## New Evidence on Spillovers Between Crypto Assets and Financial Markets — Working Paper No. WP/2023/213

## Source details

**Canonical URL:** [New Evidence on Spillovers Between Crypto Assets and Financial Markets — Working Paper No. WP/2023/213](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023213-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023213-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023213-print-pdf.pdf.json)

---

### Growth, characteristics, and policy concerns
- Crypto assets expanded from a single Bitcoin in 2009 to over 5,000 currently, with total market capitalization in excess of USD 3 trillion towards the end of 2021; the market subsequently declined to around $1.1 trillion.  
- Markets display significant volatility with cycles of rapid growth followed by dramatic collapses, raising policymakers’ concerns about contagion risks to other financial markets and broader macro-financial implications.  
- Focus of analysis: unbacked crypto assets, i.e., assets whose value is not linked to that of another asset and whose prices fluctuate freely driven by supply and demand.  
- Crypto functionalities covered: means of payment, store of value, speculative asset, smart-contract support, fundraising, asset transfer, decentralized finance, privacy, digital identity, governance.

### Channels of interaction with financial markets
- Identified interaction channels:
  - "Flight-to-safety channel": investors allocate to crypto assets during economic uncertainty if cryptos are perceived as safer and offering a hedge; implies potential diversification benefits if correlations are low.  
  - "Speculative demand channel": demand for crypto assets increases during periods of high financial market risk appetite because cryptos offer potential for high returns due to volatility.  
  - Market liquidity, information spillovers, and investor sentiment can produce additional comovement and amplify cross-market linkages.  
- Implication: crypto markets can be either sources of shocks or amplifiers of market volatility, with financial-stability implications.

### Contribution, scope, and methodology
- Time horizon: 2014 to the end of 2022; main sample period for combined analysis: October 14, 2015 to January 1, 2023.  
- Empirical approach: Diebold and Yilmaz connectedness framework (generalized VAR, forecast horizon H = 10 days, lag order p = 3).  
- Rolling-window dynamic analyses: 120-day windows (robust for 100 to 200).  
- Samples:
  - Long sample: starts October 15, 2014; 8 crypto assets.  
  - Short sample: starts August 3, 2017; 23 crypto assets (BTC, BCH, BNB, DASH, DCR, DGB, DOGE, ETH, ETC, GNO, GNT, LTC, MAID, NEO, OMG, REP, SNP, VTC, XEM, XLM, XMR, XRP, ZEC).  
- Financial variables: 15 series including 10Y US Treasury yield, sovereign bond indices (AE/EM investment-grade and high-yield), corporate bond indices, MSCI World, S&P500, MSCI EM, USD effective exchange rate, gold spot price, oil price, DJ Commodity Index, BoFA MOVE index, VIX.  
- Data sources: CoinMetrics for crypto prices; Bloomberg for financial variables.

### Key statistical properties (prices, returns, volatilities)
- Price-level examples and distributional features:
  - Bitcoin price range reported as USD 0.05 to USD 67,541; average price USD 8,674 (in descriptive text) and Mean 13034.39 in Table 8; Standard deviation around USD 16,000 (Table 8: 16164.42).  
  - Cryptos exhibit positive skewness (more extreme positive returns) and many show leptokurtic distributions.  
  - Financial assets generally exhibit lower kurtosis; some are platykurtic (bond yields, commodity prices, MSCI indices).
- Returns and volatilities summary (selected exact figures preserved):
  - Crypto mean daily returns example range up to 0.43 for Binance Coin and as low as -0.19 for Neo (text summary).  
  - Financial asset mean returns typically range between 0.1 to 0.3, with one negative mean return in the sample (corporate bond index AEHYC).  
  - Cryptos’ return standard deviations are several times larger than those of financial assets.  
  - ADF tests: returns and volatilities series are stationary (ADF p-values often 0.01 in tables).

### Static connectedness: within crypto markets
- Total connectedness:
  - Long sample: total connectedness = 59% for returns and 68% for volatilities.  
  - Short sample (23 cryptos): total connectedness = 88% for returns and 89% for volatilities.  
- Directional and net senders (examples):
  - Long sample: largest sender of return spillovers: Bitcoin (10.7%).  
  - Long sample volatility largest senders: Dash, Litecoin, and Bitcoin (net spillovers around 9% each).  
  - Short sample (2017 start): Ethereum largest sender of return spillovers (5.3%); Neo, ZEC, Dash and Bitcoin also with net spillovers exceeding 4.5%.  
- Network structure:
  - Many significant bilateral crypto-to-crypto links; some coins consistently net receivers (Vertcoin, Augur, MaidSafeCoin, Doge, Binance Coin, Decred).  
  - Market capitalization is not the primary determinant of a coin’s role in transmitting spillovers.

### Dynamic connectedness: crypto markets over time and event associations
- Rolling-window ranges:
  - Return spillover indices ranged between 25% to 92%.  
  - Volatility spillover indices oscillated between 18% and 90%.  
- Two broad phases:
  - Prior to September 2017: generally low spillovers, typically below 50%.  
  - From September 2017 onward: significant and sustained increases in spillovers.  
- Major peaks linked to events:
  - 2018 “Crypto Winter”: return/volatility spillovers peaked about 80% (and around 85% and 90% in richer samples).  
  - March 12, 2020 (COVID shock): rapid price declines with high connectedness spikes.  
  - May 2021: Bitcoin decline ~50% associated with resurgence in spillovers.  
  - November 2022: spike linked to the collapse of FTX.  
- Cyclicality: spillover cycles align with three major crypto price cycles and notable negative news events.

### Static connectedness: within financial markets
- Total connectedness (long sample):
  - Returns total connectedness slightly surpassed 63%.  
  - Volatilities total connectedness reached 54.7%.  
- Key senders and receivers:
  - Equity indices (MSCI World and S&P500) are largest senders of returns and volatility spillovers.  
  - MSCI World net emitter examples: net emitter 2.9% for returns; 1.57% for volatilities.  
  - Risky assets with large net contributions: MSCI Emerging Markets (EM) equity index and EM high-yield sovereign bond market index (EMHYS).  
  - AE and EM investment-grade bonds appear net receivers.  
  - Gold acts as a net receiver (safe-haven behavior) with least directional connectedness.

### Combined connectedness: crypto and financial markets
- Total connectedness (combined samples):
  - Long sample: around 63% for returns and 60% for volatilities.  
  - Short sample: 82% for returns and 77% for volatilities.  
- Cross-asset intensity and asymmetries:
  - Cross-asset (crypto ↔ financial) spillovers generally lower than within-asset-class spillovers; cross-asset blocks rank among the lowest values in connectedness matrices.  
  - Stronger crypto links with global equities, the VIX, and gold; spillovers with bond indices, the USD, and other commodities are comparatively modest.  
  - Cryptos transmit relatively larger returns spillovers to gold than they receive.  
- Volatility cross-links:
  - Notable crypto ↔ MSCIEM volatility links; cryptos → MSCIEM often larger than reverse.  
  - Significant volatility spillovers between cryptos and VIX, and cryptos and commodity prices (including gold).  
  - In many volatility bilateral pairs, cryptos act as net senders.
- Rankings (selected exact table figures):
  - Short-sample crypto returns connectedness: "To" row totals sum to 88.36; Net row sums to 100.00 (Net examples from Table 11: BTC Net 0.63; DOGE Net -1.23; ETH Net 1.35).  
  - Short-sample crypto volatilities: "To" row totals sum to 89.45; Net totals 100 (Net examples from Table 12: BTC Net 0.01; DASH Net 0.63; DOGE Net -0.39; ETH Net 0.43).  
  - Long-sample crypto+financial returns (Table 13): "To" row totals aggregate 63.38; Net totals 100.00 (Net examples: BTC Net 0.91; SP Net 0.67; GOLD Net -0.51).  
  - Long-sample crypto+financial volatilities (Table 15): grand "From" totals 59.61; Net totals 100.00 (Net examples: BTC Net 0.52; MOVE Net 1.45; GOLD Net -0.80).

### Dynamics of crypto–financial spillovers and the COVID-19 episode
- Rolling dynamics and joint spikes:
  - Volatility spillovers generally lower than returns spillovers across samples.  
  - Major joint spikes associated with: August 2015 (Flash Crash); end-2017/early-2018 reversal and February 6, 2018 stock turbulence; March 2020 COVID-19 onset (connectedness indices jumped by approximately 20 percent); January–May 2021 Gamestop squeeze and May 19, 2021 crypto crash (Bitcoin ~30% drop); November 2022 FTX collapse.  
- Pandemic sub-sample (March 1, 2020 to January 31, 2021):
  - Significant increase in cross-asset spillovers; more bilateral crypto↔financial links exceed significance thresholds.  
  - Enhanced transmission from bond markets (including MOVE index and bond yields) to crypto assets during pandemic onset — a reversal relative to long-run weak bond↔crypto linkage.  
  - Dynamic net connectedness: MOVE index and EM high-yield bonds (EMHYC and EMHYS) showed persistent increases in net sending during the pandemic window; Gold and US 10Y yield behaved as net-receivers.

### Main conclusions and policy-relevant implications
- Integration and risk transmission:
  - Crypto markets display high integration with substantial spillovers in returns and volatilities; connectedness increased over time, especially after 2017, peaking in the early COVID-19 phase.  
  - Early dominance by Bitcoin (and Litecoin) as net senders gave way to a less concentrated transmitter structure in the broader crypto universe, with Ethereum prominent in many recent spillovers.  
- Cross-asset relationships and stress amplification:
  - On average, crypto↔financial connectedness is weaker than within-class connectedness, but notable links exist with global equities, VIX, and gold.  
  - Spillovers with bond indices or USD are modest in the full sample but can strengthen markedly during stress (e.g., pandemic onset).  
  - Spillover magnitudes amplify during market stress; increased comovement in stress periods implies cryptos may not reliably serve as portfolio diversifiers during risk-off episodes.  
- Policy-relevant recommendations and monitoring priorities:
  - Investors, risk managers, and policymakers should monitor evolving crypto–financial linkages given their potential role in systemic risk transmission.  
  - The analysis excludes stablecoins; volatility spillovers from stablecoin events to financial markets may be material and warrant separate study with methodologies tailored to stablecoin features (including transaction-volume-based approaches).

*Source: wpiea2023213-print-pdf — New Evidence on Spillovers Between Crypto Assets and Financial Markets — Working Paper No. WP/2023/213*

### Introduction

### Introduction

### Growth, characteristics, and policy concerns
- Crypto assets have grown from just Bitcoin in 2009 to over 5,000 currently, reaching a total market capitalization in excess of USD 3 trillion towards the end of 2021.  
- The market has since that peak substantially declined to around $1.1 trillion.  
- Growth has been accompanied by significant volatility: most crypto coins have experienced cycles of rapid growth followed by dramatic collapses, reminiscent of historical private-money episodes in the absence of adequate government regulation.  
- Policymakers and regulators are concerned about potential contagion risks to other financial markets and broader macro-financial implications.  
- Crypto assets serve diverse technological attributes and functions: means of payment, store of value, speculative asset, support for smart contracts, fundraising, asset transfer, decentralized finance, privacy, digital identity, governance, among others.  
- The analysis in this paper focuses on unbacked crypto assets, defined as assets whose value is not linked to that of another asset and whose prices fluctuate freely driven by supply and demand.

### Channels of interaction with financial markets
- Potential interaction channels include:
  - "Flight-to-safety channel": investors allocate to crypto assets during economic uncertainty if cryptos are perceived as safer and offering a hedge; implies potential diversification benefits if correlations are low.  
  - "Speculative demand channel": demand for crypto assets increases during periods of high financial market risk appetite because cryptos offer potential for high returns due to volatility.  
  - Market liquidity, information spillovers, and investor sentiment can produce additional comovement and amplify cross-market linkages.  
- These channels imply that crypto markets can be either sources of shocks or amplifiers of market volatility, with implications for financial stability.

### Contribution, scope, and methodology
- Time horizon: 2014 to the end of 2022.  
- Empirical approach: spillover approach developed by Diebold and Yilmaz (2009, 2012).  
- Sample: daily price data on 23 crypto assets (unbacked) and 15 financial variables, covering the period between October 14, 2015 to January 1, 2023.  
- Two CoinMetrics subsets used:
  - "Long" sample starts on October 15, 2014 and includes 8 crypto assets: Bitcoin, Dash, Dogecoin, Litecoin, MaidSafeCoin, Vertcoin, Monero, and Ripple.  
  - "Short" sample starts on August 3, 2017 and comprises 23 crypto assets publicly traded until now: Bitcoin (BTC), Bitcoin Cash (BCH), Binance Coin (BNB), Dash (DASH), Decred (DCR), DigiByte (DGB), Dogecoin (DOGE), Ehereum (ETH), Ethereum Classic (ETC), Gnosis (GNO), Golem (GNT), Litecoin (LTC), MaidSafeCoin (MAID), Neo (NEO), OMG Network (OMG), Augur (REP), Status (SNP), Vertcoin (VTC), NEM (XEM), Stellar (XLM), Monero (XMR), Ripple (XRP) and Zcash (ZEC).  
- Financial market variables include: 10Y US Treasury yield; sovereign bond indices for advanced and emerging economies (investment grade and high-yield); corporate bond indices; MSCI World, S&P500, MSCI EM; USD effective exchange rate; gold spot price; oil price; DJ Commodity Index; BoFA MOVE index; VIX.  
- Data sources: CoinMetrics for crypto asset prices; Bloomberg for financial variables.

### Key findings (returns and volatility spillovers)
- On average, interconnections between crypto assets and financial assets are lower compared to within their respective asset classes, for both sending and receiving returns and volatility spillovers.  
- Crypto assets primarily transmit spillovers to financial markets, although during periods of financial sector stress the reverse transmission (from financial markets to crypto) may occur.  
- Stronger interconnectedness is found between crypto assets and:
  - global equities,
  - the VIX,
  - gold.  
- Spillovers with bond indices, the USD, and other commodities are comparatively modest.  
- Spillover magnitudes have increased over time, particularly during the COVID-19 pandemic (this conclusion is cautioned given the relatively short history).  
- Increases in spillovers during heightened turbulence are linked to economic-financial events, events in crypto markets, or exogenous events.  
- Heightened correlation during risk-off episodes suggests crypto assets may not function as effective diversifiers and could serve as conduits transmitting shocks across financial markets.  
- Overall implication: increased interdependency between crypto assets and global financial markets highlights potential risks for financial stability going forward.

### Relationship to existing literature
- Empirical literature shows mixed findings:
  - Some studies find a few large crypto assets dominate transmission of return and volatility spillovers.  
  - Several papers document time-varying connectedness and growing interdependence among crypto assets, implying higher contagion risk over time.  
  - Evidence on crypto–traditional asset linkages ranges from weak/insignificant correlations (suggesting diversification or hedge properties) to significant spillovers with equities, currencies, bonds, and commodities depending on study, sample, period, and methodology.  
- Methodological heterogeneity in the literature: VAR analysis, DCC-GARCH, VAR-GARCH, wavelet coherence, copula-based approaches, and connectedness approaches of Diebold and Yilmaz (2012, 2014).  
- The literature’s inconclusive nature motivates further research on crypto–financial asset interlinkages.

### Data specifics and descriptive statistics
- Sample composition:
  - 23 crypto assets not backed by other assets.  
  - 15 financial variables.  
  - Period covered between October 14, 2015 to January 1, 2023 (with longer/shorter subsamples as noted).  
- Crypto assets are classified by primary functionality into five categories:  
  - means of exchange and payments (Bitcoin, Bitcoin Cash, Dogecoin, Litecoin, Dash, Ripple, Stellar, Zcash),  
  - smart-contracts (Ethereum, Ethereum Classic, Gnosis, Golem, Neo, OMG Network),  
  - privacy (Monero),  
  - utility (Binance Coin, MaidSafeCoin, Status, Vertcoin),  
  - other (Decred, DigiByte, Augur).  
- Market-cap coverage:
  - In the "long" sample (starts October 15, 2014) the selected coins accounted for 100% of total market capitalization at the time; by the end of the sample their combined market capitalization had decreased to 43.2%.  
  - In the "short" sample (starts August 3, 2017) the 23 coins represented 84.6% of total market capitalization at the beginning of the sample and 61.7% at the end.  
- Selected descriptive entries from Table 1 (crypto assets ranked by market capitalization as of January 1, 2023):
  - BTC (Bitcoin): Start Date 7/18/2010; Avg Price 8,674.5; Min Price 0.1; Max Price 67,541.8; Stdev Price 14,486.4; Avg MktCap (USD) 160,313,561,428; Latest MktCap (USD) 458,297,300,768.  
  - ETH (Ethereum): Start Date 8/8/2015; Avg Price 812.9; Min Price 0.4; Max Price 4,811.2; Stdev Price 1,102.9; Avg MktCap (USD) 93,133,546,881; Latest MktCap (USD) 196,753,350,116.  
  - XRP (XRP): Start Date 8/15/2014; Avg Price 0.3; Min Price 0.02; Max Price 2.8; Stdev Price 0.4; Avg MktCap (USD) 33,581,607,280; Latest MktCap (USD) 41,344,897,789.  
  - DOGE (Dogecoin): Start Date 1/23/2014; Avg Price 0.0; Min Price 0.0; Max Price 0.7; Stdev Price 0.1; Avg MktCap (USD) 4,676,035,116; Latest MktCap (USD) 12,435,771,388.  
  - Additional assets and statistics are reported in Table 1 of the source.

### Structure of the paper
- Concise literature review.  
- Data and methodology.  
- Examination of returns and volatility spillovers within crypto markets (static and dynamic).  
- Analysis of financial market assets in isolation.  
- Combined-sample analysis of crypto and financial assets to identify static and dynamic connectedness for returns and volatilities, including the network of directional spillovers with specific emphasis on the COVID-19 pandemic.

*Source: wpiea2023213-print-pdf - Introduction*

### 2017. All the price statistics (average, minimum, maximum, standard deviation) and the average market

### wpiea2023213-print-pdf - 2017. All the price statistics (average, minimum, maximum, standard deviation) and the average market

### Data, transformations, and VAR specification
- Price statistics and average market capitalization are based on the entire history of each crypto asset from each crypto’s Start Date until January 1, 2023.
- Returns definition:
  - r_{i,t} = 100×[ln(P_{i,t})−ln(P_{i,t−1})]
  - P denotes the closing price.
- Volatility construction:
  - Inter-day variance: σ^2_{i,t} = 0.361×[ln(P_{i,t})−ln(P_{i,t−1})]^2
  - Annualized/log transformation with inverse hyperbolic sine to handle zeros:
    - ̃σ_{i,t} = sinh^{-1}( sqrt{252×σ^2_{i,t}} )
  - MOVE and VIX indices are not transformed.
- Stationarity and distributional tests:
  - Jarque-Bera tests confirm non-normality of price data.
  - ADF tests generally point to non-stationarity for price levels; returns and volatilities series are stationary (suitable for VAR modelling).
- VAR and spillover methodology:
  - Diebold and Yilmaz (2012) connectedness framework using generalized VAR (Koop, Pesaran, Potter 1996; Pesaran and Shin 1998) — invariant to VAR ordering.
  - Forecast horizon H = 10 days and lag order p = 3 are used for main estimates.
  - Rolling-window dynamic analyses use 120-day windows (results robust for window sizes 100 to 200).
  - Directional, net, and total connectedness indices computed from normalized forecast error variance contributions θ^H_{ij}.

### Key statistical properties: cryptos versus financial assets
- Price-level dispersion and tail risk (example and general pattern):
  - Bitcoin price range: USD 0.05 to USD 67,541.
  - Bitcoin average price: USD 8,674.
  - Bitcoin standard deviation: around USD 16,000.
  - All analyzed cryptos display positive skewness (more extreme positive returns than negative).
  - Many crypto assets display leptokurtic distributions (kurtosis larger than 3).
  - Financial assets: generally lower kurtosis; some platykurtic (bond yields, commodity prices, MSCI indices).
- Returns and volatility summary (selected figures preserved exactly as reported):
  - Mean daily returns for cryptos vary from very high values up to 0.43 for Binance Coin, to large negative values e.g. -0.19 for Neo.
  - Financial asset mean returns typically range between 0.1 to 0.3, with one negative mean return in the sample (corporate bond index AEHYC).
  - Crypto returns exhibit standard deviations several times larger than financial assets.
  - Return skewness: cryptos generally positively skewed except Bitcoin; most financial asset returns display negative skewness.
  - ADF tests indicate returns and volatilities series are stationary.

### Static connectedness: within crypto markets (long and short samples)
- Sample definition:
  - Long sample starts in 2014 (example matrices: Tables 2 and 3).
  - Short sample starts in 2017 and includes 23 crypto assets (estimates in Appendix; short-sample results highlighted in text).
- Total connectedness:
  - Long sample: total connectedness = 59% for returns and 68% for volatilities.
  - Short sample (23 cryptos): total connectedness = 88% for returns and 89% for volatilities.
- Directional and net senders (long sample highlights):
  - Largest sender of return spillovers: Bitcoin (10.7%).
  - Largest senders of volatility spillovers: Dash, Litecoin, and Bitcoin (all with net spillover around 9%).
- Short sample highlights (2017 sample):
  - No single coin dominates net spillovers; Ethereum becomes the largest sender of return spillovers (5.3%), followed by Litecoin (4.9%).
  - Neo, ZEC, Dash and Bitcoin also have net spillovers exceeding 4.5% in the 2017 sample.
- Network structure and heterogeneity:
  - Returns and volatility networks show many significant bilateral crypto-to-crypto links.
  - Some coins are consistent net receivers: Vertcoin, Augur, MaidSafeCoin, Doge, Binance Coin, Decred (often lower market capitalization).
  - Market capitalization is not the primary determinant of a coin’s role in transmitting spillovers.

### Dynamic connectedness: crypto markets over time and event associations
- Rolling results (120-day windows):
  - Return spillover indices ranged between 25% to 92%.
  - Volatility spillover indices oscillated between 18% and 90%.
  - Two broad phases identified:
    - Prior to September 2017: generally low spillovers, typically below 50%.
    - From September 2017 onward: significant and sustained increases in spillovers.
  - Major peaks and associations:
    - 2018 “Crypto Winter”: volatility and return spillovers peaked about 80% (and around 85% and 90% in the richer 23-cryptos sample).
    - March 12, 2020 (COVID shock): rapid crypto price declines with high connectedness spikes.
    - May 2021: Bitcoin decline ~50% associated with resurgence in spillovers.
    - November 2022: spike in spillovers linked to the collapse of FTX.
  - Cyclicality:
    - Spillover cycles align with three major crypto price cycles and with notable negative news events.

### Static connectedness: within financial markets
- Total connectedness (long sample):
  - Returns total connectedness slightly surpassed 63%.
  - Volatilities total connectedness reached 54.7%.
- Key senders and receivers (long sample):
  - Equity indices (MSCI World and S&P500) are the largest senders of both returns and volatility spillovers.
  - MSCI World dominates as the most important net-emitter of shocks (net emitter: 2.9% for returns; 1.57% for volatilities).
  - Risky assets with large net contributions: MSCI Emerging Markets (EM) equity index and EM high-yield sovereign bond market index (EMHYS).
  - AE and EM investment-grade bonds appear net receivers.
  - Commodities and dollar index: moderate contributors; commodity index a net sender; dollar index a net receiver.
  - Gold: least directional connectedness, acting as a net receiver (safe-haven behavior).
- Financial volatility dynamics and events:
  - Financial returns and volatility spillovers peaked during COVID-19 initial sell-off:
    - Spillovers reached 81% (volatility) and 93% (returns) at the pandemic peak.
  - Other major events with pronounced effects: 8/24/2015 Flash Crash, 6/24/2016 Brexit, 2/5/2018 stock market crash, February 2022 Russia-Ukraine invasion (commodity-driven spillovers).

### Combined connectedness: crypto and financial markets
- Total connectedness (combined samples):
  - Long sample (smaller asset set starting in 2014): around 63% for returns and 60% for volatilities.
  - Short sample (broader crypto coverage starting in 2017): 82% for returns and 77% for volatilities.
- Cross-asset intensity and asymmetries:
  - Cross-asset (crypto ↔ financial) spillovers are generally lower than within-asset-class spillovers; cross-asset blocks in the connectedness matrix rank among the lowest values.
  - Among cross-asset links, equity indices show the strongest interactions with cryptos (MSCIW and S&P500 strongest for returns in both directions).
  - Cryptos transmit relatively larger returns spillovers to gold than they receive from gold.
  - Spillovers between cryptos and bond indices or USD are relatively modest.
- Volatility cross-links:
  - Crypto ↔ MSCIEM (EM equities) volatility links notable; spillovers from cryptos to MSCIEM tend to be larger than the reverse.
  - Significant volatility spillovers between cryptos and VIX, and cryptos and commodity prices (including gold).
  - In many volatility bilateral pairs, cryptos act as net senders.
- Rankings of transmitters (high-level):
  - Long sample returns: MSCI World largest sender (5% total; 1.45% net), followed by S&P500 and high-yield bonds. Bitcoin and Litecoin also significant net senders.
  - 2017 sample returns: Ethereum most influential transmitter (3.65% total; 1.10% net), followed by Litecoin, Neo, Bitcoin, Dash, and global equity markets.
  - Volatility: MSCI World largest sender of total spillovers in the long sample; VIX largest sender of net spillovers. In 2017 sample, many cryptos (Neo, Litecoin, Ethereum, Dash, NEM) are important total and net senders.
- Network summaries (thresholded snapshots):
  - Within-class spillovers dominate network density; cross-asset links fewer but non-negligible.
  - Returns network (2017 sample): strongest crypto-to-financial links include Ethereum↔S&P500; many financial assets are net-receivers from cryptos.
  - Volatility network: more crypto→financial spillovers exceed high thresholds; cryptos consistently net senders to gold, VIX, and MSCIEM in many bilateral cases.
  - Some smaller cryptos (e.g., GNT, DGB, XMR, XLM) exhibit notable spillovers to financial assets (in examples cited, to gold).

### Dynamics of crypto–financial spillovers and the COVID-19 episode
- Rolling dynamics (combined sample):
  - Volatility spillovers generally lower than returns spillovers across samples.
  - High comovement between returns and volatility spillover indices and between long and short samples.
  - Major joint spikes:
    - August 2015 (Flash Crash) produced sharp rise in combined spillovers.
    - End of 2017/early 2018 reversal and persistent elevated connectedness following stock market turbulence (February 6, 2018).
    - March 2020 onset of COVID-19: connectedness indices jumped by approximately 20 percent; simultaneous sell-offs across traditional and crypto markets.
    - January–May 2021: Gamestop squeeze (1/28/2021) temporarily raised returns spillovers; May 19, 2021 crypto crash (Bitcoin ~30% drop) associated with increased spillovers.
    - November 2022: FTX collapse produced a detectable spike in spillovers.
- Pandemic sub-sample (March 2020 to January 2021) specifics:
  - Network analysis shows significant increase in cross-asset spillovers; more bilateral crypto↔financial links exceed significance thresholds.
  - Evidence of enhanced transmission from bond markets (including MOVE index and bond yields) to crypto assets during the pandemic onset — a reversal relative to the long-run weak bond↔crypto linkage.
  - Dynamic net connectedness series:
    - MOVE index and EM high-yield bonds (EMHYC and EMHYS) showed persistent increases in net sending spillovers during the pandemic window.
    - Gold and the US 10Y yield behaved as net-receivers during the pandemic sub-sample.
  - Interpretation: forceful monetary easing and bond yield moves during policy response plausibly transmitted into crypto markets; cryptos also acted as amplifiers and net emitters for parts of the system.

### Main conclusions and policy-relevant implications
- Integration and risk transmission:
  - Crypto asset markets exhibit a high level of integration with substantial spillovers in both returns and volatilities; connectedness has increased over time, especially after 2017, peaking in the early COVID-19 phase.
  - In earlier years Bitcoin (and Litecoin) often dominated as net senders; in the more recent and broader crypto universe dominance is less concentrated, with Ethereum prominent in number of spillovers.
- Cross-asset class relationships:
  - On average, crypto↔financial connectedness is weaker than within-class connectedness, but notable links exist with global equities, VIX, and commodity/gold prices.
  - Spillovers between cryptos and bond indices or USD are relatively modest in the full sample, but can strengthen markedly during stress (e.g., pandemic onset).
- Stress amplification and diversification implications:
  - Spillover magnitudes amplify during market stress (crypto events, financial crashes, exogenous shocks).
  - Increased comovement in stress periods implies cryptos may not reliably serve as portfolio diversifiers during risk-off episodes.
- Monitoring and further research:
  - Results underscore the importance for investors, risk managers, and policymakers to monitor evolving crypto–financial linkages and to consider the role of cryptos in systemic risk transmission.
  - The analysis excludes stablecoins; volatility spillovers from stablecoin events to financial markets may be material and warrant separate study with methodologies tailored to stablecoin features (including transaction-volume-based approaches).

*Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023213-print-pdf.pdf*

### References

### References and Appendices (Selected content unit)

### References (bibliographic list)
- Contains cited works on cryptocurrencies, spillovers, contagion, valuation, and regulation, including articles, working papers, IMF notes, Fintech Notes, and technical reports.
- Notable entries (exact formatting preserved from source):
  - Adrian, T., Iyer, T., and Qureshi, M. S. (2022). Crypto prices move more in sync with stocks, posing new risks. Imf blog, available online at https://www.imf.org/en/blogs/articles/2022/01/11/crypto-prices-move-more-in-sync-with-stocks-posing-new-risks.
  - Bains, P., Ismail, A., Melo, F., and Sugimoto, N. (2022a). Regulating the crypto ecosystem: The case of stablecoins and arrangements. Fintech Notes, 2022/008.
  - IMF (2021). The crypto ecosystem and financial stability challenges. Global Financial Stability Report. October 2021 – COVID-19, crypto, and climate navigating challenging transitions. International Monetary Fund.
  - IMF (2023a). Elements of effective policies for crypto assets. IMF Policy Paper No. 2023/004, February 23.
  - IMF/FSB (2023). IMF-FSB synthesis paper for crypto assets. 7 September 2023. Available at https://www.fsb.org/wp-content/uploads/r070923-1.pdf. Technical report.
- The references list includes empirical studies on return-volatility relationships, hedge/safe-haven properties, network causality, volatility connectedness, and macrofinancial risk assessments.

### Appendix 1: Diebold-Yilmaz Methodology (key definitions and formulas)
- VAR representation:
  - y_t = ν + A_1 y_{t−1} + A_2 y_{t−2} + ... + A_p y_{t−p} + ε_t
  - y_t is a K-dimensional vector of endogenous variables; A_p is a K-by-K matrix.
- Companion VAR(1) form:
  - Y_t = v + A Y_{t−1} + E_t, with Y_t ≡ (y_t, y_{t−1}, ..., y_{t−p+1})′ and selection matrix J ≡ [I_K, 0_{K×K(p−1)}].
- MA(∞) (Wold) representation:
  - y_t = A(L)^{-1} ν + A(L)^{-1} ε_t = μ + Σ_{i=1}^{∞} Φ_i ε_{t−i}
  - Φ_0 = I_K, and Φ_i = Σ_{j=1}^{i} Φ_{i−j} A_j for i = 1,2, ... with A_j = 0 for j > p.
- Generalized H-step ahead forecast error variance decomposition (FEVD):
  - θ^H_{ij} = σ^{-1}_{jj} Σ_{h=0}^{H−1} (e′_i Φ_h Σ e_j)^2 / Σ_{h=0}^{H−1} (e′_i Φ_h Σ Φ′_h e_i)
  - Σ is the variance matrix of ε, σ^{-1}_{jj} is the standard deviation of ε_j, and e_i is the selection vector.
- Normalization of FEVD rows:
  - e θ^H_{ij} = θ^H_{ij} / Σ_{j=1}^K θ^H_{ij} such that Σ_{j=1}^N e θ^H_{ij} = 1.
- Directional spillovers:
  - Total directional spillovers from others to variable i: C_{i←*} = Σ_{j=1, j≠i}^N e θ^H_{ij}
  - Total directional spillovers to others from variable i: C_{*←i} = Σ_{i=1, i≠j}^N e θ^H_{ij}
  - Net spillovers for variable j: C^H_i = C_{*←i} − C_{i←*}
  - Pairwise directional connectedness between variable i and j: C^H_{ij} = C^H_{j←i} − C^H_{i←j}
- Total spillover index (TCI):
  - TCI_H = Σ_{i,j=1, i≠j}^N e θ^H_{ij} / Σ_{i,j=1}^N e θ^H_{ij}

### Appendix 2: Tables and Charts — data coverage, samples, and key summary statistics (selected exact figures)
- Crypto assets sample and descriptions:
  - Table 6 lists 23 crypto assets (Numbered 1 to 23) including Bitcoin (BTC), Ethereum (ETH), Dogecoin (DOGE), Ripple (XRP), Litecoin (LTC), Monero (XMR), Zcash (ZEC), and others.
- Financial assets included (Table 7) with index symbols and descriptions such as:
  - EMIGS: ICE BofA US Investment Grade Emerging Markets External Sovereign Index.
  - USTNX: US Generic Govt 10 Yr.
  - MSCIEM: MSCI Emerging Markets Index.
  - SP: S&P 500 INDEX.
  - VIX: Chicago Board Options Exchange Volatility Index.
- Sample periods and notes (preserved verbatim):
  - "The table shows the summary statistics for all the crypto and financial assets for the longest sample included in the analysis (e.g. from October, 15, 2014 for most series, with the exception of the cryto assets marked with ”*”, for which the sample starts in August 3, 2017)."
- Summary statistics for prices (Table 8) — selected entries (exact values preserved):
  - BTC: Period (years) 8.29; Mean 13034.39; Standard deviation 16164.42; Skewness 1.52; Kurtosis 1.25; Jarque-Bera 972.94; ADF -2.03; ADF (pval) 0.56.
  - ETH*: Period (years) 5.49; Mean 1094.67; Standard deviation 11671.27; Skewness 0.55; Kurtosis 400.92; Jarque-Bera -1.78; ADF (pval) 0.67.
  - GOLD: Period (years) 8.29; Mean 1459.83; Standard deviation 275.26; Skewness 0.46; Kurtosis -1.36; Jarque-Bera 241.25; ADF -2.43; ADF (pval) 0.39.
- Summary statistics for returns (Table 9) — selected entries:
  - BTC: Count 2163; Mean 0.14; Median 0.19; Standard Deviation 4.15; Min -47.06; Max 22.41; Skewness -0.83; Kurtosis 10.97; Jarque-Bera 11089.94; ADF -12.34; ADF (pval) 0.01.
  - ETH*: Count 1432; Mean 0.01; Median 0.01; Standard Deviation 5.49; Min -56.56; Max 30.06; Skewness -0.99; Kurtosis .87; Jarque-Bera 6004.64; ADF -10.11; ADF (pval) 0.01.
  - EMIGS: Count 4713; Mean 0.02; Median 0.03; Standard Deviation 0.34; Min -4.71; Max 3.58; Skewness -2.07; Kurtosis 35.48; Jarque-Bera 250507.4; ADF -15.42; ADF (pval) 0.01.
- Summary statistics for volatilities (Table 10) — selected entries:
  - BTC: Count 2163; Mean 0.25; Median 0.16; Standard Deviation 0.25; Min 0.0002; Max 2.21; Skewness 1.83; Kurtosis 4.6; Jarque-Bera 3112.81; ADF -8.63; ADF (pval) 0.01.
  - ETH*: Count 1432; Mean 0.33; Median 0.26; Standard Deviation 0.3; Min 0.0002; Max 2.39; Skewness 1.49; Kurtosis 3.13; Jarque-Bera 1114.84; ADF -8.63; ADF (pval) 0.01.
  - MOVE: Count 4713; Mean 0.2; Median 0.16; Standard Deviation 0.14; Min 0.0000; Max 1.00; Skewness 1.64; Kurtosis 3.37; Jarque-Bera 4354.67; ADF -3.87; ADF (pval) 0.02.
- Connectedness / spillover matrices and totals (exact values preserved from tables):
  - Table 11 (Full-sample Connectedness Matrix for Crypto Asset Returns: Short Sample, August 3, 2017 to January 1, 2023): The "To" row totals sum to 88.36; the Net row sums to 100.00 (Net examples: BTC Net 0.63; DOGE Net -1.23; ETH Net 1.35).
  - Table 12 (Crypto Asset Volatilities: Short Sample): The "To" row totals sum to 89.45; the Net row sums to 100 (Net examples: BTC Net 0.01; DASH Net 0.63; DOGE Net -0.39; ETH Net 0.43).
  - Table 13 (Crypto and Financial Asset Returns: Long Sample, October 15, 2014 to January 1, 2023): "From" column for BTC totals 3.32; "To" row totals aggregate 63.38 for crypto and financial assets subset; Net totals sum to 100.00 (Net examples: BTC Net 0.91; SP Net 0.67; GOLD Net -0.51).
  - Table 15 (Crypto and Financial Asset Volatilities: Long Sample): The grand "From" column totals in bottom-right show 59.61; Net row sums to 100.00 (Net examples: BTC Net 0.52; MOVE Net 1.45; GOLD Net -0.80).
  - Table 14 and Table 16 present analogous short-sample connectedness matrices with "To" totals 81.63 (Table 14) and 77.56 (Table 16), and Net totals 100.00.
- Figures (network and time-series connectedness visualizations):
  - Figures 4–13 present net pairwise directional connectedness visualizations for crypto-only and cross crypto-financial asset spillovers in returns and volatilities, including specific pandemic-period panels (COVID-19 pandemic: March 1, 2020 till January 31, 2021).
  - Figure 14 shows Total Directional Connectedness, Returns and Volatilities, Selected Financial and Crypto Assets across Jan-18 to Jan-23; the horizontal dash lines mark the beginning of the COVID-19 pandemic sample (March 1, 2020 till January 31, 2021).
- Notes on interpretation (preserved from source):
  - Each (i,j)-th value represents the contribution of innovation in asset j returns (or volatility) to the variance of the forecast error in asset i.
  - "From" aggregates cumulative contributions to asset i from all other assets; "To" summarizes the impact of asset j on all other assets; "Net" captures the net spillover transmitted by each asset (positive indicates net transmitter, negative indicates net receiver).
  - Color scales and thresholds used in figures: top 95% percentile of bilateral spillovers for some plots; top 90% percentile for selected cross-asset spillover plots. Black corresponds to crypto-crypto, blue to financial-financial and red to crypto-financial spillovers in specified figures.

*New Evidence on Spillovers Between Crypto Assets and Financial Markets — Working Paper No. WP/2023/213*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023213-print-pdf.pdf_