- The paper analyzes the economics of non-fungible tokens (NFTs) by constructing an NFT market index and studying its properties.
- Using a repeat sales regression method, the analysis finds an average weekly return of 2.5% for the NFT market index, with a volatility of 19.2% and a Sharpe ratio of 0.939.
- The results show some exposure between the NFT market and cryptocurrency markets, but little exposure to traditional asset markets. Various factors like volatility, valuation ratios, and volume show some degree of return predictability in the NFT market.
Solution Manual for Financial Accounting, 11th Edition by Robert Libby, Patri...
The Economics of Non-Fungible Tokens: Understanding Returns and Properties of the NFT Market
1. The Economics of Non-Fungible Tokens
Nicola Borri Yukun Liu Aleh Tsyvinski
Luiss Rochester Simon Yale
Algorand Fintech Lab kick-off
September 15, 2022
2. Introduction Data & Methodology Exposures Return Predictability Conclusion
This Paper
• Understand the returns of the overall NFT market
• Construct a NFT index
• Study the properties of the index
1
3. Introduction Data & Methodology Exposures Return Predictability Conclusion
Results Preview
• Repeat sales regression method (RSR)
• 2.5% average weekly return, 19.2% volatility
• 0.939 Sharpe ratio
• Exposures
• Cryptocurrency market: some
• Traditional asset market: no much
• Time-series predictability
• Volatility
• Valuation ratio
• Cross-section predictability
• Size/masterpiece effect
• Reversal effect
2
5. Introduction Data & Methodology Exposures Return Predictability Conclusion
Data
• Major exchanges:
• CryptoKitties
• Gods Unchained
• Decentraland
• OpenSea
• Atomic
• Cross-checked with blockchain
• Aggregate to weekly data
• Trades in ETH, WETH, MANA, USD
3
6. Introduction Data & Methodology Exposures Return Predictability Conclusion
Repeat Sales Regression Method (I/II)
• There are T time periods and sales can occur in any of the periods from 0
to T. We denote t as the subscript for the time period (in our baseline
specification, a time period is a week).
• For a pair of sales of a given NFT i, prices and the NFT market index are
assumed to be related as in the following equation:
Pit0
Pit
=
Bt0
Bt
Uitt0
where Pit is the transaction price of NFT i at time period t.
• For a pair of sales, t is the time at the purchasing transaction and t0
is the
time at the sale transaction, and t0
> t. Bt is the general NFT market price
index at time t and Uitt0 is the multiplicative error term for the price pair as
discussed above.
4
7. Introduction Data & Methodology Exposures Return Predictability Conclusion
Repeat Sales Regression Method (II/II)
The model can then be converted to the logged scale, which is the basis of the
estimation:
rit0t = pit0 − pit = bt0 − bt + uitt0
5
28. Introduction Data & Methodology Exposures Return Predictability Conclusion
Size
Full Sample CryptoKitties CryptoPunks
(t0
− t) ln P -0.004*** -0.001*** -0.011**
(-10.553) (-8.904) (-2.232)
Bored Ape Sup Ducks Decentraland
(t0
− t) ln P -0.003 -0.029** -0.012***
(-0.976) (-2.415) (-21.073)
• Doubling the logged NFT purchase price is associated with a 0.4% decrease in the weekly
return (20.8% annually)
23
29. Introduction Data & Methodology Exposures Return Predictability Conclusion
Serial Dependence
• Testing for serial dependence
rit0t = pit0 − pit = bt0 − bt + γ (t0
− t) ri,b + uitt0
24
30. Introduction Data & Methodology Exposures Return Predictability Conclusion
Serial Dependence
Full Sample CryptoKitties CryptoPunks
(t0
− t) ri,t -0.014*** -0.004*** -0.108***
(-6.517) (-6.970) (-4.072)
Bored Ape Sup Ducks Decentraland
(t0
− t) ri,t -0.027*** -0.051** -0.020***
(-3.537) (-2.150) (-7.078)
• A 10% increase in average weekly return in the NFT’s previous repeat sale is associated
with an 0.14% decrease in the weekly return (7.3% annually)
25