SSRN Id3915140
SSRN Id3915140
SSRN Id3915140
∗
Christos A. Makridis, Michael Fröwis, Kiran Sridhar, and Rainer Böhme
December 4, 2022
Abstract
The market capitalization of cryptocurrencies sat at over $2 trillion in 2021. While most of
the financial activity remains on centralized systems, decentralized finance (DeFi) has ex-
perienced a particular surge with roughly 90,000 users at the start of 2020 to 4.28 million
by the end of 2021. Based on data collected from a popular crypto-asset data aggrega-
tion service and manually collected data, we document the rapid growth in decentralized
exchanges and their differences in volume and price dynamics from centralized exchanges.
Next, we investigate the role of airdrops and governance tokens as mechanisms for expanding
the base of users and driving up the value of an exchange. While our results do not have a
causal interpretation, they provide preliminary evidence that both mechanisms are effective
for expanding and strengthening networks, particularly for decentralized exchanges. We also
exploit two event studies that suggest the growth in decentralized exchanges is not driven by
speculation, but at least partially by value-creating cybersecurity benefits.
∗
Christos: Columbia Business School and Stanford University, cmakridi@stanford.edu; Michael Fröwis
michael.froewis@uibk.ac.at; Kiran Sridhar: kirsrid@gmail.com; Rainer Böhme: University of Innsbruck,
rainer.boehme@uibk.ac.at. These views are our own and do not reflect those of any affiliated institutions.
1 Introduction
Between 2020 and 2021, the number of unique decentralized finance (DeFi) users grew from 90,559
to 4.28 million.1 Furthermore, the value of DeFi assets surged past $100 billion in October
2021, up from $15 billion in 2020.2 Earlier this year, a famous economist and blogger remarked
that the “killer app” for cryptocurrencies has “finally been found” in DeFi (Cowen, 2021). DeFi
technologies leverage the blockchain—most frequently, the Ethereum blockchain—to enable users
to exchange and borrow cryptocurrencies and even to obtain specialized insurance through peer-
to-peer interchange without middlemen.3 The primary contribution of this paper is to document
the rise of decentralized exchanges (DEXs) and uncover plausible mechanisms behind the growth.
DEXs, like Uniswap, have fundamentally different architectures than centralized exchanges,
such as Coinbase and Binance. Centralized exchanges require users to deposit currencies into “hot
wallets,” which are centralized repositories. The centralized exchange maintains private crypto-
graphic keys which unlock these wallets. By contrast, DEXs are non-custodial: cryptocurrency
remains in each user’s wallet. DEXs perform transactions through smart contracts, computer
programs which automatically execute when certain conditions are met. Since the operational
footprint of many DEXs is reduced to the program code stored on a public blockchain, they often
do not hold any legal identity. Therefore, unlike CEXs, they do not currently interface with the
banking sector and do not support trade between cryptoassets and conventional fiat currency.
Despite this limitation, DEXs have enjoyed meteoric growth.4 Three principal rationales have
1
https://dune.xyz/rchen8/defi-users-over-time
2
https://consensys.net/blog/news/trends-in-defi-economy-a-snapshot-of-q3-2021/
3
See Iansiti and Lakhani (2017) for a broader discussion of blockchain technology for organizations and society.
Furthermore, see Cong et al. (2022) for evidence about the features of the Ethereum ecosystem that can inhibit
access to finance.
4
However, there is also a lot of concern about the role of speculation and fraudulent activity. For example,
been posited. First, proponents have claimed that, because cryptocurrencies are distributed in
users’ wallets and not stored in a hot wallet, DEXs are more resilient against large scale theft5 than
centralized exchanges (Bowater, 2018). Second, because DEXs fiercely compete for business, they
charge lower fees. This competitive landscape is likely to hold since DEXs are largely interoperable;
users can move to other exchanges if they do not like a DEX’s policies and practices (Cong and
He, 2019; Alkurd, 2020; Boudjemaa, 2020). Third, because they operate on a global peer-to-peer
network without formal registration in any jurisdiction, they can evade regulation and face lower
compliance costs (Aspris et al., 2021). Unfortunately, empirically evaluating these theoretical
predictions has been challenging due to the complexities of acquiring the necessary data.
Using new data that covers both centralized and decentralized exchanges over time, we doc-
ument the rise of DEXs, relative to centralized exchanges, since 2019 (see Figure 1). We also
explore the rise of DEX “governance tokens,” or tokens that give users ownership that resembles
shareholder rights for publicly traded firms. These governance tokens are often valuable assets
and tend to appreciate in value as an exchange flourishes. Since its introduction in September of
2020, the governance token for Uniswap, the most popular DEX, has more than quadrupled in
The second part of the paper estimates the difference in trading volume dynamics for centralized
Cong et al. (2021) explore the incidence of “wash trading,” which refers to strategies that exchanges may use to
fake transactions and create the perception of momentum to improve the ranking of the exchange, temporarily
distort prices, and ultimately influence their market capitalization. Makarov and Schoar (2021) conducts a similar
analysis for the Bitcoin market, arguing that only a minority of the transactions constitute illicit trades. Jahani
et al. (2018) take an alternative approach, mining the textual content of posts on cryptocurrency discussion boards,
finding that there is genuine “truth seeking” (versus hype) discussion around the more serious coins.
5
Most recent hacks of CEXs involve gaining access to their centralized infrastructure for example their hot
wallets. See https://www.hedgewithcrypto.com/cryptocurrency-exchange-hacks/.
6
Source: https://coinmarketcap.com/, August 31, 2021
and decentralized exchanges. Although centralized exchanges process larger volume overall, DEXs
have a nearly 15 percentage point higher growth rate in monthly exchange volume. Our estimates
are robust to controlling for aggregate fluctuations in the demand for cryptocurrency, the trust
rating of different exchanges, and the year that each exchange was established.
The third part of the paper investigates another potential mechanism driving DEX growth:
the use of airdrops—that is, the distribution of a native exchange token or a governance token
to current or potential users as part of a “digital asset giveaway.” On one hand, airdrops reward
early supporters who reap less value by backing a new DEX compared to an established one. Since
DEXs begin with few users who generate lower volume, they accrue less revenue in transaction
fees and cannot remunerate peer lenders as generously as established exchanges. Airdrops can also
serve as marketing opportunities to reach users who are active in other DEXs to encourage them
to experiment with another platform. On the other hand, airdrops can backfire, because they put
governance tokens in the hand of individuals who do not believe in the long-term viability of the
exchange and want to maximize their short-run returns. Moreover, airdrops may unintentionally
signal that the tokens are lower quality, influencing expectations about the exchange’s longevity.
Using a sample of 51 exchanges over time, we find that airdrops are positively associated with
growth in market capitalization and volume, but these benefits are concentrated among DEXs and
exchanges that offer governance tokens. In particular, DEXs that conduct an airdrop exhibit a 16.1
percentage point rise in their growth rate of their token’s market capitalization (significant at the
5% level). We also find some evidence, although the estimates are not statistically significant at
conventional levels, that DEXs who airdrop governance tokens experience higher volume growth
than those who do not. These results suggest that airdrops may not only serve as marketing
devices and sources for entrepreneurial funding, but also as a vehicle for ensuring that a DEX
stays true to its values. By airdropping governance tokens, DEXs reward and create skin-in-the-
game by giving some control over a protocol’s policies to some of the earliest adopters. However,
we caution that these estimates are noisy and more data is required; time will tell.
Despite our controls, we caution that these results are not causal. Some have suggested that
the DEX growth is evidence of a bubble driven by rampant speculation. Other have posited that
DeFi capitalizes on regulatory arbitrage: criminals use DEXs to hide their ill-gotten gains away
from the prying eyes of the government (Foley et al., 2019). To investigate whether growth in
DEXs can be tied to their inherent advantages, speculation or regulatory arbitrage, we conduct
two event studies. The first event we look at is the KuCoin breach in September 2020, where
cybercriminals stole cryptocurrency located in the exchange’s centralized hot wallet. This breach
exploited a security flaw in centralized exchanges that isn’t present in DEXs. Second, we examine
the ramifications of a rare joint letter written by three US regulatory agencies, released in October
2019, that urged digital currency exchanges to adhere to regulations that proscribed financial
institutions from abetting criminal activity. Many centralized exchanges responded to the letter
by fortifying their compliance procedures, while DEXs largely resisted such measures. We find
that growth in DEXs was not affected after the announcement, whereas DEX growth increased
after the KuCoin breach. These results suggest that growth in DeFi is linked to the value-creating
properties of DEXs (i.e., security benefits), rather than speculation or regulatory arbitrage.
Related Literature
Our paper contributes to an emerging literature on blockchain and DeFi (Yermack, 2017; Harvey
et al., 2021). In particular, initial coin offerings (ICOs) have emerged as an effective tool for financ-
ing new ventures by offering unique tokens with the promise that they will operate as a medium for
exchange to access the venture’s future projects/services and/or governance. According to Catalini
et al. (2018), blockchain startups have raised over $7 billion through ICOs, relative to $1 billion
through traditional venture capital in the space. These tokens differ from equity in that they are
not claims on future profits, but rather a flow of future revenues. Furthermore, Catalini and Gans
(2019) provide a theoretical framework for valuing the benefits of token offerings over traditional
venture financing, finding that there are some advantages for hybrid arrangements where ventures
raise equity initially from a narrow group before going towards the public or accredited investors.
Similarly, Cong and Wang (2021) explain how tokens (through smart contracts) provide a solution
to a time inconsistency problem with investors, thereby providing an effective way of financing
investment. Our results are also closely related with Chod and Lyandres (2020) who show that
tokens can strictly dominate equity financing for information services startups where tokens can
signal quality to outside investors and Howell et al. (2020) who show that ICOs predict firm per-
formance and expansion. Besides the prospect for enabling greater funding for entrepreneurial
ventures, there is also some evidence that smart contracts can reduce information asymmetries
and promote competition, which could indirectly raise entrepreneurship (Cong and He, 2019).
Our paper is also directly related to the study of cyptocurrency exchanges. For example,
Moore et al. (2018) study the emergence and closure of Bitcoin exchanges between 2010 and
2015, finding that higher-volume exchanges and those that derive their value from trading less
popular currencies are less likely to close. Moreover, they show that experiencing a breach is
highly predictive for whether the exchange closes. Similarly, Pieters and Vivanco (2017) show
that governance structures surrounding Bitcoin matter for centralized exchanges, especially the
establish an account. Our paper is also closely related with Catalini and Tucker (2017) who
conduct an experiment with Bitcoin, finding strong evidence of path dependence in the initial
availability of technology among early adopters. On the other hand, Cong et al. (2022) study
the emergence of the Ethereum blockchain and the factors that can inhibit financial inclusion
(e.g., high gas fees). In this sense, we build on an emerging literature that examines the effects of
Centralized and decentralized exchanges also differ in their market design. In particular, to
avoid frequent updates and prohibitive transaction costs on the blockchain, Angeris and Chitra
(2020) discuss the use of constant function market makers (CFMMs), which is an automated
market maker rule that specifies what constitutes a valid trade. These CFMMs have emerged as
a critical component of DEXs since they ensure that smart contracts can be credibly executed.7
Closely related to our work is literature on crypto-asset creation, distribution and marketing, in
particular airdrops (Arslanian, 2022). Harrigan et al. (2018) analyze the privacy risks of cross-chain
airdrops, Wahby et al. (2020) design a privacy preserving airdrop approach where recipients are
not publicly visible, and Frowis and Bohme (2019) explore the technical design space of airdrops
on the Ethereum platform. Liebi (2022) studies the effects of airdrops and hard-forks on the price
of the parent asset8 . The authors find that on average the prices of the parent assets drop by
roughly 5% after the distribution event. Li et al. (2021) study success factors of airdrop campaigns.
Their findings suggest an increased investment probability, especially for investor that where not
already convinced by the projects agenda. Our work extends on the previous works on the success
7
Numerous publications explore the design space of Automated Marked Maker (AMM) based exchanges (Engel
and Herlihy, 2021) and their use (Gudgeon et al., 2020; Lehar and Parlour, 2021; Qin et al., 2020; Lo and Medda,
2021). Automated market makers require less interactions than limit order book approaches and are thus cheaper
to operate on decentralized ledger platforms. We are not aware of any study trying to assess success criterion of
exchanges.
8
The asset used for selecting the recipients of the newly distributed asset
of airdrops by focusing on the impact of airdrops of governance tokens on the user engagement
and market capitalization of DEXs in particular. Our results are also consistent with Cong et al.
(2022) who study the effects of the OmiseGo airdrop on Ethereum, which led to an increase in
The majority of the current data systems in finance are centralized, meaning they typically depend
are designed to work without a single powerful party. This means all information stored and
The consensus protocols leading to this agreement are often based on proof-of-work (POW)
hard” mathematical problems to disambiguate potentially conflicting transactions and ensure that
only legitimate transactions are added to the ledger. In POS mechanisms, there are validators
who do so in proportion to their level of holdings. These consensus mechanisms contrast with
contract incompleteness in conventional markets where sellers are unable to charge prices that are
contingent on the success of the goods or services: decentralized consensus mechanisms executed
through smart contracts can allow for contracting and enforcement (Cong and He, 2019). However,
smart contracts in DEXs are no panacea. Dating back to theoretical work by Shleifer and Vishney
(1997), arbitrage opportunities can emerge in DEXs through a phenomenon called front-running.
These are settings where arbitrage transactions are executed by “running in front of” other trades
based on the information made public through the DEX (Daian et al., 2020).
There are several advantages of decentralized systems. First, decentralization means, by defini-
tion, that no single party has full control over the system. An ideal decentralized system would
not require a dedicated supervisory authority as every user could convince herself that every
transaction on the system is valid with respect to the rules laid down in a computer program
open to everyone. While many practical systems fall short of this vision despite employing some
decentralized technology (Böhme et al., 2020; Gencer et al., 2018), and empirical studies fail to
identify decentralization as a strong driver of Bitcoin adoption (Abramova and Böhme, 2016), the
structure of these decentralized systems can still provide users with greater autonomy.
Second, decentralized finance can help reduce private knowledge and compliance costs by
empowering users to trade a greater variety of tokens based on their local knowledge and without
as large platform fees, which can in turn promote greater competition. In centralized exchanges,
the platform decides what tokens are traded and user fees, which can create monopoly power.
Third, since users have more autonomy under open and decentralized systems, they can copy
and modify the computer program at any point in time (Reibel et al., 2019). This facilitates
continuous improvement and ever improving efficiency. For example, the Sushiswap DEX is a
copy of Uniswap’s code adding reward mechanisms for users, which in some sense redistributes
However, as in conventional finance, markets and platforms are network goods whose success
is contingent on a critical mass of users (Economides, 1993). Unlike the technology, a user base
cannot be copied. In the absence of technological leaps, attracting new users is the non-trivial
part of establishing a decentralized platform. This applies to cryptocurrencies and DEXs alike.
Airdrops refer to a “marketing strategy” that involves sending small sums of new currencies (i.e.,
tokens) to active cryptocurrency traders in exchange for promotions (e.g., social media). By
increasing awareness of new currencies, and consequently strengthened the base of active users.
More recently, this practice has been adopted by DEXs and may help normalize them.
In 2010, Gavin Andresen, a Bitcoin developer, launched a thread on Reddit to give away five
Bitcoin with the sole intent of driving greater interest and visibility towards it.10 After the “Bitcoin
faucet” demonstrated at least some success, others adopted the idea. In 2014, a new cyptocurrency
tried to bootstrap its user base by giving its tokens to every active Bitcoin, Litecoin, and Dogecoin
user (Harrigan et al., 2018). Airdrops became mainstream with the rise of Ethereum around 2016,
which made it particularly easy for user to create their own cryptoasset, a so called token. At this
stage, airdrops were increasingly given in exchange for social media services that drive visibility
and momentum in the market. That is, in exchange for greater publicity, early users are rewarded
with tokens (and encouraged to retain them) that may appreciate over time.
However, airdrops can also signal a currency’s weakness. For example, investors could interpret
the distribution of tokens as evidence that there are certain technological or design flaws in the
newly minted coins pocketed by the developers—has been launched shortly after the introduction of Zcash 2016.
10
https://cointelegraph.com/news/reddit-reminisces-defunct-bitcoin-faucet-website-that-gave-
away-19-700-btc-for-free
currency that would otherwise prevent them from attracting new users. Much like the theoretically
ambiguous effects of coupons with physical goods, airdrops may not necessarily help the currency.
Furthermore, airdrops can also get in the hands of “yield farmers:” highly speculative traders who
are merely trying to maximize return and may not have the long-term welfare of the protocol in
mind. These token holders may rapidly cash out after a run-up in the price of tokens, leading
to volatility. We investigate whether airdrops can explain the growth in exchange volumes–or
whether it is merely creating froth and increasing the volatility of the market.
Decentralized exchanges can avoid maintaining any legal identity, helping insulate them from
regulation and enforcement actions. To gain user trust given the lack of a corporate body in charge
of governance, they leverage distributed control mechanisms through smart contracts. Holders of
exchange issued governance tokens can vote on decisions—e.g., adding or removing trade pairs,
the fee and vote structure of the exchange, or whether or not to conduct an airdrop—much as
common stock holders of a publicly traded company are empowered to vote on the firm’s decisions.
Because of the potential benefits of avoiding a legal identity, all DEXs are implicitly expected
to eventually issue governance tokens, although the timing varies. A critical governance question
for decentralized exchanges relates to the initial distribution of governance tokens. A popular
option is to reward early adopters with governance tokens relative to their contribution to the
success of the exchange. However, some DEXs have opted to keep authority centralized in an
exchange’s incipient stages and to only later decentralize governance. For instance, Uniswap,
which was founded in 2018, only airdropped its governance tokens in September 2020.
Our data-set is derived from CoinGecko, a popular data aggregation site for cryptocurrency and
exchange data. We use their API to collect information about exchanges and their associated
tokens, acquiring information on a total of 477 exchanges that are listed on CoinGecko at the time
of writing.11 We subsequently restrict our sample for our statistical analysis to 336 exchanges
that have non-missing control variables, leading to 71 that are decentralized and 261 that are
centralized. For every exchange, we additionally obtain the daily trading volume. CoinGecko
reports all volume and price data in BTC. We use daily price data acquired from CoinDesk to
convert the BTC prices to USD using daily and monthly mid-point prices to create monthly
quantities. Since cryptocurrency assets are quite volatile (Liu and Tsyvinski, 2021), this helps
ensure that our measure of market capitalization is not as contaminated by volatility in prices.
Recall from Figure 1, which plots the log of the average trade volume per exchange for DEXs
and centralized exchanges over time, that DEXs have a much more rapid growth rate in volume
even though centralized exchanges trade a higher volume overall. Furthermore, Figure 2 plots the
creation of new exchanges by year, showing that the bulk of DEXs emerged since 2020.12 We
will examine whether these empirical patterns are also present in conditional correlations shortly,
To analyze the impact of exchange tokens and airdrops, we compile a list of exchanges that
have an associated token. The list was created in a semi-automatic fashion. We first obtained the
11
The list was fixed on the 15th of July, 2021. We plan to publish the data-set and code after publication.
12
Since the year an exchange was established is not always available, we impute using the date that the first data
point is available.
list of all tokens available on CoinGecko. Then, we match exchanges and tokens based on their
name and symbol using a fuzzy text matching approach (Chang and Lampe, 1991). To avoid false
negatives, we set the matching threshold conservatively such that we could remove false matches
through a manual quality control process that mitigates the possibility of measurement error.
We also investigate whether a token was ever airdropped through an online search: we looked
for airdrop announcements on exchange websites and blogs; popular news sources, like CoinDesk,
CoinMarketCap, and CoinTelegraph; and sites like Airdrop Alert, which informs crytocurrency
traders about impending giveaways. Based on our research, we created a dummy variable which
takes on the value of 1 if an exchange has offered an airdrop and zero otherwise. We record the
date of an exchange’s first airdrop so this variable is dynamic for our panels.
CoinGecko delineates tokens between exchange tokens and governance tokens. Exchange to-
kens are native to a particular DEX or centralized exchange; they are designed to increase that
exchange’s liquidity. Often, exchange tokens are awarded to particularly active users. By using
exchange tokens, traders usually receive discounts on transaction fees. Governance tokens provide
DEX users with many of the same benefits, but they also accord holders with voting power. In
our analysis, we include two dummy variables: one that takes on the value of 1 if a token is an
exchange token; and one that takes on the value of 1 if a token is a governance token.
We turn to a smaller sample of 51 exchanges that have consistent data on the year the exchange
was launched and both market capitalization and volume over time. Of these 51 exchanges, 36
are centralized and 15 are decentralized. Within these set of 51 exchanges, 5 have released a
governance token, 28 of them have done an airdrop, and 2 of them have done both an aidrop
with a governance token. Some exchanges share the same token, e.g. the Uniswap token serves as
token for three different versions of the Uniswap exchange listed on CoinGecko. We group together
exchanges that use the same governance token into one entity for convenience.13
To understand the difference in volume and price dynamics between centralized and decentralized
where yit denotes an outcome for exchange i in period t, DEX denotes an indicator for whether
the exchange is decentralized, X denotes a vector of controls about the exchange, yt−1 denotes a
one-month lagged value of the outcome variable, and λ denotes month-by-year fixed effects.14 We
cluster standard errors at the exchange-level to allow for autocorrelation in our errors across time.
Our primary concern in estimating Equation 1 is that γ might be biased downwards due to
unobserved heterogeneity. For example, trading activity will tend to be lower in DEXs since they
are newer. However, we address this concern by controlling for the date that the exchange was
created, the trust rank, and by exploiting the panel structure of the data with an autoregressive
component. The trust score is calculated by CoinGecko using a combination of several factors: (i)
web traffic of the exchange, (ii) orderbook spread & +/- 2% depth, (iii) overall trading volume,
(iv) trade frequency, and (v) outlier checks.15 We also control for the year that the exchange was
13
We have also experimented with a more narrow restriction for non-missing trust scores and that only has a
minor effect on our quantitative results by making them less statistically significant. A challenge here is the small
sample, which remains a limitation that should be overcome in future research with more data.
14
The coefficient on DEX could reflect a handful of factors that differentiate DEXs from centralized exchanges,
including fewer know-your-customer (KYC) requirements and lower fees.
15
https://blog.coingecko.com/trust-score-explained/
Moreover, our inclusion of time fixed effects purges aggregate variation, like Federal Reserve
policy, that may affect the volume and price dynamics of both centralized and decentralized ex-
changes. Finally, the autoregressive component controls for some of the unobserved heterogeneity
that could be present. We nonetheless recognize that there could be unobserved factors correlated
with whether an exchange is decentralized and the growth and/or number of transactions.
Table 1 documents these results under varying layers of controls. Starting with the raw volume
differences in column 1, we see that DEXs have nearly 330% fewer trades per month. Controlling
for the year that an exchange was established linearly or through fixed effects brings the differences
down to roughly 180-200% (columns 2 and 3). Trust rank, not surprisingly, enters negatively
since a higher trust rank implies that an exchange is less trustworthy and, therefore, attracts
fewer buyers. However, both of these regressions contain a significant amount of unobserved
producing a substantial drop in the coefficient estimate to a 12.6% difference in volume remains
regression when the outcome variable is the month-to-month growth rate. Column 5 shows that
DEXs grow 16.6 percentage points faster per month than their CEX counterparts. The coefficient
declines to a 7.8pp difference when we control for the trust score and age of the exchange (column
6), although it is not statistically significant at conventional levels. Column 7 instead uses fixed
effects on the year the exchange was established, suggesting that DEXs grow 15pp faster. Finally,
column 8 controls for an autoregressive component as a proxy for any persistence and unobserved
Of note, the invariance of our coefficient estimates to the inclusion of additional controls
suggests that omitted variables are an unlikely culprit behind our results. In particular, Pei
et al. (2019) introduce the concept of the “coefficient comparison test” where the coefficient on the
variable of interest is compared across specifications that vary in their extent of controls. That our
coefficient only changes from a 16.6pp marginal effect to 14.8pp as we move from an unconditional
But, to what extent are our results contaminated by omitted variables? While we do not have
a more exhaustive set of controls, we can apply the selection on unobservables approach in Oster
(2019) who shows how to compute the amount of selection required to qualitatively change the
result. Consider our preferred estimate in column 8, which shows that DEXs exhibit 14.8pp higher
volume growth than their counterparts (R-squared = 0.11). Suppose, however, that the maximum
R-squared is 0.20, which is reasonable in light of the incredible volatility and idiosyncracy inherent
in the cryptocurrency market. Then, we find that the corresponding δ is 0.68, which tells us that
the unobservables have less effect on the coefficient of interest than the observables.17
Our results are also related with those from Moore et al. (2018) who examine closures of
roughly 80 Bitcoin exchanges between 2010 and 2015. They find, for example, that higher-volume
exchanges and that those derive their value from trading less popular currencies are less likely to
close. They also show that experiencing a breach is highly predictive for whether the exchange
16
These coefficients are somewhat sensitive to where the distribution is winsorized. Since only 10% of the sample
is a DEX, if we winsorize at, say, the 5th and 9th percentiles, we start to cut out the key variation on DEXs.
17
Convention is to use a maximum R-squared that is 1.3 times as large as the R-squared in the baseline, i.e. 0.13
here. If we instead assume that selection on unobservables is equal to that of observables, we find a δ equal to 0.11.
closes. To the extent that decentralized exchanges also are able to maintain better security, we
As we discussed earlier, theoretical evidence highlights the role of decentralized governance tokens,
which have increased dramatically in value, and airdrops in explaining the rising prominence of
DEXs. First, governance tokens reward users with the authority to influence how the protocol
develops, which is one of the major attractions of DeFi over centralized exchanges. Second, airdrops
further reward early users with tokens, which can serve as a signal and generate momentum based
To quantify the effect of these twin forces over our monthly sample of exchanges between 2018
where yit denotes the growth of the token’s market capitalization or the exchange’s volume
over the course of month t, DEX denotes an indicator for whether the exchange is decentralized,
α denotes an indicator for whether the exchange had an airdrop, X denotes a vector of controls,
and λ denotes year and month fixed effects. Our vector of controls includes dummies on when
the exchange was created (normalized to 2014), an indicator for distributing a governance token,
and logged volume (when the outcome variable is market capitalization growth). Controlling for
vintage year is particularly important. Older exchanges are more likely to have conducted an
airdrop than newer exchanges. If we did not control for the year an exchange was established, we
may erroneously attribute vintage year heterogeneity to the effect of airdrops. When our outcome
is market capitalization growth, we restrict the sample to observations where the exchange is
strictly positive level of market capitalization. We cluster standard errors at the exchange-level to
significant difference between centralized and decentralized exchanges. Moreover, in column 2, the
indicator on airdrops is positive and suggests these exchanges have 2.2% higher market capital-
ization growth, but it is not statistically significant. However, when we allow for heterogeneity in
airdrops, we find meaningful effects: column 3 shows that DEXs that do an airdrop have a 13.1pp
higher market capitalization growth rate and column 4 shows that airdrops with a governance
token have a 14.9pp higher market capitalization growth rate (both significant at the 10% level).
Governance tokens without an airdrop, however, are linked with lower growth (column 4).18
Next, we focus on the monthly growth rate in volume as our outcome. Column 5 replicates
our main result that DEXs have higher volume growth, but those with governance tokens have
lower volume growth, although the coefficient is statistically insignificant. Column 6 adds airdrop
as a control, finding that there is a positive 9.3pp increase in the growth rate of volume. However,
columns 7 and 8 now allow for heterogeneity: DEXs that do an airdrop have a 8.6pp additional
increase in the growth rate of volume, although it is not statistically significant at conventional
levels, and airdrops with a governance token have a 25.4pp increase in the growth rate of volume,
which is significant at the 10% level. In sum, we find that airdrops are positively associated with
growth in volume and there is suggestive evidence that they are most effective for DEXs.19
18
These results can be somewhat sensitive to how the sample is winsorized, but they are qualitatively robust.
For example, winsorizing at the top and bottom percentile produces a similar estimate, but it is not statistically
significant at the 10% level.
19
While it is too early to say, in future work we will explore how the timing of airdrops and disbursement of
To further investigate the importance of governance rights, we estimate these regressions sep-
arately for CEXs and DEXs. Since governance rights generally are much more limited in scope
and airdrops have less potential impact on governance for CEXs, by construction, they serve as a
useful control group. We find a coefficient of 0.091 (p-value = 0.206) on airdrop for DEXs and a
coefficient of 0.007 (p-value = 0.859) for CEXs, conditional on all our standard controls and time
fixed effects. Furthermore, our results are consistent with the results from Cong et al. (2022) on
Like before, however, there could still be selection effects that cause us to overestimate or
underestimate the coefficient on the interaction effect, as well as the three indicators (DEX, an
airdrop, and governance). For example, we recognize that the decision to conduct an airdrop
that strategy could be correlated with other determinants of market capitalization and volume
growth. Following a similar strategy as in Section 4, we adopt the approach by Oster (2019) and
compute the amount of selection that would be needed to bias our results. In particular, using
an R-squared of 0.72, which is twice that of 0.36 in the baseline specification when the growth in
market capitalization is our outcome variable, we find a value of δ equal to 0.57 corresponding
with the airdrop × DEX interaction. Again, that tells us that the unobservables have less effect
A final concern with these results is that we may fail to take into account heterogeneity in the
type of airdrop, particularly whether it uses bounty criteria to determine who receives the airdrop,
governance tokens influences the dynamics of market capitalization and volume for different types of exchanges,
along the lines of Catalini and Tucker (2017) who explore how the distribution of Bitcoin among early users affects
the diffusion of the cryptocurrency.
thereby creating potential upwards bias on “pump and dump” tokens. Such criteria generally
prescribe a set of steps that the user must follow on social media, such as retweeting or posting
about the exchange, in order to “earn” their bounty (i.e., the airdrop). We collected additional
data on when the airdrop was released and whether there was a bounty attached to the airdrop.
First, only 7% of our sample has a bounty in it and having a bounty is negatively correlated with
being a DEX. Second, including an indicator for having a bounty as a control does not alter the
results. If anything, airdrops with bounties have slightly less growth in market capitalization.
Third, we examine the potentially persistent effect of airdrops. If the positive effects are driven
by bounties, then we should see a transient or null effect after controlling for exchange and time
fixed effects. Instead, we find that there is a 7.9pp increase in market capitalization growth after
the airdrop even though volume growth declines. In sum, we conclude that heterogeneity in the
type of airdrop is likely to create attenuation bias, but does not appear to be biasing our results.
Some have speculated that the growth of DEXs is due to unchecked speculation and not their
inherent characteristics. For example, Jay Powell acknowledged that the Federal Reserve’s easy
money policies had contributed to a “frothy” cryptocurrency market (Greifeld and Hajric, 2021).
Other detractors, such as David Jevans, the CEO of blockchain analytics firm CipherTrace have
argued that many users of DEXs are trying to evade reuglations and “launder money through the
system.” On the other hand, DEX evangelists like Uniswap CEO Hayden Adams, maintain that
the growth of DEXs can be attributed to their “inherent advantages” over centralized exchanges,
such as their better security and greater autonomy (Alloway and Weisenthal, 2021). Particularly
about the underlying drivers of DEX growth, and whether it is driven by regulatory arbitrage, is
To investigate, we exploit two plausibly exogenous events that highlight the differences between
where yit denotes the volume growth rate of exchange i over the course of day t, ξ denotes
the growth rate of DEXs in the period after (relative to before) the breach, and ζ and λ denote
exchange and time fixed effects. Unlike before, our inclusion of exchange fixed effects purges all
time-invariant heterogeneity across exchanges, isolating the effect of the events on decentralized
versus centralized exchanges. We also present the results without these exchange fixed effects.21
The first event we examine is the breach of KuCoin, a centralized exchange, in September 2020.
Hackers discovered the private keys used to manage KuCoin’s centralized hot wallets and stole
$280 million in cryptocurrency (Hui and Zhao, 2020). The attack is emblematic of a key security
flaw of centralized exchanges—money can be stolen if an exchange’s private keys are discovered—
that does not beset DEXs, which have non-custodial architectures that rely on smart contracts,
20
We only examine volume growth for the 10 days before and the 10 days after the shock for two reassons. First,
cryptocurrency markets are heavily traded and react very quickly to new information (Bleher and Dimpfl, 2019).
Second, compressing our pre- and post-periods reduces the chances that there is a concomitant shock that biases
our results.
21
Admittedly, one limitation of these results is that there has been tremendous growth in 2021 and we do not
examine any of the latest large-scale events. We leave this for future work that draws on more recent data.
not hot wallets. The second event we study is a rare joint letter written by three US regulatory
agencies, the Securities and Exchange Commission (SEC), the Commodity Futures Trading Com-
mission (CFTC), and the Financial Crimes Enforcement Network (FinCEN), which was released
on October 11th, 2019. The letter urged digital currency exchanges to adhere to regulations that
proscribed financial institutions from enabling money laundering and terrorist financing. Both of
these events were unanticipated and reflect the dynamic and fast-paced cryptocurrency environ-
ment, allowing us to assess the potential drivers behind abrupt changes in growth.
Starting with column 1 in Table 3, we find that the KuCoin breach increased DEX growth by
13 percentage points (pp) in our specification with time fixed effects but without exchange-level
fixed effects. This coefficient is statistically significant at the 10% level. With time and exchange
fixed effects in column 3, we estimate that the breach increased growth by 14pp; this coefficient
is statistically significant at the 5% level. These results provide evidence that DEX growth has in
The second event we examine is the joint-letter from the three U.S. regulatory agencies urging
digital exchanges to adhere to regulations, like know-your-customer (KYC) standards and make
a “reasonable effort” to verify that their clients are not engaging in illegal activity before they
are allowed to open an account and complete transactions. However, because DEXs do not take
custody of cryptocurrency assets and often lack order books, they are beyond the jurisdiction of
many regulators. Furthermore, most governance token holders of DEXs have proved resistant to
implementing KYC compliance checks. Consequently, a study by CipherTrace found that in Oc-
tober of 2020, a full year after this joint letter, 81% of DEXs had weak or no KYC practices. More
Whether we include or exclude time and exchange fixed effects, the joint letter had no dif-
ferential effect on DEXs. One possibility is that DEX users feared that the letter portended
greater regulatory scrutiny of all cryptocurrency exchanges. Another possibility is that regulatory
arbitrage is not the primary driver behind crypto transactions. While more research is needed,
these results are consistent with the interpretation that growth in DEX adoption is not necessarily
driven by regulatory arbitrage, but rather the features (or at least perception) of these decentral-
ized institutions, including: greater privacy and security, more agency in determining the future
of the exchange, and an opportunity to share in the rewards of the exchange through airdrops.
Our results have important implications for macroeconomic and monetary policy. In particular,
we find little evidence that growth in DEXs is driven by regulatory arbitrage. Rather, the growth
is largely explained by systematic differences in strategy (e.g., the use of governance tokens and
airdrops) and security (e.g., avoiding the use of hot wallets). Attempts to counter the heightened
demand for cryptocurrency through regulation or monetary policy would likely be ineffective.
Nonetheless, our analysis has several limitations. First, our time series is inherently limited
to only a few years. Time will tell whether DeFi continues to grow and whether airdrops have
positive effects on token market capitalization and exchange volume, but if central banks pursue
more inflationary and volatile policies, it is likely that DeFi will only become more important
and widely used. Second, our results are not fully causal. While we have controlled for the most
obvious threats to identification, such as aggregate shocks (i.e., the overall upward trajectory of
cryptocurrency) and the age of an exchange, full causality in this setting is not currently possible.
In particular, the exchanges that conduct airdrops may be different from the exchanges that don’t
in unobserved ways that we cannot detect. Finally, the CoinGecko data might contain errors. They
are self-reported by market participants via APIs, without the quality assurance of audited corpo-
rate disclosures let alone official statistics. While some peer review takes place, small exchanges
and tokens may over-state their business without being discovered. Nonetheless, CoinGecko is
currently the best data aggregator for cryptocurrencies and exchanges. Raw blockchain records,
the alternative source of verified data used in technical measurement studies (e.g., Victor and
Lüders (2019); Daian et al. (2020)), are insufficient to answer the research questions posed here.
7 Conclusion
Decentralized finance has rapidly expanded over the past few years. Decentralized exchanges, a
major application of DeFi, are beginning to surpass centralized exchanges in transaction volume, at
least in terms of their growth rates. Drawing upon the full suite of distributed ledger technologies
(DLTs) that arguably have transformational potential for organizations and societies (Iansiti and
Lakhani, 2017), decentralized exchanges seem to offer at least four advantages over centralized
ones: (i) fewer cybersecurity risks through incorruptible program code and the absence of “hot
wallets,” (ii) greater privacy through laxer or non-existent know-your-consumer (KYC) laws, (iii)
lower fees and wider support for new coins and tokens that are often not supported by centralized
changes, and (iv) greater user flexibility for those who do not like DEX policies and wish to switch.
These advantages come with the downsides of less efficient market mechanisms which leave users
exposed to front-running attacks (Daian et al., 2020; Qin et al., 2020) and the lack of support for
Motivated by these differences, this paper provides the first comprehensive quantitative assess-
ment of DEXs relative to their counterparts, by assembling the most comprehensive time series
data available on exchanges between 2018 to 2021. Using the new data, we document that DEXs
have experienced more growth than their centralized exchange counterparts, despite processing a
lower volume of transactions, even controlling for differences in trust and age of the exchange. To
understand this rapid growth, we investigate the role of airdrops, or the disbursement of tokens
to existing and potential users, and finding that they are associated with growth in capitaliza-
tion and volume primarily for DEXs. DEXs experience greater volume growth when they airdrop
governance tokens which provide holders with the right to vote on the exchange’s policies. This
suggests that cryptocurrency users value when governance rights are transferred to loyal cryp-
tocurrency users. It suggests that airdrops are not just marketing tools; they are also means of
delegating decision-making authority to active cryptocurrency users who will serve as custodians
of exchanges. Nonetheless, we caution that more data is needed to see whether these patterns
remain, especially over the boom and bust cycles of the crypto market.
We also examine whether the growth in DEXs could be a result of regulatory arbitrage or fears
about the cybersecurity of centralized exchanges. We find evidence that DEX growth accelerated
in the wake of a cyber attack on a centralized exchange; we find no evidence that DEX growth
increased after an unexpected letter indicating that centralized exchanges would be subject to
more regulatory scrutiny. In sum, we provide the first comprehensive empirical evidence behind
Our paper nonetheless leaves many questions open for future research. First, how will different
countries regulate DeFi? Since the smart contracts can be executed from anywhere and are fully
interoperable, legal statutes will have to catch up to technology. Second, what will the composi-
tion of the user base look like, particularly as the use of airdrops and governance tokens attract
a possibly more diverse set of users? Third, how will organizations potentially adopt cryptocur-
rencies into their payment systems, and what benefits will these confer in heightened productivity
and/or usability? Fourth, what new security risks might emerge with DEXs, particularly with the
potential for front-running at scale? Only time will tell what the future will hold for DeFi, but
our results here highlight its growth and the various factors at play.
10
DEX
log10 (Total Volume in USD / # Exchanges )
CEX
9
5
2019-01 2019-05 2019-09 2020-01 2020-05 2020-09 2021-01 2021-05 2021-09
Date
Source: CoinGecko, 2019-2021. The figure plots the monthly logged (base 10) exchange-level average traded volume across centralized
and decentralized exchanges from January 2019 to August 2021.
100
50
0
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
est_year_established
Source: CoinGecko, 2019-2021. The figure plots the number of newly established cryptocurrency exchanges by year.
Notes.—Sources: CoinGecko, 2018-2021. The table reports the coefficients associated with regressions of logged (base 2) transaction
volume on an indicator for whether the exchange is decentralized, a one-day lag of the outcome variable, the trust score rank, either
a linear trend in the year the exchange was established or fixed effects on it (normalized to 2011-12), and month-by-year fixed effects.
Monthly growth in volume is winsorized at the first and last percentiles. Standard errors are clustered at the exchange-level and
observations are unweighted. *** denotes significant at the 1% level, ** denotes significant at the 5% level, * denotes significant at the
10% level.
Established in 2017 -.003 .020 .056 .027 .117 .043 .061 .050
[.016] [.046] [.046] [.044] [.090] [.096] [.098] [.096]
Established in 2018 -.050∗ -.004 .013 .003 .087 .051 .059 .061
[.026] [.043] [.045] [.044] [.092] [.094] [.095] [.095]
Established in 2019 .083∗∗ .095∗∗∗ .124∗∗∗ .111∗∗∗ .312∗∗∗ .246∗∗ .262∗∗ .268∗∗
[.034] [.034] [.038] [.038] [.107] [.114] [.118] [.115]
Established in 2020 .101∗ .116∗∗ .182∗∗∗ .159∗∗ .316∗ .270∗ .307∗ .337∗∗
[.055] [.051] [.053] [.063] [.161] [.147] [.162] [.167]
R-squared .34 .35 .36 .36 .16 .16 .16 .16
Sample Size 940 940 940 940 1327 1327 1327 1327
Time FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes.—Sources: CoinGecko, 2018-2021. The table reports the coefficients associated with regressions of monthly growth in market
capitalization and transaction volume on an indicator for whether the exchange is decentralized, an indicator for whether there is a
governance token disbursed, indicators for whether the exchange had one air drop, an interaction between decentralization and airdrops,
as well as an interaction of airdrops and governance token in columns 4 and 8, and a series of controls, including: logged volume (when
the outcome is market capitalization growth), and dummies on the year the exchange was created (normalized to 2014). The growth
rates are winsorized at the 1st and 97th percentiles. Our sample excludes observations where the market capitalization is zero. Standard
errors are clustered at the exchange-level and observations are unweighted. *** denotes significant at the 1% level, ** denotes significant
at the 5% level, * denotes significant at the 10% level.
Note: —Sources: Coingecko, 2019-2020. The table documents the event study results associated with regressions of the daily trading
volume growth on an indicator for whether the exchange is a decentralized exchange (DEX), an indicator for whether the day is on or
after the specific event has happened, their interaction, and controls. The first shock we explore is the KuCoin breach, which highlighted
a security flaw of centralized exchanges that does not plague decentralized exchanges. The second shock is a joint letter written by the
SEC and two other U.S. regulators notifying centralized exchanges that they must follow customer compliance verification measures.
We allow for a window of 10 days before and after the shock. We control for the trust score rank of the exchange, which measures the
trust based on trading volume, orderbook spread, trade frequency, and more, computed by CoinGecko. Scale from 1-10 where 10 is the
best. *** denotes significant at the 1% level, ** denotes significant at the 5% level, * denotes significant at the 10% level.
References
Abramova, S. and Böhme, R. (2016). Perceived benefit and risk as multidimensional determinants
Alloway, T. and Weisenthal, J. (2021). Hayden Adams Explains Uniswap and the Rise of DeFi.
https://www.bloombergquint.com/onweb/transcript-hayden-adams-explains-uniswap-
Angeris, G. and Chitra, T. (2020). Improved price oracles: Constant function market makers. In
Proceedings of the 2nd ACM Conference on Advances in Financial Technologies, pages 80–91.
Arslanian, H. (2022). Crypto-asset creation and distribution. The Book of Crypto, (277-289).
Aspris, A., Foley, S., Svec, J., and Wang, L. (2021). Decentralized exchanges: The “wild west” of
Bleher, J. and Dimpfl, T. (2019). Today i got a million, tomorrow, i don’t know: On the pre-
Bohme, R., Christin, N., Edelman, B., and Moore, T. (2015). Bitcoin: Economics, technology,
Böhme, R., Eckey, L., Moore, T., Narula, N., Ruffing, T., and Zohar, A. (2020). Responsible
Boudjemaa, A. (2020). Cross-chain interoperability: Enabling the future of defi. Hacker Noon.
Bowater, J. (2018). Decentralized exchanges are the solution to hackings and customer privacy.
CityA.M.
Catalini, C., Boslego, J., and Zhang, K. (2018). Technological opportunity, bubbles and innovation:
Catalini, C. and Gans, J. S. (2019). Initial coin offerings and the value of crypto tokens. NBER
working paper.
Catalini, C. and Tucker, C. (2017). When early adopters don’t adopt. Science, 357(6347):135–136.
Chang, W. I. and Lampe, J. (1991). Theoretical and empirical comparisons of approximate string
California, Berkeley.
Chod, J. and Lyandres, E. (2020). A theory of icos: Diversification, agency, and information
https://ciphertrace.com/wp-content/uploads/2020/10/CipherTrace-2020-Geographic
Cong, L., Tang, K., Wang, Y., and Zhao, X. (2022). Inclusion and democratization through web3
and defi? initial evidence from the ethereum ecosystem. Working paper.
Cong, L. W. and He, Z. (2019). Blockchain disruption and smart contracts. Review of Financial
Studies, 32(5):1754–1797.
Cong, L. W., Li, X., Tang, K., and Yang, Y. (2021). Crypto wash trading. working paper.
Cong, L. W. and Wang, N. (2021). Token-based platform finance. Journal of Financial Economics,
forthcoming.
Cowen, T. (June 9th, 2021). Crypto finally has a reason to exist: Decentralized finance. Bloomberg.
https://www.gemini.com/cryptopedia/sushiswap-uniswap-vampire-attack#section-w
Daian, P., Goldfeder, S., Kell, T., Li, Y., Zhao, X., Bentov, I., Breidenbach, L., and Juels, A.
(2020). Flash Boys 2.0: Frontrunning in decentralized exchanges, miner extractable value, and
consensus instability. In IEEE Symposium on Security and Privacy, pages 910–927. IEEE.
Engel, D. and Herlihy, M. (2021). Composing networks of automated market makers. arXiv
preprint arXiv:2106.00083.
Foley, S., Karlsen, J. R., and Putnins, T. J. (2019). Sex, drugs, and bitcoin: How much illegal
Frowis, M. and Bohme, R. (2019). The operational cost of ethereum airdrops. International
Gencer, A. E., Basu, S., Eyal, I., Van Renesse, R., and Sirer, E. G. (2018). Decentralization in
Greifeld, K. and Hajric, V. (2021). Powell Breaks Out the “Froth” Word When Asked About
Markets.
https://www.bloomberg.com/news/articles/2021-04-28/powell-breaks-out-the-froth
Gudgeon, L., Werner, S., Perez, D., and Knottenbelt, W. J. (2020). Defi protocols for loanable
funds: Interest rates, liquidity and market efficiency. In Proceedings of the 2nd ACM Conference
Harrigan, M., Shi, L., and Illum, J. (2018). Airdrops and Privacy: A Case Study in Cross-
Harvey, C. R., Ramachandran, A., and Santoro, J. (2021). DeFi and the Future of Finance. John
Howell, S. T., Niessner, M., and Yermack, D. (2020). Initial coin offerings: Financing growth with
Hui, A. and Zhao, W. (2020). Over $280M Drained in KuCoin Crypto Exchange Hack.
https://www.coindesk.com/markets/2020/09/26/over-280m-drained-in-kucoin-crypto
Iansiti, M. and Lakhani, K. R. (2017). The truth about blockchain. Harvard Business Review.
Jahani, E., Krafft, P. M., Suhara, Y., Moro, E., and Pentland, S. (2018). Scamcoins, s*** posters,
and the search for the next bitcoin: Collective sensemaking in cryptocurrency discussions. Pro-
Lehar, A. and Parlour, C. A. (2021). Decentralized exchanges. Technical report, working paper,
Li, J., Wan, X. S., Cheng, H. K., and Zhao, X. (2021). Operation dumbo drop: To airdrop or not
Liu, Y. and Tsyvinski, A. (2021). Risks and returns of cryptocurrency. Review of Financial
Studies, 34(6):2689â2727.
Lo, Y. C. and Medda, F. (2021). Uniswap and the emergence of the decentralized exchange. SSRN
working paper.
Makarov, I. and Schoar, A. (2021). Blockchain analysis of the bitcoin market. NBER working
paper.
Moore, T., Christin, N., and Szurdi, J. (2018). Revisiting the risks of bitcoin currency exchange
Oster, E. (2019). Unobservable selection and coefficient stability: Theory and evidence. Journal
Pei, Z., Pischke, J.-S., and Schwandt, H. (2019). Poorly measured confounders are more useful on
the left than the right. Journal of Business & Economic Statistics, 37(2):205–216.
Pieters, G. and Vivanco, S. (2017). Financial regulations and price inconsistencies across Bitcoin
Qin, K., Zhou, L., Livshits, B., and Gervais, A. (2020). Attacking the DeFi ecosystem with flash
Reibel, P., Yousaf, H., and Meiklejohn, S. (2019). Short paper: An exploration of code diversity in
the cryptocurrency landscape. In Goldberg, I. and Moore, T., editors, Financial Cryptography
and Data Security, volume 11598 of Lecture Notes in Computer Science, pages 73–83. Springer.
Shleifer, A. and Vishney, R. W. (1997). The limits of arbitrage. Journal of Finance, 52(1):35–55.
Goldberg, I. and Moore, T., editors, Financial Cryptography and Data Security.
Wahby, R. S., Boneh, D., Jeffrey, C., and Poon, J. (2020). An airdrop that preserves recipient
privacy. In International Conference on Financial Cryptography and Data Security, pages 444–
463. Springer.