Governance and Risk Meeting: Ep. 43 (July 11 - 2019)¶
Video | Audio | Discussion
| References | Person | Text |
|---|---|---|
| # / 00:00:03 | Richard Brown | Hello everyone, welcome to the July 11th edition of the Scientific Governance and Risk meeting. My name is Richard Brown, I'm head of community developments. We're going to be talking about risk and governance in this call today. We have a special presentation from Cyrus, he's continuing his presentation of the general model for us. I want to... sorry, I shouldn't skip people. |
| # / 00:00:30 | Richard Brown | Vishesh will be presenting his Visheshalytics, when he tells us about the state of the peg. And then Matthew Rabinowitz will be giving us his weekly recap, sorry, or the weekly narrative towards the end of the hour possibly afterwards. I want to keep momentum up though, so let's scramble through the preamble here. |
| # / 00:00:52 | Richard Brown | If you have something to say, please say it. Don't be shy. If you have a microphone let's leverage that capacity. If you don't have a microphone please type your questions in the chat. And either David and I will be monitoring that and we'll ask the questions for you when we get a chance. Or you can just storm up in there and then after the hour we're going to hanging around for another 30 minutes to just a do a general Q&A. Everyone's more than welcome to hang around for that. |
| # / 00:01:20 | Richard Brown | I think that's going to be it for the preambles right now. I'm going to jump in to what I want to talk about, and hopefully not burn up too much time. We have a new forum and that forum was long overdue. There's all kinds of reasons why we waited a bit. I think that that worked out well. There was a critical mass of content that was ready to go, and so we have people jumping into the forum and I could not be happier with how well it has turned out. The content in there is amazing. The way that people were coming together and figuring out what needs to be done, and how we should be doing it is extremely impressive. |
| # / 00:01:59 | Richard Brown | If anybody in this call hasn't checked it out yet, I highly recommend that you do. I'm typing in the URL that's basically discoverable forum at MakerDAO. Have a look and contribute wherever you can. What I want to talk about today though is merchant processes and workflows and communities. Those are three different things, when you combine them they get even more complicated. |
| # / 00:02:31 | Richard Brown | One of the things that I want us as a group to begin to think about is, how do we surface valuable information, how do we collect that information, where does that information live, and how do we all work together? These are all super hard questions. To give you some context though we've been solving them slowly over time and introducing them one by one, and now it's time to start putting those pieces together. |
| # / 00:02:58 | Richard Brown | The pieces that we have are these calls. We have... yeah, let's break them down as we go. So we have these calls. These calls are very small time window, they're all held in English. They're in North American friendly time zones. I repeat these caveats every single time. That's why we don't make decisions in these calls. We don't set governance, this isn't where governance happens or governance decisions happen. This is where governance debates and education and community building and social consensus occurs. Where we're all going to figure out whether we're on the same page or not. |
| # / 00:03:31 | Richard Brown | That's great and it's required and it's super valuable. So valuable that in the near future, I hope, we're going to be expanding these calls to let's do one for the APAC region, let's have another one that's going weekly for the Latin American region. We need more than just one call and we need to expand these things because they're so valuable. But there's all kinds of downsides too so that there's no opportunity for deep debates or deep discussions. We're not going to start pulling apart algorithms in these calls. |
| # / 00:04:00 | Richard Brown | We can't give everyone an opportunity to have a back and forth because there's just not enough time. They're very temporal, so the last for an hour or an hour and a half. If you miss that you're out of luck, you have to watch YouTube. So, it has its pros and has its cons. The pros and cons for each one of these things I'm going to list are basically the same. There's the immediacy, there's sort of this dynamic interaction aspect. And those are always traded off their time to live basically. |
| # / 00:04:30 | Richard Brown | Calls last for an hour, they last for a week. Potentially, I guess, they could last forever if somebody wants to go back in to the archives and watch 49 straight hours of community calls. They're more than welcome to do it, but that's unlikely to happen. |
| # / 00:04:45 | Richard Brown | So, in order to allow us as a group to design workflows and have debates and figure out what our process is, we have these other forums. This other forums or venues are designed to allow for less mediacy, but more long tail utility and discoverability and audibility. Those other locations are the chats. The chats are all... actually, I think that we probably have the best chat in crypto, I'm gonna go on record of saying that. The chat is amazing, but it's also very temporal. |
| # / 00:05:20 | Richard Brown | We have these calls less an hour for every week. We have these really, really cool chats in Rocket that last 20 minutes, an hour at a time, and then they cycle out of visibility after 24 hours or so. It's good for rapid fire debates getting to the heart of an issue. And then coming up with an answer or something, and then having that answer slowly disappear through the history as you scroll up. |
| # / 00:05:48 | Richard Brown | To solve that problem we have a forum. And this forum, like I said earlier, is turning out swimmingly. In this forum we have deeper debates, this debates last... they could last forever if you're familiar with the way the debates happen in Ethresear.ch and some other venues that they could potentially happen for months and months at a time, and then go quite and then resurface. That's what we need because there's a lot of complicated things happening right now. |
| # / 00:06:16 | Richard Brown | What we're doing is actually serious business. There's lots of super smart people that need to be heard. And that needs to happen in a venue that there's no fear that chats or these calls or Reddit, which cycles out after 24 hours or seven days. We have this location where we can all get together, have our say, compose some well thought out arguments and then get feedback from everybody else in the community. |
| # / 00:06:45 | Richard Brown | It's amazing. But it's also temporal and it's also a bit messy. We need to be concerned about getting to the situation where we start devising canonical documents in the forum. We don't want to be in a position where six months down the line people are constantly referencing this epic thread. And that's where we expect the world to go to discover what it is that happened at MakerDAO at that point. |
| # / 00:07:09 | Richard Brown | The forum should be there to arrive at a decision, to formulate a document to discover a process. And once that happens that information needs to move out of the forum and it needs to go somewhere even more canonical than the forum. What we do and what we have been doing for the last year or so, MakerDAO community group, that's just been working out very well is we have a workflow. And that workflow is we iterate with community members on specific projects or ideas. |
| # / 00:07:41 | Richard Brown | We generally end up in HackMD where engaged stakeholders are working together to iterate on a doc. We discuss those changes in various venues like chat or the forum. And once a final draft has been achieved and this doc has utility to the wider ecosystem is probably going to last, what happens is that document gets taken out of HackMD. They were copied out of forum or out of a chat or out of a Reddit post and it goes into GitHub because that's the world that we live in. |
| # / 00:08:17 | Richard Brown | We might be writing documentation but it has all kinds of pros for us as a community. It's a flat file content database, it's easily distributed to multiple people or X number of people. It's versionable, it has public audit trail and there's lots and lots of reasons why GitHub is a great place to put a document once that document is finished. So, in the community group we put those documents into GitHub and then after that... another reason why GitHub is so great is that it is easy doing just content there to anything you could possibly imagine. |
| # / 00:08:55 | Richard Brown | So, once that content is in GitHub it could get pulled into the CDP portal, or the voting portal or it can get pulled into various others flat file marketing, our primary website or even to third party websites are welcome to start ingesting that content as they please. And that's a discussion that I want to start having with the community here, as well, as we start digging into governance and risk more deeply now that we have this forum. |
| # / 00:09:21 | Richard Brown | I'll be making a post about this, recapping some of these thoughts, and then putting out some suggestions about what these emergent workflows could look like. The discussion starts in Rocket. It'd be nice if we got to have it saying, “Okay, this is a great discussion. Can somebody just recap this for the forums? And then let's take it into the forum.” And if it turns into something perhaps we can say, “All right, let's throw this into HackMD and let's find people that want to own this process or this idea.” |
| # / 00:09:53 | Richard Brown | And then we can iterate on that and once it's done it goes into GitHub, and then it becomes canonical or it becomes “official-ish.” And then from there we can display it wherever we like. And if it comes time to review or revise we have a very clear path that's well understood. You submit an issue on GitHub, and then you submit a PR and then now we have a new doc. It's a technical business we're in, full of technologist. I don't think it's going to be too freaky to expect that kind of behavior, but I would like to explore it. |
| # / 00:10:28 | Richard Brown | And that is my speechifying for the week. There'll be a forum thread soon that encapsulates some of these ideas that's basically looking for feedback from the community about, “How are we going to capture this cool work that's happening? How do we formalize it a bit? And how do we make sure that it doesn't get lost?” |
| # / 00:10:47 | Richard Brown | Okay, I'm going to stop it there. |
| # / 00:10:48 | Speaker 2 | I have a quick question for you. So the information that you see coming out of the forum and to GitHub is stuff that is important to decision making from Maker holders. Am I understanding correctly that the types of docs would be like collateral onboarding applications, due diligence stuff? Things of that nature? Or are you also thinking about interesting threads, important conversations, important debates? |
| # / 00:11:19 | Richard Brown | Well, yeah, and this is why I'm going to make a forum thread so we can all hash this out together. My initial thoughts are, and I don't want to embarrass people and or pick favorites, but I'll do it anyway. Longforwisdom is doing a lot of really cool stuff in the forums. And one of the reasons why I wanted this forum to be out there is because of this continuing heartbreak that I've been experiencing with Reddit over the last year or two where somebody would sit down, put a great deal of time and effort and thought into a forum post, and it would be completely ignored. |
| # / 00:11:52 | Richard Brown | And the reason it gets completely ignored is because people go to Reddit to get themselves raged up and or to just figure out what's going on. And as soon as they see a long of text they don't even engage at all. There's not even enough will to complain about things. And I wanted to get around that problem. So in the forums, you've already seen deep engagement, but the second part of that heartbreak was having the stuff age out and not actually go anywhere. And I think that's making sure that it ends up in a place that's discoverable, centrally understood, like, “This is the repo where things go to and this is how we update those things,” goes a long way to making or converting some of these documents into canonical things. |
| # / 00:12:37 | Richard Brown | But there's also going to be a few different buckets like, “Is this going to be an idea for improvements in the system? Is this an idea for some public resource for how to engage with governance? Is this going to be tools and tips for collateral and best super sleuths,” or whatever they call them. So I don't know, we need to figure out some buckets and whether these things live, and whether they actually become official, we need to figure out. But I want to make sure that we start collecting this stuff and it doesn't get lost, because there's just much good work happening. |
| # / 00:13:14 | Richard Brown | Okay, I am going to stop talking now. Cyrus, I'm going to hand it off to you for risk. |
| # / 00:14:00 | Cyrus | Okay, cool. Are we good on screen? |
| # / 00:14:04 | Richard Brown | We good. |
| # / 00:14:08 | Cyrus | Okay. So, just a reminder, where we are on the process, the collateral onboarding guide did come out this week. But I don't want to break the flow of what we're talking about right now, so I'll probably come back to that in, I think, one or maybe two weeks. Today, we're going to mainly finish up and recap the stability fee model. And, yeah. |
| # / 00:14:36 | Cyrus | And then first, I want to talk a little bit about the discussion at the end of last week that Matthew brought up regarding the DSR and risk premium rate. I want to clarify a couple things and open up for discussion. Essentially, the way I'm thinking about it, and I think others are as well, is that we know that borrowers pay stability fee, which is comprised of a few different costs that are associated with the loan. |
| # / 00:15:14 | Cyrus | One of those is this fee associated with expanding the DAI supply. And is essentially a value transfer from borrowers to the counter parties who end up holding the DAI. Right? So let's say someone goes out and borrows 100 million DAI and then they want to just flood the market with it, there needs to be some sort of compensation to the counterparty to incentivize them to hold on to that DAI. |
| # / 00:15:52 | Cyrus | Now, as we all know and we've discussed before, single collateral DAI did not implement this optimally. The DSR isn't actually live, and so this value transfer doesn't work as we'd like it to. And this has led to, potentially, a higher than normal stability fee. Now, in addition to this fee associated with helping keep DAI stable, there's also this element of credit risk associated with the CDP. And this is a risk compensation, also paid by the borrower, but to the MKR token holders. And we call this the risk premium or the risk premium rate. |
| # / 00:16:44 | Cyrus | The collateral risk models which we've been discussing the last couple weeks, and the next couple weeks, are all designed to cover this latter risk premium, not so much the DSR. And so very simply, we can think of the total stability fee as being the sum of this monetary supply expansion and this credit risk. And then there are a few others as well, such as risk team compensation and some others, but that's not important for right now. |
| # / 00:17:18 | Cyrus | Right now the stability fee being 18.5%, we should be asking, "Is that due to the expansion of the DAI supply or because of the risk premium associated with Ether?" Well, I think it's pretty clear that it was actually due to the supplier expansion, right? As a community, we've never really even talked about the potential collateral risk of ETH DAI launched with 0.5% stability fee. We only started raising the stability fee as the DAI supply grew, and the DAI peg started to weaken. |
| # / 00:18:02 | Cyrus | Right. So, as we approach MCD launch, we got to start thinking about how that DSR portion of the stability fee will evolve once the DSR is actually functioning, and how we're going to handle the risk premium rate. This is what I imagine the setup to look like right now, where we have the bulk of the stability fee being paid for by the monetary expansion and pretty much nothing to the risk premium rate. |
| # / 00:18:46 | Cyrus | And after MCD launch, once the DSR is live, the supply rate will almost certainly drop considerably. But then we will have to add back what is the problem reflection of credit risk for the CDP due to the collateral backing it. It's also interesting to note that this right hand side was never really discussed by anyone over the months. Given that the media was always harping on the increasing rates, but technically speaking, there was also somewhat of an offset with the lack of any risk premium being assessed. |
| # / 00:19:34 | Cyrus | It also highlights that in the end borrowers, and really everybody, just cares about this total stability fee. And sometimes breaking it down into the individual components can maybe potentially complicate analysis a little bit. This is even more true when we start talking about how different collateral types impacting the DSR might negatively impact... Well, let me just get to the next slide. |
| # / 00:20:11 | Cyrus | But basically what Matthew was talking about last week as a result of the conversation is that, if the risk premiums are not properly set, then potentially certain collateral types impacting the DSR might unfairly impact other collaterals and in essence might end up pricing them up. So let's say the DSR needs to be raised and a price is a certain collateral asset out of the portfolio. Is this something that we care about? Is this something that we should actively stop from happening? My view is that this is not really a risk decision in that point, but more of a business decision by the MKR community on how they want to proceed. |
| # / 00:21:08 | Cyrus | So for example, you could raise the DSR but then adjust a certain collateral types risk premium back downwards to offset it and help keep it safe. Yeah, does anybody have any thoughts or questions on this? I know there was a tiny bit of confusion last week. |
| # / 00:21:36 | Alex Evans | Yeah, I do Cyrus. So when we think about the risk premium rate that's going to apply to each collateral type, correct? |
| # / 00:21:44 | Cyrus | Right. |
| # / 00:21:45 | Alex Evans | And then the DSR is obviously going to apply across all of those collateral types. So it's this one larger figure that's going to move to hopefully help us expand and contract supply? |
| # / 00:21:59 | Cyrus | Right. |
| # / 00:22:00 | Alex Evans | And so then every CDP holder will have a stability fee based on their collateral and that DSR percentage rate? |
| # / 00:22:11 | Cyrus | Right. |
| # / 00:22:12 | Alex Evans | Okay, that's it. |
| # / 00:22:14 | Cyrus | Right. So, yeah, hopefully that clarifies that. Matthew, do you have any comments on this? |
| # / 00:22:27 | Matthew Rabinowitz | No, it's really to just coming back, and Vishesh, I think you're on the line, but it's to compliment him on his comment probably two or three weeks ago, when he said, basically, if we just set the risk premium rates correctly, that really should adjust it. And so in my mind the risk premiums are just purely conditions based on a given collateral, whatever the risk associated with that collateral would be. And trying to segregate risk policy and monetary policy. |
| # / 00:22:57 | Cyrus | Exactly. Yeah. I think it's fairly intuitive to segregate the monetary policy and risk policy from a risk perspective. But then if the community wants to go back and manually override certain parameters for the benefit of the collateral type, then that's perfectly doable, possible. Maybe even recommended in some cases, but just want to highlight that that's- |
| # / 00:23:31 | Matthew Rabinowitz | We should try to avoid it because at the end of the day, that can also turn into an attack, which is Vishesh's point. What he mentioned was, if we just set them correctly in the first place, and that may mean that the DSR should be elevated, and some of the risk premiums actually should be significantly increased. Even if it's not politically appetizing if there's more associated risk, then that's the consequence of risk. It's a risk premium increase. |
| # / 00:24:01 | Cyrus | Right. |
| # / 00:24:02 | Vishesh Choudhry | Yeah, just this is, I think, the place. This subject area, the reason that I think you and Matthew are both getting involved in the same issue is because this is that intersection between the risk premiums, the risk modeling, and the monetary policy going forward and MCD because... Yeah, Matthew is right. I said that, in the event that we set the risk premiums properly, you wouldn't have this sort of discriminatory effect. But if they are improperly set, there needs to be a plan in place for how could you correct them. |
| # / 00:24:38 | Vishesh Choudhry | And I think the number one way that they would get improperly set other than just, people not doing their jobs correctly, and anybody doing math, etc, misses something that can happen. But also, a lot of that risk modeling is going to be based on certain behavioral patterns, historical data assumptions. If the system changes significantly, mid flight, you need to be able to have a lever to make adjustments. |
| # / 00:25:06 | Vishesh Choudhry | And I think that would be the case, if you model risk under certain assumptions and those assumptions change, you need a path to correcting that. The DSR won't be that path was my point, because the DSR is abroad tool. |
| # / 00:25:22 | Cyrus | Right. We're going to need to develop tools to evaluate or analyze when things are out of line. And for example, hopefully we can converge somewhat quickly on what the correct DSR level should be. So that if we see large deviations in the DSR, then we can try to figure out if something is wrong on the other side of the equation. |
| # / 00:25:51 | Vishesh Choudhry | Yeah, and I think as we go forward, it's going to be really easy to tell if something is wrong, I think it's going to become really hard to tell which thing that is wrong is causing the overall outcome, in terms of multiple different assets and attributing different behaviors to the CDPs from a given asset. |
| # / 00:26:09 | Vishesh Choudhry | I think that's going to become a really tricky problem. |
| # / 00:26:13 | Cyrus | Yeah. It's also going to be really difficult if a bunch of new collateral types start flooding in at once, right? If, say, every cycle, something like 10 collateral types are added and then the peg adjusts poorly, then it's going to be tough to figure out what was the cause of it. |
| # / 00:26:38 | Matthew Rabinowitz | I think some of that also will be just a progression where you start off with a given "collateral package" and you intentionally add a stability fee that is ridiculously conservatively high, with the full intent over the period of, say, six months to release an identical collateral package with a lower stability fee, and drop the debt ceiling on A in favor of B and slowly roll out... I guess what I'm saying is to slowly introduce a very conservative approach, which is very de-risked and naturally progressed downward. |
| # / 00:27:20 | Cyrus | Right. |
| # / 00:27:21 | Vishesh Choudhry | I want to echo that. I think design principles are going to become really important to hammer down here. We talk about scientific governance, well, a big part of scientific governance is experimentation and controlling for variables. And like we saw this with the stability fee, if you change too much too quickly you don't know what the effect was. |
| # / 00:27:42 | Vishesh Choudhry | So one is, I think you don't want to add 10 things at once, that scenario you described. I would recommend finding a way to phase things out as much as possible. I agree with Matthew as well, starting off safer and then easing the dials as you go is also going to help control for potential issues. |
| # / 00:28:03 | Cyrus | Yeah, I imagined this will be one of the first major things we would work on after MCD launch, at least as part of risk and research. |
| # / 00:28:14 | Unknown | I mean, realistically, you're not going to rip the band aid off until... like for big time new collateral coming in after maybe a couple [inaudible 00:28:23] on the term. But when security tokens when that happens, that's when we'll have a flood of collateral to come in when the SCC starts allowing that to come out. |
| # / 00:28:35 | Cyrus | Right. |
| # / 00:28:38 | Alex Evans | The danger with just letting it starting really high and using it over time, and sending that as a rule that you're going to do that is, you observe the collateral's behavior over a time interval that you don't have a choice over. |
| # / 00:28:53 | Alex Evans | In other words, it could be that you're observing it over a period that is under collateral by switching that is favorable to the asset. And you believe that there is less credit risk there, because you haven't seen a bad environment for it. And so, you leave it open for a year and it's a good year, and over that year you're defining that really the next year is where the problems begin. Which is why I think it also makes us think, a little bit more first principle with some of these and saying, "Well, how much do we want to ease the dial here." |
| # / 00:29:26 | Alex Evans | Because that's how you get situations where your own euphoria allows you to take more and more risk, when in fact, you should be doing the opposite. Just waiting things out over time doesn't always give you the right answer. |
| # / 00:29:45 | Vishesh Choudhry | And I'm not sure if I'm agreeing or disagreeing with you here, Alex, but a classic component of a lot of uncertainty mapping, like algorithms in A.I. Is you sort of up your confidence interval with more data that you've seen. And so initially, you should actually have a very low confidence, for most assets. |
| # / 00:30:03 | Vishesh Choudhry | ETH might be a slight exception, but for most assets, you should have a very low confidence gain dial. And then increase that, I think, as you start to see more behavior, more data that you can actually draw inferences from. |
| # / 00:30:22 | Alex Evans | Yeah, I don't know if you're agreeing or disagreeing with me either, but that sounds correct to me. |
| # / 00:30:32 | Cyrus | Okay, cool. Let's move on. So, back to the model, which is designed to find out what that correct risk premium rate should be. Okay, so last week we went through a very simple model on how to calculate loss for CDP. I made a few assumptions introduced in options analogy towards the end. I think a lot of it was a bit confusing, so I'm going to try to give it another shot and do a review of it and hopefully, be more clear today. |
| # / 00:31:14 | Cyrus | In general, it's not particularly easy covering the full scope of these models in 30 minute increments. So please get involved on the Rocket chat and on the forum if you want to dive into these a little bit deeper. |
| # / 00:31:37 | Cyrus | Okay, so we know that premium rate has something to do with the credit risk of the CDP. By credit risk I mean the potential for losses and uncertainties for MKR holders. And so we're uniquely concerned with what these losses look like, how they occur, why they occur, how badly they are. |
| # / 00:32:03 | Cyrus | Simple way to think about losses for CDPs is that they occur when collateral prices gap down suddenly, or when the options don't clear due to liquidity and slippage concerns. So if you think about it, if you were to completely remove black swans and huge gaps down and be able to liquidate infinite collateral at spot price you'd never have any loss in a CDP no matter what. |
| # / 00:32:46 | Cyrus | Our goal in trying to describe these losses is that we want to convert this notion of loss into the stability fee or the risk premium rate. There's a very general model of how to do this, which involves looking at the likelihood of default, as well as the severity of the losses or the loss given default. But these general models are sometimes just intuition driven. |
| # / 00:33:24 | Cyrus | When you actually dive deep to do the actual analysis you'll see that it's highly dependent on the type of loan it is. So for example, calculating loss metrics for an ETH backed CDP is much different than if you were trying to figure out your risk on a credit card loan. And the reason being should also be somewhat intuitive in the sense that for credit cards you could probably look back at 100 years or whatever of credit card data and estimate what these frequencies should be, because they probably don't change dramatically over time. But with Ether we don't have a lot of historical data, we don't have great intuition and how it evolves. |
| # / 00:34:22 | Cyrus | We can identify a handful of latent risk factors that could cause some pretty catastrophic price movements, things that haven't actually ever occurred, but we can imagine they occur. And so as we go from this general risk model, we have to apply our own understanding of Ether to make the model be realistic. |
| # / 00:34:55 | Cyrus | So this is how we've, more or less, arrived at how to do the Ether backed credit risk model. So first step is you collect your inputs, which is your due diligence report, your risk rating, the onboarding application. You have your trading history, you have historical CDP stats, and you have the current CDP distribution. We define our loss event as when the collateralization ratio reaches liquidation ratio. |
| # / 00:35:30 | Cyrus | Our next step is to identify what can cause this collateralization ratio to go up or down. And collateralization ratios essentially change based on the underlying asset price, as well as user behavior. And so, when we dive into these risk factors we want to know, "What does ETH do? How does it move?" And for behavior, "What do users do in response to... for one, ETH price drops?" So when it rallies or falls and locking and unlocking collateral, but potentially user behavior can be impacted by other factors as well. |
| # / 00:36:22 | Cyrus | And our risk rating, which is one of [inaudible 00:36:26]. So we're looking for fundamental information that can impact asset prices. I mean, typically, things that aren't immediately clear just by looking at the price chart. And that can be just... A simple classification is, you can have fundamental risks, protocol risks, exchange risks and regulatory. And obviously, these are pretty important to understand as we create CDPs. |
| # / 00:37:11 | Cyrus | We want to be able to convert these risks into a risk rating, which will then somehow make its way into the actual calculations. And so once you've defined and analyzed your risk factors, you essentially want to simulate them to see how they intermixed with each other and how they evolve over time. And so what we're doing is, we start with the current CDP distribution, and at every time increment forward we apply a function that is a change in the asset price. And we model this using a stochastic asset pricing model, which we discussed a little bit last week. |
| # / 00:37:59 | Cyrus | In here we include factors such as the volatility, and we're tracking the risk rating as well and how it impacts any sorts of jump risks. Obviously, we know that liquidations can occur slowly or can get downwards. We also apply behavior function that increases or decreases the amount of DAI drawn, or the amount of ETH locked up. This is a whole aspect of research in itself. And then we also apply a slippage function to deal with liquidation. |
| # / 00:38:47 | Cyrus | Essentially, you run the simulation a whole bunch of times, and for each one of those paths you essentially track how much collateral is liquidated and you record how much loss was incurred. Right. So let's say you do this thousands of times and you end up with a distribution of losses. And some of these paths nothing happens, right? ETH price just goes up, no liquidations fine. Some of these paths you'll have liquidations but... Some you'll have liquidations but the collateral is fine. Some will have liquidations, but the collateral is not fine because of liquidity. |
| # / 00:39:35 | Cyrus | And then sometimes you'll just see the asset just completely collapse downwards. Right? And so with this simulation, we've more or less achieved what we were trying to get at which is a pseudo potential distribution of losses. And we can... I mean, a simple characterization of this distribution is the expected loss or the average of those losses. |
| # / 00:40:08 | Cyrus | This is a visual representation of that simulation, right? So you have some paths that are fine, and plenty that reach the collateralization ratio threshold. And for those you want to apply a slippage analysis. Now we can have a handle on for a particular CDP or group of CDPs, what can we expect to lose for MKR holders? How much bad debt is expected to be triggered? |
| # / 00:40:54 | Cyrus | And of course, MKR holders are not bearing this risk for nothing. So there needs to be some sort of risk compensation for this. And of course, that is what the risk premium rate is in the first place. And so you convert that expected loss into this credit spread, which is actually quite simple calculation at this point, once you've done the simulations. We talked a little bit about this last week as well. |
| # / 00:41:30 | Cyrus | Right. And then, so back to what I was talking about last week with the whole options analogy. And essentially, if you were to simplify that simulation, with a bunch of assumptions, such as no behavioral changes, no jump risk, and a few others, your loss distribution is equivalent to that of a put option on the underlying collateral Ether. And intuitively, this should make sense that MKR holders, they are behaviorally short a put, right? When the Ether price collapses their losses increase the further the price goes down, but they receive stability fees as compensation for that. |
| # / 00:42:26 | Cyrus | And so, in this simplified example you could actually just apply the Black-Scholes formula. But, of course, we can't exactly do that analytically, because we do have to factor in behavior and jump risk and a whole bunch of other stuff. Another way to describe this example is, if you had a CDP with 100 DAI debt and a $200 liquidation price. And then you constructed a put option with the same parameters, then the CDP would be completely riskless. There'd be no way for the CDP to lose money, because as Ether price dips below 200 the put option will perfectly offset any losses. |
| # / 00:43:27 | Cyrus | You can glean that a CDP that, hypothetically, has this put option embedded into it, which of course there isn't, but if it was, then the CDP shouldn't actually command any sort of risk premium, because there is no risk to it. And so, now that we've removed this hypothetical put, then you can reason about what that risk premium should be. Essentially it's dependent on what the price of this put is. And the- |
| # / 00:44:09 | Alex Evans | [inaudible 00:44:09]. |
| # / 00:44:09 | Cyrus | Yeah. |
| # / 00:44:09 | Alex Evans | Sorry to interrupt. Just so I understand something, the put doesn't purely and completely offset one to one with this fee unless there's an additional assumption about how people are managing their debt- |
| # / 00:44:23 | Cyrus | Right. |
| # / 00:44:23 | Alex Evans | ... correctly, right? |
| # / 00:44:24 | Cyrus | Right. |
| # / 00:44:24 | Alex Evans | So, in other words, put will always be so that the value on the other side for the seller is always... Sorry, of the buyer is always greater than or equal to the value of the put. Right? Because in... sorry, to the seller. Third correction. The seller can get back at least that value or they can get more if they can exercise and keep the difference. Right? Insofar as you believe that a CDP owner has properly rebalanced. So, not only does the default not happen, i.e loss didn't materialize, you also now have a put option. |
| # / 00:45:06 | Alex Evans | So it's not a pure equality it's a greater than or equal to relationship. That's assuming the MKR... So implicitly, if the MKR holder is selling a put, implicitly you're saying the CDP owner is buying that put, right? Somebody has to buy it if you're selling. |
| # / 00:45:28 | Cyrus | Right. |
| # / 00:45:28 | Alex Evans | And so- |
| # / 00:45:28 | Cyrus | I'm not sure I understood your question, though. |
| # / 00:45:30 | Alex Evans | You're assuming here that there's no rebalancing happening, right? |
| # / 00:45:36 | Cyrus | Right. |
| # / 00:45:37 | Alex Evans | I don't [inaudible 00:45:37] CDP and so equality hold strictly. But in reality, people do manage their debt sometimes well or sometimes poorly- |
| # / 00:45:46 | Cyrus | Right, because- |
| # / 00:45:46 | Alex Evans | This option is just basically the worst case scenario if they manage it quite poorly. It perfectly offsets. |
| # / 00:45:57 | Cyrus | Right. So the behavior can add to the risk as well. Right? So I think- |
| # / 00:46:02 | Alex Evans | Oh, yeah, that's right. That's right. |
| # / 00:46:06 | Cyrus | Vishesh was... if I'm understanding your point correctly, I think me and Vishesh spoke about this recently where if the Ether price were to rally a bunch, and then a user would withdraw a whole bunch of collateral, then they have added risk to their CDP. Right? |
| # / 00:46:25 | Alex Evans | Right, right. You're right. And so- |
| # / 00:46:28 | Vishesh Choudhry | [inaudible 00:46:28]. |
| # / 00:46:28 | Alex Evans | Yeah, so the assumption here is a little bit weaker than it's like it's not just that there's no recapitalization happening at all is that the recapitalization if it's positive then it's negative to the system, like offset. |
| # / 00:46:43 | Cyrus | Yeah. You'd have to dynamically manage your put options, right? You'd always have to be going in and out of different strikes, and different notional amounts. |
| # / 00:46:54 | Alex Evans | Yeah. So you're dynamically replicating, I suppose. |
| # / 00:46:56 | Cyrus | Right. |
| # / 00:46:56 | Alex Evans | That makes sense. Thank you. Vishesh were you about to say something? Sorry. |
| # / 00:47:02 | Vishesh Choudhry | No, I mean, I was roughly going to say the same thing in similar to if you have many different... because it's not just one CDP, right? It's really a pool of CDPs that are entering and exiting all these different states. And so, basically, the idea is net equilibrium. If ETH price goes up in the short term collateralization does tend to increase, but then it does balance out a bit. |
| # / 00:47:24 | Vishesh Choudhry | And then similarly, when ETH price depreciates, people do re-collateralize. And so these things like... the hypothesis or the theories that they roughly balance out to some sort of steady state. |
| # / 00:47:36 | Alex Evans | Got it. Got it. Okay, so that makes perfect sense. Thank you. |
| # / 00:47:42 | Cyrus | Okay, cool. One of the interesting insights from looking at this from an options perspective is that aspects that are important to options pricing are also relevant to us, and in particular the volatility. When we talk about the volatility of the CDP, and in particular, the default risk and all that we're pretty interested in the tails, right? And tail risk is, obviously, one of the most important aspects of risk management for Maker. And while there's no simple way to read in about the tails for a bunch of reasons that I've mentioned before, we are looking at three components of the volatility here. |
| # / 00:48:47 | Cyrus | One being standard market volatility, standard deviation of returns, another being the liquidity analysis for auctions. And then, finally, probably the most important one being these fundamental risk factors that caused the sudden gap, sudden gaps downward. And so some of the research we've been doing here is, we've been going back and collecting data on all of the major losses and draw downs of the Ether price for the past couple years. And classifying them by what caused them and a whole bunch of different metrics. |
| # / 00:49:34 | Cyrus | This is where we glean that the main source of draw downs are typically protocol related, exchange related, and regulatory related. And using these metrics you can have an approximation of, one, how frequent these big draw downs are, and two, the severity of these losses. And that way we can input these these metrics into our simulation to help create more realistic tails for the losses. |
| # / 00:50:26 | Cyrus | I would say that this is just the beginning of this research, there's a lot of different ways to go from here in terms of collecting information. So for example, redoing this analysis, but for Bitcoin, and then making the assumption that the frequency and severity of ETH draw down is somewhat correlated to Bitcoins as well. And by doing so you can maybe get more data into your model. But essentially, the goal here is, essentially... I mean, the ultimate goal is we want to know how bad these losses can get, right? |
| # / 00:51:10 | Cyrus | Our estimation for losses from our simulation has to factor in somehow that these events do occur. Not frequently, but they do occur and they can be damaging. This may seem like a fairly imprecise manner of doing so, but, again, this is ongoing research and I think we can all... I mean, it's just an approximation that will have to suffice for now. I mean, going into some questions of complexity and for the research about this entire model... |
| # / 00:52:00 | Cyrus | Yeah, I mean, as it's listed on the slide, this is work that will be iterated on, hopefully, for potentially years and years to come, right? As we sharpen up our data and our models. For now, though, as we get near to MCD launch, just having a rough awareness that these types of big draw downs occur, particular so that we don't underprice our tails, which is a particular concern with standard asset pricing models. |
| # / 00:52:44 | Cyrus | But I would say that this is the fun part of risk work. It's the interesting part of risk work for sure. So yeah, again, pop into the forums if you want to discuss more about this stuff. And there's actually a lot of other aspects of the model as well. So far, we've just hit on the high points, but we can do more analysis into... That's a typo, that should say "liquidation penalties" not "ratios". Oh, no, sorry. That's at the bottom. |
| # / 00:53:25 | Cyrus | Yeah, higher liquidation ratios do result in different amounts of expected losses. So for example, if you have your liquidation ratio at 200%, then losses would be significantly lower and stability fees would also be lower. We'll talk about this next time with liquidity analysis. |
| # / 00:53:54 | Matthew Rabinowitz | Quick question. This is for when we're basically creating and needing to determine a risk premium completely vanilla when we don't have any background behind it. How would we compare... I think we've mentioned this one. So it's like a credit default swap or a bond for a given company, doesn't matter, that's got a credit rating and a CDS that's a known CDS. How much of our risk work would we... I wouldn't use the word outsource but, get guidance from, inspiration from to get some of those metrics. |
| # / 00:54:33 | Cyrus | A lot. As much as possible in terms of guidance and outsources, of course. I mean, I don't see why we wouldn't borrow as much efficient public market information as possible. I mean, in fact, the only reason we couldn't do that right now is because I don't trust the Deribit ETH options market for anything, and certainly no CDS market on CDPs. |
| # / 00:55:09 | Cyrus | So that public market information just isn't available for our particular choice of collateral backing right now. |
| # / 00:55:18 | Matthew Rabinowitz | No, correct. More commenting like just if we took JP Morgan debt, and somehow whatever legal construct would allow us to use it as collateral. And then we were to decide what should be that stability fee for that, where it's an investment grade cash flowing asset and there's a known CDS for that security. Right? |
| # / 00:55:41 | Matthew Rabinowitz | How much of that would that... I wouldn't say directly, we wouldn't want to just cut and paste it, but we should be able to take a good 95% of the work has already been done on a debt security. |
| # / 00:55:56 | Cyrus | Sure, yeah. I don't know. I guess we'll see when we get there. I definitely agree with the sentiment. Right. As Rich mentioned earlier, the user by the name of Longforwisdom has started a community due diligence initiative right now going through the collateral onboarding guide and collecting information. Over the next few weeks we will start releasing further frameworks for qualitative due diligence and basically, getting ready for collateral evaluations. Should be interesting stuff. And that, is it. |
| # / 00:56:53 | Cyrus | Any questions about anything? Hopefully, that made things a little bit clear from last week. All right, cool. There are no questions then Vishesh, I guess, if you want to jump into the monetary policy stuff. |
| # / 00:57:26 | Vishesh Choudhry | Sure. So a lot to cover, but I'll still try to do it pretty quickly. Okay, so hopefully everyone can see my screen all right. Fun fact, we've crossed 360 million loans originated from Maker, so yay. |
| # / 00:57:52 | Vishesh Choudhry | All right, so the ETH price, I think we noticed that for about a week there, was going down a bit. That was following a tail of a huge volume of ETH for DAI traded. And then successively a fair amount of leveraging behavior with ETH price coming down. And I don't think there was a lot of contention around that explanation, it seemed to track with a lot of what we'd seen before. |
| # / 00:58:22 | Vishesh Choudhry | Obviously, things had been pretty stable around the 16.5% level, because there had not been as much leveraging all at once. So when the stability is a little bit lower, and you have a lot of leveraging all at once, it is pretty reasonable to expect that the DAI price would take a little bit of a hit. |
| # / 00:58:40 | Vishesh Choudhry | There was a fair amount of discussion around whether this should be resolved on its own or whether a stability fee increase was necessary. I was ambivalent on that, it could have recovered on its own, I think, but the community I think chose to take a more cautious approach, I guess, and increased the stability fee to try and right the shift relatively manually. |
| # / 00:59:05 | Vishesh Choudhry | So that stability fee was increasing 17.5%, things were pretty stable, it was increased to 18.5% again. So it'll be interesting to see if... sometimes the stability fee increases are timed hilariously poorly with ETH price movements. So it's hard to tell, which had an effect, but it will be interesting to get more data points to see if this actually pushes up the DAI price at all. |
| # / 00:59:34 | Vishesh Choudhry | So, just to touch on what's been going on the last 24 hours, pretty much same state as what had been shown in that graph, roughly hovering around the 0.98 level. Pretty solid spreads that we've seen recently, and I think pretty solid volume. So when you have that increased amount of volume, you do tend to see a more smooth or less discreet spread continuous [inaudible 01:00:04]. |
| # / 01:00:05 | Vishesh Choudhry | So yeah, as I mentioned, the ETH price for the majority of June has been running up a bit, and then just around the tail end of June started to take a hit. Similarly, the DAI price did a similar thing where it was holding pretty steady and then took a hit. This is primarily leverage in this area. And then as ETH price started to recover a little bit, DAI price starting to recover a little bit and then ETH took another nosedive. And even the most recent nosedive is cut- |
| # / 01:00:59 | Richard Brown | Yeah, let's give him a second to return. I'll use the dead air while we're waiting. Matthew happily reminded us that there's been a discussion in the forums recently about the debt ceiling. We should talk about it today in this call because it's an interesting... probably the first example of a signal being raised in the forum that could potentially precede a pole in the governance system. And we're still trying to figure out what that mechanism looks like. |
| # / 01:01:37 | Richard Brown | Please visit the link that I posted in the sidebar, or in the chat. Have a look at what's going on in there and let's have a discussion about this. I'm assuming Vishesh will- |
| # / 01:01:46 | David Utrobin | The one thing that's missing in that thread Rich, is the option for no change. Just as a general, I think most people are in agreement that that ceiling should be raised. But I know somebody had brought it up there. |
| # / 01:02:01 | Richard Brown | Yeah, and this ties back to my emerging process, emergent behavior meme that I've been pushing is that we need to figure out what that actually looks like. So this is our first take at the can, but I would like to have some kind of a standard, so here's... actually, I think that Longforwisdom is already going down that road, that there needs to be some kind of a standard. |
| # / 01:02:21 | Richard Brown | So here's an official title, here's an official set of options. Here's how we do public or non public. Here's how we make sure that there is no change or a lack of confidence in the pole. There's things that we need to understand. And that's probably a good example of one of those sort of theoretical GitHub documents that I was talking about, because that would be a good example. An official doc about what does signal generation look like in the forums. |
| # / 01:02:47 | Richard Brown | And then people can just reference this GitHub page that says, "Make sure all the options are there, make sure that it's public or private, or whatever the community decides." And then we'll know what to refer people to the next time they come up with a signal request. |
| # / 01:03:12 | Matthew Rabinowitz | What is your next step with that? We know we're just going to debate it on the forum and then we... That the race- |
| # / 01:03:19 | Richard Brown | Where the debt ceiling was specifically? |
| # / 01:03:20 | Matthew Rabinowitz | Yeah. |
| # / 01:03:22 | Richard Brown | I don't know, this is where I wish I had more time in the day so I could have come back with a great plan for you guys, but I don't have one. So we need to figure this out. The forum is there in order to collect signals and to have great debates and for us to figure out as a group what we should be doing. And it happened sooner than I expected. And so the first example of that is the debt ceiling thing. But how we move ahead with this thing, I'm still in the dark about. |
| # / 01:03:50 | Richard Brown | We're hobbled right now with... So yeah, let me set some context here. And this should probably all be captured in the forums, so I'll promise to put these thoughts in there. But we're hobbled right now in that we are... The governance is completely ruled by stability fees, because we have severe limitations in the capacity of our voting mechanism. |
| # / 01:04:15 | Richard Brown | Those limitations I have been promised... well, I haven't been promised, but I'm assuming it's a promise. That's what I heard. That there's going to be a new portal coming out very, very soon. Potentially, this month or early the first week of next. Which will allow us to do a significant number of arbitrary polls. |
| # / 01:04:31 | Richard Brown | So the question in my mind is, do we put our ability to alter the stability on hold while we think about debt ceilings? So we need to figure out, I think, the severity of the requirements for this change. Do we need to fix the debt ceiling in the next three weeks? Or is something bad going to happen if we don't? Then yeah, we can say, "Okay, we're going to put the stability fee on hold next week, or the week after." |
| # / 01:04:58 | Richard Brown | But as I'm not shy to point out, I'm not smart enough to know. So I require the analists and the monetary policy experts and the risk teams to sort of lead that discussion, I guess, or at least frame that discussion. And interpret the value of the arguments that showed up in that forum thread, I think. If it turns out that it is critical, then yeah, we can say, "All right, as a group, everybody needs to go to this thread. They need to pull a vote in that poll." And depending on the results, next week or week afterwards we're going to replace the stability fee with the debt ceiling raise. |
| # / 01:05:38 | Richard Brown | I think that was an unsatisfying answer, but did that cover what you're asking Matthew? |
| # / 01:05:42 | Matthew Rabinowitz | Yes and no. I mean, I guess the part I'm trying to bring out is that we have so much to try and gain rough consensus around on... I guess I'll be presenting some of it here in the next three minutes or five minutes. If Vishesh doesn't come back. How we're going to go from A to B, and where we're going to be in, say, six 8, 10 months doesn't matter the number. |
| # / 01:06:02 | Matthew Rabinowitz | We have so much consensus to get through between now and then. Right now we're seeing the price degrade a little bit and we've just raised the stability fee. So maybe there's a wisdom in saying we need to have another week between the two. So in my mind, raising the debt ceiling doesn't really present a material risk unless we change the debt ceiling to some huge number, which we've discussed on the forum. |
| # / 01:06:27 | Matthew Rabinowitz | But for me, personally, I'd love to see at least the poll this week, even if it's not an executive vote, but a poll to try to increase that number. Because we've got so many items that we have to get rough consensus on related to MCD, the debt ceiling is just some brain damage we have in the middle. It's a distraction that we need to put on the side and put it to bed. |
| # / 01:06:53 | Richard Brown | It's technical debt. And so technical debt, we could talk for hours about the philosophies around technical debt, but it needs to be addressed. The question that I'm posing though is, doesn't need to be dressed in the next three weeks? Because sure, cleaning up the number of decisions that we need to make, it would be useful. But is that utility more than the expense of losing flexibility on the stability features? |
| # / 01:07:22 | Richard Brown | There's obviously interest in what's happening with stability right now. We just moved to 18 and I just took the opportunity because I forgot to look at the pole before we started this call, and the community signaling for 20.5 now. So, obviously, there's interest in the stability fee. It is hard to predict what the future is going to hold next week and the week after, right? |
| # / 01:07:42 | Richard Brown | So, I guess this is my question. No, I have two questions, because I don't know the answer, so they're not rhetorical. Do we need to do this now is the question and here's from my neophyte financial analysis. I thought that we have a debt ceiling because of liquidity risk, right? Cyrus correct me if I'm wrong. Isn't the idea here that we have 100 million because somebody did math somewhere that said that over a 24 hour period in this market we can comfortably liquidate X amount through the mechanisms, right? |
| # / 01:08:15 | Cyrus | I mean, that's the way you're supposed to do it. I think it's part of the alpha of SCD. They were just somewhat arbitrarily chosen. So it's not like moving upwards would pose a significant liquidity risk beyond what risk we're already taking right now. |
| # / 01:08:36 | Richard Brown | Yeah, I guess, and that's the question in my mind is that... you're right, it is arbitrary and that was kind of a leading question, because I knew it was arbitrary. But how much more arbitrary? How do you measure arbitrary at this point? So we don't know for sure whether 100 million was good or not, so do we know for sure whether 150 is going to be as good? Or is that the horrible? |
| # / 01:08:58 | Cyrus | I mean, we will know as soon as the models are fully finished and we can run them. But in any case, I don't think... I mean, just from a pragmatic standpoint, I don't think they should be raised 50 million at a time in this environment on the precipice of- |
| # / 01:09:19 | Richard Brown | 120 is [inaudible 01:09:20]. |
| # / 01:09:21 | Cyrus | Huh? |
| # / 01:09:23 | Richard Brown | Well, looking at- |
| # / 01:09:23 | Cyrus | Just that 10 or 20 million I think probably makes more sense to me. |
| # / 01:09:31 | Matthew Rabinowitz | [inaudible 01:09:31] a negative tone on the criticism, but without knowing when MCD is actually going to launch "soon." Right? Each time we go through a poll and then an executive votes, there's a two week cycle between those two points. And knowing the quantity of things we have to vote on and the community not knowing when soon is. I'm trying to take the more conservative approach of let's raise, get rid of the headache, get rid of the brain damage earlier than later. Or at least push it off for enough time to allow us to get through the rest of the consensus that we need to be able to launch. |
| # / 01:10:10 | Richard Brown | Yeah, I'm completely on board with that. And let me clarify a couple of things. So when we say the same thing is coy and it's cute, and all the rest of it. But one of the primary reasons we say soon is because we don't know exactly, and so that's even more unsatisfying. But here's why we don't know because Cyrus and I, and a bunch of other people have been spending... you'd be shocked to learn how much time we've been spending in meetings over the last two weeks trying to figure out how we can optimize to speed up the process to get us to MCD launch. |
| # / 01:10:44 | Richard Brown | And the reason, so as we pull apart these processes and the way that the system could potentially work and how we can optimize it in front of a complexity here and then increase agility there. The takeaway over and over again, is that when MCD launches has nothing to do with... Well, I shouldn't say nothing. It's not up to MakerDAO, it's up to the community. And it's up to governance. And it's up to the risk teams. |
| # / 01:11:11 | Richard Brown | It's what's decided in these calls and in the forums, and in the governance portal is what determines when MCD launches. We have a significant amount of work that needs to be completed. And I've touched on this in the blog posts that have gone out recently. But we need to get the general model in, we need to evaluate these asset classes. We need to find out how much debate is going to be kicked out by these things. We need to figure out the communities, or the governance Maker token voters appetite for risks, or their appetite for complexity in the system. |
| # / 01:11:47 | Richard Brown | We could find that we get into this cycle where we're talking about these asset classes, and people are just waiting for us to stop talking so they can just say yes. Or, we could find out that the model goes out or quantitative analysis go out and then thousands of shells show up in the forums. And we just sort through all of it. And that may take far longer than anybody expects. We don't know how long it's going to take to iterate through these things. |
| # / 01:12:16 | Richard Brown | That was sort of my speech about the "soon" thing. I probably lost the thread here a bit. I guess my question still remains. So let's assume that in three weeks, we have a new governance portal. And that governance portal allows us to put up one, two or 1000 polls at once. Do we need to delay the stability fee votes between now and then in order to handle the debt ceiling of poll? |
| # / 01:12:47 | Matthew Rabinowitz | Yeah, I would say it could wait three weeks. From my sense of things it could wait three weeks, we're not that close to the limit right now. If it was at 99 million, I would say it would be urgent. |
| # / 01:12:59 | Richard Brown | Yeah. And it's why I understand... I'm not trying to create leading questions, because I literally don't know. And so there's obviously risks here that, who knows, maybe a couple of whales just show up, or maybe the market turns on us. And maybe we rapidly approached this in the next couple weeks, it's impossible to guess. |
| # / 01:13:18 | Richard Brown | There's those concerns and the other concern too it loops back to my initial thing that's at the top of the hour where we have these emergent processes, and we need to achieve consensus. We need to surface signals, we need to have debates and we're just now taking the first baby steps about what that process actually looks like. |
| # / 01:13:37 | Richard Brown | And this forum post that I've put in the link there is a great first step, but maybe I'm being overcautious. I'm still unclear about at what point do we decide, "Okay, something has happened here." This is where it gets hard, right? Like, we have the poll, we have... Well, here's an example, and I'm not calling out tbone, because tbone's one of my favorite people. But the framing of this poll it says based on various amounts of suggestions in our chat and forum. |
| # / 01:14:10 | Richard Brown | I look at that and I go, "Well, what suggestions? Which subreddits And what were those suggestions? And were they good or not? Are they logical or not?" So in my mind, this is just a popularity poll, it's not exactly scientific governance. What I would love to see would be somebody coming into the forum saying that, "Okay, we had a situation with the debt ceiling, because of X, Y, and Z, look at this graph. Here's my arguments for." and then somebody can say, "Here's my arguments against," and then after people digest this stuff, they go, "Okay, I'm going to vote for this number." |
| # / 01:14:47 | Richard Brown | That would be the ideal situation. But we haven't seen that in this thread. And I'm sure that's coming very soon. But there was that question, how considered are these options? Do they need to be considered? Is there a certain bar that needs to be met? And at what point... oh, yeah, sorry. I was going to say, we don't know when it closes, but I stand corrected. It says that it closed five hours ago. So I guess we do have time limits on these things. |
| # / 01:15:19 | Richard Brown | Let me wrap this up. When is Vishesh coming back? Oh, my God. So let me wrap this up, we need to figure out what this process looks like. And I'm not going to tell people what this process looks like. So we need to figure it out in the forum. So there's another thread that's like, "How how do we achieve signal? And how do we determine that the signal has started and the signal has ended and that we're all comfortable with the quality of that signal?" |
| # / 01:15:46 | Richard Brown | And once we figure that out, that's going to be the pipeline where... and this is... it had actually been another forum thread part talks about, okay, we have loose chats here there, Reddit, Twitter, whatever. It goes into a forum thread, it goes into discussion. At some point that discussion ends or a poll ends. We review it in a governance call saying like... well, actually we're doing it right now, so this is more acceptable. |
| # / 01:16:14 | Richard Brown | In a governance call, we say, "Hey, look at this signal request." We have a little chitchat and then at some point that just turns into a poll. That turning into a pole thing is a significant events right now because we can only have one poll at a time. |
| # / 01:16:32 | Richard Brown | In three weeks, it will not be a significant event, because we can have arbitrary numbers of polls. All right, I feel like... Sorry, Matthew, I feel like for the second or third time in response to one of your questions I just began drifting aimlessly. I think I might be burned out. Does that answered your question. |
| # / 01:16:55 | Cmooney | So Vishesh did just return. |
| # / 01:16:58 | Richard Brown | Yeah, I saw that. Was about to... |
| # / 01:17:01 | Vishesh Choudhry | I'm sorry, there's a storm going, so I totally lost all Wifi. I'm almost afraid to ask. How far did I get in before I was cut off? |
| # / 01:17:13 | Cyrus | Were you talking for a while and not realizing? Could you been talking for like 10 minutes? |
| # / 01:17:18 | Vishesh Choudhry | A few minutes. |
| # / 01:17:21 | David Utrobin | The last thing I remember was that you were talking about the ETH price dropping with DAI towards the end of June. That's the last thing I remember. |
| # / 01:17:33 | Cmooney | You said our model doesn't have that drop off yet. And then you dropped off? |
| # / 01:17:42 | Vishesh Choudhry | Well, if you're interested, I'm happy to finish running through otherwise, I know that ate up time. |
| # / 01:17:50 | Richard Brown | Yeah. We need to know what would Vishesh do. So let's go through it. |
| # / 01:17:54 | Vishesh Choudhry | Okay, I'll power through it. My apologies. Ironically, the last thing that I have actually said in person was, there was a lot and a little time, so hopefully I went through it quick enough, but... |
| # / 01:18:11 | Vishesh Choudhry | Okay, so roughly, I was just mentioning supply had run up fairly significantly in that time period. There was some mention of big, chunky $1 million DAI transactions in the last couple days. But in the grand scheme of things, that's really not very huge. I mean, really, at the end of June we saw a 4.3 million DAI minted in one day. So that significantly puts things in perspective. |
| # / 01:18:47 | Vishesh Choudhry | We saw that supply increasing despite the increase in the stability fee, although there was a slight dip after the 17.5% increase which doesn't make sense. When you increase the stability fee, you do tend to see a little bit more of a refinancing over to secondary lending platforms. But even through the 18.5% increase, overall, the supply did continue to increase. And it's leveled out around that 91.3 million level. Some of this is also potentially explained by what's been going on with secondary lending platforms. But overall, it has been increasing that circulation. |
| # / 01:19:28 | Vishesh Choudhry | So basically, the age of open debt has consistently decreased, the age of closed debt is held pretty steady. So essentially, this is due to new debt being created, and not much more old debt is being paid back. Although there was a slight decrease through the most recent stability fee increases. More DAI is being drawn from new CDPs than old over the course of the month, which does make sense with a lot of new DAI being minted. And some of this may be due to the fact that the same potential owners at the end of the day have opened new CDPs because they previously closed old ones. |
| # / 01:20:13 | Vishesh Choudhry | So what we saw, to basically echo what Cyrus was saying, with the collateralization ratio is over the month of June that collateralization ratio ran up with the ETH price and then back down. Although, far more dramatically than the ETH price actually moved, if you look at nominal percentage terms. So this echoes that idea that, as the ETH price decreases, collateralization in the short term decreases, but then people do tend to increase their collateral to "top up". |
| # / 01:20:44 | Vishesh Choudhry | And then as the ETH price peaked around that June 25 timeframe, collateralization did run up as well, but then ran down more dramatically than the ETH did in that same time period. So again, as the collateralization shoots up due to ETH price in the short term, people do tend to pull collateral out commensurately. And so you saw this increase in leverage right at the end of June, beginning of July, which corresponded with some of the "damage to the DAI price." |
| # / 01:21:16 | Vishesh Choudhry | So I'd already gone through the spread, but essentially, that same volume weighted average has held steady, even though there has been a more dramatic drop in the ETH price in that same timeframe. So that cut off of ETH dropping from 80s to 70s. But during that same time frame, the DAI price is held around that 0.98 level, which... Sorry, DAI [inaudible 01:21:45] cipher takes time to learn. |
| # / 01:21:47 | Vishesh Choudhry | In the meantime, the secondary lending rates, so there is some new stuff on here. So the weighted average borrow and supply rates, the weighted average borrow rate tends to hold just below that stability fee, even though dYdX tends to hold a bit of a premium over Compound. On that average level, it does just hover below that line. |
| # / 01:22:10 | Vishesh Choudhry | So what's interesting is over the month of June... May was kind of an explosion for the secondary lending platforms. June has been a steady increase in borrow volume, and supply volume as well. Hard to see from these graphs, but the supply volume has outpaced the borrow volume a little bit more. So that you see in this excess supply graph. |
| # / 01:22:28 | Vishesh Choudhry | This is the amount of supply that is not being borrowed on the secondary lending platforms. And so this does tend to pretty closely correspond with those rates. Because a lot of these platforms actually just algorithmically set rates based on what this "cashflow" is. So that rate has... that excess supply being increasing through the month of June and into early July, with a bit of a dip around that same June 25 time period where the ETH price had peaked. |
| # / 01:23:00 | Vishesh Choudhry | So this is interesting because you see the excess supply has increased recently, even though the overall DAI supply has increased and the stability fee has increased. Although a little bit of dip now recently, as people heard that 18.5% was coming around the bend. So the excess has decreased as DAI got more expensive, which made sense. But initially, it was running up with that DAI supply. That was interesting, potentially people purchasing DAI from others who were selling it for leverage, and then relending it on secondary lending platforms just to get a little bit extra interest. |
| # / 01:23:46 | Vishesh Choudhry | So this last graph, just basically, makes this one around the utilization rates. So I had looked at this initially to see if there is a relationship between the stability fee and the utilization rates on the secondary lending platforms. To basically circle back to an old conversation that we had with Rich around to what extent can use secondary lending platforms to learn what to do with the stability fee or other monetary policy things on Maker. |
| # / 01:24:10 | Vishesh Choudhry | But I think we tend to see what's roughly expected, which there is a relationship when the stability fee is decreased or increased to what happens to these utilization rates. But during the time period when the stability fee is flat, you also do see some pretty solid fluctuation in these rates. So what I'm basically saying is, when the stability fee has being changed, or other significant monetary policy components are being changed on Maker, you need to wait, I think, for some of these metrics to settle on the secondary lending platforms. |
| # / 01:24:42 | Vishesh Choudhry | But whilst all is quiet on the Western front, then you do tend to see some potentially leading indicators that you can listen to around what's going on in terms of basically oversupply. And so, it's been a complex, I think, point to touch on. Some people talking about, "To what extent is this supply, people who've leveraged out of DAI and sold it and other people who are just purchasing it and lending. Versus people who are actually drawing DAI out, and then holding on to it, and in the meantime, actually lending it out of these secondary lending platforms." |
| # / 01:25:18 | Vishesh Choudhry | One of the data points that I think will be helpful to illuminate that, which... Sorry, I was looking for a tab that I recently closed. Okay, I'll also go back to it in a second, but I'll just verbally describe. So the transaction volume on DAI has actually been pretty huge lately. So over the course of the last month or so, here, the transaction volume is been... Actually since May really, the transaction volume has been significantly higher than it had been pretty much through the end of 2018 and the beginning of 2019, with the exception of around that November, December time period for 2018. |
| # / 01:26:11 | Vishesh Choudhry | But that was an atypical point for everything in Maker. So since then you've seen a pretty significant increase in that transaction volume. Which is, I think, a positive indicator for people who are... It's significantly outpaced the increase in trading volume during that time. So it's not just attributable to trading volume. So I think that's a pretty positive indicator for people who are wanting to look at, to what extent is diet being used and transacted on outside of just being traded for leverage? Because it does appear to have picked up fairly significantly. |
| # / 01:26:43 | Vishesh Choudhry | So the other last point that I want to touch on, and sorry for taking up extra time here, but the liquidations. Some people have talked about, "Oh, there was a significant amount of liquidations recently." Just want to put that in perspective. The liquidations recently have been, yes present, but overall in the grand scheme of things not huge. And have actually been at fairly high collateralization ratios. So this can potentially be attributable to less volatility in the ETH price, although I think we know that's not true. |
| # / 01:27:16 | Vishesh Choudhry | The ETH price has drop significantly recently, but I think also attributable to the fact that there was a higher overall collateralization ratio recently than there had been in the past. And I think potential more efficiency in how these liquidations are being carried out. So you do see a bit of an- |
| # / 01:27:37 | Cyrus | Do you have any long term? I mean, is that all of the history there for all of the liquidations? Do we know what the lowest... have we ever had anything go lower than 1.4 whatever that number is? Four three? Where [inaudible 01:27:50]. |
| # / 01:27:51 | Vishesh Choudhry | So, the one caveat I will say is, there is sometimes transactions, and this is what makes liquidations hard, in the same block. And there are sometimes changes to the collateralization ratio, and liquidations in the same block in the same transaction hash. And so, ordering those events properly is sometimes a little bit tricky. And ordering them also in the same block price tick because there tends to be liquidation in the block that there is a price tick. So getting that level of precision and being 100% confident is hard. |
| # / 01:28:29 | Vishesh Choudhry | But that being said, I think there are some rounding errors, where once in a blue moon you'll see a collateralization ratio around like 154, which is not possible. But pretty much every data set, there's just one liquidation I've kept my eyes on. Happened early on, and it happened at 154% collateralization. But pretty much every data set I've seen around Maker shows that same blip. But it's very tiny. |
| # / 01:28:56 | Vishesh Choudhry | And so that's my point as I will say, there is sometimes outliers in the data sets, either slightly above the line or fairly, fairly low. But I think those are very small CDPs were that rounding error can make things seem a little bit more extreme than it is. So accepting those cases, buying large, yeah, I think the collateralizations have all been in this range, roughly. So the takeaway is, overall collateral sessions have been pretty safe, and they've gotten safer recently. |
| # / 01:29:30 | Vishesh Choudhry | And I think this is primarily due to the fact that collateralization overall, if I flip back over to this, has... Collateralization overall, has come up hugely from a year ago, six months ago. But also, I think, a little bit of improvement in efficiency of how the keepers are operating and how quickly they're getting to those liquidations, but the latter is really, really difficult to show on the data. |
| # / 01:30:03 | Vishesh Choudhry | Okay. Sorry I took extra time. But hopefully that was worth it for the people that stayed. |
| # / 01:30:16 | Richard Brown | All right, we're at an hour and a half, but I don't have a hard stop so I'm fine with sticking around. Cyrus, can you stick around too? For a bit? |
| # / 01:30:27 | Cyrus | For a little bit, yeah. |
| # / 01:30:29 | Richard Brown | All right, if there's questions for Vishesh, we can address those. Or, Matthew, do you still want to dig into the narrative? |
| # / 01:30:51 | Richard Brown | Did we lose him? |
| # / 01:30:54 | Matthew Rabinowitz | And I had been talking on mute. It's the process of going through rather than after an hour and a half, it's probably best if we just put a link in the chat group for going to discuss the points from the forum. Going through the consensus in a logical sequence about how we can float, basically, DAI when we reinvent it or relaunch it. |
| # / 01:31:14 | Matthew Rabinowitz | It's absolutely not a statement that we have to do it that way, but let's beat the crap out of it and make it better. |
| # / 01:31:27 | Richard Brown | Okay. Vishesh, do you see any questions in the chat that you wanted to address? Haven't been following them. |
| # / 01:31:33 | Vishesh Choudhry | I just answered one real quick. And then one point that was touched on earlier, I think, is the debt ceiling. I just want to make sure that that conversation is moving forward because as the supply- |
| # / 01:31:44 | Cyrus | We talking about it while you were cut out. |
| # / 01:31:47 | Vishesh Choudhry | Okay. |
| # / 01:31:47 | Cyrus | Earlier. |
| # / 01:31:49 | Vishesh Choudhry | So, I'll just leave it at that. And then, yeah, I answered. 18.5, again, I would have waited from the 17.5% increase to see what would happen before even doing the 18.5% increase. That happened rather quickly, in my opinion. A further potential increase, I think, is a little bit premature. So I am generally going to recommend more wait and see and collection of data mentality before making potentially rash decision. Because, I think some of the effects of these changes to take time to realize. |
| # / 01:32:22 | Richard Brown | All right, I think that's... Well, maybe we were running out of stand, but it was a great call though. We got through a lot. So thank you Vishesh, thank you Matthew for your input and your gracious willingness to move things to the forum after this long call. Cyrus and... Thank you to me for my deep wisdom and insights. |
| # / 01:32:49 | Richard Brown | All right. This post will go up. I'm moving the conversation away from Reddit and into the forum. So the Reddit summaries for these agendas have been locked and I'm directing people to the forum. So if you're interested in looking at the summary, keep an eye on the forums. David will get something up there for you soon and I'll be posting links to the socials and Twitter. All right, thanks everybody. |