Sunday, January 14, 2018

Floors v corridors

David Beckworth argues that the U.S. Federal Reserve should stop running a floor system and adopt a corridor system, say like the one that the Bank of Canada currently runs. In this post I'll argue that the Bank of Canada (and other central banks) should drop their corridors in favour of a floor—not the sort of messy floor that the Fed operates mind you, but a nice clean floor.

Floors and corridors are two different ways that a central banker can provide central banking services. Central banking is confusing, so to illustrate the two systems and how I get to my preference for a floor, let's start way back at the beginning.

Banks have historically banded together to form associations, or clearinghouses, a convenient place for bankers to make payments among each other over the course of the business day. To facilitate these payments, clearinghouses have often issued short-term deposits to their members. A deposit provides clearinghouse services. Keeping a small buffer stock of clearinghouse deposits can be useful to a banker in case they need to make unexpected payments to other banks.

Governments and central banks have pretty much monopolized the clearinghouse function. So when a Canadian bank wants to increase its buffer of clearinghouse balances, it has no choice but to select the Bank of Canada's clearing product for that purpose. Monopolization hasn't only occurred in Canada of course, almost every government has taken over their nation's clearinghouse.

One of the closest substitutes to Bank of Canada (BoC) deposits are government t-bills or overnight repo. While neither of these investment products is useful for making clearinghouse payments, they are otherwise identical to BoC deposits in that they are risk-free short-term assets. As long as these competing instruments yield the same interest rate as BoC deposits, a banker needn't worry about trading off yield for clearinghouse services. She can deposit whatever quantity of funds at the Bank of Canada that she deems necessary to prepare for the next day's clearinghouse payments without losing out on a better risk-free interest rate elsewhere.  

But what if these interest rates differ? If t-bills and repo promise to pay 3%, but a Bank of Canada deposit pays an inferior interest rate of 2.5%, then our banker's buffer stock of Bank of Canada deposits is held at the expense of a higher interest elsewhere. In response, she will try to reduce her buffer of deposits as much as possible, say by reallocating bank resources and talent to the task of figuring out how to better time the bank's outgoing payments. If more attention is paid to planning out payments ahead of time, then the bank can skimp on holdings of 2.5%-yielding deposits while increasing its exposure to 3% t-bills.

Why might BoC deposits and t-bills offer different interest rates? We know that any differential between them can't be due to credit risk—both instruments are issued by the government. Now certainly BoC deposits provide valuable clearinghouse services while t-bills don't. And if those services are costly for the Bank of Canada to produce, then the BoC will try to recapture some of its clearinghouse expenses. This means restricting the quantity of deposits to those banks that are willing to pay a sufficiently high fee for clearing services. Or put differently, it means the BoC will only provide deposits to banks that are willing to accept an interest rate that is 0.5% less than the 3% offered on t-bills.

But what if the central bank's true cost of providing additional clearinghouse services is close to zero? If so, the Bank of Canada should avoid any restriction on the supply of deposits. It should provide each bank with whatever amount of deposits it requires without charging a fee. With bankers' demand for clearing services completely sated, the differential between BoC deposits and t-bills will disappear, both trading at 2.5%.

There is good reason to believe that the cost of providing additional clearinghouse services is close to zero. It is no more costly for a central bank to issue a new digital clearinghouse certificate than it is for a Treasurer to issue a new t-bill. In both cases, all it takes is a few button clicks.

Let's assume that the cost of providing clearinghouses is zero. If the Bank of Canada chooses to  constrain the supply of deposits to the highest bidders, it is forcing banks to overpay for a set of clearinghouse services which should otherwise be provided for free. In which case, the time and labour that our banker has diverted to figuring out how to skimp on BoC deposit holdings constitutes a misallocation of her bank's resources. If the Bank of Canada provided deposits at their true cost, then her employees' time could be put to a much better use.

As members of the public, we might not care if bankers get shafted. But if our banker has diverted workers from developing helpful new technologies or providing customer service to dealing with the artificially-created problem of skimping on deposits, then the public directly suffers. Any difference between the interest rate on Bank of Canada deposits and competing assets like t-bills results in a loss to our collective welfare.


Which finally gets us to floors and corridors. In brief, a corridor system is one in which the central bank rations the number of clearinghouse deposits so that they aren't free. In a floor system, unlimited deposits are provided at a price of zero.

When a central bank is running a corridor system, as most of them do, the rate on competing assets like t-bills lies above the interest rate on central bank deposits. Economists describe these systems as corridors because the interest rate at which the central bank lends deposits lies above the interest rate on competing safe assets like t-bills and repo, and with the deposit rate lying at the bottom, a channel or corridor of sorts is formed.

For instance, take the Bank of Canada's corridor, illustrated in the chart below. The BoC lets commercial banks keep funds overnight and earn the "deposit rate" of 0.75%. The overnight rate on competing opportunities—very short-term t-bills and repo—is 1%. The top of the corridor, the bank rate, lies at 1.25%. So the overnight rate snakes through a corridor set by the Bank of Canada's deposit rate at the bottom and the bank rate at the top. (The exception being a short period of time in 2009 and 2010 when it ran a corridor floor).

Let's assume (as we did earlier) that the BoC's cost of providing additional clearinghouse services is basically zero. Given the way the system is set up now, there is a 0.25% rate differential (1%-0.75%) between the deposit rate and the rate on competing asset, specifically overnight repo. This means that the Bank of Canada has capped the quantity of deposits, forcing bankers to pay a fee to obtain clearing services rather than supplying unlimited deposits for free. This in turn means that Canadian bankers are forced to use of time and energy on a wasteful activity like trying to skimp on BoC deposit holdings. All Canadians suffer from this waste.

It might be better for the Bank of Canada (and any other nation that also uses a corridor system) to adopt what is referred to as a floor system. Under a floor system, rates would be equal such that the rate on t-bills and repo lies on the deposit rate floor of 0.75%--that's why economists call it a floor system. The Bank of Canada could do this by removing its artificial limit on the quantity of deposits it issues to commercial banks. Banks would no longer allocate scarce time and labour to the task of skirting the high cost of BoC deposits. In theory at least, all Canadians would be made a little better off. All the Bank of Canada would have to do is click its 'create new clearinghouse deposits'  button a few times.


The line of thought I'm invoking in this post is a version of an idea that economists refer to as the optimum quantity of money, or the Friedman rule, first described by Milton Friedman back in the 1960s. Given that a central bank's cost of issuing additional units of money is zero, Friedman thought that any interest rate differential between a monetary asset and an otherwise identical non-monetary asset represents a loss to society. This loss comes in the form of people wasting resources (or incurring shoe leather costs) trying to avoid the monetary asset as much as possible. To be consistent with the zero cost of creating new monetary assets, the rates on the two assets should be equalized. The public could then hold whatever amount of the monetary asset they saw fit, so-called shoe leather costs falling to zero.

In my post, I've applied the Friedman rule to one type of monetary asset: central bank deposits. But it can also be applied to banknotes issued by the central bank. After all, banknotes yield just 0% whereas a t-bill or a risk-free deposit offers a positive interest rates. To avoid holding large amounts of barren cash, people engage in wasteful behaviour like regularly visiting ATMs.

There are several ways to implement the Friedman rule for banknotes. One of the neatest ways would be to run a periodic lottery that rewards a few banknote serial numbers with big winnings, the size of the pot being large enough that the expected return on each banknote as made equivalent to interest rate on deposits. This idea was proposed by Charles Goodhart and Hugh McCulloch separately in 1986.

Robert Lucas once wrote that implementing the Friedman rule was “one of the few legitimate ‘free lunches’ economics has discovered in 200 years of trying.” The odd thing is that almost no central banks have tried to adopt it. On the cash side of things, none of them offer a serial number lottery or any of the other solutions for shrinking the rate differential between banknotes and deposits, say like Miles Kimball's more exotic crawling peg solution. And on the deposit side, floor systems are incredibly rare. The go-to choice among central banks is generally a Friedman-defying corridor system.

One reason behind central bankers' hesitation to implement the Friedman rule is that it would threaten their pot of "fuck you money", a concept I described here. Thanks to the large interest rate gaps between cash and t-bills, and the smaller gap between central bank clearinghouse deposits and t-bills, central banks tend to make large profits. They submit much of their winnings to their political masters. In exchange, the executive branch grants central bankers a significant degree of independence... which they use to geek out on macroeconomics. Because they like to engage in  wonkery and believe that it makes the world a better place, central bankers may be hesitant to implement the Friedman rule lest it threaten their flows of fuck you money, and their sacred independence. 

That may explain why floors are rare. However, they aren't without precedent. To begin with, there is the Fed's floor that Beckworth describes, which it bungled into by accident. At the outset of this post I called it a messy floor, because it leaks (George Selgin and Stephen Williamson have gone into this). The sort of floor that should be emulated isn't the Fed's messy one, but the relatively clean floor that the Reserve Bank of New Zealand operated in 2007 and Canada did from 2009-11 (see chart above). Though these floors were quickly dropped, I don't see why the couldn't (and shouldn't) be re-implemented. As Lucas says, its a free lunch.

Wednesday, January 10, 2018

XRP and bitcoin as bridge currencies

Eshima Ohashi Bridge, Japan

The value of all outstanding XRPs recently surpassed that of bitcoin, hitting $300 billion or so last month. XRPs are a cryptocurrency issued by Ripple, a company that is trying to shake up the business of cross border payments. Ripple has a number of strategies for doing this, but the one that has caught people's imaginationespecially as the price of XRPs rocket higheris to have banks and other financial institutions use XRP as a 'bridging asset' for moving value across borders. The idea of using a cryptocurrency as a bridge isn't a new idea. Bitcoin remittance companies have been trying to do this for several years now, without very much success.

So what do I mean by using a cryptocurrency like bitcoin or XRP as a bridge asset? Does it make any sense? To answer these questions, let's dissect a hypothetical cross border payment.

Straddling two universes

As users of banks and other financial institutions, we rarely think about what is going on underneath the hood of a money transfer. If I send money from my account to yours, the language of this transaction implies that money is flowing from my bank account to bank account. But moving funds from one bank to another bank is physically impossible. If my account is at Bank A, and yours is at Bank B, I cannot send value directly from my account to your account. Our two accounts may as well exist in entirely separate universes.

The only way I can make a bank-to-bank payment to you is indirectly, by turning to a third-party who straddles both universes. Say my hair dresser has accounts at both my bank and your bank, and for a small fee she'll do the transaction for us. I tell my bank (Bank A) to credit my hairdresser's account at Bank A by $10, and my hair dresser in turn tells her bank (Bank B) to debit her account and credit you account at Bank B by $10. The payment is done. I have $10 less, you have $10 more, and my hair dresser is flat, her $10 having been erased from Bank B's ledger with a new $10 deposit appearing in her account at Bank A.

The same principle is at work in cross border payments, except the person who is doing the straddling between the two bank—my hair dresserwill need to have a domestic bank account, say in Canada, and an international account, say Philippines. And instead of crediting you $10, she will have her Filipino bank send you the peso equivalent of $10, which is around ₱400 at the current 40:1 exchange rate. But apart from that, the concept is the same.

In principle, a cross border payment like this could go very fast. Assuming that it only takes the Canadian bank a few moments to transfer $10 from my account to my hair dresser's account, then she can quickly start the Philippines leg of the transaction. And if the Filipino bank is just as fast, you'll have the ₱400 just a few moments after she clicks the send button. This whole chain needn't take more than twenty minutes of fiddling with bank websites.

The benefits of queues

But there are factors militating against speed. Say that I need to send you money several times a day. It would be a hassle for my hair dresser to log in to her Philippines bank account and process each payment as it arrives in her Canadian accountshe has to cut hair, after all. Instead, she chooses to wait till the end of the day when several of my payment requests have accumulated, upon which she batches the payments into one large payment and clicks the send button.

There is a trade-off here between speed and cost. Putting transaction requests into a queue slows down each of my payments to you, but it imposes less costs on my hair dresser. Slow speeds aren't necessarily a bug. If we all want to save some money, sluggishness may be the best solution for all of us.

Pre-funding: expensive but speeds things up

Imagine that over the course of a few weeks I make so many payments to you that my  hairdresser's Filipino account runs out of funds. When this happens she will no longer be able to make outgoing payouts to your Philippines bank account. To keep the system up and running, she will have to replenish her account with pesos. One of her options would be to withdraw cash from her Canadian account, fly it to Philippines in a suitcase, trade it at the airport for peso banknotes, and deposit these into her Filipino account. This would be slow, expensive, dangerous, and potentially illegal, but it's a theoretical option.

A more realistic option would be to sell her Canadian dollar deposits to a foreign exchange dealer and get Filipino peso deposits in return. This dealer, who will have accounts in both Philippines and Canada, will execute this trade for a commission. My hairdresser will have to ask her Canadian bank to credit the dealer's dollar account while the dealer asks his Filipino bank to credit my hair dresser's peso account. There will be some lag as the dealer processes the transaction, say because helike my  hairdresseruses a queue to batch payments together so as to save on fees. But once my hair dresser's Filipino account has been topped up, I can once again make payments to you.

Instead of allowing her Filipino account to periodically run down to zero, my hairdresser may try to maintain a permanently-funded peso account. After all, if she doesn't prefund the account, then you and I will have to cope with constant delays as she waits for the foreign exchange dealer to refill her account. Prefunding her Philippines account isn't without  cost, however. Instead of being able to invest the money in bonds or upgrading her salon, my hair dresser must tie her capital up in a low-yielding bank account as she awaits my payments requests.

Thus, as in the case of queuing on the Canadian side, prefunding on the Philippines side involves a trade-off between cost and efficiency. Reducing the amount of pesos held in anticipation of incoming payment request will allow my hairdresser to reduce costs, but it will simultaneously slow down payments from you to me since the odds of having to wait for a refill increase.

To sum up, for cross border payments to occur someone must straddle the divide between isolated banks. This straddler uses techniques like queuing and prefunding in order to make cross border payments proceed as fast as possible without costing too much.

Cryptocurrency as a bridge

So where do XRP and bitcoin come in? The two of us want little more than a flow of recurring peso payments to arrive in your Filipino bank account as fast and cheaply as possible, but what goes on underneath the hood doesn't concern us. If she can increase payment speeds without having to pay a higher cost (or, alternatively, if she can reduce costs without sacrificing speed), it may make sense for my hairdresser to incorporate an asset like XRP or bitcoin in the payments process.

Say at the end of Monday my hair dresser has amassed four $10 payments in her queue, or $40. She logs into her Filipino bank account, and sends you one ₱1600 payment. She wants to rebuild her Filipino account balance in preparation for the rest of the week's incoming payments. Normally she would do so by asking her foreign exchange dealer to swap her some Filipino deposits for her Canadian dollar deposits.

Instead, she decides to pre-fund by turning to the market for cryptocurrencies. One option is to take the $40 I've transferred her and buy $40 worth of XRPs from her foreign exchange dealer, then sell these XRPs to another dealer for ₱1600 in deposits. Alternatively, she may turn to an organized exchange to complete the refunding. She sends the $40 to a Canadian cryptocurrency exchange, buys some bitcoin or XRP, quickly sends these coins to a Filipino cryptocurrency exchange, and then sells them for pesos. At which point she will transfer the pesos to her bank. Voila, her Filipino bank account has been reloaded using cryptocurrency as a bridge.

Comparing fees and speed

Let's compare these two routes. By exchanging dollars directly for pesos via a foreign exchange dealer, only one transaction had to be completed, and thus one set of hassles and fees incurred. By going through the cryptocurrency market, my hairdresser must make two transactionsa purchase of XRP or bitcoin on the Canadian crypto exchange (or from a dealer) using Canadian dollars, and a sale of XRP/bitcoin on the Filipino crypto exchange (or to a dealer) for pesos. If the sum of these two sets of hassles and fees is less than the traditional single set of hassles and fees, then going the cypto route may make some sense for her. But I confess that I think it is highly unlikely that two sets of fees beat one.

It could very well be quicker for my hair dresser to reload her Filipino account via XRP/bitcoin than the traditional route. For instance, the dealer who is buying her Canadian dollars and selling her pesos may delay the peso leg of the transaction for twenty-four hours. But this sluggishness isn't inherent to a fiat-to-fiat transfer. If she asks nicely, there is no reason the dealer can't expedite the transaction so that the pesos appear in her account within the hour. By queuing her request with many others over a twenty-four hour period the dealer reduces his overall costs, these benefits flowing back to my hair dresser in the form of reduced fees. Likewise, my hair dresser could choose to queue her XRP/bitcoin payment into a big chunk along with other people's cryptocurrency payments. This would slow things down, but reduce fees.

US dollars a bridge currency

Not all traditional cross border payments involve one transaction. Canada-Philippines is a relatively popular payments route, but rarely used payments corridors, say like Canada to Uzbekistan, will incorporate a third fiat currency—probably U.S. dollarsas a bridge currency. For this payment to proceed, my hairdresser will need both a Canadian dollar account and a U.S. dollar account. She will have to find a counterpart who straddles the U.S. and Uzbek banking systems by maintaining a U.S. dollar account and an Uzbekistani Som account. Once she transfers her counterpart some U.S. dollars, then he can execute the Uzbek leg of the payment.

Even on exotic corridors I have troubles seeing how XRP/bitcoin can compete as a bridge. The dollar is the world's most entrenched currency. The CADUSD market will always be deeper than the XRP-to-CAD or bitcoin-to-CAD market, and same on the Uzbek Som side. This depth means that transaction costs on U.S. dollar trades will be lower than on crypto trades. For this calculus to change, bitcoin or XRP will somehow have to displace the U.S. dollar as the world's most liquid medium of exchange. But this is unlikely to happen due to the incredible volatility of cryptocurrencies.


Which leads into the next defect of crytocurrencies as bridge assets. XRP and bitcoin are inherently volatile assets, so using crypto as a bridge means the risk of encountering a plunge in value.  In the case of XRPs, my hairdresser will have to hold them for at least a few moments (or even minutes), but that could be enough time to cause some damage. As for bitcoin, which is slower, she will have to carry them for an hour or two before they can be sold in Philippines. That's an eternity in cryptoland. To top it off, crypto exchanges are notoriously risky, outages and thefts being a regular occurrence. These are pretty big risks for my hair dresser to take, so using crypto markets will only make sense if they provide her enough compensating efficiencies.

Where might these come from? Traditional cross border payments have typically offered very little in the way of transparency. If my hair dresser's payment is stuck, it'll be hard for her to get information on its status. To cope with this informational gap, she may choose to constantly over-fund her peso account, which hurts her pocket book. One advantage of something like Ripple is that all XRPs are recorded on the public Ripple ledger, and thus my hair dresser should have a better idea about what stage her payment has progressed. And this may give her the confidence to reduce the amount by which she pre-funds her peso account, the freed up capital being invested in her salon.

That's a nice feature, but I don't quite see how increased transparency can possibly make up for 1) the inherent risks of holding cryptocurrencies, even if just for a few moments, and 2) the aforementioned transactions costs involved in running the bridge. Furthermore, the transparency advantage is being eroded as traditional payments systems respond to the competitive threat posed by players like Ripple. SWIFT, the communications network that is relied on to facilitate traditional cross border payments, has recently incorporated a tracking number to all payments, thus allowing users to get a real-time end-to-end view on the status of their payments.

So for now, I don't think there is much merit to using crypto as currency bridge in cross border payments. That doesn't mean XRP must crash because it has no use case. Dogecoina parody cryptocurrency that recently rose above $1 billion in valuedemonstrates that coins don't need a fundamental use case to justify their price. But I've been wrong many times about cryptoland, so let's see what happens.