877 21,215 0.0055 -6,401*** -0.0018***
(58,611) (0.0314) (63,389) (0.0357) (54,620) (0.0278) (283) (0.0002)
SUSHI 192,334 27,046 0.0016 101,516 25,655 0.0015 90,818 28,601 0.0017 2,946*** 0.0002***
(79,659) (0.0044) (84,249) (0.0045) (74,161) (0.0042) (364) (0.00002)

All 2,769,664 15,697 0.0020 1,259,698 17,265 0.0018 1,509,979 17,166 0.0013 -98**** -0.0006***
(58,550) (0.0081) (50,501) (0.0073) (92,517) (0.0067) (579) (0.0002)

2.2 Empirical approach

The analysis comprises two parts. We first investigate the link between relative swap size and
returns and then examine the extent to which particularly large and unexpected buys and sells
trigger a market reaction. For this purpose, we define several time intervals based on swaps to
test the effect of information propagation. These are 1) single-trade windows of t = -1 (the
preceding swap), 0 (the event itself) and +1 (the subsequent swap) to quantify short-run effects,
and 2) multi-swap windows of t = -15 to -1, 0 to 15, 0 to 50, 0 to 100 and 0 to 500, for each of
which we calculate the cumulative returns. Using Spearman rank correlation analysis, we first
identify the extent to which a relationship exists between relative swap size and future returns

8
for sells and buys. This is followed by an event study to calculate abnormal returns around
exceptionally large buys and sells.
In line with MacKinlay’s (1997) essential steps for conducting event studies, we first identify the
events as those “outlier swaps” that rank in the top 1% in terms of relative swap size of each of
the 14 tokens. The same percentile-events are also identified for the sales and purchases
subsamples. In line with, e.g., Demir et al. (2022) or Diaconaşu et al. (2022), the model’s
expected return is calculated over the full sample period (excluding the individual event window
or each event), as swaps occur irregularly and we would be unable to prevent event-specific
estimation windows of the various events from overlapping. We calculate the expected return
(ERit) as the average log return over the estimation period: 𝐸𝑅𝑖𝑡 = ̅̅̅̅̅̅̅̅
𝑅𝚤𝑡 + 𝑒𝑖𝑡 . The term i refers
to a specific event and t denotes the swap within the estimation period. Rit is the token’s observed
log return over t for i. eit is the error term. The bar over Rit denotes the mean over the estimation
window. The difference between the expected and the observed return is the abnormal return
(AR) that can be attributed to the event (Brown and Warner 1985): 𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡 − 𝐸𝑅𝑖𝑡 .
Aggregated across multiple swaps, the cumulative abnormal return can be calculated as:
𝐶𝐴𝑅(𝑡1 , 𝑡2 ) = ∑𝑡𝑡=𝑡2
𝐴𝑅𝑖𝑡 . Since the statistics from Table 2 and the financial literature in
1

general (Brown and Warner 1985) suggest that such data is non-normally distributed, we test the
significance of the results using the non-parametric Wilcoxon sign-rank test (Wilcoxon 1945).

3 Results

Table 4 shows results for the Spearman rank correlation coefficient between the relative swap
size of buy and sell orders and token returns over the eight different event windows. The results
for the two subsamples – buy orders versus sell orders – are almost diametrically opposed, which
suggests that the effects in the overall sample cancel each other out.
With a few exceptions, such as ENS, we find significant negative correlations between the
relative size of sales and future token returns. The opposite holds for purchases, where the
relationship is almost exclusively positive. This finding of widespread significant correlation
between the metrics suggests that very large swaps from the tail of the distribution have a
particularly large impact on token returns. This hunch is tested in the following event study.

9
Table 4. Spearman rank correlations between relative swap size and future token log returns. This table shows Spearman correlation coefficients for sell and buy
orders. Token returns are based on single swaps (t = -1, 0, and +1); five returns relate to aggregate windows of multiple swaps: t = -15 to -1, 0 to +15, 0 to +50, 0 to +100,
and 0 to +500. For the number of observations, see Table 3. An asterisk (*) indicates statistical significance at the 0.1% level.

Sell orders Buy orders

Single swap windows Multi-swap windows Single swap windows Multi-swap windows
Token -1 0 +1 -15 +15 +50 +100 +500 -1 0 +1 -15 +15 +50 +100 +500
swap (event) swap swaps swaps swaps swaps swaps swap (event) swap swaps swaps swaps swaps swaps
AAVE -0.042* -0.023* -0.018* 0.132* -0.063* -0.048* -0.028* -0.031* 0.033* 0.020* 0.014 -0.114* 0.079* 0.093* 0.074* 0.003

ALCX -0.017* -0.007 -0.018* 0.076* -0.040* -0.031* -0.036* -0.037* 0.020* 0.024* 0.008 -0.089* 0.053* 0.083* 0.063* 0.019

COMP -0.033* -0.034* -0.009 0.079* -0.063* -0.046* -0.039* -0.036* 0.024* 0.019* 0.008 -0.085* 0.054* 0.044* 0.020 0.051*

ENS -0.001 -0.003 -0.009 0.014* 0.004 0.018* 0.026* 0.044* 0.009 0.001 0.008 -0.019* 0.036* 0.038* 0.039* 0.041*

LINK -0.007 -0.005 -0.003 0.081* -0.051* -0.053* -0.039* 0.002 0.007 0.006 0.006 -0.093* 0.061* 0.078* 0.073* 0.047*

LOOKS -0.005 -0.004 -0.000 0.005 -0.009* -0.011* -0.013* -0.007 0.002 0.004 -0.002 -0.009* 0.003 0.001 -0.003 0.007*

MANA -0.017* -0.014* -0.011 0.057* -0.037* -0.026* -0.012 0.057* 0.014* 0.015* 0.012 -0.101* 0.062* 0.066* 0.065* 0.082*

MKR -0.028* -0.023* -0.021 0.048* -0.030* -0.009 -0.005 -0.012 0.015 0.013 0.005 -0.045* 0.037* 0.048* 0.039* 0.005

OHM -0.007 -0.007 0.004 0.026* -0.007 -0.028* -0.037* -0.028* 0.001 0.004 0.001 -0.006 0.017* 0.009 -0.000 -0.001

RBN -0.011 -0.017 -0.020* 0.040* -0.055* -0.049* -0.027* -0.000 0.010 0.009 0.009 -0.005 0.042* 0.039* 0.038* 0.030*

SAND -0.015* -0.018* -0.010* 0.053* -0.038* 0.035* -0.007 0.048 0.024* 0.021* 0.011 -0.091* 0.057* 0.074* 0.076* 0.084*

SHIB -0.001 -0.000 -0.002 0.009* -0.006* -0.006* -0.008* 0.009* 0.002 0.002 0.005 -0.016* 0.012* 0.016* 0.019* 0.022*

SPELL -0.013* -0.009 -0.002 0.016* -0.028* -0.010 -0.001 0.040* 0.009 0.004 0.005 -0.029* 0.022* 0.031* 0.026* 0.019*

SUSHI -0.014* -0.014* -0.003 0.073* -0.056* -0.058* -0.047* -0.017* 0.007 0.005 0.005 -0.091* 0.051* 0.073* 0.073* 0.029*

10
The cumulative swap-to-swap abnormal returns for sell and buy order swaps from t = -15 to +15
are presented graphically in Figure 1. Before the actual event (from t = -15 to -1), the cumulative
abnormal return is much lower for sell orders (-0.11%) than for buys (0.009%). This suggests that
some adverse information contained in a sell order is already impounded before the actual swap,
likely due to MEV trading or arbitrage. For buy orders, the cumulative abnormal return remains
relatively stable until the last individual swap before the actual event. At that point, the mean
abnormal return amounts to 0.019% (p < 0.01), which is much higher than at the time of the actual
event (0.008%). It is also striking that the cumulative abnormal returns after the actual event are
also higher: 0.015% in t = +1, 0.016% in t = +2 and 0.018% in t = +3 (all p < 0.01). For a large
buy order, the cumulative abnormal return over the next 15 swaps is 0.12%. For sell orders, it is
-0.01%; however, as the figure shows, the abnormal return drops between the swaps and t = +7
(by 0.07% in total) before rising again by 0.06.

0.20%

Buy orders
0.15%
Sell orders
0.10%

0.05%
Mean CAR

0.00%

-0.05%

-0.10%

-0.15%

-0.20%
-15 -10 -5 0 5 10 15
Swaps pre / post large swap in the 1% percentile (in t = 0)

Figure 1. Mean cumulative abnormal returns (CARs) attributable to large sell or buy orders on a
transaction-by-transaction basis from t = -15 to 15 swaps. The data comprise all 2,769,677 swaps of the 14
crypto assets shown in Table 1. The number of events is 16,586 for buy orders and 13,848 for sell orders (top
1% percentile of each token).

On average, the order volume in the top 1% percentile is $0.747 million for sell orders and $0.824
million for buy orders. In relation to the market capitalization of the respective assets, this amounts
to an average of 0.059% for sales and 0.043% for buys. With an average market capitalization of
about $3 billion (cf. Table 1), a cumulative abnormal return of -0.11% in the swaps before the
actual event already means a loss of $3.3 million, which is 4.4 times the economic value of the
average sell order. This value drops further in the subsequent swaps to a minimum of -0.18%
(factor of 7.37) before recovering slightly. In the case of buy orders, in t-1 there is already an
abnormal increase in value of $0.266 million, which corresponds to about 32% of the event

11
volume. Subsequently, the market capitalization increases by another $3.62 million on average,
which corresponds to a factor of 4.39.
Table 5 shows event study results for the market reaction to the occurrence of buy and sell order
swaps of the largest one percent by crypto asset. For sell orders, the average swap size per asset is
between $120,000 (SPELL) and $1,134,000 (RBN) in absolute terms and between 0.003% (SHIB)
and 1.185% (RBN) in relation to market capitalization. For buy orders, the average transaction
size is between $113,000 (SAND) and $660,000 (RBN). In relative terms, the swaps average
between 0.002% (SHIB) and 0.735% (RBN).
Looking at the individual swap windows of t = -1, 0 and +1 for sale swaps, it is immediately clear
that the effects are very similar across all crypto assets. With a single exception (COMP in +1
swap), we identify significant negative abnormal returns in all three periods. The results for t = -1,
at -0.006% (SHIB) to -0.3% (SPELL), suggest that traders have moved ahead of the transaction as
part of Mempool monitoring (cue MEV, sandwich attacks and frontrunning). At the time of the
swap, abnormal returns on sells range from -0.002% (SAND) to -0.699% (SPELL). In t = +1, the
range is -0.001% (ENS) to -0.56% (RBN). A critical observation here is the relationship between
the relative size of the swap and the market reaction (or rather anticipation) in t-1. For eight of the
fourteen crypto assets, the abnormal and statistically significant token return in t = -1 is so strongly
negative that it exceeds the economic value of the swap. For example, an average swap in the top
1% of AAVE amounts to about 0.017% of the asset’s market capitalization, but the significant
abnormal return already equals 0.015% in t-1 (similarly for COMP, LINK, MANA, OHM, SHIB
and SPELL). At the time of the swap and in t = +1, this phenomenon is amplified, with ten of the
fourteen projects (71%) exhibiting a significantly higher economic market reaction than assumed.
The differences are particularly significant for the MKR and SHIB tokens, whose abnormal returns
are -18.8 and -17.66 times as large, respectively, as the swap size. The average factor across all 14
tokens is -4.32, i.e., for every dollar sold in a swap, the market capitalization of the tokens declines
by $4.32. Clearly such averaging can only serve as an indication. It can be interpreted as evidence
of inefficiency and suggests that the (reported) market capitalization of the tokens is not a
meaningful metric for assessing their liquidity or value.
A similar effect, but with the reverse sign, is identified for buy orders, which consistently lead to
significant positive abnormal returns in t = -1, 0 and +1. In t = -1, for 64% of the tokens, the return
effect exceeds the relative size of the swap. The effects range from 0.007% (SHIB) to 0.495%
(SPELL). At the time of the swap, they range from 0.004% (SHIB) to 0.429% (SPELL), and in
t = +1, from 0.003% (LOOKS) to 0.180% (RBN). Accumulated over all three time points, again
for 71% of the tokens, the abnormal return exceeds the relative size of the swap. The differences
are particularly large for AAVE (factor of 11.9), COMP (10.6), MANA (16.6) and SHIB (17.0).
Using SHIB as an example, this means that an average swap in the 1% percentile worth $157,000
raises the market capitalization of the crypto asset by $2,669,000.

12
Table 5. Effects of large DEX swaps (top 1% percentile) on token returns. The table shows event study results for returns around swap events that range in the top 1% in terms of
swap size relative to the asset’s market capitalization. The panels include swaps whose tokens were sent to the liquidity pool, i.e., sell orders, and ones whose tokens were withdrawn
(buy orders). The swap size columns show means for the size of the swap in relative (Mcap, in percent) and in absolute ($k, in thousands of dollars) terms. The mean (cumulative)
abnormal return ((C)AR) in percent and results of the Wilcoxon sign-rank test (z-test) are shown for the single swap periods t = -1, 0, and +1, and for aggregated series of t = -15 to -1, 0
to +15, 0 to +50, 0 to +100, and 0 to +500 swaps after the event. The asterisks ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Single swap windows Multi-swap windows
Swap size -1 swap Event +1 swap -15 swaps +15 swaps +50 swaps +100 swaps +500 swaps
Type N Mcap $k AR z-test AR z-test AR z-test CAR z-test CAR z-test CAR z-test CAR z-test CAR z-test
Panel A: Sell orders
AAVE 344 0.017 637 -0.051 -10.95*** -0.016 -12.98*** -0.017 -14.97*** -1.053 -2.53*** 0.131 3.07*** 0.516 1.73* -0.197 -0.57 -1.513 -1.09
ALCX 388 0.300 694 -0.080 -10.66*** -0.023 -10.80*** -0.124 -10.29*** -0.679 -2.75*** -0.290 -3.88*** -0.594 -2.85*** -0.231 -0.72 -0.549 -1.01
COMP 357 0.016 280 -0.036 -10.42*** -0.077 -11.01*** 0.096 -1.94 -0.465 -3.21*** -0.531 0.20 -0.332 -1.02 -0.605 -0.19 -0.096 -1.96**
ENS 1,072 0.032 259 -0.025 -27.03*** -0.004 -27.03*** -0.001 -28.26*** 0.177 1.02 -0.199 -18.37*** 0.036 7.89*** 0.144 5.38*** 1.090 2.39**
LINK 1,050 0.005 527 -0.008 -24.78*** -0.002 -23.84*** -0.005 -24.79*** 0.039 0.32 0.110 7.22*** 1.615 4.38*** 2.302 6.08*** 2.634 5.15***
LOOKS 1,681 0.093 284 -0.003 -39.99*** -0.006 -40.09*** -0.001 -40.20*** -0.117 0.86 -0.606 -31.98*** -0.164 -14.32*** -0.297 -9.80*** -0.313 1.95*
MANA 502 0.005 139 -0.026 -16.09*** -0.008 -16.54*** -0.018 -18.17*** -0.223 -2.66*** -0.099 -2.34** 0.595 1.66* 0.977 2.07** 0.414 4.03
MKR 298 0.015 347 -0.078 -8.63*** -0.067 -9.32*** -0.152 -10.04*** -0.451 -4.15*** 0.143 -0.52 -0.416 1.14 0.089 1.09 -0.191 0.01
OHM 526 0.047 445 -0.057 -17.82*** -0.074 -19.24*** -0.022 -20.27*** -0.084 -3.02*** -0.304 -4.15*** -0.694 -3.92*** 0.902 -4.55*** -0.992 -1.23
RBN 174 1.185 1,134 -0.287 -6.40*** -0.221 -6.49*** -0.560 -5.30*** -0.195 -2.99*** -0.384 -5.00*** -2.392 -5.03*** -2.306 -3.73*** 0.040 -0.48
SAND 1,012 0.016 137 -0.006 -28.14*** -0.002 -26.63*** -0.012 -27.68*** -0.407 -4.56*** -1.611 -5.31*** 0.535 4.46*** 1.203 4.48*** 4.475 8.94***
SHIB 4,510 0.003 217 -0.009 -19.29*** -0.006 -20.20*** -0.041 -19.33*** -0.001 -2.89*** -0.259 -47.31*** -0.217 -14.12*** -0.087 -7.14*** -0.093 -0.89
SPELL 818 0.235 120 -0.300 -5.31*** -0.699 -7.12*** -0.006 -65.15*** -0.139 -7.23*** -0.028 -3.85*** 0.719 0.56 2.261 3.44*** 10.429 8.80***
SUSHI 1,116 0.030 494 -0.024 -8.17*** -0.025 -6.61*** -0.005 -28.59*** -0.036 -2.09** -0.111 -2.11** -0.117 -0.99 0.247 1.39 0.482 1.65*
Panel B: Buy orders
AAVE 299 0.011 422 0.075 9.11*** 0.039 10.47*** 0.017 13.43*** -0.295 -4.32*** 0.318 6.11*** 0.254 1.90* -1.007 -0.75 -1.717 -1.84*
ALCX 287 0.313 728 0.149 11.96*** 0.156 13.88*** 0.105 14.00*** -0.566 -6.11*** 0.720 6.89*** 0.420 2.58** 0.220 1.29 0.136 0.07
COMP 297 0.014 251 0.070 7.98*** 0.055 10.43*** 0.023 12.05*** -0.633 -3.76*** 0.574 -0.84 -0.408 -0.17 -0.327 1.62 -0.062 -2.31**
ENS 526 0.046 322 0.077 19.63*** 0.023 19.62*** 0.008 20.27*** -0.097 -7.33*** 0.031 14.52*** 0.433 9.44*** 1.001 9.17*** 3.160 5.65***
LINK 944 0.005 528 0.003 24.10*** 0.023 24.93*** 0.009 25.31*** 0.089 0.29 0.158 10.87*** 0.375 5.80*** 0.587 6.25*** 0.855 3.23***
LOOKS 1,812 0.104 235 0.013 42.04*** 0.006 42.19*** 0.003 42.28*** -0.071 -3.88*** 0.052 38.28*** 0.112 25.00*** 0.214 17.49*** 0.864 5.89***
MANA 549 0.005 130 0.026 15.24*** 0.017 18.45*** 0.040 18.92*** -0.165 -4.44*** 0.064 1.77* -0.264 -0.17 -0.003 -1.25 1.552 2.09**
MKR 273 0.012 271 0.084 10.09*** 0.011 10.43*** 0.025 11.32*** -0.234 -8.34*** 0.163 2.39** 1.370 3.35*** 1.614 3.27*** 2.027 1.96**
OHM 651 0.036 354 0.008 23.85*** 0.001 24.10*** 0.056 23.53*** -0.463 -11.01*** 0.021 13.03*** 0.297 4.28 0.307 2.51** 0.172 1.72*
RBN 221 0.735 660 0.046 9.77*** 0.022 8.38*** 0.180 10.93*** -0.322 -2.98*** 1.063 5.90*** 0.540 1.32 -0.969 -1.07 1.537 0.27
SAND 1,436 0.014 113 0.035 31.35*** 0.031 32.37*** 0.032 34.21*** 0.052 5.43*** 0.208 8.29*** 0.817 7.35*** 1.285 8.20*** 3.112 8.52***
SHIB 7,195 0.002 157 0.007 82.26*** 0.004 82.66*** 0.023 12.05*** -0.100 -12.89*** 0.006 65.39*** 0.119 23.17*** 0.269 7.85*** 0.412 2.91***
SPELL 1,098 0.136 118 0.495 3.43*** 0.429 4.63*** 0.136 21.69*** -0.014 -5.17*** 0.857 4.87*** 1.572 5.78*** 2.270 6.43*** 2.700 3.97***
SUSHI 998 0.030 497 0.123 8.26*** 0.004 7.42*** 0.006 28.17*** 0.008 0.78 0.032 7.64*** 0.181 2.73*** 0.312 3.04*** 0.757 3.33***

13
A similar effect, but with the reverse sign, can be identified for buy orders, which consistently
lead to significant positive abnormal returns in t-1, 0 and +1. In t-1, the return effect exceeds
the relative size of the swap for 64% of the tokens. The effects range from 0.007% (SHIB) to
0.495% (SPELL). At the time of the swap, they range from 0.004% (SHIB) to 0.429%
(SPELL), and in t+1, from 0.003% (LOOKS) to 0.180% (RBN). Accumulated over all three
time points, the abnormal return exceeds the relative size of the swap for 71% of the tokens.
The differences are particularly large for AAVE (factor of 11.9), COMP (10.6), MANA (16.6)
and SHIB (17). Again using SHIB as an example, this means that a swap worth $157,000 (the
average swap size in the top 1%) raises the crypto asset’s market capitalization by $2,669,000.
The average factor across all tokens for these three swap dates is 7.02, i.e., for every dollar
invested in a token, its market capitalization grows by $7.02. As with sell orders, this suggests
that the market capitalization metric poorly reflects the views and valuation of the market.
Looking at the aggregate period of 15 swaps before the large transactions, a similar picture
emerges for sells and buys. In both subsamples, eleven of the fourteen crypto assets show
significant negative abnormal returns. Only one token (SAND) experiences a significant
positive abnormal return of 0.052%. Twelve of the tokens show significant negative abnormal
returns after fifteen swaps with large sell orders, five after 50 swaps, four after 100 and two
after 500. However, significant positive abnormal returns also occur. Already 15 swaps after a
large sell order, AAVE (0.13%) and LINK (0.11%) show significant positive abnormal returns.
After 50 swaps, ENS, MANA and SAND join the group. Particularly striking are the extremely
high significant abnormal returns of SAND (4.48%) and SPELL (10.43%) after 500 swaps.
For the larger multi-swap windows of buy orders, we obtain more consistent results. There are
significant positive abnormal returns for all periods considered, but also two negative returns
at +500 (AAVE and ALCX). In the short term (+15 swaps), ALCX, RBN and SPELL feature
the highest abnormal returns, ranging from 0.72% to 1.06%. In relation to the relative volume
of the swap, AAVE and LINK are clearly ahead. The market capitalization of AAVE increases
by $28.91 per dollar of purchasing power within 15 swaps, and that of LINK by $31.6. For
LINK, this value even increases to $75 by the 50th swap, to $117.4 by the 100th, and to $171
by +500. The highest ratios of abnormal returns to relative swap size are found for SHIB ($206
per dollar spent on buying into the crypto asset), SAND ($222.29), and MANA ($310.4).

4 Discussion

DeFi is a rather recent phenomenon but has quickly grown into a multi-billion-dollar market,
making it a significant aspect of cryptocurrency or crypto asset markets. This study uses an
event study approach to examine price and liquidity shocks in DeFi markets and contributes to
understanding the market efficiency of crypto assets in general and DeFi markets in particular.
It also questions the validity of the ubiquitous "market capitalization" metric.

4.1 Reflection on the main results

Dividing DEX trades into sells and buys (or in and out) yields statistically significant
differences between the dollar equivalent of buys and sells on DEXes for all crypto assets

14
examined. For example, $49 million more ALCX were sold than bought over the period, which
translates to a significant difference of $5,451 per swap between sells and buys. This suggests
that DeFi users were on average acquiring ALCX via DEXes—be it just to hold ALCX, to
stake it for yield, lend it out or to send it to centralized exchanges. Similar results apply to
COMP or SUSHI—both of which can also be staked by DeFi users for yield. The largest
discrepancy between buys and sells is observed for the ENS token ($13,370), which is
distributed by airdrop. Many ENS holders received the tokens quasi-free (for historical
actions), which explains why many smaller sales took place.
In contrast to ALCX or ENS, some of the examined crypto assets, however, feature
significantly larger sell swaps. For example, the average sell swaps exceed the average buy
swaps of RBN and SPELL by $11,209 and $6,401, respectively. Overall, $50 million more
RBN was sold than bought, which can potentially be explained by a drop in the token’s price
(cf. the negative mean return in Table 2). SPELL, however, has positive average token returns
over the considered period, and $2b billion in sales are offset by $2.1 billion in buys. The
smaller average size of buys may indicate buying from individual / retail investors and selling
by larger investors, or that traders are more likely to use DEXes to sell SPELL—while possibly
buying it on centralized exchanges.
On average, buy orders of AAVE are slightly larger than sell orders ($26,879 vs. $27,140).
However, the standard deviations of these two measures differ significantly: $389,943 for sells
versus $70,964 for buys. This indicates that sell swaps tend to be more extreme in size (both
very large and very small) than buy swaps. It seems that individual investors or DeFi users buy
AAVE for smaller amounts while very large investors use DEXes to liquidate AAVE positions.
This result of significantly larger variability in the size of sell compared to buy orders similarly
applies to COMP, MKR, OHM, and RBN.
Using an event-study approach, we estimate cumulative swap-to-swap abnormal returns for
sell and buy order swaps that are in the top 1% based on swap size in USD. For the 15
transactions before the large and unexpected swap, we identify abnormal returns that are
significantly negative for sell orders and positive for buy orders. Sell orders show a semi-linear
negative trend until t = 0. This abnormal price effect may suggest that large sells usually occur
in extremely negative market conditions (herding or cascading). Herding, overreactions,
overconfidence and sell-offs have already been identified for crypto assets in different settings
(e.g., Ajaz and Kumar, 2018; Ballis and Drakos, 2020; Bouri et al., 2019; Youssef, 2022).
Another explanation would be that traders (i.e., bots) pick up the information about initiated
but not yet confirmed large swaps in the Ethereum Mempool and place themselves in front of
and/or behind the transaction in order to arbitrage it. This process is known as MEV and
represents a rapidly growing and quite opaque market. Scientific studies show that MEV
arbitrage and frontrunning are common on the Ethereum network (e.g., Daian et al., 2020; Qin
and Gervais, 2021).
For buy orders, we also identify a significant effect before very large swaps take place.
However, abnormal returns from t = -15 to -2 are insignificant and hardly worth mentioning.
Yet in t = -1, the abnormal return increases by 0.009%. The fact that there is no quasi-linear
trend as in the case of sell orders rules out the explanation of fundamentally rising prices. A

15
much more likely cause is that here, too, we are looking at informed trading in the form of
MEV and traders are pushing their own transactions in front of the swap. After the large swap
takes place, abnormal returns increase continuously over the following swaps, suggesting that
the market takes the information about the increased demand for the token to be a positive price
signal. Another major result of this study pertains to the discrepancy between liquidity and
valuation of crypto assets in DeFi markets, which can be seen as a sign of market inefficiency.
We show that large swaps on DEXes have a significantly higher impact on crypto asset prices
than should be the case if market capitalization were a meaningful metric. This applies to both
sell orders and buy orders, although the effect is significantly larger on average for sells (factor
of -7.4 for sells compared to +4.4 for buys). The results call for some skepticism regarding the
suitability of market capitalization as a metric for valuing crypto assets.
The analysis of individual crypto assets shows that significant abnormal returns around very
large sell swaps are already greater than the economic value of the swap by t = -1 for more than
half of the crypto assets considered. If t = 0 to 1 are added, this applies to 71% of the assets. In
some cases, the extent of these differences is astonishing. For example, a large sale of the crypto
assets MKR and SHIB results in a price drop of $18.8 and $17.66, respectively, per dollar of
sale value. This means that in the short term, it would be quite cheap to shift the price and
market capitalization of these crypto assets. Of course, such behavior could be exploited very
quickly by arbitrageurs, and prices on other (centralized) exchanges need not follow suit.
Accordingly, the results are particularly relevant for tokens that are either traded only in DeFi
markets or for which a DEX is the leading or most liquid market.

4.2 Limitations and future research paths

This study is subject to a number of limitations, some of which have already been outlined in
the methodology section. DeFi markets cover thousands of assets, only 14 of which were
selected for this exploratory study. In particular, since inclusion in the sample was conditioned
on an asset featuring extremely large swaps, it is unclear to what extent the results can be truly
generalized. Future studies could seek to validate the results based on the same crypto assets
and methodology or use different assets and methodologies to address similar questions.
Basically, we have dealt only with the tip of the iceberg, so the possibilities for follow-up
studies are vast.
Many potentially significant explanatory variables have not been considered in this study. The
addition of detailed trading volumes, liquidity, and additional explanatory factors such as
market sentiment or momentum could help to better understand the phenomena investigated
here. In the context of MEV as a possible explanation of ex-ante abnormal returns, future
studies could collect Mempool and on-chain data to examine when a transaction first became
public and to what extent MEV bots reacted to it. Furthermore, it can be an exciting research
avenue to supplement the analysis with data from centralized exchanges to evaluate
information or volatility transmission between DeFi markets and centralized trading venues
(for related studies, see, e.g., Alexander et al., 2021; Barbon and Ranaldo, 2021; Makarov and
Schoar, 2020). The fact that Ethereum on-chain transactions are settled only every 12-14

16
seconds while centralized exchanges have no significant time lag may entail significant
differences between these markets.
The data basis of this study is twofold: on the one hand, blockchain data on tokens with
timestamps on a per second basis, and on the other hand, daily data on circulating supply.
Calculating the market capitalization metric from these two asynchronous sources logically
leads to the limitation that intraday fluctuations in the number of tokens are not taken into
account. While this leads to an overweighting or underweighting of the metric depending on
the time of day, the size of the samples should serve to curb this limitation at least on average.
However, it may be helpful for future studies to improve on this form of calculation.
A final fundamental point for future research is to further challenge the ubiquitous metric of
market capitalization and develop new metrics that better incorporate the illiquidity and
inefficiency of crypto assets.

4.3 Implications

Probably the most significant implication of this study is that ‘market capitalization’ is quite
unsuitable for determining the "true value" of crypto assets. The main reason for this is that the
metric completely disregards the underlying liquidity of the assets. An example shall serve to
illustrate this shortcoming. Unlike in traditional markets, it is an extremely simple process to
create, say, a million tokens in no time at all, to trade only one of them on a decentralized
exchange and to deposit the rest in one or more blockchain wallets. The value of the token can
then be increased via wash trading, so that a public market price of, say, $1,000 is now used as
the basis for calculating the market capitalization of all tokens. That way, in theory, a crypto
asset with a market cap of $1 billion can quickly be created. As long as the major data providers
like CoinGecko or Coinmarketcap rank all cryptocurrencies by market cap, it will always be
possible to drive up the listed value of illiquid and highly centralized crypto assets to attract
investors.1 However, these investors may then find it difficult to liquidate their crypto holdings.
While the above example is clearly fictitious and extreme, phenomena such as so-called “rug-
pulls”, scam tokens or “pump-and-dumps” show that the cryptocurrency market, and the DeFi
market in particular, faces significant challenges of fraud and investor deception (e.g.,
Scharfman, 2022; Wronka, 2021; Xia et al., 2021). Research and practice should strongly
question the market capitalization metric and generate alternative approaches to objectively
valuing crypto assets.
This study contributes to existing research in multiple ways: It studies a transparent blockchain-
based market and ecosystem in which information about market orders, liquidity and pricing is
openly available and to some extent known even prior to actual execution. The results can thus

1
For the sake of completeness, it should of course be noted that these sites, which are listed here as examples,
also have listing criteria against which they evaluate the inclusion of projects. For example, Coinmarketcap reports
an evaluation framework that includes, 1) trading volume and liquidity, 2) community interest and engagement,
3) traction/progress, 4) team, 5) product/market fit, 6) impact & practicality, 7) uniqueness & innovation, and 8)
project longevity & activity (Coinmarketcap 2022). It remains unclear to what extent these criteria are consistently
checked. It is important to keep in mind that there is a potential conflict of interest, as customers could potentially
pay for a listing. However, Coinmarketcap states that “We don’t ask for payment on listings, period.”
(Coinmarketcap 2022).

17
contribute to the topic of market transparency and efficiency and question to what degree
transparency is even desirable in financial markets (Bloomfield and O’Hara 2000). While
transparency offers significant informational benefits to market participants and market
makers, research in equity markets has already shown that higher transparency can reduce
liquidity, which is in turn associated with lower stock prices (Madhavan, Porter, and Weaver
2005). Individual investors, who generally underperform the market (Barber and Odean 2013;
Ante et al. 2022), are discouraged from participating in the market, as professional traders have
an information advantage over them (H.-K. Chen, Hsieh, and Ma 2011). Accordingly, it seems
reasonable to assume that full transparency is not optimal for financial markets; instead, an
optimal tradeoff must be found. The results described here and the assumption of frontrunning
attacks on transparently initiated DEX swaps are in line with the findings for stock markets.
Professional traders likely benefit from their information advantage, which tend to dissuade
individual investors from participating in such a market.
The study analyzes the price behavior of crypto assets based on abnormal returns on an intraday
trade-by-trade (or rather swap-by-swap) basis. In particular, we investigate the market reaction
to, and anticipation of, large unexpected trades on DEXes. The results thus contribute to
research on herding, overreaction and information cascades in cryptocurrency markets.
Accordingly, they may inform the development of crypto asset pricing models, risk metrics
and models, or trading strategies (e.g., Liu et al., 2022; Petukhina et al., 2021; Ren et al., 2022).
Dividing the samples into sell and buy orders (or rather swaps in and out of a pool) permits a
differentiated view of crypto assets that can potentially proxy retail investor attention or sell-
offs from large holders. The results suggest that the user or investor base of different DEX
trading pairs varies substantially. For example, some tokens predominantly attract individual
investors (e.g., SHIB), whereas others may be rather or additionally be used to liquidate larger
positions (e.g., AAVE). While this study cannot clarify why these differences exist, it
contributes to a better assessment of liquidity pools and their users (e.g., Heimbach et al., 2021),
on which little research has been conducted as yet.

5 Concluding remarks

This study applies quantitative analysis to explore the interplay between DEX swaps, i.e., buys
and sells of crypto assets, and token returns in DeFi. Based on a sample of 2.77 million swaps
of 14 crypto assets on three major DEXes, we identified that the size of sell (buy) orders is
negatively (positively) correlated with future token returns. Using an event study approach, we
quantified how the market reacts to extremely large unanticipated swaps, finding significant
overreaction. On average, this amounts to a factor of -7.4 (+4.4) for sells (buys) in relation to
the reported market capitalization of the tokens, with much larger effects accruing to individual
tokens. This suggests that some of the crypto assets are overvalued and that market
capitalization is only of limited value in determining the actual recoverable value of (possibly
illiquid) crypto assets. We also identify that the market reacts before the actual execution of
the swap, suggesting that market participants monitor unconfirmed transactions in order to
generate returns via so-called sandwich or frontrunning attacks. This is by no means illegal
activity, but it is likely to deter individual investors from participating in the market.

18
The results point to various challenges for investors and their protection in DeFi markets.
Besides price, market capitalization is likely the most important metric for ranking and
evaluating crypto assets. Major data aggregators such as CoinGecko or Coinmarketcap (both
of which are among the top 500 websites globally based on Alexa ranking) sort crypto assets
by market capitalization and rank them accordingly. The top 100 assets in terms of market
capitalization are shown directly on the main page and accordingly receive the most attention
from website visitors. This approach needs to be questioned, and more suitable metrics should
be developed that integrate liquidity as an essential factor, among other things. Ultimately, the
protection of investors should be paramount so that crypto markets become as safe a place to
invest in as possible.

References

Adams, Hayden, Noah Zinsmeister, and Dan Robinson. 2020. “Uniswap v2 Core.” 2020.
https://uniswap.org/whitepaper.pdf.
Adams, Hayden, Noah Zinsmeister, Moody Salem, and Dan Robinson. 2021. “Uniswap v3 Core.” 2021.
https://uniswap.org/whitepaper-v3.pdf.
Aitken, Michael J., Alex Frino, Michael S. Mccorry, and Peter L. Swan. 1998. “Short Sales Are Almost
Instantaneously Bad News: Evidence from the Australian Stock Exchange.” Journal of Finance 53
(6): 2205–23. https://doi.org/10.1111/0022-1082.00088.
Ajaz, Taufeeq, and Anoop S Kumar. 2018. “Herding in Crypto-Currency Markets.” Annals of Financial
Economics 13 (02): 1850006. https://doi.org/10.1142/S2010495218500069.
Alexander, Carol, Daniel Heck, and Andreas Kaeck. 2021. “The Role of Binance in Bitcoin Volatility
Transmission,” July. http://arxiv.org/abs/2107.00298.
Ante, Lennart. 2020. “Bitcoin Transactions, Information Asymmetry and Trading Volume.” Quantitative
Finance and Economics 4 (3): 365–81. https://doi.org/10.3934/QFE.2020017.
———. 2021. “Smart Contracts on the Blockchain – A Bibliometric Analysis and Review.” Telematics
and Informatics 57: 101519. https://doi.org/10.1016/j.tele.2020.101519.
Ante, Lennart, and Ingo Fiedler. 2021. “Market Reaction to Large Transfers on the Bitcoin Blockchain -
Do Size and Motive Matter?” Finance Research Letters, no. 39: 101619.
https://doi.org/10.1016/j.frl.2020.101619.
Ante, Lennart, Ingo Fiedler, Marc von Meduna, and Fred Steinmetz. 2022. “Individual Cryptocurrency
Investors: Evidence from a Population Survey.” International Journal of Innovation and Technology
Management 19 (4): 2250008. https://doi.org/10.1142/S0219877022500080.
Ante, Lennart, Ingo Fiedler, and Elias Strehle. 2021a. “The Impact of Transparent Money Flows: Effects
of Stablecoin Transfers on the Returns and Trading Volume of Bitcoin.” Technological Forecasting
and Social Change 170: 120851. https://doi.org/10.1016/j.techfore.2021.120851.
———. 2021b. “The Influence of Stablecoin Issuances on Cryptocurrency Markets.” Finance Research
Letters 41: 101867. https://doi.org/10.1016/j.frl.2020.101867.
Ballis, Antonis, and Konstantinos Drakos. 2020. “Testing for Herding in the Cryptocurrency Market.”
Finance Research Letters 33: 101210. https://doi.org/10.1016/j.frl.2019.06.008.
Barber, Brad M., and Terrance Odean. 2013. “The Behavior of Individual Investors.” In Handbook of the
Economics of Finance, 2:1533–70. Elsevier B.V. https://doi.org/10.1016/B978-0-44-459406-
8.00022-6.
Barbon, Andrea, and Angelo Ranaldo. 2021. “On The Quality Of Cryptocurrency Markets: Centralized
Versus Decentralized Exchanges,” December. http://arxiv.org/abs/2112.07386.

19
Beaver, W. H. 1968. “The Information Content of Annual Earnings Announcements.” Journal of
Accounting Research 6 (1968): 67–92.
Black, Fischer. 1986. “Noise.” The Journal of Finance 41 (3): 528–543. https://doi.org/10.1111/j.1540-
6261.1986.tb04513.x.
Bloomfield, Robert, and Maureen O’Hara. 2000. “Can Transparent Markets Survive?” Journal of
Financial Economics 55 (3): 425–59. https://doi.org/10.1016/S0304-405X(99)00056-2.
Bouri, Elie, Rangan Gupta, and David Roubaud. 2019. “Herding Behaviour in Cryptocurrencies.” Finance
Research Letters 29: 216–21. https://doi.org/10.1016/j.frl.2018.07.008.
Brown, Stephen J., and Jerold B. Warner. 1985. “Using Daily Stock Returns. The Case of Event Studies.”
Journal of Financial Economics 14 (1): 3–31. https://doi.org/10.1016/0304-405X(85)90042-X.
Cao, Charles, Oliver Hansch, and Xiaoxin Wang. 2009. “The Information Content of an Open Limit-Order
Book.” Journal of Futures Markets 29 (1): 16–41. https://doi.org/10.1002/fut.20334.
Chae, Joon. 2005. “Trading Volume, Information Asymmetry, and Timing Information.” Journal of
Finance 60 (1): 413–42. https://doi.org/10.1111/j.1540-6261.2005.00734.x.
Chen, Hsiu-Kuei, Shu-Fan Hsieh, and Tai Ma. 2011. “Who Wins and Who Loses in Transparent Markets?
Daily and Intraday Analysis of Taiwan Stock Market.” Taiwan Economic Forecast and Policy 41
(2): 127–78.
Chen, Ying, Wee Song Chua, and Wolfgang Karl Härdle. 2019. “Forecasting Limit Order Book Liquidity
Supply–Demand Curves with Functional Autoregressive Dynamics.” Quantitative Finance 19 (9):
1473–89. https://doi.org/10.1080/14697688.2019.1622290.
CoinGecko. 2022. “Olympus DAO Price.” 2022.
https://www.coingecko.com/en/coins/olympus/usd#panel.
Coinmarketcap. 2022. “Listings Criteria.” Coinmarketcap. 2022. https://support.coinmarketcap.com/hc/en-
us/articles/360043659351-Listings-Criteria.
Cornelli, Francesca, and David Goldreich. 2003. “Bookbuilding: How Informative Is the Order Book?”
The Journal of Finance 58 (4): 1415–43. https://doi.org/10.1111/1540-6261.00572.
Cunado, Juncal, and Fernando Perez de Gracia. 2014. “Oil Price Shocks and Stock Market Returns:
Evidence for Some European Countries.” Energy Economics 42: 365–77.
https://doi.org/10.1016/j.eneco.2013.10.017.
Curve.fi. 2022. “Curve Documentation.” 2022. https://curve.readthedocs.io/.
Daian, Philip, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach,
and Ari Juels. 2020. “Flash Boys 2.0: Frontrunning in Decentralized Exchanges, Miner Extractable
Value, and Consensus Instability.” In 2020 IEEE Symposium on Security and Privacy (SP), 910–27.
https://doi.org/10.1109/SP40000.2020.00040.
DeFi Sniper. 2022. “DeFi Sniper Twitter Account.” Twitter. 2022. https://twitter.com/DefiSniper.
Demir, Ender, Oguz Ersan, and Boris Popesko. 2022. “Are Fan Tokens Fan Tokens?” Finance Research
Letters, 102736. https://doi.org/10.1016/j.frl.2022.102736.
Diaconaşu, Delia-Elena, Seyed Mehdian, and Ovidiu Stoica. 2022. “An Analysis of Investors’ Behavior in
Bitcoin Market.” Plos One 17 (3): e0264522. https://doi.org/10.1371/journal.pone.0264522.
ENS. 2022. “Ethereum Name Service Airdrop Claim.” 2022. https://claim.ens.domains.
Etherscan. 2022a. “Ethereum Average Block Time Chart.” 2022. https://etherscan.io/chart/blocktime.
———. 2022b. “Etherscan Transaction.” 2022.
https://etherscan.io/tx/0x8f45cc66eba60229085440505c84e063366b4e3d71fd98c1121018a7978e197
8.
Griffin, John M, and Amin Shams. 2020. “Is Bitcoin Really Un-Tethered?” The Journal of Finance 75 (4):
1913–64. https://doi.org/10.1111/jofi.12903.

20
Heimbach, Lioba, Ye Wang, and Roger Wattenhofer. 2021. “Behavior of Liquidity Providers in
Decentralized Exchanges,” May. http://arxiv.org/abs/2105.13822.
Hu, Chunyan, Xinheng Liu, Bin Pan, Bin Chen, and Xiaohua Xia. 2018. “Asymmetric Impact of Oil Price
Shock on Stock Market in China: A Combination Analysis Based on SVAR Model and NARDL
Model.” Emerging Markets Finance and Trade 54 (8): 1693–1705.
https://doi.org/10.1080/1540496X.2017.1412303.
Kang, Wensheng, Ronald A Ratti, and Joaquin Vespignani. 2016. “The Impact of Oil Price Shocks on the
U.S. Stock Market: A Note on the Roles of U.S. and Non-U.S. Oil Production.” Economics Letters
145: 176–81. https://doi.org/10.1016/j.econlet.2016.06.008.
Koutmos, Dimitrios. 2018. “Bitcoin Returns and Transaction Activity.” Economics Letters 167: 81–85.
https://doi.org/10.1016/j.econlet.2018.03.021.
Kristoufek, Ladislav. 2021. “Tethered, or Untethered? On the Interplay between Stablecoins and Major
Cryptoassets.” Finance Research Letters, 101991. https://doi.org/10.1016/j.frl.2021.101991.
Liu, Yukun, Aleh Tsyvinkski, and X I Wu. 2022. “Common Risk Factors in Cryptocurrency.” The Journal
of Finance 77 (2): 1133–77. https://doi.org/10.1111/jofi.13119.
MacKinlay, A Craig. 1997. “Event Studies in Economics and Finance.” Journal of Economic Literature 35
(1): 13–39. http://www.jstor.org/stable/2729691.
Madhavan, Ananth, David Porter, and Daniel Weaver. 2005. “Should Securities Markets Be Transparent?”
Journal of Financial Markets 8 (3): 265–87. https://doi.org/10.1016/j.finmar.2005.05.001.
Makarov, Igor, and Antoinette Schoar. 2020. “Trading and Arbitrage in Cryptocurrency Markets.” Journal
of Financial Economics 135 (2): 293–319.
https://doi.org/https://doi.org/10.1016/j.jfineco.2019.07.001.
Martinelli, Fernando, and Nikolai Mushegian. 2019. “A Non-Custodial Portfolio Manager, Liquidity
Provider, and Price Sensor.” 2019. https://balancer.fi/whitepaper.pdf.
Milgrom, Paul, and Nancy Stokey. 1982. “Information, Trade and Common Knowledge.” Journal of
Economic Theory 26: 17–27.
Pennec, Guénolé Le, Ingo Fiedler, and Lennart Ante. 2021. “Wash Trading at Cryptocurrency Exchanges.”
Finance Research Letters, 101982. https://doi.org/10.1016/j.frl.2021.101982.
Petukhina, Alla, Simon Trimborn, Wolfgang Karl Härdle, and Hermann Elendner. 2021. “Investing with
Cryptocurrencies – Evaluating Their Potential for Portfolio Allocation Strategies.” Quantitative
Finance 21 (11): 1825–53. https://doi.org/10.1080/14697688.2021.1880023.
Qin, Kaihua, and Arthur Gervais. 2021. “Quantifying Blockchain Extractable Value: How Dark Is the
Forest?” https://arxiv.org/abs/2101.05511.
Ren, Rui, Michael Althof, and Wolfgang Karl Härdle. 2022. “Financial Risk Meter for Cryptocurrencies
and Tail-Risk Network Based Portfolio Construction.” The Singapore Economic Review, no. in press.
https://doi.org/10.1142/S0217590822480010.
Saggu, Aman. 2022. “The Intraday Bitcoin Response to Tether Minting and Burning Events: Asymmetry,
Investor Sentiment, and ‘Whale Alerts’ on Twitter.” Finance Research Letters 49: 103096.
https://doi.org/10.1016/j.frl.2022.103096.
Scharfman, Jason. 2022. “Decentralized Finance (DeFi) Compliance and Operations.” In Cryptocurrency
Compliance and Operations : Digital Assets, Blockchain and DeFi, edited by Jason Scharfman, 171–
86. Cham: Springer International Publishing. https://doi.org/10.1007/978-3-030-88000-2_9.
Spence, Michael. 1973. “Job Market Signaling.” The Quarterly Journal of Economics 87 (3): 355–74.
https://doi.org/10.1055/s-2004-820924.
Strehle, Elias, and Lennart Ante. 2020. “Exclusive Mining of Blockchain Transactions.” In Scientific
Reports 2020 - Conference Proceedings of the Scientific Track of the Blockchain Autumn School
2020, 87–95. https://doi.org/10.48446/opus-11870.

21
Wei, Wang Chun. 2018. “The Impact of Tether Grants on Bitcoin.” Economics Letters 171: 19–22.
https://doi.org/10.1016/j.econlet.2018.07.001.
Wilcoxon, Frank. 1945. “Individual Comparisons by Ranking Methods.” Biometrics Bulletin 1 (6): 80–83.
Wronka, Christoph. 2021. “Financial Crime in the Decentralized Finance Ecosystem: New Challenges for
Compliance.” Journal of Financial Crime. https://doi.org/10.1108/JFC-09-2021-0218.
Xia, Pengcheng, Haoyu Wang, Bingyu Gao, Weihang Su, Zhou Yu, Xiapu Luo, Chao Zhang, Xusheng
Xiao, and Guoai Xu. 2021. “Trade or Trick? Detecting and Characterizing Scam Tokens on Uniswap
Decentralized Exchange.” In Proceedings of the ACM on Measurement and Analysis of Computing
Systems. Vol. 5. Association for Computing Machinery. https://doi.org/10.1145/3491051.
Youssef, Mouna. 2022. “What Drives Herding Behavior in the Cryptocurrency Market?” Journal of
Behavioral Finance 23 (2): 230–39. https://doi.org/10.1080/15427560.2020.1867142.

22
Declarations

Availability of data and materials

The datasets used for the study are available from the corresponding
author on reasonable request.

Conflicts of interest

Not applicable.

Funding

Not applicable.

Acknowledgements

Not applicable.

About the Blockchain Research Lab

The Blockchain Research Lab promotes independent science and

research on blockchain technologies and the publication of the results
in the form of scientific papers and contributions to conferences and
other media. The BRL is a non-profit organization aiming, on the one
hand, to further the general understanding of the blockchain
technology and, on the other hand, to analyze the resulting challenges
and opportunities as well as their socio-economic consequences.

www.blockchainresearchlab.org

23