Key Takeaways
- • Selection bias fundamentally distorts NFT returns: Observed transaction prices reflect only completed sales where sellers actively chose to transact, systematically excluding losses held by underwater investors—a statistical illusion that has plagued illiquid asset markets (housing, art, venture capital) but proves particularly severe in NFTs due to extreme infrequency of trading and asset uniqueness.
- • Portfolio scale is critical to positive returns: Simulations demonstrate that an investor requires a minimum portfolio of approximately 400 NFTs to achieve a 90% probability of earning positive returns, indicating the market's extreme right-skewed return distribution and concentration of gains among large holders.
- • Blockchain data reveals market structure: Goetzmann et al. (2026) leverage complete blockchain transaction records to correct for the disposition effect—investors' documented tendency to hold losers and sell winners—applying methodologies previously validated in fine art markets (Korteweg et al. 2016) to quantify actual NFT investment performance.
- • The 2021 boom masked underlying economics: The explosive 2021 NFT market, characterized by record sales at Christie's and Sotheby's, celebrity endorsements, and media hype, collapsed by late 2022, but the fundamental performance metrics remained obscured by this selection bias problem throughout the cycle.
- • Cryptocurrency-specific factors drive performance: Liu et al. (2022) and Tsyvinski and Liu (2018) document that momentum and investor attention—cryptocurrency-specific variables—strongly predict NFT returns, yet these factors cannot account for the returns distribution once selection bias is corrected.
Summary
The non-fungible token market's explosive growth in 2021, followed by its dramatic collapse by late 2022, has left a critical analytical gap: what did NFT investors actually earn after accounting for market mechanics? Research published by Goetzmann et al. (2026) addresses this fundamental question by identifying and correcting for a statistical distortion that has long plagued illiquid asset markets but proves particularly acute in NFTs.
The core problem is a selection bias known as the disposition effect—the well-documented tendency of investors to hold losing positions while selling winners. In frequently traded markets, this behavior has modest impact because transaction data captures representative samples of all holdings. However, in NFTs, where assets are unique and trading is infrequent, only completed sales generate observable prices. This means researchers see a curated dataset: transactions that reflect active seller decisions, systematically excluding losses held by investors underwater on their positions. Borri et al. (2022) documented the NFT market's extreme return characteristics, while Liu et al. (2022) demonstrated that cryptocurrency-specific factors like momentum and investor attention strongly predict performance. Yet beneath these surface-level findings lies the more fundamental measurement problem: the returns investors observe are not the returns investors actually received.
Researchers studying illiquid assets—venture capital, private equity, real estate, and fine art—have long recognized this distortion. Korteweg et al. (2016) showed that correcting for selection bias in fine art investment dramatically altered apparent returns. The Goetzmann team applies and extends this methodology to NFTs using blockchain data, which records transaction history with unusual completeness, providing a more granular picture than traditional art markets offer.
The results are sobering for retail NFT investors. Portfolio simulations indicate that an investor needs at least 400 NFTs to achieve a 90% probability of positive returns—a threshold that reveals the extreme right-skewed return distribution in the market. This concentration of gains reflects the reality that while some early investors or large-scale collectors captured outsized returns during the 2021 peak, the median investor faced significant losses once the selection bias is corrected for. The hype cycle of 2021, driven by record auction sales and celebrity endorsement, masked this underlying distribution throughout the market's evolution.
As regulators worldwide develop frameworks for digital assets, this research underscores that policy decisions must be informed by accurate performance data. The apparent returns visible on public markets do not reflect what most investors earned—a distinction with material consequences for investor protection and market confidence in this emerging asset class.
The core problem is a selection bias known as the disposition effect—the well-documented tendency of investors to hold losing positions while selling winners. In frequently traded markets, this behavior has modest impact because transaction data captures representative samples of all holdings. However, in NFTs, where assets are unique and trading is infrequent, only completed sales generate observable prices. This means researchers see a curated dataset: transactions that reflect active seller decisions, systematically excluding losses held by investors underwater on their positions. Borri et al. (2022) documented the NFT market's extreme return characteristics, while Liu et al. (2022) demonstrated that cryptocurrency-specific factors like momentum and investor attention strongly predict performance. Yet beneath these surface-level findings lies the more fundamental measurement problem: the returns investors observe are not the returns investors actually received.
Researchers studying illiquid assets—venture capital, private equity, real estate, and fine art—have long recognized this distortion. Korteweg et al. (2016) showed that correcting for selection bias in fine art investment dramatically altered apparent returns. The Goetzmann team applies and extends this methodology to NFTs using blockchain data, which records transaction history with unusual completeness, providing a more granular picture than traditional art markets offer.
The results are sobering for retail NFT investors. Portfolio simulations indicate that an investor needs at least 400 NFTs to achieve a 90% probability of positive returns—a threshold that reveals the extreme right-skewed return distribution in the market. This concentration of gains reflects the reality that while some early investors or large-scale collectors captured outsized returns during the 2021 peak, the median investor faced significant losses once the selection bias is corrected for. The hype cycle of 2021, driven by record auction sales and celebrity endorsement, masked this underlying distribution throughout the market's evolution.
As regulators worldwide develop frameworks for digital assets, this research underscores that policy decisions must be informed by accurate performance data. The apparent returns visible on public markets do not reflect what most investors earned—a distinction with material consequences for investor protection and market confidence in this emerging asset class.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.