Ripple and contingent liability
Martin Brock
For example, Peter owes me $10, and I want a $10 good from you. You
agree to exchange the $10 good for Peter's obligation to me. I then
have the $10 good, Peter owes me nothing, and Peter owes you $10.
If Peter defaults on his $10 obligation after this transaction, you
wish me to share the loss with you. You essentially paid Peter's debt
to me by giving me the $10 good, but you are left with nothing. You
agreed to trust Peter, but I also agreed to trust him, and I agreed to
trust him first.
In Ripple's system of settlement, in this scenario, following Peter's
default, do I owe you anything?
If I don't share the cost of Peter's default with you in this
scenario, then I may game the system by creating a bogus obligation
from Peter to me and transferring this obligation to you in exchange
for valuable goods, knowing all along that Peter will default. Peter
might not even exist.
This is basically what happened in the mortgage backed security
debacle. Creditors wrote dubious mortgages and then sold them to
securitizers. The securitizers then created mortgage backed securities
and resold the securities, and buyers of the securities ultimately
swapped them for taxpayer obligations. Ideally, a mutualist system of
exchange should avoid this problem.
Ripple addresses this problem by transferring obligations only to
people who trust the obligor, but contingent liability addresses the
problem more fully. Is accounting for contingent liability practical?
Melvin Carvalho
In this scenario you need honest ratings agencies
http://www.wikirating.org/wiki/Main_Page
Has potential
>
> --
> You received this message because you are subscribed to the Google Groups "Ripple Project" group.
> To post to this group, send email to rippl...@googlegroups.com.
> To unsubscribe from this group, send email to rippleusers...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/rippleusers?hl=en.
>
Kurt Padilla
But "you" already trust Peter. Otherwise "you" wouldn't have given him at least $10 of credit.
Sent from my HTC Incredible 2.
Martin Brock
There is no more central ratings agency. You and I extend credit to
Peter because we know and trust Peter, not because a more central
authority tells us to trust him and/or guarantees the credit by
socializing losses.
You and I really know Peter, and we really trust him, but he could
default anyway. He could be part of a fraudulent scheme, or he could
default honestly. He could die or become disabled for example. The
question is: who bears the loss? Only the last person to trust him in
a sequence of transactions or everyone who trusted him in the
sequence?
In the system of contingent liability that I imagine, everyone in a
sequence of transactions, in which Peter's promissory note is used as
money, shares the risk of Peter's default. If you accept the note, you
retain some of the risk of default when you pass the note to someone
else.
Martin Brock
> least $10 of credit.
trust him first to return a good comparable to a good that he
(presumably) receives from me. You then trust him to return a good
comparable to a good that I receive from you. Peter will return this
good to you on my behalf rather than returning a comparable good to
me.
In other words, I hold Peter's promissory note, and I spend it in a
transaction with you.
Ultimately, Peter doesn't return a good to anyone. Without contingent
liability, if you're the last person holding the note, you bear the
loss exclusively.
You could allow me to spend Peter's promissory note with you only if I
agree to share the risk of Peter's default.
Favorati keeps records of all transactions. It never deletes or
updates a transaction, so it could account for contingent liability,
but contingent liability raises questions about creditworthiness. If I
am contigently liable for many defaults, after holding and then
spending many notes, how creditworthy am I?
Melvin Carvalho
Perhaps you need a mutual credit club of say 10 people. If someone
default the community covers the lost and they are ejected from the
club.
Martin Brock
> default the community covers the lost and they are ejected from the
> club.
small group, it might work, but in larger groups, it can encourage
reckless credit.
With contingent liability, you bear the cost of Peter's default only
if you exchange a good for Peter's promise to repay you. If Paul then
accepts Peter's promissory note from you in exchange for another good,
you share the default risk with Paul. If Mary accepts the note from
Paul in exchange for another good, you, Paul and Mary all share the
risk. As more people accept Paul's note, more people share the risk,
but no one shares the risk without a making personal decision to trust
Peter.
Perhaps anyone sharing the risk should be entitled to call the note,
i.e. to ask Peter to settle his obligation with the current note
holder.
Melvin Carvalho
>> Perhaps you need a mutual credit club of say 10 people. If someone
>> default the community covers the lost and they are ejected from the
>> club.
>
> This arrangement is essentially what I mean by "socialized loss". In a
> small group, it might work, but in larger groups, it can encourage
> reckless credit.
Slight aside, isnt this how Grameen Bank works?
>
> With contingent liability, you bear the cost of Peter's default only
> if you exchange a good for Peter's promise to repay you. If Paul then
> accepts Peter's promissory note from you in exchange for another good,
> you share the default risk with Paul. If Mary accepts the note from
> Paul in exchange for another good, you, Paul and Mary all share the
> risk. As more people accept Paul's note, more people share the risk,
> but no one shares the risk without a making personal decision to trust
> Peter.
>
> Perhaps anyone sharing the risk should be entitled to call the note,
> i.e. to ask Peter to settle his obligation with the current note
> holder.
>
Kurt Padilla
But what if there are other people besides Peter between you and me?
I'm not convinced that Ripple needs to spread losses associated with defaults beyond whomever issued credit to the defaulter.
Kurt
Sent from my HTC Incredible 2.
Apostolis Xekoukoulotakis
Sincerely yours,
Apostolis Xekoukoulotakis
Martin Brock
valuable good in exchange for his IOU. You then accept Peter's IOU
from me for another valuable good. Mary accepts Peter's IOU from you
for a third valuable good.
At this point, Peter and I and you all have valuable goods, and Mary
has the IOU. Mary asks Peter for a comparable good, but he defaults.
With contingent liability, in this scenario, you and I both owe Mary a
third of the value of Peter's IOU. If Peter only defaults partially,
we owe Mary a third of her loss.
If Luke accepts the IOU from Mary before Peter defaults, then you and
I and Luke each owe Mary a fourth of her loss.
Also, if you accept only a fraction of Peter's IOU from me, you may
owe Mary less than a fourth of her loss while I owe her more than a
fourth.
> I'm not convinced that Ripple needs to spread losses associated with
> defaults beyond whomever issued credit to the defaulter.
scenario, several different people accept the same IOU from Peter.
Everyone exchanging a valuable good for Peter's IOU extends the same
credit to Peter. Without contingent liability, the last person
accepting the IOU bears the entire loss, even though everyone in the
sequence trusts Peter to return the same favor at some point.
A system like Ripple can offer contingent liability or not. The
question is: are you more willing to accept Peter's IOU from me if I
share the default risk with you? If so, then more commerce occurs in a
system offer contingent liability.
Suppose the system gives you an option. You can accept Peter's IOU
from me and accept all of the default risk along with it, or you can
accept the IOU from me knowing that I share the risk. Which option
would you choose? This seems like a no-brainer. Practically everyone
would choose the latter.
Martin Brock
> whoever accepts Peters notes is the only liable person.
before Peter defaults.
> I think that all the transactions of Peter should be seen by those that
> trust him.
you also see summary information, like Peter's total debt. I may add
other summary information.
I see your point, but I'm not sure that people will play the game if a
single transaction exposes all of their transactions. Some anonymity
is attractive, and existing monetary systems offer much more
anonymity. A new system must compete with the existing systems.
> Secondly, the trustor could be paid a small amount to act as
> intermediary, that amount which corresponds to the risk taken.
persuade me to accept his IOU, and I could pay you a bit more, and you
could pay Mary a bit more, but if Mary still bears the entire loss
when Peter defaults, the other bits are little comfort to her. She
might receive similar bits in other transactions in which she
experiences no loss, but these gains aren't related to Peter's default
risk specifically. The idea is to ensure that everyone passing Peter's
note at every step really does trust Peter when s/he passes the note.
If I discover that Peter's note is worthless after I accept it, I can
pass it to someone who doesn't know its value yet. That's the problem.
People could pay in advance for default risk, and the payments could
accumulate in an insurance fund until a default occurs, but in this
system, people must estimate default risk in advance, to determine an
insurance premium, and if people underestimate the risk, the insurance
fund itself defaults. Contingent liability avoids this problem.
Martin Brock
>> I and Luke each owe Mary a fourth of her loss.
then you and I and Mary each owe Luke a fourth of his loss.
Kevin
> Right. I agree to trust Peter, and you also agree to trust him. I
> trust him first to return a good comparable to a good that he
> (presumably) receives from me. You then trust him to return a good
> comparable to a good that I receive from you. Peter will return this
> good to you on my behalf rather than returning a comparable good to
> me.
>
> In other words, I hold Peter's promissory note, and I spend it in a
> transaction with you.
>
> Ultimately, Peter doesn't return a good to anyone. Without contingent
> liability, if you're the last person holding the note, you bear the
> loss exclusively.
I could be wrong (in multiple ways), but I think this points out a major
difference between Ripple and Favorati that hasn't yet been clearly
explained.
Let's say Ann buys a trinket from Bob for $10. Then Bob buys a widget
from Carly for $18.
In Favorati, if I understand correctly, the system factors out the chain
of debt, resulting in Ann directly owing Carly $10, and Bob owing Carly
$8. So Carly now has to trust Ann, and if Ann defaults, Carly takes the
$10 loss. I think that is the "contingent liability" you want to handle.
But in Ripple, Ann still owes Bob $10, and Bob owes Carly $18. Only Bob
is on the hook for Ann's default. Carly doesn't even know that Ann was
involved in the transaction, and relies only on Bob, who she earlier
decided to trust with a $10 debt.
Is that right for both systems?
Kevin
Martin Brock
>>from Carly for $18.
>>In Favorati, if I understand correctly, the system factors out the chain
>>of debt, resulting in Ann directly owing Carly $10, and Bob owing Carly
>>$8.
favor. Carly may ask Bob to ask Ann to do Carly a favor worth $10 (or
less). After this $10 favor, Ann owes Bob nothing, because she did
Carly the favor as a favor to Bob, thus returning Bob's $10 favor to
her. Bob then owes only $8 to Carly, because Ann did a $10 favor for
Carly on Bob's behalf.
>>So Carly now has to trust Ann, and if Ann defaults, Carly takes the
>>$10 loss. I think that is the "contingent liability" you want to handle.
defaults, Bob still owes Carly an $18 favor. He bears the entire loss,
because Carly never agreed to trust Ann.
>>But in Ripple, Ann still owes Bob $10, and Bob owes Carly $18.
Contingent liability occurs in a different scenario. Ann buys the $10
trinket from Bob, so she owes Bob a $10 favor in return. Bob then
approaches Carly for a favor. Carly does not trust Bob, but she does
trust Ann. In this scenario, Bob may ask Ann to do Carly a $10 favor
in the future, and Carly trusts Ann to do this favor, so Carly will do
Bob a $10 favor now. Ann then owes Carly a $10 favor and owes Bob
nothing.
In this scenario, Ann owes the same $10 favor before and after the
transaction. Before the transaction, she owes the favor to Bob. After
the transaction, she owes the favor to Ann. Bob trusts Ann to return
the favor first. Carly then trusts Ann to return the same favor. Carly
never trusts Bob, and Bob never owes Carly a favor.
Though Bob never owes Carly a favor, he arguably should retain some
liability if Ann defaults, because he agreed to trust her for this
favor before passing the obligation to Carly. He could know that Ann
is not trustworthy (if only because she is dead) before passing the
obligation to Carly, so his expenditure of Ann's obligation could be
fraudulent. Even if he doesn't know, why should the last person
accepting Ann's note bear all of the cost of Ann's default?
Kevin
> Contingent liability occurs in a different scenario. Ann buys the $10
> trinket from Bob, so she owes Bob a $10 favor in return. Bob then
> approaches Carly for a favor. Carly does not trust Bob, but she does
> trust Ann. In this scenario, Bob may ask Ann to do Carly a $10 favor
> in the future, and Carly trusts Ann to do this favor, so Carly will do
> Bob a $10 favor now. Ann then owes Carly a $10 favor and owes Bob
> nothing.
>
> In this scenario, Ann owes the same $10 favor before and after the
> transaction. Before the transaction, she owes the favor to Bob. After
> the transaction, she owes the favor to Ann. Bob trusts Ann to return
> the favor first. Carly then trusts Ann to return the same favor. Carly
> never trusts Bob, and Bob never owes Carly a favor.
>
> Though Bob never owes Carly a favor, he arguably should retain some
> liability if Ann defaults, because he agreed to trust her for this
> favor before passing the obligation to Carly. He could know that Ann
> is not trustworthy (if only because she is dead) before passing the
> obligation to Carly, so his expenditure of Ann's obligation could be
> fraudulent. Even if he doesn't know, why should the last person
> accepting Ann's note bear all of the cost of Ann's default?
>
Ok. Thanks for the clarification.
It seems like there are two possibilities here:
1. Ann's obligation to Bob is legitimate, and verifiable. In that case,
when Carly agreed to accept that obligation, she has decided to trust
Ann, and therefore has agreed to take on the risk of Ann defaulting. I
don't see a need for contingent liability in this case.
2. Ann's obligation to Bob is illegitimate and/or unverifiable. In that
case, Carly is taking Bob's word for it that Ann's obligation exists.
But since Carly doesn't trust Bob, she should not accept this alleged
obligation. Therefore, I don't see a need for contingent liability here
either.
Of course, I may (still) be missing something.
In Favorati and in Ripple, is the actual process of shifting a debt
verifiable? Is it possible to forge a debt and pass it along? Is that
shift even possible in Ripple without Ann's involvement?
Kevin
Martin Brock
> when Carly agreed to accept that obligation, she has decided to trust
> Ann, and therefore has agreed to take on the risk of Ann defaulting. I
> don't see a need for contingent liability in this case.
liability. She could also accept Ann's promise from Bob with
contingent liability. Either transaction is legitimate and verifiable.
The question is: what is Carly's preference? What is in her best
interests? If I offer an accounting service with contingent liability,
and you offer a service without it, which service does Carly choose?
Is Bob willing to use the same service? These decisions are not mine,
but I want to offer the most attractive service I can offer.
> 2. Ann's obligation to Bob is illegitimate and/or unverifiable. In that
> case, Carly is taking Bob's word for it that Ann's obligation exists.
> But since Carly doesn't trust Bob, she should not accept this alleged
> obligation. Therefore, I don't see a need for contingent liability here
> either.
transaction. Either way, contingent liability protects Carly, because
three other people share the default risk with her.
If Bob is not defrauding Carly, contingent liability also protects
Bob, because Ann could also die five minutes before the transaction,
in which case two other people share the default risk with Bob.
> In Favorati and in Ripple, is the actual process of shifting a debt
> verifiable? Is it possible to forge a debt and pass it along? Is that
> shift even possible in Ripple without Ann's involvement?
I join Favorati as myself, and I also create an account for a
fictitious member, say Peter. The service tries to avoid this
deception, but it cannot be omniscient.
By some deception, "Peter" (who is really me) persuades you to trust
him. "Peter" acknowledges a favor from me, though this favor never
occurred. I then pass Peter's bogus IOU on to you. I do the same with
many other people, gaining many valuable favors this way. Then Peter
defaults.
Peter could also be a real person, other than me, with whom I conspire
in this scheme.
Here's a somewhat less egregious fraud.
Peter is a real, honest person, and I am a real, mostly honest person.
I do Peter a favor, so he owes me a favor legitimately. Carly really
trusts Peter, because Peter is really trustworthy.
A drunk driver kills Peter before he can return the favor he owes me.
Carly still trusts Peter, because she doesn't know he is dead yet, so
I spend Peter's IOU on a favor from Carly before she finds out.
These are extreme cases to illustrate the problem. Real cases could
involve any bits of information that I know but Carly does not know.
Is it "moral" for someone to spend notes with Carly that he knows to
be worth less than their face value? I don't know. I'm not Moses. I'm
not sure I accept a universal "morality" at all. The only question is:
will Carly and others be more willing to use my service if it accounts
for contingent liability?
Martin Brock
> But since Carly doesn't trust Bob, she should not accept this alleged
> obligation.
to extend them credit, then negotiable IOUs are not necessary at all,
but this requirement severely limits the possibility of mutually
beneficial exchange. More conventional money, with all of its faults,
is still more valuable than money that Carly hardly ever uses.
Conventional money, these days, is essentially a system of negotiable
IOUs (including IOUs created by states imposing debts involuntarily),
but losses are socialized. Escaping the weaknesses of this system is
the point of Ripple and Favorati, but any alternative must provide
most of the convenience of conventional money.
Apostolis Xekoukoulotakis
>him. "Peter" acknowledges a favor from me, though this favor never
>occurred. I then pass Peter's bogus IOU on to you. I do the same with
>many other people, gaining many valuable favors this way. Then Peter
>defaults.
Contingent liability is something that goes against the whole structure of ripple.
--
You received this message because you are subscribed to the Google Groups "Ripple Project" group.
To post to this group, send email to rippl...@googlegroups.com.
To unsubscribe from this group, send email to rippleusers...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/rippleusers?hl=en.
Sincerely yours,
Apostolis Xekoukoulotakis
Martin Brock
> the victim (through a chain) and one from you to the victim.
transaction, requires only one false edge. In general, the fraud
problem requires only one person knowing more about the value of a
note than one other person knows.
But fraud is only one concern. Default risk concerns everyone
accepting a note, whether or not fraud is involved. Maybe I don't know
that Peter has died before passing Peter's note to Carly. Carly still
bears the loss exclusively, so she still has reason to value
contingent liability.
Contingent liability is an alternative to default insurance, systems
like deposit insurance, credit default swaps and the like. The problem
with default insurance is that we buy it before defaults occur, and
after we buy it, we don't really care about the defaults anymore, so
we don't bother to learn knowledge that might limit the circulation of
a note with a high risk of default. The face value of a note doesn't
change when its expected value changes. All notes trade at face value,
regardless of default risk, because all losses are socialized. Ripple
addresses this problem, but people still want to hedge default risk.
The question is: how?
> Contingent liability is something that goes against the whole structure of
> ripple.
within a system like Ripple is feasible. When Peter defaults in the
scenario above, others who have held Peter's note owe a favor to the
last person holding the note. All of these people, from Peter onward,
have exchanged the note for valuable goods, so I don't see the
problem.
What you and I believe about the virtue of contingent liability is not
the issue here. What Carly and others holding Peter's note believe
about it is the only issue. I may offer a service accounting for
contingent liability, and Carly and others may choose to use this
service instead of a service that does not account for contingent
liability. In a free market, nothing else matters.
Ultimately, to be competitive with conventional money, a system like
Ripple must offer some sort of default insurance, or people won't use
it. Conventional deposit insurance spreads risk and socializes loss
too much. Credit default swaps pay people with knowledge of default
risk to keep the knowledge to themselves. Carly values knowledge of
this risk before accepting a note, and the person passing her the note
likely has more of the knowledge than she does. The solution should
encourage dissemination of this information rather than discouraging
it.
> I havent got enough time to do research on this matter, I will eventually,
> but you should read research papers on trust networks and then give us some
> info here in this mailing list.
read.
Jorge Timón
> Does Ripple implement any sort of contingent liability?
>
> For example, Peter owes me $10, and I want a $10 good from you. You
> agree to exchange the $10 good for Peter's obligation to me. I then
> have the $10 good, Peter owes me nothing, and Peter owes you $10.
>
> If Peter defaults on his $10 obligation after this transaction, you
> wish me to share the loss with you. You essentially paid Peter's debt
> to me by giving me the $10 good, but you are left with nothing. You
> agreed to trust Peter, but I also agreed to trust him, and I agreed to
> trust him first.
>
> In Ripple's system of settlement, in this scenario, following Peter's
> default, do I owe you anything?
No.
> If I don't share the cost of Peter's default with you in this
> scenario, then I may game the system by creating a bogus obligation
> from Peter to me and transferring this obligation to you in exchange
> for valuable goods, knowing all along that Peter will default. Peter
> might not even exist.
Why does "you" trust Peter if he doesn't even exist? Participants are
supposed to only accept IOUs from nodes they know and trust.
If Alice pays Bob with 5 bigmac tickets and mc donalds goes bankrupt,
Bob will have to take all the losses. He should have not accepted the
tickets if he wasn't confident about mc donald's solvency and
trustworthiness. Alice is not responsible for mc donald's default in
any way, she did nothing wrong and doesn't have to share the losses.
If you pay me with dollars and next week there's panic hyperinflation
and usd goes to zero, should you take any losses?
In your words, I trusted the federal reserve but you did first.
--
Jorge Timón
Martin Brock
>Participants are supposed to only accept IOUs from nodes they know and trust.
him is very credible. He can die unexpectedly for example.
>If Alice pays Bob with 5 bigmac tickets and mc donalds goes bankrupt, Bob will have to take all the losses. He should have not accepted the tickets if he wasn't confident about mc donald's solvency and trustworthiness.
bigmac ticket, but if Alice shares the default risk with him, as she
accepts Bob's valuable good for the ticket, Bob might be more willing.
Again, our opinion of Bob's liability isn't the issue here. The terms
of Bob's agreement with Alice are the issue. Bob may refuse Alice's
bigmac ticket if Bob refuses to share liability for the default. Only
Bob and Alice matter. They're the ones agreeing to trade.
If you want to dissuade me from accounting for contingent liability,
tell my why Bob and Alice will be less likely to use Favorati if I do.
>Alice is not responsible for mc donald's default in any way, she did nothing wrong and doesn't have to share the losses.
person holding the ticket. Alice also held the ticket. She also
trusted McDonalds.
>If you pay me with dollars and next week there's panic hyperinflation and usd goes to zero, should you take any losses?
More to the point, if I write a mortgage on a house that you sell to a
second person and then sell the mortgage to a third person who
packages the mortgage into a mortgage backed security and sells the
security to fourth person, I also think that you and I and the second
and third persons should retain some liability for the default.
I'm not suggesting that anyone be compelled to accept this sort of
liability in sales of this sort. I'm saying that I would accept notes
with this sort of liability if I had a choice, because I prefer not to
bear all of the liability if I happen to be the last person holding
the hot potato. I want the person passing the potato to me to have
some skin in the game.
>In your words, I trusted the federal reserve but you did first.
look for a monetary model less susceptible to its weaknesses. That's
what we're doing here.
Martin Brock
Jorge Timón
>>Why does "you" trust Peter if he doesn't even exist?
>
> Because I believed a con man.
Only your fault.
>>Participants are supposed to only accept IOUs from nodes they know and
>> trust.
>
> Participants are imperfect, but Peter can default even if my trust in
> him is very credible. He can die unexpectedly for example.
Those things happen. Maybe Peter linked his IOUs to a legal contract
and his heirs must pay you from his legacy.
>>If Alice pays Bob with 5 bigmac tickets and mc donalds goes bankrupt, Bob
>> will have to take all the losses. He should have not accepted the tickets
>> if he wasn't confident about mc donald's solvency and trustworthiness.
>
> Lecturing Bob this way doesn't make him any more likely to accept a
> bigmac ticket, but if Alice shares the default risk with him, as she
> accepts Bob's valuable good for the ticket, Bob might be more willing.
But if he still does, he has to responsibly accept the risks he's talking.
> Again, our opinion of Bob's liability isn't the issue here. The terms
> of Bob's agreement with Alice are the issue. Bob may refuse Alice's
> bigmac ticket if Bob refuses to share liability for the default. Only
> Bob and Alice matter. They're the ones agreeing to trade.
Yes, he can refuse the big mac tickets. He should only accept them if
he trusts mc donalds enough.
> If you want to dissuade me from accounting for contingent liability,
> tell my why Bob and Alice will be less likely to use Favorati if I do.
Maybe it's a good way to differentiate Favorati from Ripple. I don't know.
The main drawback I see is that something I thought I had pay may
suddenly appear as unpaid. Also I don't see how this would work with
longer paths.
Consider this payment A -> B -> C -> D
C dies when D was holding C's IOUs. Must B take some losses? He was
only acting as intermediary. I think that people will be less likely
to act as intermediary, which is the very basis of Ripple
>>Alice is not responsible for mc donald's default in any way, she did
>> nothing wrong and doesn't have to share the losses.
>
> Bob isn't responsible for the default either. He's only the last
> person holding the ticket. Alice also held the ticket. She also
> trusted McDonalds.
But he was the one taking the risk of accepting the tickets.
>>If you pay me with dollars and next week there's panic hyperinflation and
>> usd goes to zero, should you take any losses?
>
> In my opinion? Yes, I should.
What? You buy me a car paying with dollars, the value of the dollar
drops and you still owe me something when I voluntarily accepted your
payment as good?
> More to the point, if I write a mortgage on a house that you sell to a
> second person and then sell the mortgage to a third person who
> packages the mortgage into a mortgage backed security and sells the
> security to fourth person, I also think that you and I and the second
> and third persons should retain some liability for the default.
As long as there's no fraud involved (like was the case with the
subprime packages), people should be responsible for the trades they
make.
> I'm not suggesting that anyone be compelled to accept this sort of
> liability in sales of this sort. I'm saying that I would accept notes
> with this sort of liability if I had a choice, because I prefer not to
> bear all of the liability if I happen to be the last person holding
> the hot potato. I want the person passing the potato to me to have
> some skin in the game.
>
>>In your words, I trusted the federal reserve but you did first.
>
> That's right, and I will suffer hyperinflation if it occurs, and I'll
> look for a monetary model less susceptible to its weaknesses. That's
> what we're doing here.
You will not suffer the hyperinflation with the money you've already
spent and I see no reason why you should. I could have sell the
dollars for euros, bitcoins or gold when you gave them to me. It's my
problem, not yours. If the dollars double in price (compared to, say,
gold) will I return half of the price of the car? No. Of course not.
--
Jorge Timón
Martin Brock
>Only your fault.
In this scenario, I receive a note from someone else who also trusted
Peter, so "only" is misleading.
Think of a Ponzi scheme. People who get out early make out like
bandits, often because they are bandits. I'm only suggesting that
these people, and only these people, share the risk of the scheme's
default, an inevitable default in this case. Of course, I don't want
this sort of thing ever to happen in Favorati.
But "fault" is irrelevant. I don't care who you blame. The question
is: will Bob accept Alice's note without contingent liability if he
has a choice, and will Alice give him a choice? That's the only
question I'm discussing.
>Those things happen. Maybe Peter linked his IOUs to a legal contract and his heirs must pay you from his legacy.
irrelevant. Bob is free to accept Alice's note only if Alice will
share liability. Nothing else is relevant to my point.
>But if he still does, he has to responsibly accept the risks he's talking.
liability in his bargains if others will agree.
>Yes, he can refuse the big mac tickets. He should only accept them if he trusts mc donalds enough.
the terms of trade. Would you forbid Bob and Alice to bargain this
way?
>Maybe it's a good way to differentiate Favorati from Ripple.
liability appeals to people, Ripple may offer the same accounting
service. We're providing a service to a free market.
>Consider this payment A -> B -> C -> D
>C dies when D was holding C's IOUs. Must B take some losses?
C holds A's note. Then D holds A's note. The default risk is always
the risk of A's default. C owes nothing to D after the last
transaction, unless A defaults. That's the "contingency".
A may owe B who owes C without C ever trusting A. In this case, C is
not liable for A's default.
>He was only acting as intermediary. I think that people will be less likely to act as intermediary, which is the very basis of Ripple
unless B somehow guarantees A's indirect obligation to C.
A bank is an intermediary between its depositors and borrowers,
because the bank's assets guarantee deposits. Bank shareholders lose
everything before a depositor loses anything.
Contingent liability is more like a default insurance pool. Everyone
passing A's note joins the pool, and only these people join the pool.
If A issues many notes, I share default risk only for the notes that I
pass, not for all of the notes.
>What? You buy me a car paying with dollars, the value of the dollar drops and you still owe me something when I voluntarily accepted your payment as good?
notes that will depreciate precipitously tomorrow, so you wish me to
accept the terms.
>As long as there's no fraud involved (like was the case with the subprime packages), people should be responsible for the trades they make.
contingent liability if they wish.
>You will not suffer the hyperinflation with the money you've already spent and I see no reason why you should.
my notes if I will not trade on this basis?
You may accept notes without contingent liability, but if you do,
people with rapidly depreciating notes will seek you out, especially
if you don't know that the notes are rapidly depreciating. In other
words, people will profit from a market inefficiency at your expense.
I'm not saying that these people are immoral. I'm only saying that you
need not trade with them.
>If the dollars double in price (compared to, say, gold) will I return half of the price of the car? No. Of course not.
circumstances, then you will honor your agreement, because you are an
honorable man. We aren't discussing universal, moral absolutes here.
We're discussing specific terms of specific contracts.
Jorge Timón
new general rule when in fact you just want to add another option.
While I don't think this would be the most common case, it makes sense
to add a new possibility that users can use or not.
2012/2/6, Martin Brock <reston...@gmail.com>:
>>Consider this payment A -> B -> C -> D
>>C dies when D was holding C's IOUs. Must B take some losses?
>
> No. In the contingent liability scenario, B first holds A's note. Then
> C holds A's note. Then D holds A's note. The default risk is always
> the risk of A's default. C owes nothing to D after the last
> transaction, unless A defaults. That's the "contingency".
So you want D to accept A's notes (which he doesn't trust) by sharing
the default risk.
Is this how the new transaction type works?
> A may owe B who owes C without C ever trusting A. In this case, C is
> not liable for A's default.
This is the general case.
>>He was only acting as intermediary. I think that people will be less likely
>> to act as intermediary, which is the very basis of Ripple
>
> If A owes B, and B owes C, B is not an intermediary between A and C,
> unless B somehow guarantees A's indirect obligation to C.
B is an intermediary in the Ripple transaction. That's how we use to
name B's role in the Ripple transaction. B doesn't warranties anything
but his obligation to C. A has no obligation to C, only to B.
Martin Brock
>the default risk.
>Is this how the new transaction type works?
A gives an IOU to B in exchange for B's valuable good.
B now wants a good from C, but C doesn't trust B; however, C does
trust A, and B has A's IOU. B exchanges A's IOU for C's good. A no
longer owes B. A now owes C.
C now wants a good from D, but D doesn't trust C, so C exchanges A's
IOU for D's good. A no longer owes B or C. A now owes D.
Everyone in this sequence trusts A. No one trusts anyone else.
A is like a bank issuing promissory notes. In a free banking system, a
bank is an organization that many people trust. Because many people
trust the bank, if it turns out to be untrustworthy, many people
suffer losses.
In a bank run, depositors suddenly stop trusting a bank, and
depositors race to the bank to claim what little the bank has before
it runs out. In this scenario, depositors at the end of the line or
the bank itself may ask a court to halt the bank's redemption of its
notes and then to determine the bank's ability to keep its promises in
an orderly hearing. If the bank cannot keep its promises, the court
may declare it bankrupt and order an equitable division of its assets
among the depositors. Being first in a line at the bank on a given day
is not decisive in this division. Other priorities, like who trusted
the bank first, may be relevant. Maybe B has higher priority than D or
vice versa. Maybe everyone has the same priority.
>This is the general case.
> name B's role in the Ripple transaction. B doesn't warranties anything
> but his obligation to C. A has no obligation to C, only to B.
be a financial intermediary. A bank is a financial intermediary but
not simply because it owes and is also owed. A bank is a financial
intermediary because of its fiduciary role, because many other people
have faith in it. Contingent liability essentially makes A a financial
intermediary in the scenario above, but A only owes things. Everyone
trusts A, but no one owes A anything.
Kevin
> Contingent liability is more like a default insurance pool. Everyone
> passing A's note joins the pool, and only these people join the pool.
> If A issues many notes, I share default risk only for the notes that I
> pass, not for all of the notes.
I think this gets closer to the heart of the matter. We are discussing a
form of insurance. I see these options:
1. Self-insurance. If you take a risk, you take the associated losses.
This would be the "default" behavior, and is how Ripple works today.
2. Pre-paid insurance. For a transaction, you would put some money into
a pool, which would pay out in the case of default. This could be
optional, or a system could make it mandatory.
3. Contingent liability ("spreading out the pain"). By participating in
a CL transaction, you agree to take on some of the risk associated with
that obligation. A system could make this mandatory for everything, or
could enforce it for obligations that have been flagged for CL, or could
enforce it for transactions involving obligations by person X, or could
make it optional on a per-transaction basis (or perhaps other
possibilities).
The default could be caused by intentional fraud, or could be caused by
death or incapacitation, or by simple reneging (including bankruptcy).
I am all in favor of having voluntary, free-market, pre-paid insurance
available, but I'm not sure I would participate in a system with
mandatory insurance. Similarly, I am fine with voluntary contingent
liability, but doubt I would participate in a system with mandatory
contingent liability. When practical, I prefer to self-insure, partly
because I tend to take steps to limit my losses, making any form of
insurance relatively more expensive for me than average.
> >If the dollars double in price (compared to, say, gold) will I return
> >half of the price of the car? No. Of course not.
>
> If you agree to return half the price of the car under these
> circumstances, then you will honor your agreement, because you are an
> honorable man. We aren't discussing universal, moral absolutes here.
> We're discussing specific terms of specific contracts.
>
At first I thought you were proposing extending contingency to cover
fluctuating asset values, which seemed very strange, and very
unappealing. If the contract was for dollars, or gold, or donuts, or
hours, then whoever accepted that asset in payment agreed to take on the
risks (and rewards) of that asset going down (or up) in value. This
seems entirely different from the case of promising $100 but only
delivering $50 (or nothing at all).
But now I see you were just saying that an individual contract could be
written to account for that. Yes, but that is entirely outside the scope
of Ripple (or presumably Favorati). The system would merely record the
debt, relying on the users to adjust it later if the contract required
that. (I hope) this tangent has nothing to do with contingent liability.
Your question is whether contingent liability is desirable, or perhaps
even necessary, in a system like Ripple. I am not yet convinced it is
worth the complexity of offering it as an optional feature. The benefits
seem minimal, so the costs (of development and of complexity) seem to
outweigh that. I'm still listening, however.
Kevin
Jorge Timón
When I say intermediary I mean "Ripple intermediary" a person that is
in the middle of a Ripple transaction. Ripple intermediaries also play
a fiduciary role so I would say Ripple intermediaries are also
financial intermediaries but the important thing is that we understand
each other. If the term is important for something, let's distinguish
between Ripple intermediaries and financial intermediaries.
>>Consider this payment A -> B -> C -> D
You didn't understood my example. A cannot pay D directly with aUSD
(A's IOUs) because D doesn't trust A.
By A -> B -> C -> D I meant a typical Ripple transaction. That could
be something like:
A buys 10 bUSD from B for 10 aUSD
10 cUSD from C for 10 bUSD
10 dUSD from D for 10 cUSD
A pay D with 10 dUSD
All this happens atomically.
So they end up
A Balance = -10 (has -10 aUSD or B has 10 aUSD and therefore she owes B 10 usd)
B balance = 0 (has 10 aUSD but -10 bUSD)
C balance = 0 (has 10 bUSD but -10 cUSD)
D balance = +10 (has 10 cUSD)
Is it possible to apply your proposal to a transaction like this?
How would it work?
By general case I mean a Ripple transaction as Ripple is now. Without
your special contingent liability case.
Martin Brock
>This would be the "default" behavior, and is how Ripple works today.
an uninsured risk.
>2. Pre-paid insurance. For a transaction, you would put some money into
>a pool, which would pay out in the case of default. This could be
>optional, or a system could make it mandatory.
collect from the pool if the risk affects us adversely.
We contribute to the pool before any adverse effect, and once we
contribute, we worry less about the effect. We no longer avoid it.
Also, we might not contribute enough to the pool.
>3. Contingent liability ("spreading out the pain"). By participating in
>a CL transaction, you agree to take on some of the risk associated with
>that obligation. A system could make this mandatory for everything, or
>could enforce it for obligations that have been flagged for CL, or could
>enforce it for transactions involving obligations by person X, or could
>make it optional on a per-transaction basis (or perhaps other
>possibilities).
fact, and 3 spreads it out after the fact.
In the system I imagine, the spreading occurs only if someone actually
defaults, after the dispute resolution process described at Favorati.
This process involves an online hearing by a jury of other Favorati
members; however, you always know your contingent liability, even if
you're never actually liable, and you can take steps to prevent the
liability, like calling notes for which you share liability before the
note issuer becomes too deeply indebted.
>The default could be caused by intentional fraud, or could be caused by
>death or incapacitation, or by simple reneging (including bankruptcy).
fraud at Favorati, because I don't want the service associated with
fraudulent schemes; however, people desire insurance for many reasons
other than fraud.
Fraudulent schemes discourage membership, and they also expose me to
legal liability potentially. A Favorati member agrees not to pursue
any dispute resolution involving Favorati outside of Favorati's
process, but state courts need not respect this agreement.
>I am all in favor of having voluntary, free-market, pre-paid insurance
>available, but I'm not sure I would participate in a system with
>mandatory insurance.
circulating notes, I will inform members in advance, of course. I
can't force you to read the terms of service, but you'll click
something saying that you've read them at least.
>Similarly, I am fine with voluntary contingent liability, but doubt I
>would participate in a system with mandatory contingent liability.
>When practical, I prefer to self-insure, partly because I tend to take
>steps to limit my losses, making any form of insurance relatively more
>expensive for me than average.
default) actually occurs. If the contingency occurs, you may owe
people from whom you have already received valuable goods in exchange
for a worthless note. You will always owe less than the value of the
good you received in exchange for the worthless note.
If you want to accept notes without contingent liability, that's up to
you, but if you want to pass notes to other people without this
liability, people aware of the option may be less willing to accept
your notes. I don't see how they could be more willing, unless you pay
them a risk premium. A risk premium in advance of a contingency has
the problems discussed above.
Without a risk premium, refusing this liability reduces your
opportunity to trade, and trading benefits you, because others produce
goods at lower cost than you can produce them.
If you accept notes without contingent liability, people know that you
do, so people with depreciated notes seek you out. If you don't know
that the notes have depreciated (if you haven't updated your credit
limit for the note issuer), you will accept the notes at face value.
You can accept notes this way automatically at Favorati, when you
accept offers (buy things), if your credit limit (called "trust
policy" at Favorati) permits it. I believe Ripple is the same.
I can make contingent liability optional, but you understand the risk
you're taking; otherwise, you could sue me as the site's proprietor.
Some state court might shut me down no matter what I do, but I can
take reasonable steps to protect myself and the business. Accounting
for contingent liability is one of these steps.
You might not appreciate the fact, but I'm taking quite a few,
substantial, legal risks by offering this service. Circulating notes
at all (outside of the state monopoly system) is possibly illegal. I'm
not sure. You and I may agree that the service should not be illegal,
but that's small comfort to me when armed Feds show up at my door.
>But now I see you were just saying that an individual contract could be
>written to account for that. Yes, but that is entirely outside the scope
>of Ripple (or presumably Favorati).
will not use the services without it or something like it. I
understand your preference for self-insurance. I'm a skydiver myself,
but the overwhelming majority of people are not.
>The system would merely record the debt, relying on the users to
>adjust it later if the contract required that.
adjusts member profiles accordingly, with or without contingent
liability. Currently, without it, obligations in default remain on the
record. All obligations remain on the record indefinitely for that
matter, including obligations that have already been satisfied.
With contingent liability, defaulted obligations will remain on the
record, and new obligations will appear. Obligations from the
contingently liable members to the defaulted note holders will
apppear. Obligations from the defaulting member to the contingently
liable members could also appear, but the defaulting member is already
in default, so these obligations presumably have little value. At
Favorati, if someone ultimately will not return a favor to you, you
may not force him. You may only refuse him further favors and inform
others of his default.
>... the costs (of development and of complexity) seem to
Thanks for your input. I'm also listening.
Martin Brock
>in the middle of a Ripple transaction.
>Ripple intermediaries also play a fiduciary role so I would say Ripple intermediaries
>are also financial intermediaries but the important thing is that we understand
>each other.
does not play a fiduciary role in the conventional sense. If A owes B
and B owes me, I have no faith in A. I have only have faith in B. I
might not even know A. I might not know that B owes anyone else.
>You didn't understood my example. A cannot pay D directly with a USD
>(A's IOUs) because D doesn't trust A.
>Is it possible to apply your proposal to a transaction like this?
>10 cUSD from C for 10 bUSD
$10.
>10 dUSD from D for 10 cUSD
>A pay D with 10 dUSD
>All this happens atomically.
automatically when D accepts a $10 good from A.
Your example can occur at Favorati, but the settlement in your example
does not involve any contingent liability. Everyone has the means to
pay his debt. You're assuming that D has something worth $10 that A
wants. Default occurs when he does not.
Your example does not involve a circulating note. Ripple uses credit
limits, so I thought that Ripple also circulates notes, but I'm not
sure.
Martin Brock
wants.
If I understand you correctly.
Jorge Timón
>>You didn't understood my example. A cannot pay D directly with a USD
>>(A's IOUs) because D doesn't trust A.
>
> O.K. At Favorati, D cannot accept A's IOU without trusting A.
But aren't transitive transactions like this possible?
>>Is it possible to apply your proposal to a transaction like this?
>
>>A buys 10 bUSD from B for 10 aUSD
>
> A exchanges his $10 IOU for something of B's worth $10.
No, no. A buys B's IOUs (bUSD) paying B with its own IOUs (aUSD).
>>10 cUSD from C for 10 bUSD
>
> B exchanges his own $10 IOU (not A's IOU) for something of C's worth
> $10.
Only IOUs exchanged here too.
>>10 dUSD from D for 10 cUSD
and here.
> C exchanges his own $10 IOU for something of D's worth $10.
>
>>A pay D with 10 dUSD
This is the only part of the transaction where an actual good is
transacted. All what happened before was only a preparation for A to
get IOUs that D can actually accept.
>>All this happens atomically.
This was important.
> A owes B who owes C who owes D. All of these debts are settled
> automatically when D accepts a $10 good from A.
In my example the debts aren't settled, are in fact created. To settle
those debts more transactions (maybe including direct payments in
cash) are needed.
> Your example can occur at Favorati, but the settlement in your example
> does not involve any contingent liability. Everyone has the means to
> pay his debt. You're assuming that D has something worth $10 that A
> wants. Default occurs when he does not.
Not sure if a Ripple-like (transitive) transaction can happen within
Favorati now. But it seems that my question has been answered and it
doesn't apply to cases where there's a "trade of currencies".
> Your example does not involve a circulating note. Ripple uses credit
> limits, so I thought that Ripple also circulates notes, but I'm not
> sure.
Well it depends on how do you want to see it. Ryan prefers to consider
the IOUs exchanged between participants accounts and balances between
two pair of participants. I prefer to consider that every participant
prints its own currency and their neighboring nodes are the only ones
who accept it. But it's really the same concept and both approaches
are equivalent.
I also would prefer to implement the distributed protocol like this
and allow transactions involving proof of work chain based currencies
like bitcoin (which is technically possible). Also, other instruments
such as bonds or shares could be represented by the self issued
tokens. But that's kind of off-topic.
So the answer to "Do currencies circulate within Ripple?" is both yes and no.
Martin Brock
Keeping track of all of these relationships is complex and tedious,
but that's why a system like Ripple is valuable, of course.
Kevin
> >1. Self-insurance. If you take a risk, you take the associated losses.
> >This would be the "default" behavior, and is how Ripple works today.
>
> I understand you, but that's not "insurance" in the usual sense. It's
> an uninsured risk.
Self-insurance is a very common term, and I think it's accurate. I set
aside money in a bank account, so I can pay myself if something bad
happens that imposes a loss on me. Just for clarity, I am taking on an
uninsured (self-insured) risk, not generating an uninsured risk for
someone else.
> 2 [pre-paid insurance] and 3 [contingent liability] both spread out
> the pain, but 2 spreads it out before the
> fact, and 3 spreads it out after the fact.
Correct. However, the cost of 2 is known up front, and the cost of 3 is
not. Also, the cost of 2 is cleared up front, whereas the cost of 3 may
come back to haunt you 10 years from now. There are emotional benefits
to knowing that once a deal is done, it is done.
It is true that the insurer/pool could default in 2, although I suppose
it's also true that the participants in the contingent liability could
refuse or be unable to contribute to covering the loss. So neither is
100% safe, although I would agree that CL is probably safer.
> I can make contingent liability optional, but you understand the risk
> you're taking;
I can see where it might make more sense for a system to either have
mandatory contingent liability, or to not have it at all.
> You might not appreciate the fact, but I'm taking quite a few,
> substantial, legal risks by offering this service. Circulating notes
> at all (outside of the state monopoly system) is possibly illegal. I'm
> not sure. You and I may agree that the service should not be illegal,
> but that's small comfort to me when armed Feds show up at my door.
I can see that.
> >But now I see you were just saying that an individual contract could be
> >written to account for that. Yes, but that is entirely outside the scope
> >of Ripple (or presumably Favorati).
>
> Contingent liability is within the scope of both insofar as people
> will not use the services without it or something like it. I
> understand your preference for self-insurance. I'm a skydiver myself,
> but the overwhelming majority of people are not.
I think you missed my point. You are proposing contingent liability for
actual defaults. That is, someone is supposed to pay X and they do not.
You are not proposing that contingent liability affect cases where X has
changed in value since the transaction occurred.
Please confirm: You are not proposing that CL have any effect if a debt
denominated in dollars now buys only half as much gold as it used to. If
users want to factor that into their contracts outside of the scope of
Favorati, they could do so.
> >... the costs (of development and of complexity) seem to
> >outweigh that. I'm still listening, however.
>
> Fortunately for you, you don't bear the costs of development. ;)
Indeed. Although there are also usability costs to users, which is why I
can see why having CL be everywhere or nowhere may be better than having
it be optional.
Kevin
Martin Brock
site's home page (www.favorati.net) presents an example. Search for
"These terms permit the Favorati to satisfy many obligations
automatically".
A circulating note is different. The home page also presents an
example of a circulating note. Search for "Luke will not increase his
trust in Mary".
A ripple is also possible in the circulating note scenario. If John
owes Peter who owes Mary, and if Luke trusts John, then Mary may spend
John's IOU in a transaction with Luke. Peter then no longer owes John
(or owes him less), and Mary no longer owes Peter, but John owes Luke.
>This is the only part of the transaction where an actual good is
>transacted. All what happened before was only a preparation for A to
>get IOUs that D can actually accept.
transaction.
Instead, you're saying that D trusts C who trusts B who trusts A, so B
accepts A's note, and C accepts B's note, and D accepts C's note.
Then A owes B, and B owes C, and C owes D, and D delivers a good to A.
Right?
>In my example the debts aren't settled, are in fact created. To settle
>those debts more transactions (maybe including direct payments in
>cash) are needed.
occurs if A owes B directly and B owes C directly. A then owes C
indirectly, and A may satisfy both obligations at once by doing C a
favor, i.e. A does a favor for C as a favor to B. A's favor to B is
satisfying B's obligation to C.
What you've described is more like a circulating note with a ripple,
but I need to think more about it and don't have the time right now.
I'll be back.
Martin Brock
>John's IOU in a transaction with Luke. Peter then no longer owes John
>(or owes him less), and Mary no longer owes Peter, but John owes Luke.
then Mary may spend John's IOU in a transaction with Luke. John then
no longer owes Peter (or owes him less), and Peter no longer owes
Mary, but John owes Luke.
This software often gives me a headache ...
Martin Brock
>... whereas the cost of 3 may come back to haunt you 10 years from now. There are emotional benefits to knowing that once a deal is done, it is done.
Because the cost could come back to haunt you, you still care about
the contingency. You take steps to prevent it. When you pay in
advance, you're done. You no longer care. You take no steps to prevent
the contingency. Who cares if anyone defaults? You get paid anyway.
Default insurance seems fundamentally flawed for this reason. I oppose
the FDIC for this reason.
>It is true that the insurer/pool could default in 2, ...
So the cost isn't really known up front.
>... I would agree that CL is probably safer.
It seems safer to me because the parties still have an incentive to
avoid the contingency. A for profit insurance company might have the
same incentive, but it can still charge premiums that are too low. The
future just isn't predictable, and credit risk is especially
unpredictable.
>You are not proposing that contingent liability affect cases where X has changed in value since the transaction occurred.
owe the current value of what you promised to pay for in the past.
I don't account for time value at Favorati. I plan to account for it
at some point, but the accounting won't assume that people always
prefer goods delivered currently to goods delivered in the future,
i.e. negative interest rates will be possible in an intuitive way if
the market is willing. At this point, all debts at Favorati are
payable on demand.
>Please confirm: You are not proposing that CL have any effect if a debt denominated in dollars now buys only half as much gold as it used to.
and you only owe the dollars you promised to pay. You have a choice of
standards, and the U.S. dollar is one of the choices. Gold, silver and
common labor are also standards. A member may accept any standard, but
no one is obliged to accept your standard. I may add Bitcoins and some
others. Adding another standard is not difficult, but each member must
agree to accept a new standard, and each member must specify an
exchange rate. Ripple handles exchange rates differently.
>Indeed. Although there are also usability costs to users, which is why I can see why having CL be everywhere or nowhere may be better than having it be optional.
Favorati permits favors with non-standard terms, but these favors
don't participate in automatic settlement. It's tough to automate
anything when people trade on different terms. Handling multiple
standards and exchange rates was difficult enough.
Martin Brock
>Not sure if a Ripple-like (transitive) transaction can happen within
>Favorati now. But it seems that my question has been answered and it
>doesn't apply to cases where there's a "trade of currencies".
at Favorati, but other transitive transactions do occur. I'm still not
sure I understand the transaction you describe.
A owes B who owes C who owes D, all for D's good delivered to A.
Right?
Everyone except A has zero net debt after this transaction, and
everyone except D has default risk. Right?
If B defaults on C, C still owes D? C suffers the loss exclusively?
A seems to benefit from C's default risk without paying C a risk
premium? Is that true?
If I understand you, then B and C are financial intermediaries after
all; however, A must pay them to accept the default risk, or the
system can't work. These payments seem impractical, so I suspect that
I still don't understand you.
>So the answer to "Do currencies circulate within Ripple?" is both yes and no.
and it's closer to the scenario I described earlier. A owes B, so B
has A's IOU. B spends A's IOU in a transaction with C, and C then has
A's IOU. C spends A's IOU in a transaction with D, and D then has A's
IOU. A is like a bank issuing banknotes, but a bank pays you to accept
the risk of holding its notes, i.e. it pays you interest when you
deposit the notes.
Jorge Timón
> @Jorge
>
>>Not sure if a Ripple-like (transitive) transaction can happen within
>>Favorati now. But it seems that my question has been answered and it
>>doesn't apply to cases where there's a "trade of currencies".
>
> If I understand you, the transaction you describe above does not occur
> at Favorati, but other transitive transactions do occur. I'm still not
> sure I understand the transaction you describe.
>
> A owes B who owes C who owes D, all for D's good delivered to A.
> Right?
Yes.
> Everyone except A has zero net debt after this transaction, and
> everyone except D has default risk. Right?
Not sure what you mean by default risk here, but D has the risk that C
doesn't pay him.
> If B defaults on C, C still owes D? C suffers the loss exclusively?
Yes to both.
> A seems to benefit from C's default risk without paying C a risk
> premium? Is that true?
If B defaults to C or C to D, A still owes B.
Intermediaries can charge fees. For example, B could sell his 10 bUSD
for 10.1 aUSD.
> If I understand you, then B and C are financial intermediaries after
> all; however, A must pay them to accept the default risk, or the
> system can't work. These payments seem impractical, so I suspect that
> I still don't understand you.
The system works even if the default risk is not taken into account.
Note that the debts are always between trusted nodes. I consider the
risk of default from a friend to me close to zero.
Ripplepay and villages are already working without taking risks
directly into account.
>>So the answer to "Do currencies circulate within Ripple?" is both yes and
>> no.
>
> "Circulating note" has an established meaning in conventional finance,
> and it's closer to the scenario I described earlier. A owes B, so B
> has A's IOU. B spends A's IOU in a transaction with C, and C then has
> A's IOU. C spends A's IOU in a transaction with D, and D then has A's
> IOU. A is like a bank issuing banknotes, but a bank pays you to accept
> the risk of holding its notes, i.e. it pays you interest when you
> deposit the notes.
Imagine B owes 10 to A.
You can see it as "the account that B has with A has a negative
balance of -10" or "A has 10 B IOUs".
Now C also trusts B and A pays C. You can see it as
"the account that B has with A had a negative balance of -10 which is
reduced to zero
AND
the account that C has with B had zero balance and is increased to -10
"
or you can see it just as "A gives the 10 B IOUs she has to C".
In this later case, you may consider the IOUs circulating notes, I don't know.
Martin Brock
> Not sure what you mean by default risk here, but D has the risk that C
> doesn't pay him.
reason. It's like dyslexia. Only proofreading saves me.
Everyone except A has zero net debt after the transaction, and
Essentially, A pays B to hold A's note, and B pays C to hold B's note,
and C pays D to hold C's note, while D pays for the good.
D is the payer in this transaction initially. D pays for the good.
Everyone else only promises to pay.
If A defaults after the transaction, A gains at everyone else's
expense. D loses the value of the good unless C pays him, so if C is
trustworthy, C bears the risk of A's default. How is C compensated for
this risk?
A pays B to hold A's note. So far so good, but if A is a high default
risk, A pays B a high price. Does B then pay C a high price, or does B
pay only what C demands? Does C know A?
In general, people can't price risk effectively even if they try, so I
don't like the whole insurance premium approach to this problem;
however, the system you describe here apparently has a more
fundamental problem. B presumably pays what the market will bear. If
he does not pay C what A pays him, C is not compensated for A's
default risk, even if everyone prices risk correctly.
> The system works even if the default risk is not taken into account.
> Note that the debts are always between trusted nodes. I consider the
> risk of default from a friend to me close to zero.
so untrustworthy nodes seek you out. Every untrustworthy node wants to
be your friend, and some untrustworthy nodes gain your confidence
before ultimately defaulting on you. These nodes are confidence men,
also known as con men.
I'm not paranoid, but con men do exist. An effective system of credit
must account for them.
> Ripplepay and villages are already working without taking risks
> directly into account.
realize that the risk is not zero and that everyone's default risk is
not the same.
Our conversation has been very informative, Jorge. Thanks.
Martin Brock
>everyone except D has default risk.
and everyone except A has default risk.
How did I manage to do that again?
Jorge Timón
> @Jorge
>
>> Not sure what you mean by default risk here, but D has the risk that C
>> doesn't pay him.
>
> You're right. I reversed the roles again. I do that a lot for some
> reason. It's like dyslexia. Only proofreading saves me.
>
> Everyone except A has zero net debt after the transaction, and
> everyone except D has default risk.
>
>>Intermediaries can charge fees. For example, B could sell his 10 bUSD for
>> 10.1 aUSD.
>
> That's a risk premium.
>
> Essentially, A pays B to hold A's note, and B pays C to hold B's note,
> and C pays D to hold C's note, while D pays for the good.
>
> D is the payer in this transaction initially. D pays for the good.
> Everyone else only promises to pay.
>
> If A defaults after the transaction, A gains at everyone else's
> expense. D loses the value of the good unless C pays him, so if C is
> trustworthy, C bears the risk of A's default. How is C compensated for
> this risk?
A gains only at B's expense. B trust A directly and he will stop to do
it after his default.
C only has risk of B default. A default only affects B.
> A pays B to hold A's note. So far so good, but if A is a high default
> risk, A pays B a high price. Does B then pay C a high price, or does B
> pay only what C demands? Does C know A?
If D wants 10.1 cUSD for his 10 dUSD, C wants 10.2 bUSD for his 10.1 cUSD
B wants 10.3 aUSD for his 10.2 bUSD, A ends up paying 10.3.
Only B knows and trusts A in this example.
> In general, people can't price risk effectively even if they try, so I
> don't like the whole insurance premium approach to this problem;
> however, the system you describe here apparently has a more
> fundamental problem. B presumably pays what the market will bear. If
> he does not pay C what A pays him, C is not compensated for A's
> default risk, even if everyone prices risk correctly.
>
>> The system works even if the default risk is not taken into account.
>
> I must doubt this theory.
It's not a theory, it's a fact. Ripple is already working in ripplepay
and villages.
In villages there's no risk premium. Not sure if ripplepay allows flat
fees (I think that was only for distributed Ripple), but it allows
different interest rates for each connection.
>> Note that the debts are always between trusted nodes. I consider the
>> risk of default from a friend to me close to zero.
>
> An untrustworthy node knows that you consider the cost close to zero,
> so untrustworthy nodes seek you out. Every untrustworthy node wants to
> be your friend, and some untrustworthy nodes gain your confidence
> before ultimately defaulting on you. These nodes are confidence men,
> also known as con men.
>
> I'm not paranoid, but con men do exist. An effective system of credit
> must account for them.
The way Ripple takes care of this is with an imperative "Don't endorse
people you don't trust". "Don't set high limits". You're supposed to
only endorse your friends and trustworthy partners.
>> Ripplepay and villages are already working without taking risks
>> directly into account.
>
> The zero risk assumption works until it stops working, until people
> realize that the risk is not zero and that everyone's default risk is
> not the same.
When someone defaults the system doesn't collapse. If I trust someone
for 10 usd and he doesn't pay me I lose 10 usd and I disconnect from
him, that's all. I think Ripple is more resilient than the current
hierarchical financial system. Maybe this convinces you:
> Our conversation has been very informative, Jorge. Thanks.
You can find more useful information here:
Martin Brock
>it after his default.
>C only has risk of B default. A default only affects B.
owes C, and B bears A's default risk exclusively, while C bears B's
risk exclusively. That's the theory. I'll tell you why I don't believe
it.
C also owes D and receives nothing of value except B's promise to pay
C, so C is strictly a financial intermediary as you say.
In the best case scenario, where no one defaults, C benefits from this
transaction only by receiving from B more than he pays D. Right? Does
C participate in this transaction if B owes him less than he owes D
(including the premiums)?
Even if C expects no compensation for his role as an intermediary, I
don't expect C to participate unless B owes him more than he owes D,
because B's trustworthiness depends upon A's trustworthiness, i.e. B
is more likely to default on C if A defaults on B.
If default risks are not independent, an intermediary trusts the
entire chain of relationships, regardless of anyone's promise. A
default is a broken promise definitively. If A can default on B, then
B can default on C, and these events are not independent. The further
from A an intermediary is, the less he trusts the entire chain of
relationships, so the risk premiums in this sequence are strictly
increasing.
>It's not a theory, it's a fact. Ripple is already working in ripplepay and villages.
Ripple can scale if it offers more value to users then their
alternatives. Alternatives compensate intermediaries for risk, so a
system like Ripple must do the same at some scale.
>In villages there's no risk premium. Not sure if ripplepay allows flat
>fees (I think that was only for distributed Ripple), but it allows
>different interest rates for each connection.
intermediary's service. It can be both at the same time. I only expect
the risk premium to increase along the sequence; however, I also
expect intermediaries to charge for their service ultimately.
>The way Ripple takes care of this is with an imperative "Don't endorse people you don't trust". "Don't set high limits". You're supposed to only endorse your friends and trustworthy partners.
can default.
If members can't pass default risk to other members with even less
knowledge of the risk, then the problem is less severe, but the "don't
endorse" rule, without a risk premium, essentially allows only two
trust levels, 0% and 100%. Credit on a larger scale, and over a longer
term, requires more discretion.
Since Ripplepay accounts for a risk premium, it addresses the problem;
however, Favorati is less like Ripple than I first thought. At
Favorati, a member never owes a favor without receiving a favor first.
No one is in the game only to the play the role of financial
intermediary.
Martin Brock
>owes C, and B bears A's default risk exclusively, while C bears B's
>risk exclusively. That's the theory. I'll tell you why I don't believe it.
to reflect the dependence of B's default risk on A's default risk, so
I'm not sure the indirect relationship is more cost effective than A
owing D directly.
Martin Brock
The relationships in your sequence A->B->C->D are credit default
swaps. B accepts A's default risk for C who accepts B's default risk
for D.
Let's shorten the sequence to A, B and C.
A needs credit to buy C's good, so A borrows from C, but because C
doesn't know A, C also buys a credit default swap from B, i.e. C pays
B to guarantee A's debt to C. Right?
How does C know how much to pay B for the default swap?
Martin Brock
default swap.
A must pay the risk premium somehow. Presumably, A owes B more than
the price of the good that C advertises to A, and A owes B the
difference.
If B receives the difference only if A does not default, this model
doesn't work. A must pay this premium in advance of the contingency.
Right?
Jorge Timón
> If default risks are not independent, an intermediary trusts the
> entire chain of relationships, regardless of anyone's promise. A
> default is a broken promise definitively. If A can default on B, then
> B can default on C, and these events are not independent. The further
> from A an intermediary is, the less he trusts the entire chain of
> relationships, so the risk premiums in this sequence are strictly
> increasing.
Let's discuss the simple case first and then the "increasing risk with
longer transactions" later. The conclusions on the simple case should
make things simpler.
2012/2/8, Martin Brock <reston...@gmail.com>:
> Since Ripplepay accounts for a risk premium, it addresses the problem;
> however, Favorati is less like Ripple than I first thought. At
> Favorati, a member never owes a favor without receiving a favor first.
> No one is in the game only to the play the role of financial
> intermediary.
Participants are expected to clear their debts by selling products
rather than in cash (most times).
For example, say we have the payment: B -> C -> D
And then the payment A <- B <- C <- D <-E
After that, B owes 10 to A and E 10 to D. The rest of the debts have
been cleared.
Well, not the example I should have used. But you can see debts clearing.
Another try
A -> B -> C
and then
A <- B <- C <- D
After this, only D owes to C, A has cleared his debt by selling a
product to another participant.
Oh...B is still only an intermediary...
3)
A <- B <- C
and then B -> C -> D
After that, B owes A and C owes D. B and C are even and all of them
have participated directly as seller or buyer on a transaction.
2012/2/10, Martin Brock <reston...@gmail.com>:
Then it has to pay with another currency, because if he pays with
aUSD, the payment is not in advance. But I don't understand why that's
a requirement.
Can't B just bet on and profit from A's solvency?
Also can't an account of credit from A to B be compensated with a
similar account from B to A as their contract? Why are fees always
needed?
The bidirectional connection between two nodes benefits both parties.
If you don't trust anyone you don't have to fear any risk but you
cannot be paid with Ripple neither. I see fees (and interest) more
necessary for unidirectional trust relationships.
There's an spanish saying that would translate to "today for you,
tomorrow for me".
Martin Brock
theory) B recovers the loss from the premiums of other high risk
borrowers who do not default; however, I don't wish to implement this
model.
Ripple apparently works as follows. C offers a good for $10. When A
sees this offer, Ripple finds a path from C to A through B.
B charges A a 10% premium to guarantee C's loan to A, so A owes B $11.
C also charges B a premium, say 5%, so B owes C $10.50. Right?
I'm still not sure I believe that C can charge B a premium that is
independent of A's default risk, but that's the theory. Right?
Maybe A never sees the $10 price. Favorati works this way with
differing exchange rates. If your standard is dollars and mine is
silver, I price my offers in silver, but you see the price in dollars
using my silver/dollar exchange rate. Your silver/dollar exchange rate
can differ from mine. You price your offers in dollars but I see the
price in silver using your silver/dollar exchange rate.
Extending this logic to different risk premia (exchange rates for each
member) is not difficult, so I'm thinking in this direction now.
At Favorati, a member has a policy of trusting other members.
Currently, this policy is only a credit limit, but I'll add a degree
of trust between 0 and 100%, so if I trust you 100%, you see the same
price I assign to my offer, but if I trust you only 90%, you see a
higher price. 90% is my estimate of the probability that you'll return
the favor, so the price I expect is the product of this probability
and the price you promise to pay, i.e. your price is my price divided
by 0.9. The difference between your price and my price is the risk
premium.
In Favorati's model, a chain of financial intermediaries, as described
above, does not exist; however, a circulating note creates similar
relationships, and securing this chain of relationships is the goal of
contingent liability.
Everyone pays risk premia into a contingency fund. When I do A a favor
(accepting A's IOU), A owes my price to me and owes the risk premium
to the contingency fund.
If I pay B with A's IOU, I see a price reflecting B's policy of
trusting A, so after the transaction, I also owe a risk premium to the
fund. [At Favorati, I may pay B with a combination of A's IOU and C's
IOU. In this case, part of the price I see reflects B's policy of
trusting A and part of the price reflects B's policy of trusting C.]
If A defaults on the IOU while C holds it, the fund pays C, i.e. the
people who spent the IOU then owe C instead of owing the fund. The
fund refunds the balance of premiums paid on the IOU, other than the
premium that A paid.
If A does not default on the IOU, the fund returns all of the
premiums.
If the fund never defaults, this scheme is similar to contingent
liability, except that members pay in advance and receive a refund if
no default occurs. People paying in advance might still worry less
about default, but the prospect of a refund keeps them interested in
the default risk.
An insurance policy that returns premiums to policy holders this way
is called mutual insurance. A mutual insurance company is owned by its
policy holders and distributes dividends to them. Many mutual
insurance companies exist, so the theory is well established.
My problem with this approach is that the fund could default. People
may pay premia too small to cover their risks. In this scenario, the
person holding a note when it defaults is not fully compensated, and
the beneficiaries of this loss are the people who paid small premia.
The person holding the note last might have trusted A's IOU least and
paid the highest premium, but he still suffers the greatest loss. This
person could sue me personally as the proprietor of the site. The
terms of service exclude this possibility, but state courts can ignore
the terms of contract.
Martin Brock
>Participants are expected to clear their debts by selling products rather than in cash (most times).
which a participant does not return a favor.
>Can't B just bet on and profit from A's solvency?
accepts the entire loss and still owes C the entire debt that A fails
to pay, but I doubt this theory, especially if the lost profit is B's
only stake in the game, if B has never received anything but the
prospect of profit in the bargain.
When A defaults, I expect B's default risk to increase. C doesn't know
A's default risk, and C doesn't know that A owes B, so C can't really
know B's default risk either.
At Favorati, C never bears A's default risk unknowingly this way. C
bears A's default risk only by accepting A's IOU, knowing that it's
A's IOU. C can accept A's IOU from B, rather than accepting it from A
directly, but C trusts A directly in this scenario. C estimates A's
default risk directly. C does not trust B's opinion of A.
>Also can't an account of credit from A to B be compensated with a similar account from B to A as their contract? Why are fees always needed?
favor in return, and this obligation stays on the record until A
returns the favor. A must return the favor on demand, when asked, but
in some cases, A does not return the favor. The fees pay people that A
owes when A defaults this way.
>There's an spanish saying that would translate to "today for you, tomorrow for me".
me ... maybe.
Jorge Timón
>>A must pay this premium in advance of the contingency. Right?
>
> Correction: A need not pay the premium before defaulting, because (in
> theory) B recovers the loss from the premiums of other high risk
> borrowers who do not default; however, I don't wish to implement this
> model.
>
> Ripple apparently works as follows. C offers a good for $10. When A
> sees this offer, Ripple finds a path from C to A through B.
>
> B charges A a 10% premium to guarantee C's loan to A, so A owes B $11.
> C also charges B a premium, say 5%, so B owes C $10.50. Right?
Yes, that would be an example.
> I'm still not sure I believe that C can charge B a premium that is
> independent of A's default risk, but that's the theory. Right?
B takes care of A's default risk. The dependency of B over A is
included in B's default risk.
> Maybe A never sees the $10 price. Favorati works this way with
> differing exchange rates. If your standard is dollars and mine is
> silver, I price my offers in silver, but you see the price in dollars
> using my silver/dollar exchange rate. Your silver/dollar exchange rate
> can differ from mine. You price your offers in dollars but I see the
> price in silver using your silver/dollar exchange rate.
Ripple also implements the proportional fees through exchange rates.
10 dollars for 10 dollars plus 10% fee, exchange rate = 1 : 1.1
1 silver ounce for 30 dollars plus 10% fee, exchange rate 1 : 33
You don't really care much about denominations unless the exchange
rate are dependent on a given market or index.
> Extending this logic to different risk premia (exchange rates for each
> member) is not difficult, so I'm thinking in this direction now.
2012/2/11, Martin Brock <reston...@gmail.com>:
> @Jorge
>
>>Participants are expected to clear their debts by selling products rather
>> than in cash (most times).
>
> The same is true at Favorati, but the site must account for cases in
> which a participant does not return a favor.
Ripple does: the endorser takes the losses.
>>Can't B just bet on and profit from A's solvency?
>
> Yes. B can also lose the bet. In the losing scenario, B theoretically
> accepts the entire loss and still owes C the entire debt that A fails
> to pay, but I doubt this theory, especially if the lost profit is B's
> only stake in the game, if B has never received anything but the
> prospect of profit in the bargain.
That case looks bad for B and every participant should know the risks
of using the system.
But you're not accounting for the benefits of B to participate in
Ripple, he's able to trade without cash. As said, if he doesn't take
any risk, his connections within the Ripple network will be weak and
his liquidity and ability to accept payments low.
> When A defaults, I expect B's default risk to increase. C doesn't know
> A's default risk, and C doesn't know that A owes B, so C can't really
> know B's default risk either.
>
> At Favorati, C never bears A's default risk unknowingly this way. C
> bears A's default risk only by accepting A's IOU, knowing that it's
> A's IOU. C can accept A's IOU from B, rather than accepting it from A
> directly, but C trusts A directly in this scenario. C estimates A's
> default risk directly. C does not trust B's opinion of A.
Say you somehow perfectly account for the transitive risk within the
system in the following transaction: B-> C -> D
But B didn't wanted D's product for himself. Outside the system, he
gives it to A and A promises to pay B back. This would be an unknown
risk for D, right?
How would you prevent that from happening?
And more importantly, I still don't see why you should. Should D know
and account all the details of C's financial situation before lending
to him? Not only that, the finances of every party that owes to C?
D should be responsible for his decision of trusting C with the data C
gave him and independently of other contracts C makes.
>>Also can't an account of credit from A to B be compensated with a similar
>> account from B to A as their contract? Why are fees always needed?
>
> The fee is for the default risk. When B does A a favor, A owes B a
> favor in return, and this obligation stays on the record until A
> returns the favor. A must return the favor on demand, when asked, but
> in some cases, A does not return the favor. The fees pay people that A
> owes when A defaults this way.
I know, I'm just saying that many times people would just bargain a "I
won't charge you fees (or interest) if you don't charge me neither".
I guess that from your point of view is a miracle that LETS work even
in small towns without proportional fees nor interest at all.
Martin Brock
>> independent of A's default risk, but that's the theory. Right?
>B takes care of A's default risk. The dependency of B over A is
>included in B's default risk.
A defaults on B, then C's five percent premium could be too low. C's
premium cannot take A's default risk into account, because C doesn't
know A and doesn't know that A owes B what B owes C.
>That case looks bad for B and every participant should know the
>risks of using the system.
play the role of financial intermediary with nothing to gain but the
possibility of profit if their trustees do not default.
>But you're not accounting for the benefits of B to participate
>in Ripple, he's able to trade without cash.
benefits.
>As said, if he doesn't take any risk, his connections within
>the Ripple network will be weak and his liquidity and ability to
>accept payments low.
badly for him.
The question is: what does he do when he loses a bet or two or three?
When A breaks his word to B, does B keep his word to C anyway? Or is
default a systemic risk? Is default a contagious disease? This
question isn't merely theoretical. Recent experience with credit
default swaps suggests that contagion is likely.
At Favorati, returning a favor is voluntary. If A doesn't return your
favor, a jury of other members may award you "Favorati money", i.e. it
may declare that favors owed to A are owed to you instead, but you
have no other recourse. Other members also know that A defaulted and
may refuse him favors, but you may not sue him in a state court. You
agree not to sue outside of Favorati, though a state court could
refuse to enforce this agreement.
>Say you somehow perfectly account for the transitive risk within
>the system in the following transaction: B-> C -> D
see how D accounts for B's default risk when he doesn't know B and
doesn't know that B owes C.
>But B didn't wanted D's product for himself. Outside the system, he
>gives it to A and A promises to pay B back. This would be an unknown
>risk for D, right?
>How would you prevent that from happening?
accountable for things I can't possibly know, but I expect it to hold
me accountable for things I can know and do know but do not tell you,
even though you need to know these things in order to make an informed
decision, particularly if I profit by not telling you.
I am the site's proprietor. Favorati's database is my memory, and the
software's actions are my actions. If I have information that you need
to know to make an informed decision, I had better show it to you,
before you make the decision, certainly before I make the decision for
you. I had better not expose you to risks that you don't understand
because I never gave you the information. I worry about the legal
risks, and I don't want to do it morally, but more to the point, I
don't expect the business model to succeed if I do it
>And more importantly, I still don't see why you should.
Favorati members want.
>Should D know and account all the details of C's financial situation before lending to him?
risk perfectly, but the system can give him all the relevant
information it has. If B owes C only because A owes B, and if A is a
high default risk, this information seems very relevant to C's
assessment of B's risk of default. If Ripple informs C of this
information before C assess B's risk of default, I worry less about C
challenging the integrity of the system itself. If Ripple extends C's
credit to B automatically without informing C of what it knows, I
expect C to be upset when B refuses to pay C because A hasn't paid B
yet.
>Not only that, the finances of every party that owes to C?
obligors, because this information is relevant to B's trustworthiness.
I would want this information myself before extending credit to B, and
I don't want a machine extening my credit to B automatically without
accounting for this information. I doubt that fully informed people
would choose the model you describe over a different model.
>D should be responsible for his decision of trusting C with
>the data C gave him and independently of other contracts C makes.
want to know who owes B and how trustworthy they are before I extend
credit to B.
>I know, I'm just saying that many times people would just bargain a "I
>won't charge you fees (or interest) if you don't charge me neither".
among close friends.
>I guess that from your point of view is a miracle that LETS work
>even in small towns without proportional fees nor interest at all.
a small town using LETS for a few, small transactions. How many car
loans? How many home mortgages?
I don't expect a system of credit to scale without accounting for
risk.
Interest is more than a risk premium. It's also a fee for doing the
research necessary to estimate risk, assuming that you want to
outsource this work.
I do believe that outsourcing this work is less valuable in the
information age and that a system like Ripple can decentralize this
function, even all the way to the individual creditor, but
decentralizing the business of estimating risk does not make risk go
away.
Jorge Timón
>>> I'm still not sure I believe that C can charge B a premium that is
>>> independent of A's default risk, but that's the theory. Right?
>
>>B takes care of A's default risk. The dependency of B over A is
>>included in B's default risk.
>
> Again, that's only a theory. If B is more likely to default on C when
> A defaults on B, then C's five percent premium could be too low. C's
> premium cannot take A's default risk into account, because C doesn't
> know A and doesn't know that A owes B what B owes C.
Yes, 5% could be too low, maybe 6%? It's just an example.
C does know B well enough to extend credit to him without knowing
anything about A.
Imagine C extends credit to B and B to C. A week later, B extends
credit to A and A to B.
Does the relationship between B and C need to change?
>>That case looks bad for B and every participant should know the
>>risks of using the system.
>
> That case that looks bad for B seems to be norm. People frequently
> play the role of financial intermediary with nothing to gain but the
> possibility of profit if their trustees do not default.
I wouldn't say that defaults between friends are the norm. But if you
say people "frequently act as intermediaries with nothing to gain but
a profit if their debitors default" What's the problem?
>>But you're not accounting for the benefits of B to participate
>>in Ripple, he's able to trade without cash.
>
> I understand the benefits. B is happy to participate while he
> benefits.
He benefits overall. It doesn't mean that he needs to profit in every
transaction.
>>As said, if he doesn't take any risk, his connections within
>>the Ripple network will be weak and his liquidity and ability to
>>accept payments low.
>
> He takes risks, and he benefits, but eventually some of the risks go
> badly for him.
>
> The question is: what does he do when he loses a bet or two or three?
> When A breaks his word to B, does B keep his word to C anyway? Or is
> default a systemic risk? Is default a contagious disease? This
> question isn't merely theoretical. Recent experience with credit
> default swaps suggests that contagion is likely.
As said, I think the Ripple network is more resilient that the current
hierarchical financial system, but I can't prove it.
>>Say you somehow perfectly account for the transitive risk within
>>the system in the following transaction: B-> C -> D
>
> Accounting for the transitive risk is possible in theory, but I don't
> see how D accounts for B's default risk when he doesn't know B and
> doesn't know that B owes C.
In Ripple D only accounts for C's default risk. I meant "Imagine that
give have a ripple-like system that meets your requirements for
accounting transitive risk."
>>But B didn't wanted D's product for himself. Outside the system, he
>>gives it to A and A promises to pay B back. This would be an unknown
>>risk for D, right?
>
> Right. No system can account for obligations outside of the system.
>
>>How would you prevent that from happening?
>
> I wouldn't, because I can't. I don't expect a court to hold me
> accountable for things I can't possibly know, but I expect it to hold
> me accountable for things I can know and do know but do not tell you,
> even though you need to know these things in order to make an informed
> decision, particularly if I profit by not telling you.
My question is. Can't the system (the ripple-like system with
contingent liability) also "go terribly wrong" because it is not
isolated with the rest of the world?
Is perfect knowledge necessary for a healthy financial system? Because
we won't ever have that.
>>And more importantly, I still don't see why you should.
>
> It's not a matter of should or shouldn't. It's a matter of what
> Favorati members want.
I think they want to be able to accept payments from members they
don't trust. And I think people want payment systems that are final.
If I buy something but a month later I have to pay more because the
IOUs I used for paying have defaulted, I would exit from that system
fast.
People want some level of privacy.
>>Should D know and account all the details of C's financial situation before
>> lending to him?
>
> If he wants to estimate the risk well, he should. No one can estimate
> risk perfectly, but the system can give him all the relevant
> information it has. If B owes C only because A owes B, and if A is a
> high default risk, this information seems very relevant to C's
> assessment of B's risk of default. If Ripple informs C of this
> information before C assess B's risk of default, I worry less about C
> challenging the integrity of the system itself. If Ripple extends C's
> credit to B automatically without informing C of what it knows, I
> expect C to be upset when B refuses to pay C because A hasn't paid B
> yet.
Yes, the system could also give any user all the information it has
from all the intermediaries involved before committing any payment.
Their home addresses of every participant so that one can make his own
research and calculate the risk more accurate, but that is against
privacy.
>>Not only that, the finances of every party that owes to C?
>
> C at least needs a statistical summary of the trustworthiness of B's
> obligors, because this information is relevant to B's trustworthiness.
> I would want this information myself before extending credit to B, and
> I don't want a machine extening my credit to B automatically without
> accounting for this information. I doubt that fully informed people
> would choose the model you describe over a different model.
C is free to not extend the credit to B if B doesn't provide him
enough information.
>>D should be responsible for his decision of trusting C with
>>the data C gave him and independently of other contracts C makes.
>
> The data that C gives D (or B gives C) is the issue here. If I'm C, I
> want to know who owes B and how trustworthy they are before I extend
> credit to B.
If I'm A I don't want D to know all my balances in the system.
I don't think you mean this seriously. Every time you lend someone
money you ask him for all the people who owe him and all the people
who owe to those who owe him, and...
You don't lend money very often, do you?
--
Jorge Timón
Martin Brock
know is the issue here. I have this knowledge of B's relationship to
A, and C wants to know what I know. If B doesn't want C to know, C
wants to know that too.
>Imagine C extends credit to B and B to C. A week later, B extends credit to A and A to B.
>Does the relationship between B and C need to change?
assets that B could otherwise use to meet his obligation to C.
If I owe half a million dollars on a house worth $300,000, this
mortgage affects my application for another mortgage on another house.
If I have this mortgage and another homeowner extends me credit on
another house anyway, my first lender want to know about it, and he
will know about it.
>I wouldn't say that defaults between friends are the norm.
friends.
>But if you say people "frequently act as intermediaries with nothing to gain but a profit if their debitors default" What's the problem?
a lot of research. They don't lend to their friends without a care for
their friends' other financial relationships.
A few small favors among friends is one scale of economic
organization. If Ripple never exceeds this scale, your assumptions
could be valid, but even with all of their supposed professionalism
and statistical expertise and research, people created a financial
meltdown by trading credit default swaps very much as you describe
here.
>He benefits overall. It doesn't mean that he needs to profit in every transaction.
>Can't the system (the ripple-like system with contingent liability) also "go terribly wrong" because it is not isolated with the rest of the world?
that is relevant to their decisions. I know that A owes B, and I know
A's history at Favorati. I know, and C doesn't know. That's the point.
>Is perfect knowledge necessary for a healthy financial system? Because we won't ever have that.
database is the issue.
>I think they want to be able to accept payments from members they don't trust. And I think people want payment systems that are final.
will adopt a system like Ripple unless the system incorporates
something akin to deposit insurance.
>If I buy something but a month later I have to pay more because the
>IOUs I used for paying have defaulted, I would exit from that system
>fast.
paid you? You lose in either case, and you know the rules in either
case. You lose more in the latter case.
I think that's how people will see the credit relationships that
Ripple creates. They agree to pay bills for their friends, with their
friends' money. They don't agree to pay bills that their friends
cannot or will not pay.
>People want some level of privacy.
key to an effective system of credit. You can't expect people to
extend you credit without knowing your business.
>Their home addresses of every participant so that one can make his own
>research and calculate the risk more accurate, but that is against
>privacy.
any change of address, and they know my credit history and a credit
rating summarizing this history and other information and my
employment history and academic credentials and marriage history and
criminal history (if I had one) and all sorts of other things about
me. They know many things that my friends don't know, unless a friend
has done a background check on me. If I don't want a creditor to know
these things, I don't ask for the credit.
>C is free to not extend the credit to B if B doesn't provide him enough information.
point. A system like Ripple must provide this information if possible.
>If I'm A I don't want D to know all my balances in the system.
statistical summary of C's balances at least, because C owes him the
debt. If credit default is contagious across a sequence of swaps, he
also like a statistical summary of the entire sequence making
reasonable assumptions about the contagion (the dependence of B's
default on A's default). A system like Ripple can provide this
information, because it knows the entire sequence. Why not inform D?
Why not enable D to base his trust on this information?
>You don't lend money very often, do you?
all sorts of information on the people they credit and offer deposit
insurance and the like. The vast majority of people do so. That's my
point. These people won't use a system like Ripple without similar
assurances.