Research

The non-fungible-token (NFT) market, long hailed for frictionless, gatekeeper-free trading, experienced a meteoric rise in 2021–2022 before collapsing abruptly in mid-2022. To identify the forces behind this crash, we analyze 15 million on-chain transactions from 867 leading NFT collections. We estimate that wash trading, a fraudulent self-trade behavior designed to fabricate demand, generated approximately US $33 billion, or 47% of recorded trade volume, and was conducted by just 4% of addresses. Our Local Projections with Instrumental Variables (LP-IV) estimations provide four key results. First, increases in wash-trade volume inflate performance metrics (trade count, trader count, capital inflow, and realized returns) during year 1, forming a speculative bubble. Second, the same activities are linked to significant declines across key performance metrics during year 2, crowding out genuine participation and eroding valuations. Third, these boom–bust dynamics are significantly dampened in collections characterized by a higher share of experienced traders, a larger proportion of long-horizon art collectors, and greater trader-network centrality, implying that informed or socially embedded participants buffer speculative shocks. Fourth, collection-level transaction fees and creator resale royalties deter the entry of wash traders, underscoring the importance of trading frictions while also benefiting creators. Taken together, our findings suggest three policy implications for NFT platforms and regulators: (i) implement real-time wash-trade detection in this largely unregulated market, (ii) disclose market-quality indicators, such as the art-collector ratio and network-centrality scores, to support informed investment decisions and enhance market resilience, and (iii) reconsider recent moves to revoke or make optional collection-level transaction fees and creator resale royalties, given their deterrent effect on wash trading.

Information Systems researchers have led efforts to understand user behaviors in emerging digital markets that often diverge from traditional economic theories (Bapna et al. 2004). However, participant behavior in the burgeoning Non-Fungible Token (NFT) markets remains largely unexplored due to a lack of analytical tools and challenges posed by their low-liquidity nature. This study addresses this gap by developing a novel analytical framework to measure market trends and individual participant performance. Using transaction data from 56,609 NFTs and 18,733 participants over 5 years, we identify a complete boom-bust market cycle and uncover distinct groups of participants. Our findings reveal that, unlike traditional art or investment markets, participants focusing on extremely high-priced NFTs underperform most others. Notably, long-term holders outperform speculators who chase popular NFTs. Furthermore, the success of participants with extensive trading experience is highly dependent on market timing: they only succeed when entering during pre-boom and boom, not bust, periods. Mechanism analyses show that successful groups avoid extrapolating past returns and instead prioritize early participation in primary sales. This study contributes a robust framework for analyzing NFT markets and provides actionable insights into participant behavior during volatile boom-bust cycles.

Blockchain-based Web3 has created a decentralized digital ecosystem that reduces reliance on traditional gatekeepers and intermediaries. This transformation raises a critical question: How does this new environment affect gender and racial minorities in the art market? To address this, we analyze approximately 27,000 sales and 90,000 offers/bids involving 40,000 NFTs created by 2,500 artists on SuperRare, a leading curated NFT art market, proposing a nuanced view of the NFT art market: (i) White male artists still dominate supply, but less than in traditional markets; (ii) non-White artists face larger disadvantages than female artists in both sale probability and price, relative to White male artists; and (iii) self-curation via verifiable quality signals materially mitigates these gaps, with minority artists benefiting disproportionately. Robustness checks exclude supply-side explanations (e.g., artist self-underpricing), pointing to demand-side factors. Despite lower entry barriers from disintermediation, structural inequities endure; however, self-curation enables minority artists to mitigate demand-side disadvantages.

Large language models (LLMs) have attracted significant attention for their potential to enhance knowledge worker productivity. In this study, we provide the first large-scale empirical evaluations of LLMs’ impact on academic research productivity. Leveraging a comprehensive dataset of 4,582 computer science scholars across 194 top U.S. universities and analyzing 218,723 research papers published between 2019 and 2024, we find that the introduction of LLMs is associated with about 8% increase in publication output; a gap that persists across alternative measures, including the first-author publications and top-tier conference papers. Our regression discontinuity in time (RDiT) analysis further reveals that LLMs not only shifted the average publication level but also accelerated the growth rate of productivity, rising to 3.2% in 2023 and 12.8% in 2024. Notably, junior scholars realize stronger gains than their senior counterparts, with the productivity benefit diminishing by roughly 1% for each additional year of experience. Recognizing that LLMs’ benefits may not be uniformly distributed, we also investigate their impact on non-native English-speaking (NNES) researchers, who have historically faced disadvantages in academic writing (Liao et al., 2024). Difference-in-differences and generalized synthetic control analyses indicate that, following LLM adoption, native English-speaking (NES) researchers produced more papers than their NNES counterparts. Overall, our findings indicate that while LLMs significantly boost scholarly productivity, they also exhibit dual effects, lowering barriers for junior scholars while potentially reinforcing linguistic inequities.

Abstract coming soon...

App-based bike-sharing platforms are rapidly transforming urban transportation. This study investigates how bike-sharing platforms influence demand for ride-sharing services, with a focus on platform interaction. To measure this effect, we use spatiotemporally staggered expansions of Citi Bike, New York City’s largest bike-sharing service, as a natural experiment. We examine the treatment effects on ride-sharing services (Uber, Lyft) and Yellow taxi demand across the city. Our novel identification strategy, geographically nearest neighbor matching, is based on 0.38 billion individual trip records. The results reveal a complementary relationship between bike‑sharing and ride‑hailing (Citi Bike stations stimulate nearby ride‑hailing trips) while simultaneously depressing Yellow Taxi demand, pointing to a substitution between ride‑hailing and taxis. This research contributes to the sharing economy literature in Information Systems (IS). To our knowledge, it is one of the first studies to explore interactions between app-based sharing platforms. Understanding how people connect different shared mobility services has never been more important. We argue that this study lays the foundation for future research on sharing-to-sharing mobility interactions.