Decrypting New Age International Capital Flows

This paper employs high frequency transactions data on the world’s oldest and most extensive centralized peer-to-peer Bitcoin market, which enables trade in the currencies of more than 135 countries. We develop an algorithm that allows, with high probability, the detection of “crypto vehicle transactions” in which crypto currency is used to move capital across borders or facilitate domestic transactions. In contrast to previous work which has used “on-chain” data, our approach enables one to investigate parts of the vastly larger pool of “off-chain” transactions. We find that, as a conservative lower bound, over 7 percent of the 45 million trades on the exchange we explore represent crypto vehicle transactions in which Bitcoin is used to make payment in fiat currency. Roughly 20 percent of these represent international capital flight/flows/remittances. Although our work cannot be used to put a price on cryptocurrencies, it provides the first systematic quantitative evidence that the transactional use of cryptocurrencies constitutes a fundamental component of their value, at least under the current regulatory regime.


Summary: Decrypting Capital Flight through Cryptocurrencies
We develop an algorithm that provides first evidence on Bitcoin being used to move capital capital across borders, and/or exchange one fiat currency for another Within the off-chain dataset we analyse, at least 11% of trades are used for such transfers.
Bitcoin appear to be used to circumvent taxes and regulations, i.e. to evade restrictions on international capital flows and foreign exchange transactions, including on remittances.
The use case we find is most prominent emerging markets.
The data The Identification Algorithm -for individual trades Each trade in sample, i, has a trade-size x i .
Define n i as the number of times that the trade size x i occurs within five hours prior to trade i.
We are interested in times when n i > 0.
But this could happen just by chance, when many trades happen in five hours, or when the trade size x i is common, exempli gratia 1.00000000 Bitcoin.For an illustration of trade size distribution To evaluate the random-match-hypothesis, we require a null hypothesis: Matches being random.
Assumption 1 -The null model: Assume trades of any size x k appear as independent Poisson processes.The Poisson process intensity being the product of p k (the probability of any new trade having the size x k ), and the number of arrivals of trades over the time period of interest.

The Identification Algorithm-for individual trades
Under the null, the probability of a trade finding a match is like that in a multinomial draw: For a detailed discussion Where we will estimate p i based on data prior to t: Definition 1 -Discovery We declare a discovery, when we find n i > 0, and can reject The Identification Algorithm -trade share estimand

Conclusion
Cryptocurrencies serve as a channel for transactions between fiat currencies, especially when capital controls aim at impeding such transfers.
This use cases for Bitcoin challenges the view of Bitcoin as a purely speculative bubble.
International capital flows through Crypto Vehicle Trades remain off the radar of any stock taking agency (similar to transactions with large denomination cash (Rogoff, 2016)).
Possible Policy Implications: Capital controls = Crypto Controls Outlook: Rise of stable coins -accelerator of crypto vehicle trades or beginning of the end of Bitcoin?
To arrive at an estimate of the estimand, we define a discovery as And control for false discoveries, with the expected matches in a random sample: The Identification Algorithm The share of trades that are crypto vehicle trades thus becomes: Under an arbitrary data generating process for (n 1 , . . ., n I ), Where we make use of the fact that for any single i: E [φi | Ni ] = 1 * P((ni > 1)|Ni ).

Back to Presentation
Graf von Luckner, Reinhart and Rogoff Decrypting New Age International Capital Flows March 28, 2023 Empirical Evidence in Support of Key Assumption

Figure :
Figure: Bitcoin Trades as parallel market is born (once again) in Argentina in 2019.
Figure: Daily Maximum Drawdown.Sources: Bloomberg and CryptoCompare USD/BTC USD/EUR USD/MXN Annualized Standard Deviation 93 % 8 % 12 % Sum of individual hypothesis tests would create an inflated share of trades.Biased equal to Θ at most.Multiple hypothesis tests => need to net false discovery rate.Allows us to control for the expected False Discovery Rate, and thus arrive at an unbiased estimate of the share of trades that are crypto vehicle trades.
Figure: The World's 25 biggest Crypto Vehicle Channels.Circles: Origin, Triangles: Destination.Line-width: Channel volume as share identified trade volume in Origin Currency Graf von Luckner, Reinhart and Rogoff Decrypting New Age International Capital Flows March 28, 2023 Figure: Event Study: Guri-Dam Power-cut in Venezuela between March 7th and March 9th 2019 (Share of id.trade volume in base currency with VES as origin or destination) 6.5% (42%) 6.6% (30%) 7.5% (38%) Diff-in-Diff Graph Counterfactual Graf von Luckner, Reinhart and Rogoff Decrypting New Age International Capital Flows March 28, 2023