PaperlayerView exact revision on arXiv

Source-linked AI summary

Crypto Wash Trading

Lin William Cong, Xi Li, Ke Tang, Yang Yang

arXiv:2108.10984v1econ.GN

TL;DR

The paper addresses limited systematic evidence on the prevalence of wash trading across cryptocurrency exchanges. It applies statistical and behavioral tests to exchange trading patterns and finds that unregulated exchanges exhibit widespread manipulation, with wash trading accounting for most reported volume and estimates constrained by test adaptability and data scope.

  • Problem

    Systematic evidence was limited on whether wash trading was widespread across cryptocurrency exchanges and how regulation related to it.

  • Method

    The study applies Benford’s law, trade-size clustering, power-law tests, hypothesis testing, and exchange-level volume estimation across regulated and unregulated exchanges.

  • Results

    Over 70% of reported volume on unregulated exchanges is estimated to be wash trading, while regulated exchanges generally display regular trading patterns.

  • Takeaways & Limitations

    Fabricated volume improves exchange rankings, relates to short-term prices and exchange characteristics, and motivates regulatory supervision and caution when using reported volumes.

  • Takeaways & Limitations

    The tests are not exhaustive, wash traders may adapt to them, and the short observation window limits conclusions about volatility-related patterns.

Abstract

from arXiv · show

We introduce systematic tests exploiting robust statistical and behavioral patterns in trading to detect fake transactions on 29 cryptocurrency exchanges. Regulated exchanges feature patterns consistently observed in financial markets and nature; abnormal first-significant-digit distributions, size rounding, and transaction tail distributions on unregulated exchanges reveal rampant manipulations unlikely driven by strategy or exchange heterogeneity. We quantify the wash trading on each unregulated exchange, which averaged over 70% of the reported volume. We further document how these fabricated volumes (trillions of dollars annually) improve exchange ranking, temporarily distort prices, and relate to exchange characteristics (e.g., age and userbase), market conditions, and regulation.

1 Introduction

The paper develops systematic statistical tests to establish that wash trading is a widespread, industry-wide problem on cryptocurrency exchanges. It finds abnormal trading patterns and estimates that unregulated exchanges fabricate most reported volume, affecting rankings, prices, and market oversight.

  • 1 Introduction: Unregulated exchanges show abnormal first-digit, round-number, and transaction-tail patterns, while regulated exchanges broadly follow established financial-market regularities.Tier-1 exchanges fail more than 20% of tests and Tier-2 exchanges more than 60%, with results robust to joint hypothesis testing.
  • 1 Introduction: Over 70% of reported volume on unregulated exchanges is estimated to be wash trading, totaling over 4.5 trillion USD in spot markets and over 1.5 trillion USD in derivatives during Q1 2020.The average estimate exceeds 70% after accounting for observable exchange heterogeneity.
  • 1 Introduction: Wash trading improves exchange rankings: 70% fabricated volume can move an exchange up 46 positions.The study also links wash trading to short-term cryptocurrency prices and price dispersion across exchanges.
  • 1 Introduction: Wash trading is more common on newer exchanges with smaller userbases, while it is positively predicted by returns and negatively predicted by price volatility.The paper reports that longer-established exchanges and those with larger userbases wash trade less.
  • 1 Introduction: The proposed tools support exchange regulation, third-party supervision, and more cautious use of reported volumes in empirical research.The authors caution that their tests are not exhaustive and wash traders may adapt their strategies.
  • 1 Introduction: The study fills an evidence gap by providing an industry-wide exchange-level analysis using Benford’s law, trade-size clustering, and Pareto-Levy tail distributions.It responds to earlier industry analyses described as imprecise, ad hoc, unscalable, and non-transparent.

2 Data and Summary Statistics

The study assembles transaction and exchange-level data for 29 cryptocurrency exchanges, classifying them by regulation and web traffic. Unregulated exchanges often report larger volumes and higher rankings than regulated exchanges, while regulatory definitions produce broadly robust findings.

  • Data sources: The study obtains transaction records through official exchange APIs and combines them with exchange characteristics, web traffic, rankings, and reported volumes.The transaction records include identifiers, timestamps, prices, traded amounts, and pair symbols.
  • Data: The sample contains 448,475,535 transactions covering BTC, ETH, XRP, and LTC across 29 exchanges.The four cryptocurrencies represent over 60% of volume and are available on almost all exchanges.
  • Exchange classification: Exchanges are classified into three regulated platforms, 10 Tier-1 unregulated platforms, and 16 Tier-2 unregulated platforms using NYSDFS status and web traffic.Tier-1 platforms rank within SimilarWeb’s top 700 investment websites, while the remainder are Tier-2.
  • Robustness: Alternative regulatory definitions leave the main findings robust, although UT1, UT5, and UT8 resemble regulated exchanges more closely than average unregulated platforms.These exchanges pass most tests, but UT5 and UT8 still differ from regulated exchanges in trade-size roundness and exceed 50% estimated wash trading on average.
  • Reported activity: Unregulated exchanges can report far larger volumes than regulated exchanges, including U4’s 50,944 million USD versus R2’s 15,212 million USD.Reported volumes vary substantially across unregulated exchanges.
  • Exchange rankings: Seven unregulated Tier-2 exchanges rank in CoinMarketCap’s top 20 by trading volume, outperforming most Tier-1 and regulated exchanges.Trading-volume rankings are widely used despite not fully representing exchange quality or liquidity.

3 Empirical Evidence of Wash Trading

The paper detects wash trading by testing exchange trade sizes against three established statistical and behavioral regularities. Because these benchmarks reflect fundamental patterns, the exchange-level tests are designed to be relatively insensitive to authentic heterogeneity across traders and platforms.

  • Testing framework: The study tests four major trading pairs against three statistical and behavioral benchmarks to identify fake transactions.The pairs are BTC/USD, ETH/USD, LTC/USD, and XRP/USD.
  • Evidence: The exchange-level tests provide robust evidence of wash trading on unregulated exchanges.The authors use multiple tests and control for heterogeneous but authentic trading behavior before quantifying wash trading.

3.1 Distribution of First Significant Digits

The paper evaluates whether trade-size leading digits follow Benford’s law, using deviations as evidence of potential manipulation. Regulated exchanges conform consistently, whereas violations are concentrated among unregulated exchanges, especially Tier-2 platforms.

  • Benford’s law: Benford’s law predicts the distribution of first significant digits, and deviations can indicate anomalous or manipulated transaction data.The law is applied to trade-size leading digits because it is robust to changes in accounting units and has been used to test data reliability.
  • Figure 1: In Figure 1, regulated R2 assigns 32.75% of BTC trades and 32.73% of ETH trades a leading digit of 1, near Benford’s 30.10% benchmark.Unregulated exchanges such as U8 and U9 show visibly different first-digit distributions.
  • Unregulated exchanges: Nine Tier-2 exchanges diverge from Benford’s law, with seven violating it for at least two cryptocurrencies.Other unregulated exchanges also show sizable cross-asset differences.
  • Exchange-level tests: Regulated exchanges and most Tier-1 unregulated exchanges pass the chi-squared tests, but UT3 fails for BTC and XRP while several Tier-2 exchanges diverge across assets.UT3’s BTC and XRP inconsistencies are significant at 1%, while U5, U7, U8, U9, and U14 diverge for most cryptocurrencies.
  • Overall results: All regulated exchanges conform to Benford’s law, while 20% of Tier-1 and 50% of Tier-2 unregulated exchanges violate it for at least one cryptocurrency.The comparison is based on 5% significance-level tests.

3.2 Trade Size Clustering

The study tests whether authentic trading exhibits round-number clustering and declining trade-size frequencies. Regulated exchanges show these regularities, while many unregulated exchanges display weak, irregular, or absent clustering.

  • Unregulated exchanges: Unregulated Tier-2 exchanges show little round-size clustering and highly irregular trade-size distributions, including gaps, cliffs, and non-monotonic frequencies.U8, U9, and U14 provide representative examples across multiple cryptocurrencies.
  • Tier-2 exchanges: For U14, several cryptocurrency distributions are approximately uniform, implying similar trade frequencies across sizes rather than the usual financial-market pattern.Uniformity appears in LTC and XRP and across larger BTC and ETH observation ranges.
  • Tier-1 exchanges: Three Tier-1 exchanges show significant clustering at some round sizes, but clustering at multiples of 500 units is not significant for the reported comparison.UT3, UT7, and UT9 exhibit positive and significant differences at the 1% level for available cryptocurrencies, with exceptions noted in the tests.
  • Overall results: Regulated exchanges show evident trade-size clustering, whereas 30% of Tier-1 and 50% of Tier-2 unregulated exchanges show no clustering across all cryptocurrencies.The regulated pattern includes prominent peaks at round sizes and a downward-sloping distribution.

3.3 Tail Distribution

Power-law tail behavior provides a benchmark for detecting anomalous cryptocurrency trading. Regulated exchanges generally match Pareto–Lévy patterns, whereas many unregulated exchanges do not.

  • The study fits trade-size tails using OLS and maximum likelihood methods, including log-log estimation and exponent calculations.
  • Regulated exchanges show stable power-law decay across all four cryptocurrencies, with estimators in the Pareto–Lévy regime.
  • 90% of unregulated Tier-1 exchanges resemble power-law tails, although UT4 and UT6 fall outside the Pareto–Lévy range across four cryptocurrencies.
  • 50% of Tier-1 exchanges display Pareto–Lévy tail exponents across all cryptocurrencies, compared with 75% of Tier-2 exchanges that fail the benchmark.
  • 75% of unregulated Tier-2 exchanges fail to follow the Pareto–Lévy power law commonly observed in financial markets.

3.4 Selection Concern, Multi-hypothesis Testing, and Conclusive Evidence

The authors combine multiple statistical tests to address selection concerns and assess whether exchange trading patterns conform to regularities in traditional financial markets. Joint testing confirms normal patterns on regulated exchanges but widespread failures on unregulated ones.

  • More than half of unregulated exchanges fail at least half of all tests, and the three tests complement one another in identifying wash trading.
  • Regulated exchanges pass all tests, while unregulated exchanges as a whole fail more than 40% of tests for each cryptocurrency.
  • Fisher’s method combines individual p-values into a global test to reduce concerns that multiple tests increase Type I error or enable p-hacking.
  • 75% of Tier-2 unregulated exchanges fail the universal trade-pattern tests, with BTC showing the highest failure rate.
  • The evidence argues that trader or algorithmic heterogeneity is unlikely to explain the exchange differences, although self-selection cannot be completely ruled out.

4 Quantifying Wash Trading

The paper estimates wash trading by comparing observed unrounded trading volume with a benchmark for legitimate activity derived from regulated exchanges. The estimates indicate extensive fabricated volume on unregulated exchanges.

  • Unregulated exchanges exhibit lower trade-size roundness than regulated exchanges, especially in Tier-2 markets, with differences significant at the 1% level for nearly all cryptocurrencies.
  • The estimator treats excess unrounded volume beyond the legitimate benchmark as wash-trading volume and provides a time-varying first-order benchmark.
  • Regulated-exchange cross-validation estimates less than 5% wash-trading volume, supporting the no-wash-trading benchmark.
  • Wash trades average over 70% of total volume on each unregulated exchange, including 53.4% for Tier-1 and 81.7% for Tier-2 exchanges.
  • After controlling for exchange characteristics, estimated wash trading averages about 61%, while Tier-1 and Tier-2 estimates are 12.9% and 18.5% in the alternative comparison.
  • Unrounded trades on most unregulated exchanges fail Benford’s-law tests, indicating they are unlikely to be predominantly authentic algorithmic trades.

5 Wash Trading Incentives, Impacts, and Implications

Wash trading is associated with exchange visibility incentives, short-term price distortions, and exchange characteristics such as age and userbase. The authors emphasize that the observational evidence does not establish causality.

  • Exchange-level evidence suggests exchanges themselves wash trade directly or indirectly, while the study makes no causal claims because of data limits.
  • 70% wash trading can move an exchange’s ranking up by more than 25 positions relative to its rank without wash trading.
  • Wash-trade volume is positively associated with contemporaneous returns and price deviations, but prices reverse in the following week.
  • Unregulated exchanges over five years old wash trade 48.12% of reported volume, compared with 82.89% for newer exchanges.
  • Wash trading is negatively associated with exchange age and unique visitors, while recent positive returns and lower volatility predict higher wash-trading volume.

6 Conclusion

Statistical and behavioral benchmarks distinguish regulated exchanges from pervasive wash trading on unregulated platforms. The study estimates substantial fabricated volume and links it to exchange rankings, prices, market conditions, exchange characteristics, and regulation.

  • 6 Conclusion: Regulated exchanges follow Benford’s law, show round-size clustering and transaction roundness, and exhibit Pareto–Lévy tail decay.These patterns are presented as regularities in financial markets and nature.
  • 6 Conclusion: Nearly 30% of unregulated exchanges violate Benford’s law, while 20% of Tier-1 and 75% of Tier-2 exchanges fail to follow Pareto–Lévy trade-size distributions.Unregulated exchanges also show less prominent trade-size clustering and generally lower transaction roundness.
  • 6 Conclusion: 53.4% of trading on unregulated Tier-1 exchanges and 81.8% on Tier-2 exchanges is estimated to be wash trading.The study reports robustness and validation tests alongside these estimates.
  • 6 Conclusion: Wash trading is suggestively associated with inflated exchange rankings, distorted cryptocurrency prices, past prices, volatility, exchange age, and userbase.The reported relationships are predictive or suggestive rather than stated as universal causal effects.
  • 6 Conclusion: The paper presents the first comprehensive academic study of crypto wash trading as an industry-wide phenomenon and highlights regulation’s screening effects.It also emphasizes wash-trading-adjusted volume and statistical and behavioral tools for forensic finance and fraud detection.

Trade Volume

The paper tests whether cryptocurrency trade volumes exhibit established statistical regularities and uses deviations from them to estimate wash trading and examine its market effects.

  • Benford’s Law: Regulated exchanges’ trade sizes conform more closely to Benford’s law, whereas unregulated exchanges show significant distributional inconsistencies.The comparison covers BTC/USD, ETH/USD, LTC/USD, and XRP/USD across regulated and unregulated exchange groups.
  • Trade-size Clustering: Round trade sizes occur more frequently than nearby unrounded sizes on regulated exchanges, providing a benchmark for testing size clustering.The clustering tests compare frequencies at round sizes with the highest nearby unrounded frequencies.
  • Power-law Fitting: Trade-size tails are evaluated with OLS and MLE power-law fits, including whether estimated exponents fall within the Pareto–Lévy range.The tests use four cryptocurrency trading pairs for each exchange.
  • Multiple Hypothesis Testing: The study combines Benford, clustering, and power-law tests to evaluate whether exchange trade patterns match universal financial-market regularities.The global null is rejected when Fisher’s combined statistic exceeds the relevant threshold.
  • Wash Trading Estimates: Over 70% of reported volume was estimated as wash trading on average across unregulated exchanges.The estimates use round versus unrounded trade volumes, exchange characteristics, and bootstrapped standard deviations.
  • Price Effects: Wash trading volume is positively and significantly associated with cryptocurrency price returns in the reported regressions.The price-impact analysis also compares unregulated exchange prices with contemporaneous averages from regulated exchanges.

U9 BTC/USD ETH/USD XRP/USD

Figures and tests compare trade-size regularities, failed-test rates, volumes, ranks, and wash trading across regulated, Tier-1 unregulated, and Tier-2 unregulated exchanges.

  • U9 BTC/USD ETH/USD XRP/USD: The figures compare regulated, Tier-1 unregulated, and Tier-2 unregulated exchanges across BTC, ETH, LTC, and XRP trading pairs.The displayed analyses include digit distributions, size clustering, tail fits, failed tests, volumes, and ranks.
  • U9 BTC/USD ETH/USD XRP/USD: Failed tests are defined as nonconformity with Benford’s law, absent clustering at multiples of 100 units, or power-law exponents outside (1, 2).Failure percentages aggregate tests at the 5% significance level.
  • U9 BTC/USD ETH/USD XRP/USD: Figure 6 relates logarithmic trade volumes to exchange ranks, with the fitted relationship reporting an adjusted R2 of 93%.The fitting uses OLS regression and reports estimated coefficients with t-statistics.
  • U9 BTC/USD ETH/USD XRP/USD: Figure 7 compares estimated wash-trading fractions with improvements in counterfactual exchange ranks based on estimated real volume.Rank improvement is the difference between the counterfactual rank and the reported CoinMarketCap rank.

Appendix A. Institutional Background of Crypto Exchanges: Development and Regulation

The appendix describes the growth, institutional role, weak regulation, and manipulation risks of cryptocurrency exchanges, including how wash trading can be conducted and detected.

  • Development and Regulation: Cryptocurrency exchanges are centralized gateways between fiat and decentralized cryptocurrency systems and hold a dominant role in the industry.The ecosystem includes exchanges alongside mining, payment companies, wallets, and decentralized applications.
  • Development and Regulation: More than 300 exchanges operated globally amid fierce competition and loose regulatory standards.Exchanges offer similar products and services while new competitors continue to emerge.
  • Development and Regulation: Unregulated exchanges need not report trading records to authorities, while data aggregators and market participants encourage greater transparency.Algorithmic trading and API-based market data create practical incentives to provide trade information.
  • Manipulation Risks: The paper reports that many exchanges engaged in wash trading, likely to improve rankings or attract customers.The appendix frames this behavior within broader information asymmetry and limited consumer protection.
  • Manipulation Risks: Exchanges may manipulate markets through fake records, self-filled orders, trading robots, or user incentives such as fee rebates and transaction mining.Combining these practices makes specific wash trades difficult to detect from transaction history alone.

Appendix B. Exchange Names

The appendix lists exchange codes and names for regulated, Tier-1 unregulated, and Tier-2 unregulated exchanges in the dataset.

  • Exchange Names: The dataset includes three regulated exchanges, ten Tier-1 unregulated exchanges, and sixteen Tier-2 unregulated exchanges.The listed exchanges include Bitstamp, Coinbase, Gemini, Binance, Bittrex, Bitfinex, and the named Tier-2 platforms.
Loading 2108.10984v1…