Neither intuition nor economic theory well explain the fact that modern economies require transfers of monetary value on the order of 50 times national income to function properly—but they do. Since average transactions balances comprise about 10 percent of GDP in advanced economies, those balances must turn over, on average, 500 times per year to enable the necessary payments (≈5,000 percent of GDP).
This fact, documented by myself and researchers such as Claudio Borio at the BIS, perplexes theorists. They commonly assume in their stylized models that monetary transfers are equal to monetary income, i.e. one dollar is needed to purchase one dollar of income. Expressed in the words of the quantity theory of money, there is a fixed relationship between money income and money expenditure, aka "velocity", that, while not equal to "1" as in simple theoretical models, is a constant determined by a static payments system technology. Back when I was in graduate school the velocity of M1 was stable at around "7".
Nor do the extraordinary turnover rates of transactions balances accord with our common daily experience unless we are corporate treasurers, equity or bond trader/brokers or work on the payments settlement side of a bank. But the data and facts are real. Well before the advent of widespread retail application of digital electronic communications, payments were flowing through advanced economics at high rates. By 1966, the average annual turnover rate of New York City demand deposits already exceeded 100X. At the interbank level, digital electronic communications have been employed to settle interbank monetary transfers for over a century. The US "FEDWIRE" system originated as a telegraphic system-using Morse code--to enable banks to transfer funds among themselves.
What does all this have to do with bitcoin? It is very important to understand the modern global payments settlement infrastructure, its capacity and cost structure, before rushing to judgement about the future role of proof-of-work cryptocurrencies therein. Modern centralized fiat money payments systems routinely handle a huge volume of daily network traffic even though it is uniform neither inter-day nor intraday. The values transferred--JP Morgan Chase alone transfers about $5 trillion per day (about 20 percent of US GDP)--are not "intuitive" because roughly 98 percent of those transfers are not being made to settle purchases of final goods and services (GDP) like buying a coffee at Starbucks. They are primarily being made among enterprises to settle wholesale transactions, among trader/brokers to settle equity, bond, foreign exchange, commodities and derivatives trading, to settle transfers of real property, and to effect tax payments and government transfers. So the "velocity", I prefer to use the term "turnover", of settlement balances is extraordinarily high when one realizes the appropriate metric of transfers is not GDP but total monetary transfers for all purposes. In 2006, US bank balances held at the Federal Reserve—the money most intensively used to settle payments, turned over an average of 11,000 times per year or 44 times per day.
Since monetary transfers are so large in relation to national income it is clear a viable payments system must be efficient. Were a modern payments system’s operational costs to average 1 percent of the value transferred, this would amount to half national income—obviously an impossible burden for the economy to support. That would be like having to set aside half your paycheck just to pay transactions fees.
Interestingly, if one looks at the total cost of effectively using bitcoin as a settlement media (taking payments to bitcoin miners as total cost) it averages out as approximately 1 percent of the transactions value. While this might seem to be a small percentage, it is 100,000 times higher than the settlement cost in the Eurozone's real time gross settlement system, TARGET2. On a per transaction basis, the average cost of settling with bitcoin is $ 44 per transaction—6,000 times more than through NACHA, the US ACH system.
Similar cost disparities are present in comparisons with RTGS systems in Canada, Switzerland, Hong Kong, and the UK.
Consequently, even were the bitcoin blockchain able to handle the massive daily transfers enabled by global RTGS, the cost of bitcoin settlement would grind trading to a halt. Thus in considering the possible niche proof-of-work validation cryptocurrency payments systems may occupy within the future global payments architecture we must keep in mind both its relatively miniscule payments capacity and its enormous relative cost. The capacity constraints are well know--the bitcoin blockchain (globally) would be at capacity, for example, if everyone in Hong Kong used bitcoin to pay for their morning coffee... . Provided centralized fiat money payments systems continue to upgrade—as intended—it is difficult to see an alternative system seizing their hold on high value, high volume payments. Low frequency payments with transactors willing to pay a high premium to avoid centralized clearing may continue to be well served by bitcoin but we clearly cannot all afford to use it for the enormous volume of payments required in a modern economy.