...

2. Three orders of magnitude cost reduction.

CyberCash's entire business model for e-cash is wrong. They ask for a 4% slice of each transaction, which is absurd. How much do issuers of real cash ask per transaction? Zero, of course. As I've mentioned before, the way you make money with money is seignorage, that is you print it and make your profit on the float and on coins that are never redeemed. Works great for travellers checks. I suspect that Bob H. would agree. Regards, John Levine, johnl@iecc.com, Primary Perpetrator of "The Internet for Dummies", Information Superhighwayman wanna-be, http://iecc.com/johnl, Sewer Commissioner Finger for PGP key, f'print = 3A 5B D0 3F D9 A0 6A A4 2D AC 1E 9E A6 36 A3 47

Bob Hettinga wrote:

At 2:19 PM -0400 on 7/7/98, John R Levine wrote:

...

CyberCash's entire business model for e-cash is wrong. They ask for a 4% slice of each transaction, which is absurd. How much do issuers of real cash ask per transaction? Zero, of course.

As I've mentioned before, the way you make money with money is seignorage, that is you print it and make your profit on the float and on coins that are never redeemed. Works great for travellers checks. I suspect that Bob H. would agree.

Pretty much. Actually, the model I like is one where transactions on the net are free, but putting money *on* the net costs some smidge of money, just like buying traveller's checks cost money, something like one or two percent, I think.

You can't get any network effects, oddly enough, if you charge to take it *off* the net, because merchants have to *pay* to participate. It would be as if you had to pay to redeem a traveller's check. If that were the case nobody would accept them. Pretty serious seignorage loss that would cause. :-).

From a *technical* standpoint, it makes sense to design an electronic cash system to allow cheap coin-for-coin exchanges. The marginal cost of a

Haven't you said in the past that the *consumer* can't be charged anything to use the system in the way of fees (consumer being the party bringing money into the system), and that all fees need to get charged to the merchant (or whoever the party is who removes money from the system)? Certainly, at first order, the fee *should* be on removing money from the system, from a selfish point of view on the part of Ivan the Issuer -- if you're making money on seignorage, you're making money when people add money to the system *and leave it there*. Thus, you want to provide incentives for people to buy money and leave it there -- charging no "on" fees and minimal "off" fees would do it. To ensure that people have faith in the cheap convertability of electronic cash to real money, you could provide a highly regressive fee structure for taking money out of the system -- if people end up only redeeming their stuff through middlemen in huge batches, the fees will go down very fast, and so will your costs. The percentage cost of shipping a billion dollars to a major customer in exchange for a billion dollar token is much lower than shipping one billion one dollar bills to a bunch of holders of one dollar tokens. (This is an argument for regressive fees, not for charging out rather than in, since it would apply just as well to those who bought a billion tokens and then resold them, of course) transaction like this *is* 3 orders of magnitude less than one involving a back-office system and wire transfer. *This* is why electronic cash is 3 orders of magnitude cheaper than existing systems. If you kludge a back office system, a debit card system, and a wire transfer system onto the backend, transactions going through that system don't get any cheaper than they wold be in an unregulated traditional offshore bank. As an issuer, you want as few back-office processed transactions as possible, and those you do have, should be as large as possible. A better example is the credit card, rather than the traveler's check. I've *never* used a traveler's check, since I'm fundamentally annoyed by the idea that I'd have to pay a fee to put my money into the system. The *vast* majority of consumers are unwilling to pay a fee to conduct a transaction they can conduct in another way for free -- credit cards, cash, personal checks, wire transfer, or whatever. Merchants *already* pay money to handle money -- it's just another cost of doing business. For a merchant, processing *cash* costs money, credit cards cost 1-4%, and I'm sure traveler's checks cost *something* to handle. I think it's a fundamental principle of financial psychology that the resistance in going from a free service to a slightly charged service is far higher than even a much larger increase in a service that already costs money. I think the ideal fee structure for an electronic cash issuer is: * No cost to exchange coin-for-coin. These cost you microcents anyway. * Minimal (zero) cost to the public to issue cash. This could be done by selling large blocks of tokens at cost to issuers who are already in contact with the public, and who get some kind of tangible benefit from processing the transaction for the customer -- a pre-existing banking relationship that wants a value-add, a retail store, a web site with advertising, whatever. Primary market cash issuing might have a highly regressive fee structure, to force secondary markets to come into being, but this is not essential. * Secondary market redemption costs which *also* go to zero due to many players, for the same reason as issuance. Primary market should charge enough to recoup redemption costs (charge for the wire transfer, or whatever), and potentially a regressive structure to help pay for operations costs. Optionally, the issuer could link through a proprietary and cheap protocol their own bank accounts to the redemption of cash, making it much cheaper for customers of the issuer's bank to redeem cash. This is related to seignorage, if these bank accounts are at the same bank which holds the backing accounts for the currency. A lot of people have discussed this fee structure, and I agree with them that it is the best. AFAIK, Mark Twain had a somewhat opposite fee structure, and look at how successful they were... This model is not really the credit card model (charge per transaction), nor it is the traveler's check model. It is the *cash* model -- banks and cash handling companies charge handling fees for dealing with large quantities of paper bills. Think of user-to-user transactions as secondary market, withdrawing cash from the bank as primary market withdrawal, and exchanging electronic cash for fiat money as the (currently hypothetical) step of redeeming bank notes for their backing, or something similar to taking physical cash from retailer tills and getting it deposited to a cash management account. It is electronic *cash*, after all. -- Ryan Lackey rdl@mit.edu http://sof.mit.edu/rdl/

Bob Hettinga wrote:

At 6:41 AM -0400 on 7/8/98, Ryan Lackey wrote: You probably can't bootstrap things on seignorage alone, I don't think<double negative?>, but if you did, again, you would kill your digital cash issue anyway, because of lack of merchant acceptance, especailly in the early phases.

I think you can bootstrap things on "value-add" to a bank, and on seignorage, taken together. It is *good business* for a bank to provide a service like this. It is one of those services, like online banking, ATMs, ubiquitous branch locations (mmm, BankBoston...), private banking personnel, etc. which make customers happier. It is a good business case for a bank.

Note, in my canonical example, cash, that people already pay a fee for "foreign" ATM withdrawls anyway, and people are used to doing things that way. That's great, because I expect that margins on a net.ATM withdrawl through an underwriter would be significantly higher than what a bank makes on their non-customer ATM transactions. Three orders of magnitude cheaper? I wouldn't be surprised.

My banks eat the foreign ATM costs as a customer value-add expense. For the cash example, banks don't generally charge users to withdraw or deposit money at a teller window, or at their own ATMs. I think an online bank dealing with electronic cash is more dealing with "internal" transactions than with "foreign" transactions -- AFAIK, a lot of the foreign ATM charges are charged by the foreign bank, which has no real interest in the user doing anything but switching banks, anyway.

...

Certainly, at first order, the fee *should* be on removing money from the system, from a selfish point of view on the part of Ivan the Issuer -- if you're making money on seignorage, you're making money when people add money to the system *and leave it there*. Thus, you want to provide incentives for people to buy money and leave it there -- charging no "on" fees and minimal "off" fees would do it.

Yup. But, oddly enough, that was Mark Twain's (final) fee structure, too. Didn't save them in the end, though.

I argue that they had already shot their feet full of holes to the extent that nothing they did would have saved them by that point, but Mark Twain is a complex enough example that it might be worth just ignoring. I don't know if their experiment will even be a footnote to the footnote to history that is DigiCash.

I just think that if merchants are the key to acceptance of digital bearer cash, much less fully anonymous blinded digital bearer cash :-), then you shouldn't charge merchants anything to accept the stuff. If a merchant can download a wallet or registerware free or very cheap, and instantly start taking cash payments for whatever they sell over the net, and, when it came time to take that cash and put it into their own bank it didn't cost them anything to do it, then they would probably accept the stuff a heartbeat. The cost of anything is the foregone alternative, three orders of magnitude, and all that and "free" comes pretty close to three orders of magnitude in cost reduction from merchant credit card fees in my book. :-).

I've already shown why it is in the best interest of the issuer to participate. The next step is to convince merchants it is worth doing. Merchants, I believe, are willing to accept a percentage redemption fee for a new payment system if their customers demand it. They fundamentally want to make their customers happy. If a lot of customers request a new payment scheme, or they are convinced customers want to use it, they will happily use a new payment scheme. I think resistance to new payment schemes on the part of a merchant is in cost and effort to set up. Having a free apache commerce plugin which handles a variety of payment systems, including electronic cash, so the systems integration costs for adding electronic cash to an operation are zero, and they end up being cheaper than credit cards to process -- a 1% primary market fee, or a going-to-zero percent secondary market fee. Or, provide non-fee incentives for merchants to accept the system -- loyalty points, tax-free accounting in offshore bank accounts, the Amex approach of marketing merchants who accept the system to their client base directly, etc. For a directed payment system, going from customer to merchant to bank, then you definitely need merchants before you can get customers, which turns this on its head. *However*, a good electronic cash system is user-to-user, the Mondex model, rather than DigiCash's model. This means the system can bootstrap as soon as a single pair of people want to exchange money. People already want to exchange money user-to-user on the net -- there's a huge unsatisfied demand. That's more than enough to get it in place. Making it anonymous adds even more incentive.

I see seignorage as a tasty long-view source of income, certainly, but I expect the half-life of a given dollar of digital cash to be measured in days, if not hours, in the early stages. Frankly, I don't see seignorage showing up significantly on the income statement of an underwriter until there are other bearer instruments (stocks, bonds, and especially mutual fund shares) to invest that cash in, keeping it on the net.

The half-life of a given dollar of digital cash will be *infinite* if there is an expanding base of users, a cheaper secondary market for getting electronic cash than the cash out and cash back in system, and there is never a big panic. The holding time for a particular user might be pretty short, but I think from the beginning a large amount of money will just be left on the net, since it's far cheaper for a consumer to buy cash from a merchant directly than for the merchant to cash out and the user to buy them on the primary market.

I think that that's one of the reasons that the Fed, among other people, aren't too worried about the immediate macroeconomic effects of digital cash. That and that the Fed's seignorage income, in the overall scheme of Fed revenues, is pretty small, and probably dwarfed by other things like printing and handling costs, etc. Greenspan himself is/was a free banking advocate, certainly, and has said publically (see last September's Official Cypherpunk Forbes issue :-)) that he thinks that private electronic banknote issue is not a scary proposition at all.

I thought seignorage was a net profit for the Fed, after printing/handling. They *do* give the Treasury a kickback for this, and I'm not sure how the budget actually works, but I think they pay all distribution/printing/etc. costs themselves, and thus this is actually spare money. It doesn't really apply here, since many of those costs are fundamentally different in the electronic cash world than in the printed cash world. Greenspan is/was a libertarian, and would probably be ok with private electronic banknotes putting him out of business, as long as it was better for the economy and people as a whole. -- Ryan Lackey rdl@mit.edu http://sof.mit.edu/rdl/