Research

I examine how personal experiences causally impact investor behaviors and market boom-bust episodes by exploiting a unique experimental setting in the non-fungible token (NFT) market. Using blockchain transaction-level data for about 1 million wallets, I find that NFT investors who randomly receive more valuable NFTs in the primary market are more likely to participate in subsequent primary market sales and trade more NFTs in the secondary market. These experience effects spill over to the cryptocurrency market as investors who randomly receive more valuable NFTs purchase more lottery-like cryptocurrencies. I also find that personal experiences and new investor inflows have contributed to the formation of bubbles in the NFT market. A model-free reinforcement learning framework best explains the empirical results.

Technological innovations in decentralized finance have reduced the cost of listing and trading cryptocurrencies, resulting in a proliferation of meme tokens. Using novel blockchain data, we document issuance of more than 300,000 meme tokens in 2021, with total trading volume of more than $30 billion. Words related to animals (e.g. Doge), cartoons (e.g. Spongebob), and celebrities (e.g. Elon Musk) have been featured in token names. Exploiting the rich heterogeneity of such keywords, we find that investors’ interests shift between distinct meme styles, which we define as tokens sharing the same meme keywords. Token issuers cater to investor demand by issuing more tokens with popular meme keywords and profit from such issuances, primarily through exit scams (e.g. “rug pulls”). Consistent with salience theory’s prediction for competitive markets, issuers compete for investor attention by issuing tokens that feature more meme keywords and lower prices. In response to Elon Musk’s tweets on DogeCoin, tokens whose names include “Doge” saw an increase in volumes, prices, and issuances relative to other non-Doge meme tokens. Our findings highlight the social aspects of meme-token investing. We also discuss how our findings are related to meme stocks.

By dynamically tracking which accounts are holding which assets in real-time, we examine the cross-sectional return profile of factors based on investor composition for cryptocurrencies. Leveraging the blockchain data, we construct several novel, observable, and tradable demand factors: investor age, investor trading frequency, and investor diversification. We show that these novel factors generate sizable and persistent return spreads between quintiles, and that such factors are not fully spanned by other existing pricing factors for cryptocurrencies. These findings suggest that an investor’s account characteristics are strongly predictive of their average realized return.