Non-fungible tokens (NFTs) have been in the spotlight for several years now following a trading frenzy and spectacular returns. An NFT is a digital record of ownership, created and stored on a blockchain, and thus tamper-proof and auditable. Blockchains are decentralized ledgers of transactions that are designed to run perpetually, hence providing the basis for imperishable NFT ownership. While NFTs are often linked to images (CryptoPunks and Bored Apes are the most famous examples), they can actually relate to any type of underlying asset, be it digital or physical. Given that NFT transaction records are stored on a publicly accessible database that is not controlled by a central authority, they permit instantaneous and auditable trading that circumvents slow and costly traditional infrastructure. In our recent paper, we explore the functionality and architecture of NFTs and examine the primary and secondary markets for this emerging asset class. We discuss NFT minting, trading and different types of NFT exchanges. We explain NFT liquidity provision through NFT staking, borrowing, and lending. Finally, based on a sample of 2 million NFT transactions from various NFT categories, we analyze the NFT risk and return characteristics during the boom and bust cycles.
NFT Markets
NFT assets are introduced to the market by the creators themselves without relying on the intermediary. The process is termed “NFT mint” or “NFT drop,” which are used interchangeably to indicate NFT production and the transfer of ownership to the first owner in the primary sale. The initial price (the “mint” price) is typically relatively low, or even zero, with the aim to boost participation (for example, CryptoPunks were given away free of charge). Thus, the measure of success of a collection does not relate to the offering price (as e.g. in a traditional initial public offering), but rather depends on the speed of sale at launch. The launch of a new collection (“NFT drop”) takes place on a specific date and hour. Usually, potential buyers have the option to register for the NFT drop in advance, in a process called “whitelisting” or “allowlist.” Whitelisting serves two main purposes: first, it allows creators to gauge the interest for the collection and second, it promotes the equitable distribution of new assets, since high net-worth buyers can crowd out ordinary retail investors by paying higher transaction fees in order to transact faster. For the retail buyer, being in the allowlist usually indicates a lower mint price vis-à-vis the price set for public sale later on.
After the drop, NFTs are traded on the secondary NFT exchanges, which are typically decentralized and interact directly with the blockchain. Each transaction takes place via a smart contract that registers the change in NFT ownership, which means that exchanges do not rely on market makers that hold NFT inventory. The largest and most successful NFT exchange to date is OpenSea. As of 31 October 2022, the lifetime volume on OpenSea amounted to $32.3 billion. OpenSea allows anyone to put up NFTs for sale and select the price (fixed or auction) and the duration of the listing. Auctions include English-type auctions (where the NFT is sold to the highest bidder) as well as Dutch-type auctions, where the price falls over time during the duration of the auction according to a predetermined schedule. As a result, the NFT is sold to the highest (lowest) paying bidder depending on the auction type, with the seller retaining the right to abstain from the sale regardless of the outcome of the auction. Many collections have their own “exchanges,” which permit NFT trading from a single specific NFT collection. Finally, traditional auction houses ventured relatively early into the NFT universe. For example, Christie’s has launched its own NFT exchange in collaboration with OpenSea (Christie’s X OpenSea). In addition, Christie’s has sold the most expensive NFT to date (“Everydays: The First 5000 Days” by Beeple for over $69 million) and has auctioned several CryptoPunks as well as NFTs from other collections and creators.
Liquidity Provision
Similar to traditional assets, NFTs can be used as collateral for loans, typically denominated in fungible tokens (cryptocurrencies). Due to high market volatility, the loan-to-value percentage is low (usually around 20%), using as reference the value of the lowest priced NFT from the specific collection. Since loan terms are typically non-standard, the claim to the pledged asset (NFT) takes the form of an NFT as well. When the loan is initiated, both the borrower and the lender receive an NFT that represents their claim to the loan collateral. This smart contract ensures that ownership of the underlying asset is transferred to the borrower (if the loan is repaid) or is turned over to the lender (if the borrower defaults on the loan). In this manner, NFTs can be used by their holders to provide liquidity to the market.
Liquidity provision for NFT holders can also be ensured by NFT “staking.” When staking an NFT, the asset is turned into a fungible token that can be instantly cashed in by the staker. This ensures liquidity by (a) adding the staked NFTs to the overall market supply and (b) providing the original holder with the option to exit the investment at any time, without reclaiming the NFT but the monetary equivalent of it. NFT staking can take the form of liquidity staking or inventory staking, with the two processes demonstrating a number of important differences. In inventory staking, the staker locks the NFT in the smart contract and receives a yield, whereas in liquidity staking, NFT gets locked in a smart contract together with some amount of ETH. Unlike inventory staking, the rewards to the staker (accrued fees) are not compounded and can be cashed-in at any time. Upon exit, the staker may redeeem less NFTs and more ETH depending on the supply and demand.
Empirical analysis
We mine the OpenSea API for data on NFT transactions. Since the number of collections is enormous, including many unsuccessful ones, we focus on the leading blue-chip NFTs including CryptoPunks, Bored Ape Yacht Club, Mutant Ape Yacht Club, Bored Ape Kennel Yacht Club, Axie Infinity, Sorare, Art Blocks, Sandbox, Decentraland, and the Ethereum Name Service (ENS). We classify them into five distinct groups. Our data include information on price, currency, type of sale (e.g., auction), token ID, collection ID, timestamp, as well as the buyer and seller (by wallet address). Summary information for our sample is presented in Table 1.
Table 1. NFT Collections Examined
Note: The table reports ten leading blue-chip NFT collections that can be classified into different categories including NFT Profile Picture (PFP), NFT Gaming, NFT Art, NFT Metaverse, and NFT “Other.” Market Cap is the NFT market capitalization calculated as the sum of the most recent transacted prices across all NFTs within the NFT collection. Market Cap denominated in USD is based on ETH/USD exchange rate as of 30 June 2022. Exp. stands for “expanding supply.” The data are derived from OpenSea API.
CryptoPunks is the most valuable NFT collection to date, worth close to half a million ETH (the equivalent of a half a billion USD at the 30 June 2022 exchange rate1), followed by Bored Ape Yacht Club (BAYC) valued at 349,738 ETH (370 million USD). Together with its two follow-up derivative collections (BAKC and MAYC), BAYC market capitalization amounts to 570,202 ETH (616 million USD). The remaining NFT collections in our sample are an order of magnitude smaller. Among them, two stand out: Sandbox (NFT Metaverse) (56 million USD) and Sorare (NFT Game) (39 million USD). Somewhat surprising is the relatively low valuation of Axie Infinity (22 million USD), especially given that it was created in 2018, long before other well-known NFT collections. Overall, CryptoPunks and BAYC dominate the entire NFT secondary market with the market capitalization dominance levels of 40% and 29%, respectively. CryptoPunks have the historical advantage of being the first NFT collection of its kind and were launched in 2017. BAYC has a strong community and focuses on developing high-quality follow-up collections as well as a metaverse.
Table 2 shows the turnover and volume of the NFT collections in our sample. The median NFT turnover (Panel A) for Profile Picture (PFP) collections is 2 (CryptoPunks, BAYC, MAYC), implying that a typical PFP NFT is traded twice during our sample period. However, given that CryptoPunks were launched in 2017 (so they have been available for trading much longer), the turnover per unit of time is significantly lower compared to e.g., BAYC, which was launched in 2021. This observation is confirmed by the total number of transactions per NFT collection (Column 7), where CryptoPunks have fewer transactions compared to all other PFP collections (the total turnover for BAYC is 42% higher – the “hard” supply for both collections is identical and equal to 10,000). In terms of value (Panel B), CryptoPunks have a higher trading volume (Column 7) and much higher mean and median valuations (Columns 2 and 3). It is important to notice that a typical CryptoPunk is about 7 times more valuable than a typical BAYC, with similar values for the maximum transaction prices in the respective price distributions (Column 6), thus confirming the dominance of CryptoPunks in the overall NFT market.
Table 2. NFT Turnover and Volume in the Secondary Market
Note: The table presents NFT turnover and NFT volume in the NFT secondary market. NFT Turnover (volume) is the number (value) of NFTs traded since the release of the NFT collection. NFT Volume is based on all available prices per NFT since the NFT collection launch. We omit the last recorded transaction on CryptoPunk #9989 that sold for 124,457.1 ETH, as this transaction has been documented to be a wash-trade. The data are derived from OpenSea API.
The remaining NFT collections follow different market activity patterns. First, quite surprisingly (given their much lower valuations), they are typically traded only once, as indicated by the median number of transactions (Panel A, Column 2). However, their total turnover is significantly higher than PFP NFTs, primarily due to their substantially larger supply and much different utility. This concerns mainly the in-game items (Axie Infinity), football cards (Sorare), land plots (Sandbox and Decentraland), and NFT art (Art Blocks). These collections also show substantially lower NFT volume, mainly due to their relatively low average values. We must note at this point that in the primary market, prices are uniformly low across all NFT collections, regardless of the category.
Table 3. NFT Collection Returns
Note: The table presents returns to NFTs. Trades are matched by NFT ID. Two is the minimum number of sales necessary to calculate the return. Values in the table can be converted to percentages when multiplying by 100%. Trade-by-trade return is the return per NFT sale. Buy-and-hold return is the return calculated by dividing the price of the last sale by the price of the first sale minus one. We omit the last recorded transaction on CryptoPunk #9989 that sold for 124,457.1 ETH, as this transaction has been documented to be a wash-trade. The data are derived from OpenSea API.
Next, we compute the returns to the NFTs within each collection and present them in Table 3. Returns are computed in two different ways: on a trade-by-trade basis (Panel A) and as buy-and-hold returns (Panel B). Since each NFT asset is distinguishable, we compute trade-by-trade returns as the average return across all trades for the specific NFT (based on all consecutive sale prices). On the other hand, we compute buy-and-hold returns as the ratio of the last sale price divided by the first price (minus unity). Consequently, the holding period in buy-and-hold returns may vary depending on the specific NFT. A buy-and-hold return can be thought of as the geometric return based on the intermediate trade-by-trade returns.
The average returns to CryptoPunks and BAYC are spectacular and amount to roughly 1,600% and 1,300% per trade, respectively (Panel A, Column 2). Among other NFT collections, Art Blocks, ENS, and Sandbox yield an average return ranging from 220% to 160% per trade. Nevertheless, average returns are sensitive to outsized returns in the upper tail of the distribution, particularly for CryptoPunks and BAYC (Column 6). Consequently, the median (Column 3) can provide more reliable information about typical returns in the NFT market. Median returns to PFP NFTs are still high, but the differences among the categories are not that large anymore. For example, BAYC yields 83% at the median, whereas Art Blocks and Sandbox 47% and 37%, respectively. The latter collections outperform CryptoPunks (31%). The worst performing NFT collection is Axie Infinity that generates -3% per NFT sale. Decentraland performs only slightly better (3%).
On the buy-and-hold basis (Panel B), BAYC outperforms CryptoPunks both at the mean and median, and both NFT collections outperform other NFT categories. The median return from holding BAYC is roughly 2,000%, with CryptoPunks generating 270%, whereas the two derivative collections of BAYC (BAKC and MAYC) yield 152% and 93%, respectively. NFTs that belong to other categories do not demonstrate such high buy-and-hold returns compared to NFT PFPs, but nevertheless, their returns remain quite substantial. For example, Art Blocks and Sandbox yield a median buy-and-hold return of 95% and 83%, respectively. The worst performers are gaming NFTs, with Sorare earning 11% and Axie Infinity -6%.
Further, we compute returns by splitting the sample into bull and bear market cycles, with the latter period starting from 9 May 2022. This point in time marks the Terra stablecoin de-peg from USD, which caused a contagion effect between various cryptocurrency start-ups and cryptocurrencies themselves. We also estimate returns on a market-adjusted basis, by benchmarking raw NFT returns against returns to Bitcoin (BTC). These approaches yield similar results, as we show that BAYC is the best performing NFT collection during the bear market period, which quite interestingly did not include any trading activity within Art Blocks and Axie Infinity. In addition, the market-adjusted analysis suggests that NFTs deliver higher returns than cryptocurrencies (as proxied by BTC) in the bull market.
Conclusion
Our work shows that NFTs can be regarded as fairly liquid and high-performing investment assets with a rapidly developing NFT markets for trading, liquidity provision, borrowing and lending. Given our categorisation of NFTs, we find that profile picture (PFP) NFTs dominate the market in terms of market capitalization, volume, and returns, with returns to other NFT categories (art, gaming, metaverse) being markedly lower but still substantial. For example, buy-and-hold returns to BAYC and CryptoPunk NFTs are close to 2,000% and 270%, respectively (calculated at the median). In addition, NFTs seem to beat the cryptocurrency market and deliver positive returns even during the bear market.
Mieszko Mazur is an Associate Professor of Finance at the ESSCA School of Management.
Efstathios Polyzos is an Assistant Professor of Finance at Zayed University.
This post is adapted based on the chapter “Non-Fungible Tokens (NFTs) as an Investment Vehicle”, from the book “A Companion to Decentralized Finance, Digital Assets, and Blockchain Technologies”, by Edward Elgar Publishing and available on SSRN; data used in this study are available from the authors upon request).
