Researchers mapped how national barriers split bitcoin prices

Research exposed a fragmented global market

On September 21, 2019, economists Igor Makarov of the London School of Economics and Antoinette Schoar of MIT presented evidence that bitcoin did not trade as one efficiently integrated global asset. In an author-written summary published by the Harvard Law School Forum on Corporate Governance, they described large, recurring price differences across cryptocurrency exchanges that sometimes persisted for hours, days or weeks.

The summary drew from their working paper, which had been revised on July 12, 2019 and was forthcoming in the Journal of Financial Economics. It used transaction-level data from 34 exchanges across 19 countries. The central finding was not simply that exchanges quoted different prices. Differences were substantially larger across national or regional boundaries than among exchanges operating within the same country.

That mattered institutionally because bitcoin could move across its own network without a bank, yet completing an arbitrage still required access to banking systems, exchange accounts and local fiat currencies. The research suggested that the supposedly borderless asset remained constrained by borders whenever traders needed to repatriate conventional money.

The largest premiums appeared across borders

For the core exchange analysis, the working paper examined data principally from January 1, 2017 through February 28, 2018. The authors calculated that the daily average bitcoin price ratio between the United States and South Korea exceeded 15% from December 2017 into early February 2018 and reached 40% on several days. They reported an average difference of approximately 10% between Japan and the United States and about 3% between Europe and the United States.

By comparison, average price deviations among exchanges within the same country generally did not exceed 1%, according to the authors’ September 21 summary. Their calculations estimated at least $2 billion in potential arbitrage profit during the period of the largest dislocations.

“Potential” is essential. The estimate was a research calculation based on observed trades and modeled execution constraints, not an audited record of profit collected by identifiable traders. Moving capital across jurisdictions could require currency conversion, banking approval and compliance with local controls. Traders also faced exchange solvency, custody, settlement, execution and price-impact risks.

Capital controls offered an explanation

Makarov and Schoar found that countries with higher average bitcoin premiums also tended to experience larger increases in those premiums when global buying pressure intensified. Prices outside the United States and Europe were generally above, rather than below, the U.S. reference price.

The authors interpreted that asymmetry as evidence consistent with capital-market segmentation. Buying bitcoin in the United States and selling it in South Korea could produce won, but the trade could not be repeated indefinitely if those proceeds could not be converted and returned efficiently. Arbitrage capital could become trapped in the expensive market while remaining scarce elsewhere.

Their comparison between fiat and crypto trading pairs strengthened that interpretation. On exchanges exhibiting large bitcoin-to-fiat differences, spreads between cryptocurrencies were approximately an order of magnitude smaller. Because a crypto-to-crypto trade did not require the same cross-border fiat movement, the result was consistent with conventional financial barriers contributing to the regional premiums.

The paper did not prove that capital controls explained every price difference. Its authors excluded some exchanges suspected of inflated volume or wash trading, and data availability varied by venue. Governance risk, fees, withdrawal limits, account restrictions and counterparty concerns could also discourage apparent arbitrage.

September 21 market context

Kraken’s venue-specific report for September 21 recorded $71 million of activity across its reported markets. It listed bitcoin at $9,982, down 1.62% over the report’s daily comparison window, with $37.1 million in reported BTC activity.

Those figures describe Kraken, not a consolidated global market, and the surviving report does not fully specify a universal closing convention. They therefore cannot test whether the historical regional spreads persisted on September 21. The research’s significance was broader: it showed that a globally transferable token could still trade through locally segmented financial infrastructure, limiting price convergence even when a theoretical arbitrage was visible.