## 2.1  Modalities of Bitcoin Cross-Border Flows

## Source details

**Canonical URL:** [2.1  Modalities of Bitcoin Cross-Border Flows](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024085-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024085-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024085-print-pdf.pdf.json)

---

### On-chain versus off-chain transactions
- On-chain transactions
  - Permanently recorded on the Bitcoin blockchain; require sender’s address, recipient’s address, amount of Bitcoin, and sender’s signature using the wallet’s private key (a 256-bit random number).
  - Lifecycle: sender initiates → signs with private key → broadcast to network → entered into mempool → miners validate by solving cryptographic puzzles in exchange for a fee → validated transactions stored in a block and confirmed by other miners → block added to blockchain → transaction completed.
  - Fees determined by transaction size (data volume) and block space (desired speed).
  - Offer a high degree of security due to blockchain immutability.
- Off-chain transactions
  - Occur outside the blockchain (e.g., via a crypto exchange); not recorded on the blockchain and not validated by miners.
  - Can lower fees and speed processing relative to on-chain; provide a lower degree of security due to lack of public record and exchanges’ vulnerability to crypto hacks.
- Empirical note: based on a sample of 17 crypto exchanges, Makarov and Schoar (2022) find that the Bitcoin off-chain volume is somewhat larger than the on-chain volume.

### Types of crypto exchanges
- Centralized exchanges
  - Manage a central order book and custody users’ crypto assets (users do not have direct control over private keys); typically regulated and subject to Know Your Customer and Anti-Monetary Laundering regulations.
  - Provide user-friendly interfaces, broader asset/product sets, and higher liquidity.
  - Continue to account for a substantial share of trading activity.
- Peer-to-peer (P2P) exchanges
  - Decentralized exchanges with some centralization; facilitate direct trading between clients with public listings and often provide escrow services.
- Decentralized exchanges
  - Use smart contracts to facilitate direct P2P trading without intermediaries; preserve privacy and reduce hack risk.

### Three measurement approaches for cross-border Bitcoin flows
- Approach 1: On-chain exchange wallets (Kondor et al. (2021) + WalletExplorer)
  - Data: on-chain transactions with sending/receiving addresses, amounts, timestamps.
  - WalletExplorer identifies about 5 percent of addresses.
  - Dataset: 1.6 million transactions on 80 exchanges; about one-third across exchanges.
  - Blockchain raw data: 44.6 million transactions over Jan 12th, 2009–Feb 7th, 2020.
  - Matched cross-exchange transactions: 592,218 over Aug 2nd, 2011–Feb 7th, 2020.
  - Limitations: difficult to assign exchanges to countries; WalletExplorer coverage limited.
- Approach 2: On-chain exchange wallets plus web traffic (Chainalysis Know Your Transaction)
  - Method: identify exchange wallet addresses and assign flows to countries via monthly web traffic patterns, distributing transaction volume across countries proportional to web traffic shares.
  - Data coverage: Chainalysis Bitcoin cross-border flows over March 2019–March 2023.
  - Key assumptions: (i) users do not mask online activity by employing virtual private networks (VPNs) and (ii) transaction amounts are, on average, broadly equal across users in different countries.
  - Pros: broader scope than Approach 1; captures larger share of on-chain volume; innovative country identification.
  - Cons: country identification imprecise if VPNs used; assumption of equal transaction amounts across countries may impact precision.
- Approach 3: Fiat-Bitcoin counterparts on P2P platforms (LocalBitcoins, Graf von Luckner et al. (2023) algorithm)
  - Data: LocalBitcoins dataset contains 40.6 million transactions in 136 fiat currencies over March 15th, 2017–February 28th, 2023; observe unique transaction ID, timestamp, amount in Bitcoin up to eight digits, counterpart fiat currency, and price in local fiat.
  - Matching rule: match pairs of transactions of the exact same size X within a conservative five-hour window; use probability framework with Θo = 0.05 for 95 percent confidence.
  - Identified 2.1 million crypto vehicle transactions; about 180 thousand occur across two different fiat currencies.
  - Matched cross-fiat transactions: 187,701 over March 15th, 2017–February 16th, 2023.
  - False positive rate average suggests 0.9 percent of crypto vehicle transactions are random.
  - Advantages: straightforward residency identification through fiat currency; transaction-level data and relatively long sample.
  - Drawbacks: not fully representative of off-chain market; imprecision for globally dominant currencies (e.g., U.S. dollar); five-hour window conservative—matched transactions likely a lower bound.

### Comparative summary of approaches
- Blockchain (Approach 1): on-chain; transaction-level data; representative of on-chain volume; cannot map exchanges to countries.
- Chainalysis (Approach 2): on-chain; uses web traffic = residency assumption; many exchanges; representative of large share of on-chain market volume; imprecise if VPNs used; web traffic assumption simplified.
- LocalBitcoins (Approach 3): off-chain (P2P); fiat currency = residency assumption; transaction-level data and relatively long sample; not representative of full off-chain volume; imprecise for dominant currencies; matching assumption valid only when fiat currencies are not widely used in third countries.

---

### 3.1  Bitcoin Cross-Border Flows and Capital Flows

### Data and definitions
- Portfolio flow sources:
  - EPFR Global (EPFR): net flows into investment funds (purchases minus redemptions), proxy for investment fund flows.
  - Institute of International Finance (IIF): tracks flows for emerging markets based on national sources broadly in line with official Balance of Payments (BoP) data.
- Bitcoin flow definitions:
  - LocalBitcoins: an inflow corresponds to fiat flowing from source to destination with Bitcoin as vehicle.
  - Chainalysis: captures flow of Bitcoin from source to destination; for comparability, Chainalysis "inflows" and "outflows" are switched so an inflow corresponds to the unobserved payment flow from source to destination.
- Comparability: EPFR and IIF flows are conceptually comparable to gross inflows in a BoP sense; Bitcoin datasets provide estimates of gross flows (both inflows and outflows) without netting.

### Stylized facts on transactions and cross-border flows
- Transaction-size differences (averages and maxima preserved exactly):
  - Average transaction sizes:
    - Blockchain: 13.3486 Bitcoin
    - LocalBitcoins: 0.0178 Bitcoin
  - At a Bitcoin price of US$10,000:
    - Average blockchain transaction value: US$133,486
    - Average LocalBitcoins transaction value: US$178
  - Maximum transaction amounts at same price:
    - Blockchain: US$300,000,000
    - LocalBitcoins: US$1,875,000
  - Temporal trend: average transaction size tends to decrease over time for both off-chain and on-chain transactions.
  - Interpretations:
    - LocalBitcoins transactions may reflect remittances and circumvention of capital controls.
    - On-chain transactions may reflect larger market participants preferring blockchain security.
- Geographic spread and intensity:
  - Bitcoin cross-border transactions are geographically widespread with relatively high intensities across regions for both off-chain and on-chain flows.
  - High relative inflows in Latin America (notably Argentina and Venezuela); large inflows in parts of Africa, Asia, and Eastern Europe.
  - LocalBitcoins particularly important for Nigeria.
- Magnitudes relative to capital flows (upper-quartile averages preserved exactly where given):
  - Chainalysis upper quartile (2019–2022 average monthly inflows): ranges from 0.1–2.5 percent of average annual GDP.
  - EPFR inflows upper quartile (2017–2022 average monthly): ranges from 0.01–0.1 percent of average annual GDP.
  - IIF inflows upper quartile (2017–2022 average monthly): ranges from 0.07–0.4 percent of average annual GDP.
  - LocalBitcoins upper quartile (2017–2022 average monthly): ranges from 0–1.7 percent of average annual GDP.
  - Examples:
    - Chainalysis: Seychelles largest monthly inflows at 2.5 percent of average annual GDP (2019–2022), Venezuela 0.8 percent, Moldova 0.7 percent (2019–2022 averages).
    - LocalBitcoins: Venezuela monthly inflows 1.7 percent of average annual GDP (2017–2022 average), Nigeria 0.0005 percent (2017–2022 average).

### Key transaction-level descriptive statistics (matched blockchain and LocalBitcoins)
- Number of matched transactions:
  - Blockchain: 1,632,049
  - LocalBitcoins: 2,107,509
- Average transaction size (BTC):
  - Blockchain: 13.3486
  - LocalBitcoins: 0.0178
- Largest transaction size (BTC):
  - Blockchain: 30,000.0
  - LocalBitcoins: 187.5
- Number of currencies/exchange countries (destination):
  - Blockchain: 86
  - LocalBitcoins: 104
- Number of currencies/exchange countries (source):
  - Blockchain: 78
  - LocalBitcoins: 106
- Cross-currency/exchange country transactions:
  - Number:
    - Blockchain: 592,218
    - LocalBitcoins: 182,204
  - Average transaction size (BTC) for cross-currency:
    - Blockchain: 8.6170
    - LocalBitcoins: 0.0205
  - Largest transaction size (BTC) for cross-currency:
    - Blockchain: 20,000.0
    - LocalBitcoins: 25.0
  - Number of currencies/exchange countries (destination) for cross-currency:
    - Blockchain: 86
    - LocalBitcoins: 94
  - Number of currencies/exchange countries (source) for cross-currency:
    - Blockchain: 75
    - LocalBitcoins: 100
- Overlapping period March 2017–February 2020:
  - Average (maximum) transaction size: blockchain 32.1284 (30,000) Bitcoin vs LocalBitcoins 0.0204 (187.5) Bitcoin.

### Empirical strategy
- Baseline panel OLS (monthly):
  - Y_{c,t} = α Y_{c,t−1} + Γ1 GLOBAL_t + Γ2 DOMESTIC_{c,t} + η_{c,y} + e_{c,t}
  - Dependent variable Y_{c,t} alternately: Bitcoin cross-border flows or capital flows, scaled by average GDP over 2017–2022, divided by their standard deviation (sample starting in 2017m1 or later), and the resulting ratio scaled by 10^6.
  - GLOBAL_t includes: VIX, broad dollar index, crypto fear & greed index.
  - DOMESTIC_{c,t} includes: inflation (year-over-year), interest differential to the U.S. (based on overnight rates), and Bitcoin parallel rate premium (percent deviation between average local currency–U.S. dollar exchange rate on LocalBitcoins and official exchange rate).
  - Country-year fixed effects η_{c,y} included.
  - Baseline sample starts in 2017m3; Chainalysis sample starts in 2019m4; U.S. excluded.
- Robustness: difference GMM used in Appendix B to address Nickell bias.

### Empirical results — traditional global and domestic determinants
- Global risk aversion and broad dollar
  - Increase in the VIX and a strengthening U.S. dollar lead to lower EPFR inflows (columns 5 and 6) and lower IIF inflows (columns 7 and 8).
  - Broad dollar also significant for Chainalysis flows (columns 1 and 2).
  - No significant impact of global drivers on LocalBitcoins flows (columns 3 and 4).
- Domestic factors
  - Higher inflation associated with lower portfolio flows but does not affect Bitcoin flows.
  - Interest rate differential to the U.S.:
    - Appears associated with lower LocalBitcoins inflows in one specification, but result not robust.
    - Does not determine Chainalysis flows or portfolio flows in main results.

### Empirical results — adding Bitcoin-specific drivers
- Specification adds Crypto fear&greed_t and BTC parallel premium_{c,t}.
- Key findings
  - EPFR and IIF portfolio inflows: increases in VIX and broad dollar continue to negatively impact inflows (columns 5–8).
  - Chainalysis flows:
    - VIX effect becomes positive and significant: increase in VIX associated with higher Chainalysis inflows and outflows (columns 1 and 2).
    - Improvement in crypto sentiment (higher crypto fear & greed index) associated with higher Chainalysis flows.
  - Crypto sentiment spillovers: better crypto sentiment associated with higher IIF equity inflows (column 8).
  - Domestic effects for off-chain flows:
    - Higher interest rate differential to the U.S. associated with higher LocalBitcoins inflows in some specifications.
    - Increase in BTC parallel premium associated with higher LocalBitcoins outflows (column 4) and lower IIF equity inflows.
  - Inflation: associated with lower EPFR inflows and, in one specification, lower LocalBitcoins outflows; not robust across specifications.
- Relative magnitudes
  - Absolute magnitude of response to a change in the VIX is larger for capital flows than for Bitcoin flows.
  - Chainalysis inflows and outflows increase by 0.05 standard deviations in response to a one standard deviation increase in the VIX (textual magnitude preserved).

### Robustness and estimation methods
- Difference GMM (per Bond et al. (2001) rule) used; Table B.1 shows global-driver results very similar to baseline.
- Robustness across flow definitions:
  - Both Bitcoin volume and value regressions used; Table C.2 shows volume regressions (columns 1–4) and value regressions controlling for BTC price (columns 5–8); results broadly similar.
  - Volume regressions: VIX 1.390*** (CA in, vol), 1.401*** (CA out, vol); BTC parallel premium 0.511*** (CA in, vol), 0.453*** (CA out, vol).
- Specific robustness notes:
  - BTC parallel premium impact remains positive and significant for LocalBitcoins outflows and negative for IIF inflows across estimators.
  - Results for inflation and interest differential show limited robustness.

### Major empirical findings and numerical estimates (preserved exactly)
- Chainalysis flows respond positively to the VIX:
  - Chainalysis VIX coefficients (Table 8): 0.66*** (CA in), 0.67*** (CA out).
  - Summary statement: Chainalysis response 0.7 standard deviations in response to a one standard deviation increase in the VIX.
- Crypto sentiment:
  - Crypto fear&greed (Table 8): 0.54*** (CA in), 0.52** (CA out).
- BTC parallel premium (Table 8):
  - 0.06** (LB out), -0.05** (IIF equity in).
  - Textual magnitudes: one standard deviation increase in Bitcoin parallel premium associated with a 0.04 standard deviation increase in LocalBitcoins outflows and a 0.03 standard deviation decline for IIF equity inflows.
- Broad dollar effects:
  - Table 9 cross-exchange blockchain flows: Broad dollar -0.23** (Blockchain in), -0.12** (Blockchain out).
- Cross-exchange blockchain VIX coefficients (Table 9): VIX -0.08 (Blockchain in), -0.01 (Blockchain out).

### Policy-relevant implications and interpretation
- Different responses to global risk and dollar dynamics:
  - Chainalysis (on-chain) flows respond positively to VIX, contrasting with traditional capital flows which decline with higher risk aversion and broad dollar appreciation.
- Circumvention and capital controls:
  - Off-chain evidence that BTC parallel rate premiums associate with higher outflows aligns with literature suggesting crypto can facilitate circumvention of capital flow restrictions.
  - IMF (2023a) implication: policymakers aiming to manage capital flows should ensure that capital flow management regulations cover crypto assets.
- Structural policy point:
  - Addressing underlying imbalances that create exchange rate pressures is important because crypto usage may reflect symptomatic behavior rather than root causes.
- Future dynamics:
  - Market developments (e.g., authorization of spot Bitcoin ETFs in the U.S.) could lead to greater convergence between Bitcoin flows and traditional capital flows as user bases converge, potentially complicating policy responses.

*Italic source: IMF Working Paper chapter "2.1  Modalities of Bitcoin Cross-Border Flows" and chapter "3.1  Bitcoin Cross-Border Flows and Capital Flows" (source PDF content).*

### 2.1  Modalities of Bitcoin Cross-Border Flows

### 2.1  Modalities of Bitcoin Cross-Border Flows

### On-chain versus off-chain transactions
- On-chain transactions
  - Permanently recorded on the Bitcoin blockchain (a distributed ledger which is public as it is visible to all network participants) and are immutable.
  - Require sender’s address, recipient’s address, amount of Bitcoin, and the sender’s signature using the wallet’s private key (a 256-bit random number).
  - Transaction lifecycle: sender initiates → signs with private key → broadcast to network → entered into mempool → miners validate by solving cryptographic puzzles in exchange for a fee → validated transactions stored in a block and confirmed by other miners → block added to blockchain → transaction completed.
  - Offer a high degree of security because of blockchain immutability.
  - Fees are determined by transaction size (i.e., data volume) and block space (i.e., the desired speed).
- Off-chain transactions
  - Occur outside the blockchain—for example, via a third party such as a crypto exchange which verifies legitimacy and facilitates completion.
  - Not recorded on the blockchain; miners are not required to validate such transactions.
  - Can lower fees and speed up processing time relative to on-chain transactions.
  - Provide a lower degree of security, reflecting the lack of a public record and exchanges’ vulnerability to crypto hacks.
- Empirical note: based on a sample of 17 crypto exchanges, Makarov and Schoar (2022) find that the Bitcoin off-chain volume is somewhat larger than the on-chain volume.

### Types of crypto exchanges
- Centralized exchanges
  - Manage a central order book; responsible for custody of users’ crypto assets (users do not have direct control over their private keys), security, maintenance, functionalities, and transaction approval.
  - Typically regulated and subject to Know Your Customer and Anti-Monetary Laundering regulations requiring collection and storage of customer information, including identity documents and addresses.
  - Provide user-friendly interfaces, broader sets of crypto assets and products, and higher liquidity via access to extensive pools of buyers and sellers (e.g., regulated market makers).
  - Despite growth in P2P and decentralized trading, centralized exchanges still account for a substantial share of trading activity.
- Peer-to-peer (P2P) exchanges
  - Decentralized exchanges with a degree of centralization.
  - Facilitate direct trading between clients who create public listings; frequently provide an escrow service to secure transactions.
- Decentralized exchanges
  - Use smart contracts to facilitate direct P2P trading without intermediaries.
  - Preserve a high level of privacy and reduce the risk of hacks.

### Approach 1: Measuring cross-exchange flows using on-chain exchange wallets
- Data and identification
  - On-chain transactions obtained from Kondor et al. (2021) with sending/receiving addresses, amounts, timestamps.
  - WalletExplorer used to identify exchange wallet addresses from public websites and internal transactions; identifies about 5 percent of the addresses in the dataset.
- Dataset scope and limitations
  - Resulting exchange-level dataset contains 1.6 million transactions on 80 different exchanges; about one-third of these transactions occur across exchanges.
  - Blockchain raw data (after merging transaction IDs, in/out transactions, and timestamps) contains 44.6 million transactions over Jan 12th, 2009–Feb 7th, 2020.
  - Matched cross-exchange transactions dataset contains 592,218 cross-exchange transactions over Aug 2nd, 2011–Feb 7th, 2020.
  - Major limitation: challenging to assign exchanges to countries to obtain cross-border flows because exchange registration location may not reflect user residency and users may transact on exchanges in countries other than where they reside.
  - WalletExplorer’s coverage is limited and cannot provide information on all wallets and exchanges.

### Approach 2: Measuring on-chain cross-border flows using exchange wallets plus web traffic (Chainalysis)
- Method
  - Identify exchange wallet addresses using blockchain public data plus exchange-provided transaction data (Chainalysis Know Your Transaction).
  - Assign flows to countries based on monthly web traffic patterns for each exchange; distribute transaction volume across countries proportional to web traffic shares.
- Example stylized allocation
  - For 100 Bitcoin moving from exchange 1 to exchange 2 in one day, distribution example: 35 Bitcoin from country X to country Z, 15 Bitcoin from country Y to country X, 15 Bitcoin from country Y to country Z, etc.
- Data coverage and assumptions
  - Bitcoin cross-border flows obtained from Chainalysis over March 2019–March 2023.
  - Key assumptions: (i) users do not mask online activity by employing virtual private networks (VPNs) and (ii) transaction amounts are, on average, broadly equal across users in different countries.
- Pros and cons
  - Pros: broader scope than Approach 1; captures a larger share of on-chain transaction volume across many exchanges; innovative country identification via web traffic.
  - Cons: country identification can be imprecise if users use VPNs; assumption of broadly equal transaction amounts across countries may impact precision.

### Approach 3: Measuring cross-border flows via fiat currency counterparts on off-chain P2P platforms (LocalBitcoins)
- Data and matching algorithm
  - Use fiat-Bitcoin transactions from LocalBitcoins to identify cross-border flows, following a probabilistic algorithm developed by Graf von Luckner et al. (2023).
  - LocalBitcoins dataset contains 40.6 million transactions in 136 fiat currencies over March 15th, 2017–February 28th, 2023.
  - For each transaction observe a unique transaction ID, timestamp, amount in Bitcoin up to eight digits, counterpart fiat currency, and price paid in local fiat currency.
- Matching intuition and parameters
  - Match pairs of transactions of the exact same size X that occur within a short period (conservative five-hour window).
  - Rationale: probability of two transactions of the same size within a short period is extremely low; vehicle trades minimize holding time given price volatility.
  - Formal parameters: define ni = number of times transaction size xi occurs within five-hour window, Ni = total number of transactions within five-hour window, pi = probability that transaction size xi occurs.
  - Under null hypothesis H0,i: ˆθ* i > Θo, i = 1,...,I, where P(ni > 1 | Ni) = ˆθ* i ≈ 1 − (1 − ˆpi)Ni. Choose Θo = 0.05, implying matches represent vehicle transactions with a 95 percent confidence level. Two matching transactions of size xi classified as a vehicle transaction if ni > 1 and ˆθ* i ≤ Θo.
  - LocalBitcoins charges a one percent fee, however, this fee does not affect the reported transaction amount and price.
- Empirical results and limitations
  - Matching identifies 2.1 million crypto vehicle transactions.
  - About 180 thousand transactions occur across two different fiat currencies.
  - False positive rate computed by averaging ˆθ* i over identified vehicle transactions suggests 0.9 percent of crypto vehicle transactions are random.
  - Matched cross-fiat transactions dataset contains 187,701 cross-fiat transactions over March 15th, 2017–February 16th, 2023 (one matched transaction involves two separate fiat-bitcoin transactions).
  - Advantages: straightforward identification of residency through fiat currency; transaction-level data and relatively long sample.
  - Drawbacks: not fully representative of broader off-chain market (captures only share of off-chain universe); preferences for exchanges differ across countries; imprecision for transactions relying on globally dominant currencies such as the U.S. dollar.
  - The five-hour matching window is conservative; matched transactions likely represent a lower bound of crypto vehicle transactions but may miss matches occurring over longer intervals.

### Comparative summary of approaches
- Each approach captures different segments and relies on different assumptions:
  - Blockchain (Approach 1): on-chain; transaction-level data; representative of on-chain volume; cannot map exchanges to countries.
  - Chainalysis (Approach 2): on-chain; uses web traffic = residency assumption; many exchanges; representative of a large share of on-chain market volume; imprecise if VPNs used; web traffic assumption simplified.
  - LocalBitcoins (Approach 3): off-chain (P2P); fiat currency = residency assumption; transaction-level data and relatively long sample; not representative of full off-chain volume; imprecise for dominant currencies; matching assumption valid only when fiat currencies are not widely used in third countries.

*Italic source: IMF Working Paper chapter "2.1  Modalities of Bitcoin Cross-Border Flows" (source PDF content).*

### 3.1  Bitcoin Cross-Border Flows and Capital Flows

### 3.1  Bitcoin Cross-Border Flows and Capital Flows

### Data and definitions
- Portfolio flow sources:
  - EPFR Global (EPFR): net flows into investment funds (purchases minus redemptions), proxy for investment fund flows.
  - Institute of International Finance (IIF): tracks flows for emerging markets based on national sources broadly in line with official Balance of Payments (BoP) data.
- Bitcoin cross-border flow sources and definitions:
  - LocalBitcoins dataset: an inflow corresponds to the flow of fiat money from the source to the destination with Bitcoin acting as the vehicle.
  - Chainalysis dataset: captures the flow of Bitcoin from the source to the destination. For comparability, Chainalysis "inflows" and "outflows" are switched such that an inflow corresponds to the unobserved flow of the payment from the source to the destination.
- Conceptual comparability:
  - EPFR and IIF flows are conceptually comparable to gross inflows in a BoP sense (non-resident purchases netted against non-resident sales) — referred to as inflows.
  - Bitcoin datasets provide estimates of gross flows (both inflows and outflows) without netting, so estimated Bitcoin flows are not fully comparable to observed portfolio inflows.

### Stylized facts on Bitcoin transactions and cross-border flows
- Transaction-size differences:
  - On-chain (blockchain) transactions are, on average, significantly larger than off-chain (LocalBitcoins) transactions.
  - Average transaction sizes:
    - Blockchain: 13.3486 Bitcoin
    - LocalBitcoins: 0.0178 Bitcoin
  - At a Bitcoin price of US$10,000 (price in summer of 2020):
    - Average blockchain transaction value: US$133,486
    - Average LocalBitcoins transaction value: US$178
  - Maximum transaction amounts at the same price:
    - Blockchain: US$300,000,000
    - LocalBitcoins: US$1,875,000
  - Temporal trend: average transaction size tends to decrease over time for both off-chain and on-chain transactions (likely related to Bitcoin price increases).
  - Possible interpretation:
    - Off-chain (LocalBitcoins) transactions may reflect remittances and circumvention of capital flow restrictions.
    - On-chain transactions may reflect market participants moving larger sums preferring blockchain security.
    - Fee structures differ: LocalBitcoins fees based on transaction amounts; on-chain fees depend on data volume and desired speed.
- Geographic spread and intensity:
  - Bitcoin cross-border transactions are geographically widespread with relatively high intensities across regions for both off-chain and on-chain flows.
  - High relative inflows in Latin America (notably Argentina and Venezuela) — these countries fall within the top inflow quartile for both Chainalysis and LocalBitcoins.
  - Relatively large inflows observed in a number of countries in Africa, Asia, and Eastern Europe.
  - LocalBitcoins is particularly important for Nigeria (noted difference between Chainalysis and LocalBitcoins patterns).
- Magnitudes relative to capital flows:
  - Bitcoin cross-border flows can be sizeable in some countries, especially those with small capital flows; countries with large traditional capital flows typically have lower Bitcoin flows.
  - Comparison of upper quartiles (2019–2022 averages for Chainalysis; 2017–2022 for EPFR and IIF unless otherwise stated):
    - Chainalysis upper quartile (2019–2022 average monthly inflows): ranges from 0.1–2.5 percent of average annual GDP.
    - EPFR inflows upper quartile (2017–2022 average monthly): ranges from 0.01–0.1 percent of average annual GDP.
    - IIF inflows upper quartile (2017–2022 average monthly): ranges from 0.07–0.4 percent of average annual GDP.
    - LocalBitcoins upper quartile (2017–2022 average monthly): ranges from 0–1.7 percent of average annual GDP.
  - Distribution skew and country examples:
    - Chainalysis: Seychelles largest monthly inflows at 2.5 percent of average annual GDP (2019–2022), followed by Venezuela at 0.8 percent and Moldova at 0.7 percent (averages over 2019–2022).
    - LocalBitcoins: Venezuela monthly inflows amounted to 1.7 percent of average annual GDP (2017–2022 average), followed by Nigeria at 0.0005 percent (2017–2022 average).

### Key transaction-level descriptive statistics (matched blockchain and LocalBitcoins, Table 6)
- Number of matched transactions:
  - Blockchain: 1,632,049
  - LocalBitcoins: 2,107,509
- Average transaction size (BTC):
  - Blockchain: 13.3486
  - LocalBitcoins: 0.0178
- Largest transaction size (BTC):
  - Blockchain: 30,000.0
  - LocalBitcoins: 187.5
- Number of currencies/exchange countries (destination):
  - Blockchain: 86
  - LocalBitcoins: 104
- Number of currencies/exchange countries (source):
  - Blockchain: 78
  - LocalBitcoins: 106
- Cross-currency/exchange country transactions:
  - Number:
    - Blockchain: 592,218
    - LocalBitcoins: 182,204
  - Average transaction size (BTC) for cross-currency:
    - Blockchain: 8.6170
    - LocalBitcoins: 0.0205
  - Largest transaction size (BTC) for cross-currency:
    - Blockchain: 20,000.0
    - LocalBitcoins: 25.0
  - Number of currencies/exchange countries (destination) for cross-currency:
    - Blockchain: 86
    - LocalBitcoins: 94
  - Number of currencies/exchange countries (source) for cross-currency:
    - Blockchain: 75
    - LocalBitcoins: 100
- Note on sample overlap:
  - When restricting both samples to overlapping period March 2017–February 2020:
    - Average (maximum) transaction size: blockchain 32.1284 (30,000) Bitcoin vs LocalBitcoins 0.0204 (187.5) Bitcoin.

### Empirical strategy
- Objective: compare determinants of Bitcoin cross-border flows and capital flows, focusing on traditional capital-flow drivers and Bitcoin-specific drivers.
- Baseline panel OLS specification (monthly frequency):
  - Y_{c,t} = α Y_{c,t−1} + Γ1 GLOBAL_t + Γ2 DOMESTIC_{c,t} + η_{c,y} + e_{c,t}
  - Dependent variable Y_{c,t} alternately: Bitcoin cross-border flows or capital flows, scaled by average GDP over 2017–2022, then divided by their standard deviation (sample starting in 2017m1 or later depending on availability), and finally the resulting ratio is scaled by 10^6.
  - GLOBAL_t includes: VIX, broad dollar index, crypto fear & greed index.
  - DOMESTIC_{c,t} includes: inflation (year-over-year), interest differential to the U.S. (based on overnight rates), and Bitcoin parallel rate premium (percent deviation between average local currency–U.S. dollar exchange rate on LocalBitcoins and official exchange rate).
  - Country-year fixed effects η_{c,y} included.
- Sample periods:
  - Baseline sample starts in 2017m3 (start of LocalBitcoins dataset).
  - Chainalysis sample starts in 2019m4.
  - U.S. excluded from sample (interest differential to the U.S. computed; LocalBitcoins residency identification issues for Bitcoin–U.S. dollar transactions).
- Robustness:
  - Difference GMM used in Appendix B to address Nickell bias from dynamic panel.

### Empirical results — traditional global and domestic determinants (Equation (4))
- Global risk aversion and broad dollar:
  - An increase in the VIX and a strengthening of the U.S. dollar lead to lower EPFR inflows (columns 5 and 6) and lower IIF inflows (columns 7 and 8) for both debt and equities.
  - Broad dollar also plays a significant role for Chainalysis flows (columns 1 and 2).
  - No significant impact of global drivers on LocalBitcoins flows (columns 3 and 4).
    - Interpretation: LocalBitcoins transactions are smaller and reflect motives (remittances, circumvention of controls) less linked to traditional capital-flow drivers.
- Domestic factors:
  - Higher inflation is associated with lower portfolio flows but does not affect Bitcoin flows.
  - Interest rate differential to the U.S.:
    - Appears associated with lower LocalBitcoins inflows in one specification, but this result is not robust.
    - Does not determine Chainalysis flows or portfolio flows in the main results (possible role of synchronized global monetary policy cycles).

### Empirical results — adding Bitcoin-specific drivers (Equation (5))
- Specification adds Crypto fear&greed_t and BTC parallel premium_{c,t}.
- Key findings:
  - EPFR and IIF portfolio inflows:
    - Increases in VIX and broad dollar continue to negatively impact EPFR and IIF inflows (columns 5–8).
  - Chainalysis flows:
    - VIX effect becomes positive and significant: an increase in the VIX is associated with higher Chainalysis inflows and outflows (columns 1 and 2).
      - Interpretation: increased activity via the Bitcoin market as investors move away from other traditional risky assets.
    - Improvement in crypto sentiment (higher crypto fear & greed index) is associated with higher Chainalysis flows.
  - Crypto sentiment spillovers:
    - Better crypto sentiment is associated with higher IIF equity inflows (column 8).
  - Domestic effects for off-chain flows:
    - Higher interest rate differential to the U.S. is associated with higher LocalBitcoins inflows (consistent with flows responding to relative interest rates), though prior result on lower LocalBitcoins inflows with higher interest differential was not robust.
    - Increase in BTC parallel premium is associated with higher LocalBitcoins outflows (column 4) — indicates Bitcoin may be used to circumvent capital controls.
    - Increase in BTC parallel premium also associated with lower IIF equity inflows — supporting use of parallel premium as proxy for exchange rate pressures.
  - Inflation:
    - Increase in inflation associated with lower EPFR inflows and, in one specification, lower LocalBitcoins outflows; these trends are not robust across specifications.
- Relative magnitude of responses:
  - The (absolute) magnitude of the response to a change in the VIX is larger for capital flows than for Bitcoin flows.
  - Monthly EPFR and IIF inflows decline in the range of 0.2– [text truncated in source] (analysis notes: source indicates larger declines for portfolio flows than for Bitcoin flows).

*Source: IMF Working Paper chapter 3.1, "Bitcoin Cross-Border Flows and Capital Flows" (from the provided PDF content).*

### 0.7 standard deviations in response to a one standard deviation increase in the VIX.

### wpiea2024085-print-pdf - 0.7 standard deviations in response to a one standard deviation increase in the VIX.

### Major empirical findings on Bitcoin cross-border flows
- Chainalysis flows respond positively to the VIX: a one standard deviation increase in the VIX is associated with a 0.7 standard deviations response (summary statement; detailed estimates in Table 8 show VIX coefficients for Chainalysis: 0.66*** for CA in and 0.67*** for CA out).
- EPFR and IIF inflows respond to traditional risk and dollar dynamics in line with expectations:
  - EPFR bond in: VIX coefficient in Table 7 is -8.52*** and in Table 8 is -8.98***.
  - EPFR equity in: VIX coefficients -2.09*** (Table 7) and -2.52*** (Table 8).
  - IIF debt in and IIF equity in show negative responses to VIX: Table 7 IIF debt in -1.99*** and IIF equity in -3.05***; Table 8 IIF debt in -2.26*** and IIF equity in -2.30*.
- Broad dollar effects:
  - Broad dollar appreciation lowers many traditional capital inflows (Table 7: Broad dollar coefficients include -2.20** and -2.15** for Chainalysis CA in/CA out; Table 9: Blockchain in -0.23**, Blockchain out -0.12**).
  - Cross-exchange blockchain flows (Table 9) show broad dollar coefficients: Blockchain in -0.23**, Blockchain out -0.12**.
- Crypto sentiment:
  - Crypto fear&greed positively correlates with Chainalysis flows (Table 8: Crypto fear&greed 0.54*** for CA in and 0.52** for CA out).
- Bitcoin parallel rate premium:
  - A one standard deviation increase in the Bitcoin parallel premium is associated with a 0.04 standard deviation increase in LocalBitcoins outflows and a 0.03 standard deviation decline for IIF equity inflows (textual magnitudes preserved).
  - Table 8: BTC parallel premium coefficients include 0.06** for LB out and -0.05** for IIF equity in.
- On-chain versus off-chain differences:
  - Cross-border on-chain (blockchain) transactions are on average considerably larger than off-chain transactions.
  - Off-chain data (LocalBitcoins) suggest higher outflows associated with an increase in the BTC parallel rate premium.

### Robustness and estimation methods
- Dynamic panel concerns:
  - To address Nickell bias, the authors implement a difference GMM estimator (difference GMM rather than system GMM per Bond et al. (2001) rule of thumb).
  - Table B.1 (GMM estimation) shows global-driver results remain very similar and qualitatively unchanged relative to baseline.
- Robustness across flow definitions:
  - Analyses use both Bitcoin cross-border flow volumes (amount of Bitcoin) and values (amount of Bitcoin times the average global Bitcoin price).
  - Table C.2 presents volume regressions (columns 1–4) and value regressions controlling for BTC price (columns 5–8). Results broadly similar to baseline Table 8.
  - For volume regressions: VIX remains significant for Chainalysis flows; broad dollar remains insignificant for volumes. BTC parallel rate premium turns significant for Chainalysis volume flows (Table C.2: BTC parallel premium 0.511*** for CA in and 0.453*** for CA out in columns 1–2).
- Specific robustness notes:
  - Among domestic drivers, the BTC parallel rate premium impact remains positive and significant for LocalBitcoins outflows and negative for IIF inflows across estimators.
  - Results for inflation and interest rate differential show limited robustness: negative relationship between interest differential and LocalBitcoins inflows in some specifications and insignificance of inflation for LocalBitcoins outflows in GMM.

### Policy-relevant implications and interpretation
- Bitcoin cross-border flows respond differently than traditional capital flows to global risk and dollar dynamics:
  - Chainalysis (on-chain) flows respond positively to VIX, contrasting with traditional capital outflows during risk aversion spikes.
  - EPFR and IIF inflows decline with higher risk aversion and broad dollar strengthening, consistent with standard capital flow behavior.
- Policy considerations:
  - Off-chain evidence that BTC parallel rate premiums associate with higher outflows aligns with literature suggesting crypto can facilitate circumvention of capital flow restrictions.
  - IMF (2023a) implication: policymakers aiming to manage capital flows should ensure that capital flow management regulations cover crypto assets.
  - Structural policy point: addressing underlying imbalances that create exchange rate pressures is important because crypto usage may reflect symptomatic behavior rather than root causes.
- Future dynamics:
  - Recent market developments (e.g., authorization of spot Bitcoin ETFs in the U.S.) could lead to greater convergence between Bitcoin flows and traditional capital flows as user bases converge, potentially complicating policy responses.

### Key numerical estimates (selected, preserved exactly)
- Chainalysis VIX coefficients (Table 8): 0.66*** (CA in), 0.67*** (CA out).
- Crypto fear&greed (Table 8): 0.54*** (CA in), 0.52** (CA out).
- BTC parallel premium (Table 8): 0.06** (LB out), -0.05** (IIF equity in).
- Table 9 (cross-exchange flows): VIX -0.08 (Blockchain in), -0.01 (Blockchain out); Broad dollar -0.23** (Blockchain in), -0.12** (Blockchain out); Crypto fear&greed -0.01 (Blockchain in), 0.01 (Blockchain out).
- Table C.2 (volume regressions): VIX 1.390*** (CA in, vol), 1.401*** (CA out, vol); BTC parallel premium 0.511*** (CA in, vol), 0.453*** (CA out, vol).
- Magnitudes reported in text: Chainalysis inflows and outflows increase by 0.05 standard deviations in response to a one standard deviation increase in the VIX; a one standard deviation increase in the Bitcoin parallel premium is associated with a 0.04 standard deviation increase in LocalBitcoins outflows and a 0.03 standard deviation decline for IIF equity inflows.

*Source: A Primer on Bitcoin Cross-Border Flows: Measurement and Drivers, Working Paper No. WP/24/85 (excerpts from wpiea2024085-print-pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024085-print-pdf.pdf_