Authors:
(1) Howard Zhong, MIT CSAIL (howardzh@mit.edu);
(2) Mark Hamilton, MIT CSAIL, Microsoft (markth@mit.edu). Table of Links Abstract and 1 Introduction 2 Methodology 3 Results 3.1 CryptoPunks 3.2 Aggregate NFT Market 4 Discussion 5. Conclusion/ Acknowledgements/ References A Appendix A.1 Implementation Details A.2 Detailed NFT Information & A.3 Google NFT Searches Map 3. Results Our goal is to investigate biases in pricing with respect to gender and race for NFT collections. We first analyze CryptoPunks, as it is one of the first NFT collections and is commonly credited for starting the 2021 NFT revolution [20]. The dataset we use is compiled from querying the OpenSea API as described in Section 2.1. 3.1 CryptoPunks While [8] identifies biases in prices for both gender and race for CryptoPunks, we find only darkerskinned CryptoPunks are sold for less than lighter-skinned CryptoPunks. 3.1.1 Gender We found price difference across gender was not significant. In Figure 3a, male and female weekly mean sale price are roughly equal with blue and red lines being roughly level. This conclusion is corroborated by the male and female box plots at similar price levels for different months in Figure 3b. We supplement the visual analysis with statistical anylsis. From 2021-01 to 2022-11, the median selling price of male CryptoPunks (64.9 eth) is greater than median price of female CryptoPunks (64.0 eth), but the mean selling price of male CryptoPunks (63.5 eth) is less than that of female CryptoPunks (63.9 eth). Because the means and medians yield different conclusions, we also analyze the p-value of different one-sided t-tests on whether male prices are greater than female prices. As shown in Table 2, we vary outlier detection percentiles, whether to run unpaired or paired t-test, and whether to apply the log transformation. Only at a 0.1% outlier percentile level do some of the t-tests imply male price is greater than female price. This is reasonable because from Figure 3b, male outliers tend to be higher priced than female outliers. As a precondition for the t-test is the absence of outliers, t-tests at other outlier levels (2.5% and 5%) may be more informative. Thus, there is not enough statistical evidence to conclude there is gender bias in CryptoPunks. 3.1.2 Race Although we do not find gender bias in the pricing of CryptoPunks, we corroborate the visual analysis in [8] with statistical results that lighter-skinned CryptoPunks are valued more than darker-skinned CryptoPunks. We first plot the weekly mean sale prices of Dark and Light CryptoPunks between 2021-01 and 2022-11 in Figure 4a and find that Light CryptoPunks are consistently sold at a higher price than Dark CryptoPunks. Examining the box plot in Figure 4b for CryptoPunks prices across different months, we observe similar results. At a 2.5% trim level, the median sale price for Light CryptoPunks (65.6 eth) is greater than that of Dark CryptoPunks (64.0 eth), and the mean sale price for Light CryptoPunks (66.1 eth) is also greater than that of Dark CryptoPunks (64.5 eth). In addition, from Table 3, every paired t-test across different outlier detection schemes supports this hypothesis with very low p-values, indicating the effect is large. Furthermore, the unpaired t-test results were above 0.05 at 2.5% and 5.0% outlier trim levels may be due to CryptoPunks’ large price variation across time that increases standard deviation and lowers t-stat. Thus, the evidence suggests that Light CryptoPunks are sold for more than Dark CryptoPunks. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Howard Zhong, MIT CSAIL (howardzh@mit.edu); (2) Mark Hamilton, MIT CSAIL, Microsoft (markth@mit.edu). Authors: Authors: (1) Howard Zhong, MIT CSAIL (howardzh@mit.edu); (2) Mark Hamilton, MIT CSAIL, Microsoft (markth@mit.edu). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Methodology 2 Methodology 3 Results 3.1 CryptoPunks 3.1 CryptoPunks 3.2 Aggregate NFT Market 3.2 Aggregate NFT Market 4 Discussion 4 Discussion 5. Conclusion/ Acknowledgements/ References Conclusion/ Acknowledgements/ References A Appendix A Appendix A.1 Implementation Details A.1 Implementation Details A.2 Detailed NFT Information & A.3 Google NFT Searches Map A.2 Detailed NFT Information & A.3 Google NFT Searches Map 3. Results Our goal is to investigate biases in pricing with respect to gender and race for NFT collections. We first analyze CryptoPunks, as it is one of the first NFT collections and is commonly credited for starting the 2021 NFT revolution [20]. The dataset we use is compiled from querying the OpenSea API as described in Section 2.1. 3.1 CryptoPunks While [8] identifies biases in prices for both gender and race for CryptoPunks, we find only darkerskinned CryptoPunks are sold for less than lighter-skinned CryptoPunks. 3.1.1 Gender 3.1.1 Gender We found price difference across gender was not significant. In Figure 3a, male and female weekly mean sale price are roughly equal with blue and red lines being roughly level. This conclusion is corroborated by the male and female box plots at similar price levels for different months in Figure 3b. We supplement the visual analysis with statistical anylsis. From 2021-01 to 2022-11, the median selling price of male CryptoPunks (64.9 eth) is greater than median price of female CryptoPunks (64.0 eth), but the mean selling price of male CryptoPunks (63.5 eth) is less than that of female CryptoPunks (63.9 eth). Because the means and medians yield different conclusions, we also analyze the p-value of different one-sided t-tests on whether male prices are greater than female prices. As shown in Table 2, we vary outlier detection percentiles, whether to run unpaired or paired t-test, and whether to apply the log transformation. Only at a 0.1% outlier percentile level do some of the t-tests imply male price is greater than female price. This is reasonable because from Figure 3b, male outliers tend to be higher priced than female outliers. As a precondition for the t-test is the absence of outliers, t-tests at other outlier levels (2.5% and 5%) may be more informative. Thus, there is not enough statistical evidence to conclude there is gender bias in CryptoPunks. 3.1.2 Race 3.1.2 Race Although we do not find gender bias in the pricing of CryptoPunks, we corroborate the visual analysis in [8] with statistical results that lighter-skinned CryptoPunks are valued more than darker-skinned CryptoPunks. We first plot the weekly mean sale prices of Dark and Light CryptoPunks between 2021-01 and 2022-11 in Figure 4a and find that Light CryptoPunks are consistently sold at a higher price than Dark CryptoPunks. Examining the box plot in Figure 4b for CryptoPunks prices across different months, we observe similar results. At a 2.5% trim level, the median sale price for Light CryptoPunks (65.6 eth) is greater than that of Dark CryptoPunks (64.0 eth), and the mean sale price for Light CryptoPunks (66.1 eth) is also greater than that of Dark CryptoPunks (64.5 eth). In addition, from Table 3, every paired t-test across different outlier detection schemes supports this hypothesis with very low p-values, indicating the effect is large. Furthermore, the unpaired t-test results were above 0.05 at 2.5% and 5.0% outlier trim levels may be due to CryptoPunks’ large price variation across time that increases standard deviation and lowers t-stat. Thus, the evidence suggests that Light CryptoPunks are sold for more than Dark CryptoPunks. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Behind the Scenes: Data Collection and Statistical Tools for NFT Bias Study

NFT Networks Mirror Social Structures with Influencer Nodes, Community Hubs, & Zipfian Distributions

Read My Stories

Too Long; Didn't Read

Make resilience your competitive advantage

Gender-Based NFT Price Differences Found to Be Insignificant

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

How Gender and Race Labels Are Applied to NFT Data Analysis

Behind the Scenes: Data Collection and Statistical Tools for NFT Bias Study

Advancing Data and Equity in NFTs and Shaping a Fairer Metaverse

Lighter-Skinned NFTs Command Higher Prices in the Broader Market

Regional NFT Trends and Insights from 44 Avatar-Based Collections

How Gender and Race Labels Are Applied to NFT Data Analysis

Behind the Scenes: Data Collection and Statistical Tools for NFT Bias Study

Advancing Data and Equity in NFTs and Shaping a Fairer Metaverse

Lighter-Skinned NFTs Command Higher Prices in the Broader Market

Regional NFT Trends and Insights from 44 Avatar-Based Collections

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps