 Welcome to the October session of our CB and DC virtual seminar series. We have a new format this month. We have four candidates who will be on the job market this year, all of whom are working on issues related to digital currencies and central bank virtual seminar series. We have a new format this month. We have four. Okay. So, um, so first some ground rules, um, what there's going to be each session, each speaker will have 25 minutes. We've asked that you only ask, uh, short clarifying questions during those 25 minutes since it's a short presentation, and then we'll have 10 minutes for Q and A during the Q and A participants can unmute themselves. Um, sorry, panelists can unmute themselves, but participants can enter questions in the Q and A box at any time. Okay. So with that, let's move straight to our first presentation, which will be by Shumiel Ooyang from Princeton University. Shumiel, take it away. Okay. So, uh, can you see the screen? Yes, looks good. Okay. Thank you very much. I understand the organizers for having me here and also my journal paper here. I'm Shumiel Ooyang from Princeton and my research interest center on FinTech, digital economy, financial intermediation and household finance. So this talk is about cash payment and financial inclusion. First, let me introduce the background about this financial inclusion challenge a little bit. It's widely known that it's quite hard to provide financial services to the underprivileged individuals and extending credit access to them is even harder because the overhead cost is really quite a large, um, compared with the small loan sizes. And also there are some severe information asymmetry between the lender and the borrowers. But recently we see a boom of big tech credit in a global level. And it is also overtaking FinTech credit in many places. And it's likely to provide some new credit to those who are not, uh, previously covered by the traditional banks. At the same time, we also see a rise of the mobile cash cash payment, especially those provided by big tech forms like AliPay, WeChat Pay, Apple Pay and so on. So these kinds of mobile cash payments naturally combine the important technological advancement such as better data, better models and more accessibility. At the same time, it has accelerated the shift from cash to cashless economy in a lot of places. And in China, even the beggars would print a QR code on their begging bowls to collect arms. So is this a coincidence or does the information flow that comes from the cash payment fuel the boom in the big tech credit, big tech credit? So here's the research question. Has cash payment facilitated the lending to the traditionally underserved? If so, then how? Although we see the joint rise of the digital payment and digital credit. It is just quite hard to establish some causal relationship between the two. Why? Because it requires an exogenous shock and cash payment activity. It also requires detailed individual level data on payment, credit and so on. So I try to combine a natural experiment with reaching administrative data from Alipay to solve these challenges. And I've actually spent more than two years physically in Alibaba just to work with the data. And here are the main findings. First, the cash payment flow facilitates credit provision and credit takeout on the extensive margin using the in person payment in a month lead to the increase of the likelihood of getting credit access by as much as 56.3%. And on the intensive margin, if we exorbitiously increase the in person payment flow by 1%, then the credit line is expected to increase by about 0.41%. So just from this number, we can see that magnitude is quite large. So why can this credit provision happen? I show that it's the big tag is because the big tag takes advantage of the information in the payment flow and it is beyond the information in the credit usage or credit repayment. It is also beyond the enforcement of power that comes from the access and management. And in my paper, I also estimate a simple structural model to quantify the information value of the payment data. And it shows that compared with the counterfactual case where the lender only knows the personal characteristics. If they also know the payment data to get more information about the borrowers, the credit line is expected to increase by about 57.7%. But due to the time limit, I will focus on empirical analysis in this talk. And I also show that it's the financially underserved who benefit more from it. We see a stronger credit provision effect after digital payment adoption, but it focuses on the less educated and the older. Okay, so before going to the data, let me introduce two phenomena that are quite related to this research. The first is the rise of the cash payment market. So here, I've brought the US card payments transaction volume of its GDP in blue and the China's mobile payments transaction volume of its GDP in red. And we can see the patterns are quite different. For the US card payment is quite a mature market and the transaction volume of the card payment is about 30% of its GDP in the past decade. But for China's mobile payment, in 2012, it was only about 4% of its GDP, but as of the end of 2018, it has become about 300% of its GDP. So we see the boom of the mobile payment in China. But this is not just for China. Actually, if we look at the other countries, it's also happening, especially when we compare it with the plot that I did about two years ago, we can see that a lot of the countries are catching up in terms of mobile payments. And the second observation is the rise of the big-time credit at the same time. So let me briefly introduce Alipay and Huawei Credit Line, the two objects that I'm going to study in this research. Alipay is the largest mobile payment in China. Alipay is the largest mobile wallet in China with more than 1 billion active users. And the Huawei Credit Line is the virtual credit card provided by Alipay to its users and it now has become the largest consumer finance product in China. In my representative sample of Alipay users, I find that about 72% of them have access to the Huawei Credit Line. And people do use it a lot. Among those with Huawei access, about 95% of them have used the credit with an average monthly usage of about 80 US dollars, which is roughly 20% of the average disposable income in China. And people do use it and also it's likely to provide some new credit to those who are not well-covered by the traditional banks. So even among those who do not have a credit card on file, we can see that about 64% of them have access to the Huawei Credit Line. So from these numbers, we can see it's quite inclusive already. In terms of the data, I get the representative random sample from the whole population of Alipay users, and it contains about 41,485 users with in-person cash payment activities. It's individual level panel data at monthly frequency, and it contains detailed information on not only their personal characteristics, but also their payment credit investment, and many other relevant digital footprints. The sample period covers about 41 months is from May 2017 to September 2020, during which period both the mobile payment and the back-sharing industries develop quite fast. So you might be wondering, why am I talking about back-sharing industries? Because it's critical to my identification strategy. So basically, I use the placement of Alipay-bounded shared bikes as the instrument for the in-person Alipay payment. And for the Alipay-bounded shared bikes, it's a bike that you can directly unlock with Alipay's scan function. You just scan the QR code on the bike with Alipay, so you can use the bike and then you can ride it anywhere. And with exactly the same scan function of Alipay, you can also scan the QR code of the local merchants and make some in-person payments. So that's why we expect to see some large effect there. So when people start to use the bikes, they will also start to use the in-person payments. And next slide, I will show some direct evidence about this large effect. So here, we're doing an event study. The event we're looking at is the adoption of the bike. And the target variable is a log 1 plus x transformed in-person non-bike payment flow. So we see that right after the bike adoption, people start to heavily increase their non-bike payment flow, although this non-bike flow has already excluded the bike-related usage fees just to avoid the mechanical effect. So we see the effect on the non-bike payment flow is both large and persistent. And next, I will show some evidence to support the validity of the instrument. First, I will show some evidence about the relevance condition, and then I will talk about the exclusion restriction condition. For the relevance condition, we are expected to see a very strong correlation between the bike placement and the in-person payment flow. And that's what we see in column one. So when there are more bikes placed in the city, people living in the city turn to increase their in-person payment flow. And I guess what's more interesting is in column two, where we further separate the Alipay users into the bike users and the non-bike users, and we're only seeing these effects of this bike placement on the bike users in-person payment, but not on that of the non-bike users. Well, it might be surprising on the full site because it's so easy to become a bike user. So in my definition, as long as you have used the bike at least once during the whole summer period, then you are a bike user. So it's quite surprising. But after second thought, it actually makes a lot of sense. Imagine you're a person who do not know how to ride a bike. Then no matter how many bikes are placed around you, your behavior might just not change. So in that sense, this result also supports the notch story. So basically, when you are a bike user, you are starting to be affected. Otherwise, you won't be affected. We can also see that this story is kind of consistent in the intensive margin on the intensive margin in column three. So here we only focus on the bike users, and we see that it's after the first bike usage that the bike placement starts to have an effect on their payment flow. And the timing of the first bike usage itself does not explain the payment flow very much. And then I will talk about the exclusion restriction. The first column just shows some reduced form result that when there are more bikes placed in the city, people living in the city tend to have higher credit line. And one common concern is that there could be some factors that are correlated with everything. For example, some local growth potential with local infrastructure plan. So it can be correlated with the bike placement decision of the back sharing companies. It can be also correlated with the credit provision decision of the big tech companies. So in column two, we again separate the adipate users into the bike users and the non-bike users. And we're only seeing the effect of the bike placement on the credit line for those bike users, but not the non-bike users. If it's about these kind of common factors that are correlated with everything, then we expect to see that it's correlated with everyone's credit line. But we're only seeing the correlation for the bike users. So it kind of help us to rule out these alternative stories. And column three is another robustness test. It just shows that it's after the bike adoption that the bike placement start to have an effect on the credit provision. I've actually done many other robustness tests, but due to the time limit, let me skip them for now. So now I hope I've convinced you that the bike placement is a good instrument for the payment flow, and then we can do the average analysis. So in this table, we use the bike placement as an instrument for the in-person payment flow, and we show the first stage result in panel B. And then we see the causal effect of the in-person payment flow and the credit provision in panel A, where we show the second stage result. And in panel C, we just shows the corresponding OIS estimations. So different columns just focus on different specifications, and we just look at the measures on different margins. For example, in columns one, two, and three, we focus on the extensive margin of this credit provision. And in columns four, five, and six, we focus on the intensive margin of this credit provision. So the credit access is a dummy variable indicating whether this person have access to the credit in the current amounts. And then the log credit line is just a log transformation of the credit line if the credit line is larger than zero. And we also look at different margins of this in-person payment flow. For example, in columns one and four, we use log one plus X transformation, which captures both the extensive margin and the intensive margin, but it's a little bit hard to explain. So in columns two and five, we only focus on the extensive margin of the in-person payment flow. And in columns three and six, we only focus on the intensive margin. But across all these specifications, the message is quite consistent. In panel B, we see that the back placement has always been a quite strong instrument for the in-person payment flow. And in panel A, we see that with some exogenous increase in the in-person payment flow, the credit line is expected to increase on both the extensive margin and intensive margin. So we see the credit provision happens, and we might be wondering why it can happen. So here we look at the potential channels. So potentially the lender knows more about the borrowers by getting more information from the payments. And the information can come from a lot of places. It can come from the transaction data itself. It can also come from the credit usage and the credit repayment. So to separate those two, we categorize the payment flow into the non-credit payment flow and the credit payment flow. So the non-credit payment flow adjusts the money that people pay with their own money, so it has nothing to do with the credit usage or credit repayment. But even if we only focus on the non-credit payment flow, it still has quite a large effect on the credit provision on both the extensive margin and intensive margin. And the effect doesn't change much if we add the credit payment flow as a control. And we might be wondering that the enforcement power of the big-time credit might arise if we use the digital wallet every day. For example, if we do not repay our time, then potentially the platform can just freeze our account or forbid us from using our assets under measurement. So in panel B, we use the assets under measurement as a proxy for the collateral and add it as a control. But even after that, we still see that the in-person payment flow has quite a large effect on credit provision, and the magnitude doesn't change very much, actually slightly higher than the previous estimate. So this table just shows that the transaction data itself can have a lot of value. So this is different from the credit usage and credit repayment information that which is a foundation of the business model of the current credit bureau. So it shows that the transaction data itself can already be quite useful. And we see we're also quite wondering like whether people use the credit and whether people use it for good. So in these slides, we try to look at the takeoff of the credit by looking at the virtual credit card share in both the in-person transaction settings and also the online transaction settings. And indeed, we see that with the increase in the in-person payment flow, people tend to have more credit usage in both the in-person and the online environments. So this can be a result of learning by doing effects. It can also be a result of the increase in the credit provision. And in terms of the whether people are using the money for good, we may be worried that when people get some cheap credit, they might use it for games, cigarettes, and so on. So since I have access to the very detailed transaction level data, so it allows me to categorize the spending into the compulsive spending and the non-compulsive spending. So the compulsive spending are the ones related with games, cigarettes, alcohol, and so on. But we do not see that the in-person payment have a significant effect on the consumption structure, although we might see a shift in the consumption level. And the last part, we focus on the heterogeneous effect of these digital payment adoption. And we might be wondering that with more precise information from the payment data, whether it will lead to more credit provision to the less credit worthy? Well, actually, it might not necessarily be the case. And I will show the intuition with a simple theoretical example. So just to be clear, this is based on a simple economy where there's a modernistic lender and also a bunch of borrowers with different credit worthiness. And this quite specific environment is also, and these plots are given by specific parameters, and I'm just using these to show the intuition. So here I plot the optimal credit line provided by the lender to the different borrowers with different credit worthiness. And we look at the three stages of the borrowers adoption of the digital money. The first stage, people are just cash users, so they're not using the digital money at all. And here we just assume that the lender does not know any information about the credit worthiness of each borrower. And it will just decide not lend to anybody. And in stage two, people have become the new digital money adopter. And here we just assume that now the lender knows whether the credit worthiness of each borrower is above a certain threshold. Now we just make it 0.25. And then in this setting, the lender will decide to lend to those who are above the threshold, but not lend to those who are below the threshold. And now let's look at the stage three, which is shown in the red line. And now people have become a digital money user. So in the stage two, people are just the new digital money adopter, so they just submitted their personal characteristics information. So that's why the lender knows more about them. But now since they are a digital money user, they use the digital money every day. So in addition to this kind of personal characteristics, the borrower, the lender also knows their consumption behavior. So here we just assume that now the lender knows the exact credit worthiness of each borrower. So the optimal credit provision was shipped from the blue dash line to the red solid line. And we can see that for the really high type of group, they will have an increase in their credit line. But for the ones who are relatively less credit worthy, previously they can get some credit line, but now they get zero. So in that sense, they're actually worse off. But I have to admit this result is given by quite specific parameters. And for this scenario, we call it financial divide because some of them are better off, some of them are worse off. And actually it can also lead to financial inclusion. So in the panel B, the only thing that I change is to change the parameter from 0.25 to 0.8. And we can see that for the really high type group, on average, their credit provision is the same. But for the ones who are relatively less credit worthy, they actually get higher credit line. So from the example, we know that actually the theoretical models can predict a very different result. It can be a good news for the less credit worthy. It can also be a good news. So we really need to look into the data and see what's happening. So firstly, we need to categorize people into different groups. And we need to know who are the financially underserved. In this slide, I just use my data to conform a traditional view in China that is the less educated and the older who tend to be financially underserved. And indeed, we see that they tend to have lower financial service usage and also lower financial literacy because they have fewer linked debit cards, lower maximum assets and management, shorter investment experience. They're also less likely to pay with their real name, less likely to use their own account, or less likely to have a complete profile. So let's separate the people by their education and see whether the effects of the digital payment adoption are different on them. So in columns one and three, we'll only focus on the less educated group. And in columns two and four, we only focus on the more educated group, meaning that they have a quality degree or above. So in panel B, we see that the first stage result is pretty strong for both groups. So the compliance is there for both the less educated and the more educated. But in panel A, we see that the effect is only there for the less educated group, but not there for the more educated group. So in essence, it helps us to support the financial inclusion story. Okay, so let me conclude here. In a nice review by Bergfuster and Puri, there is two open questions. The first is, is the information from the payment flows a causal factor behind the credit expansion? And the second is, does it benefit customers previously underserved by the traditional financial institutions? And my paper tried to combine unique data and a new identification strategy to answer yes to both questions. And this is also the first paper showing the payment information fuels the big credit to the household. It also has important policy implications. The mobile payment can potentially be the new financial infrastructure and provide a new opportunities for sustainable and inclusive finance. Okay, thank you. Okay, thanks very much. Very interesting. So we've got a sizable group of panelists. So why don't we raise your hand if you want to ask a question? Yeah, you go ahead. This is very interesting. I have a question. How can you distinguish information from punishment? So if you see, I have more transactions that reveals information about me, but at the same time, that means I really like this payment method a lot. If you exclude me from it, that's a big punishment to me. So there is a question. How would you distinguish these two? Is that even possible? Okay, so thank you. This is a great question. And I have to admit I can only partially address it. So this is kind of, so the punishment is kind of related with the enforcement of power of these big tech forms. So what I did is just to use the assets and measurement as a proxy for the collateral. And we see that even when added as a control, the effect doesn't change very much. Actually, the effect is slightly larger. So meaning that at least it's not come, it does not come from the enforcement of power in terms of the assets and the measurement. But you are right. The platform can potentially punish the consumers in other different ways. And it might be the case that the consumers just feel that they just kind of leave without the platform. So this can kind of increase the enforcement of power, but it's something that we cannot directly observe in the data. So yeah, so that's, there are some limitations. And I can only partially address this concept. Okay, thank you. So in the interest of time, maybe we'll bundle the next set of questions. So, so I'll have people ask questions. You may you can take some notes and then try to address them together. So, so you go ahead. Yeah, so my question is what exactly the mechanism here you showed, I think it's funny and very sensible that you show the non banks access to payment data can improve their decision of making credit and improving, improve the financial inclusion. My question is really about the key thing is here is the information on non bank. So, if the payment information is so valuable, would that you would say if the bank have access to the same information, would that do the same decision to extend credit to the underserved population? Oh, because of like Ali pay is a non bank, so it's less regulated. So the cost is much lower. So they can do that. So that has the implication how your result can be generalized to other scenarios, right? Okay, thanks. So let's get to Charlie. Thanks. You have used question the issue of the using the the monies for for making bad or inadvisable purchases. I'm wondering whether you can turn that around and do part of the information gathering on that same thing. Is it the case that people who use their Ali pay for cigarettes and gambling are also going to get lower lines of credit? And because that would help untangle whether it really is an informational story here. Okay, thanks, Scott. Thank you. It's very interesting. I just have three comments to which I think have already been said but three things that I would appreciate hearing more about in the presentation of paper are one is the issue of the endogenity of the bike use. I know the sample is not exactly representative but there seems to me there might also be a selection effect as to who's riding the bikes, at least in my perception that might not be a select group that may have an impact on the credit provision. Second thing just to follow, just to affirm what Zhu said is the results are strikingly similar to what we've known for decades about early monitoring and credit screening by banks and their monitoring of credit of payments. And so it would be helpful to know what do we learn that's different or better or insightful beyond what banks have already kind of known proprietary. And the third thing is to think about whether the credit provisions, how does that fit into an optimal lifetime consumption path? Is this providing more credit that is allowing them to buffer against income shocks and therefore raising lifetime utility by smoothing it? Or is it enticing people who are maybe impatient to overborrow? So some discussion of those three things I think would strengthen that presentation and paper. Okay, and Russell? Yeah, I think it's super promising data set, it's super interesting. Thank you. So I will be quick. I think for people, lots of people are in the central bank in this audience, it might be good to know the real impact about having access to credit and credit lines. So I think one thing might be we can measure is whether the guy have access to the buy and they have access to credit line. So it is that they have also consumed more later on. So it's something we can measure. So it might be good to see something like that. So it is also the amplification mechanism. So after they spend more, they may also get more credit. So whether we see this amplification mechanism is happening in the data. Now, and the second question is about execution. So how exactly 4B is work? So it's an early pay, they partner with one of the banks, it's like Amazon and JP Morgan Chase. So Amazon don't issue credit card, JP Morgan trade issue credit card. So maybe a loss of screening or having access to credit actually go through the bank, but not the payment side. I don't know. So just some institutional data, maybe you can explain a little bit. So the last thing is similar to what Charlie mentioned, maybe not just the type of the payment matter, but maybe the size of payment matter, right? Because now if I see you a big spender, so I may be more likely to extend credit to you. So do we see something like that in the data? Thank you. Scott, did your hand go back up or? No, okay, okay. Sorry, no, I forgot to do it. Just wanted to be sure. So shoot me out, you've got like a minute and a half to respond to all of that. So you can, you can, you can say whatever you like at this point, however you want to wrap things up. Okay. Thank you, Todd. So for those point, so about whether it can be generalized. So I guess the mechanism can be the same for many other traditional banks. For example, if one bank provide debit card to the consumers other than the other than the other banks, then potentially these banks can provide better credit than the other banks who have the same access to the credit bureau data. And for China's point, that's definitely a great point about whether we can use the more detailed transaction data to see whether the credit line responds to the different types of consumptions. And I will look into it later because my data can allow me to do this. And for Scott's concern about the endogeneity issues there, I've actually done some tests in the paper but not presented here. So I've done some selection issues about who are the back users and also whether the same thing would happen for those who have used the bikes only once. And actually the result is basically there for the one-time back users. So it can help rule out many alternative stories. And for the lifetime consumption pass, that's definitely a good point. And I'll try to address it more in the future. And thank you for bringing this up. And for Russell's point, yes, we do see that when people get the credit, they will consume more. And you are right, potentially we can also link the size of the individual transactions to the credit provision to them. And that's something that I haven't looked into for the future research. That's definitely a good direction. So thank you all for all the great suggestions. Okay, thanks, Xiumiao, for an interesting presentation. Thanks everyone for those great questions. And I hope that you'll feel free to anyone else who has questions to contact Xiumiao directly. So with that, I'll turn over the moderating duties to Jonathan for our next paper. Yeah, thank you. So our second presenter is Remo Diffenega from the University of Basel. So Remo, please share your slides. Remo is going to talk about central bank digital currency and bank intermediation. Yes, great. So first, I want to thank the organizers for putting my paper to the program. And I followed the seminar quite closely over the last two years. I think that's one of the good things of COVID that we have these online seminars. So I'm very happy to be here and present. So as Jonathan said, I'm going to present the paper about central bank digital currencies and bank intermediation with heterogeneous bank deposits. So what I talk about in the paper is that CBDC that is can be held by the general public. So I'm going to talk about the retail CBDC. As most of you probably are aware of, there are also caveats when we talk about the introduction of the CBDC and one caveat being the effect it might have on bank intermediation. What's kind of the line of argument here? Well, deposits currently are a cheap way of funding for banks. And if the central bank introduced CBDC and the people in the economy might see it as a substitute for bank deposits, then we might see a crowding out from deposits to the CBDC. And then there will be less funding for banks. And if there's less funding, there might be less bank lending which might reduce credit availability, could increase credit costs and hence might have real effects on the economy. Hence we might be really interested to think about this problem and analyze it how big it is. And so that is my first research question. Dust introduction of the CBDC leads to disintermediation of banks. And I answer this question by building a tractable general equilibrium model based on the one hand on a new monetarist model in line of logs and right. And on the other hand, ONLG environment in line of logs. So I'm obviously not the first one who thought about CBC and bank intermediation. There is also people in the panel Jonathan has a paper, Todd and Daniel have a paper on that as well. So let me tell you what my main contribution of this paper is and then come to my second research. So if we think about the effects of the CBDC, they might differ depending on the how the central bank designs the CBC. We might think about different designs that could limit the amount of the CBDC that people can help central bank could limit the number of transactions, or there's also a proposal by Binsai that we could have different remunerations pending on the amount. And for example, here Fabio Panetta, who is an executive director of the ECB board, talks about this remuneration example and then mentions that we should ensure that CBDC is a means of payment that is as attractive as cash, but we should reduce the attractiveness of the CBDC as a store of life. So it seems like policymakers take care about whether the CBDC is held only as a payment vehicle or also as a saving vehicle. So why is that relevant when we think about CBDC and banking intermediation? Well, if the CBDC is only used as a payment vehicle, it might mostly be a substitute for liquid transaction deposits. And if it's also held as same vehicle, it might also be a substitute for saving deposits. And this effect, this difference between whether people hold the CBDC as a payment vehicle or also as a saving vehicle has not been studied that deeply yet. So I want to bring more depth to the analysis here and have a look at this question. So my second research question is, how does the introduction of a CBDC differ depending on whether it is held only as a payment vehicle or also as a saving vehicle? And then I first have an analytical part to get some qualitative results where I solve my model. And then I also want to get some numbers. I want to get quantitative results. And for that, I calibrate the model to US data. So that's basically what I do. And let me give you a preview of the results. So what do we find with respect to the first research question? I find that there is no effect on bank lending if banks hold voluntary reserves. So if we are in an excess reserve environment, and that's kind of straightforward, if we have an outflow in deposits, then what happens is basically just that first the excess reserves are reduced and hence with no effect on bank lending. However, I find that if we have an introduction of a CBDC and we have some shift from private money to public money, so from deposits to the CBDC, and we have no excess reserves, then bank lending will decrease. With respect to the second research question, I find that if there is a preference shift such that 10% of the agents switch from transaction deposits to the CBDC, that decreases bank lending by about 1.2%. And the effect is almost three times stronger if CBDC also crowdsides same in deposits where I find a decrease in bank lending of about 3.0%. So as I said, there is already some literature on CBDC and bank intimidation, but my main contribution is adding this depth with respect to whether the CBDC is used as payment vehicle or also as a saving vehicle. Okay, so let me give a quick outline of the presentation. So I'm now going to give you an intuition of the model. I'm not going to show you the math and equations. If you want to see something, I have everything in the appendix, but for today I'm just going to give you an intuition. Then I'm going to tell you how I introduced the CBDC in the model. I'm going to show you some analytical results and then I'm going to show you one collaboration results. There's more in the paper for today. I'm just going to show you one of the main major results. Then I'm going to conclude, but let's start with the model. So in the model, there's different types of agents. On the one hand side, there is entrepreneurs and the entrepreneurs, they have an investment opportunity, but they cannot work and they have no endowment. So somehow they need to acquire a good that they can then invest. And this investment then yields a return rate. Then on the other hand, there is consumers. The consumers, they can work when young, but there is some time in friction. So actually what they want to do is they want to consume later in life, but later in life they cannot work anymore. So they can only work when they're young but do not want to consume there and want to consume later but cannot work anymore. So what do they want to do? They want to get some assets to shift consumption over time. Now there's two types of consumers in the model. There is an early consumer and this early consumer will demand as an asset a medium of exchange. And there's a late consumer who consumes later and this late consumer will demand as an asset the same week. You can think of that a bit as a like this type of the model. You can think of it a bit as a dynamic environment, not the whole model, but only this part, but there is no uncertainty here. So this kind of organically gives me a demand for a medium of exchange and the same. Then on top of that, agents in the model have a preference over public and private money. So in other models, how it is generated for both is sometimes that we assume we have different types of meetings and in different types of meetings we can only pay with cash or only pay with public money etc. and private money etc. So what I do is I assume that they're heterogeneous preferences of the agents such that someone to hold private money, which is bank deposits, and someone to hold public money, which is cash or then later CDs. These heterogeneous preferences are also dependent on the interest rate differential of the two types of assets. So you have a preference over one of the assets, but if the interest rate differential is big enough, you might still choose the other asset. So let me give you a quick example. If for example you slightly prefer public money over private money, so you slightly prefer for example cash over deposits, but deposits pays like the 4% interest rate and cash does not pay anything, then you still might want to hold, you still might want to hold deposits. What's nice about that is that it kind of shifts agents from one asset to the other one depending on how the interest rate is. Exactly. So I have these two types of major ingredients in the model. There is the natural demand for a medium of exchange and the saving vehicle, and then a demand for private money and for public money. Now what do we have so far? We have these entrepreneurs. These entrepreneurs, they want to get this good, but they cannot produce themselves, but they want to invest it and the consumers, they can't produce this good. The thing is that the consumers cannot lend directly to the entrepreneurs because by assumption they cannot enforce repayment, and hence that gives rise to intermediaries, which is banks. So we have banks in the model and they basically intermediate between entrepreneurs and consumers. So what happens? An entrepreneur goes to the bank, at the bank he says, hey, I want to have a loan. So the bank creates a loan by creating a loan. He credits the deposit account of the entrepreneur, and the entrepreneur can then use these deposits to go to the consumer and purchase the good from the consumer. He can then invest the good and this year it's a return and with the return he quickly consumes it and also pays back the loan. So that's basically what banks do. Then by assumption here, I assume that there is a perfectly competitive deposit market and an imperfectly competitive loan market. I do that because I want to have some imperfect competition in the model because you see that the results are quite interesting if I do that. And I assume this perfectly competitive deposit market because it keeps the model more tractable because I have. Okay, so now let's come to these two types of deposits. In the model, I exorbitantly assume that there is a minimum reserve requirement on all the payments that are done from these early consumers. So we only have the mini reserve requirement for the early consumers. That's also how it's done in reality, right? We have only a minimum reserve requirement on liquid deposits, but not on the liquid. Now, because we have that only on liquid ones, it is actually profitable for the bank to offer two types of deposits. The bank then wants to offer liquid and liquid transaction deposits that are going to be held by the early consumers. And the bank will want to offer illiquid saving deposits that are going to be held by the late consumers. Why is that? Because saving deposits, they can be invested one-to-one into loans, which they have some profit marching on. But they also want to offer the liquid transaction deposits because there is some demand from the early consumers. Okay. So that's basically the setup of the model. And now let me explain you how I introduce CBDC. So as I said before, first, I want to say that I only talk in this presentation about non-interest-bearing CBDC. In the paper, I also talk about interest-bearing CBDC, but for now, for this presentation, I only talk about non-interest-bearing CBDC. So what I basically assume is that basically the policy experiment is that upon the introduction of a CBDC, there is a preference shift between private money and public money. So the prior assumption would I guess be that upon the introduction of a CBDC, some agents would switch from private to public money. You can think of that as some person wanted to hold public money, but also likes to pay digitally. That was not possible so far, but now we see CBDC as possible. So this agent now might have a strong preference for public money. So basically the policy experiment is you have a preference shift from private money to public money upon the introduction of a CBDC. What I do not do, I want to emphasize that I do not state how many people will switch from private to public money. I just state even that there is a certain preference shift and there is a certain switch from private to public money. What is the effect on bank lending and what is different effect on bank lending depending on whether people only switch from liquid transaction deposits or whether they also switch from illiquid saving deposits to CBDC. Now, first two analytical results. Here we need to differentiate two cases. The first case is whether the constraint is non-binding. So the minimum reserve requirement constraint is non-binding. That means we are in an excess reserve environment. Here I find that there is no effect on bank lending. Why is that? I already mentioned that in the introduction. The banks here hold an optimal loan amount and if there is a shift in deposits, so for example if there is an outflow in deposits, first the excess reserves will adjust and hence there is no effect on bank lending as long as we are in this excess reserve environment. Now the more interesting case is if the minimum reserve requirement constraint is non-binding. What I want to give the example here is if we think about a preference shift from private money to public money upon the introduction of the CBDC. So we have this underlying preference shift and that means that that basically triggers a switch from private money to public money and that means that this fraction here alpha D which is the fraction of early consumer holding transaction deposits decreases. So the fraction of agents holding deposits decreases and here in the first row I'm only talking about the case where we only have a shift in the preferences with respect to the payment vehicle. So people only switch from the liquid transaction deposits to the series. Let me talk you through the table to explain to you what happens and what I show in a little bit. So first straightforward because there is some people switching from private money to public money from liquid deposits to CBDC, there is on the extensive margin a decrease in the transaction deposit amount. By assumption there is no effect on the saving on the saving deposits but now because the banks they have some profit margin they will increase their interest rates. They will increase their interest rates and why do they do that? They do that because by increasing their interest rates they attract some deposits on the intensive margin. Now what's different to other papers that have thought about this as well so far is that here the banks can not only have this intensive margin effect on this transaction deposits but they also have this intensive margin effect on the saving deposits and now because they have this intensive margin effect on the saving deposits they can partly compensate this outflow in transaction deposits. Now what I can show is that bank lending here always decreases but we have because we have this compensating effect the outflow in total might not be that bad. Then the second row is if we have both, if we have a preference shift from both liquid deposits and illiquid deposits to see what happens additionally it's just that we also have an extensive margin effect on the saving deposits. Exactly now we have some analytical results but we don't know yet how big these effects are so now to get some idea how big the effects are going to be I want to calibrate them all to the data. So for that what do I do? I calibrate them model to the US economy from 1987 to 2006. I choose this time period because I want to analyze the model in the binding constraint invite right because I mean sure we are currently in excess reserve regime but as you've seen before in excess reserve regime the results are quite straightforward there is no effect on bank lending so what I want to do to also differentiate the effects with respect to the payment vehicle or the saving vehicle I want to analyze the model when the reserve requirement was right. So what do I do? I consider only cash held in the US because we know that a lot of dollar is held abroad right and since I want to match some fractions between cash and deposits, deposits held domestically etc. I want to control that. I assume that a hundred dollar bills are used as savings and all small bills are used as payment vehicle and then as I said I match some fractions and some of them. With respect to the data to get some bank data on interest rates etc. I derive that from data if you call reports and from most standards time series. Now I'm only going to show you one slide today with the vibration knife most often in the paper but for today let me just show you this main result and what we have here is we have a preference shift from or we have upon the introduction of CBDC we have this preference shift between private and public so and we saw before that this preference shift that changes this fraction of agents holding this fraction of early consumers holding transaction. Now here at the red vertical line we are in the calibrated equilibrium. Now if we here go to the left that means we have an underlying shift from private money to public money upon the introduction of CBDC and that decreases the fraction of agents holding transaction deposit. If we go to the right we have exactly the opposite we have an introduction of CBDC and just to keep it general we could assume that this also triggers a shift from public money to private money and that would mean that here a fraction of early consumers holding transaction deposits increases. Okay so that's the setup here and now what do we see? So here this red line is the amount of transaction deposit. As expected the amount of transaction deposits decreases upon the shift from private money to public money because CBDC is introduced. Now interesting now and again what's different to the other papers is that additionally because we have these two types of deposits and the banks increase the interest rate on the saving deposits we will actually see an increase on the intensive margin so the banks increase the interest rate it can raise more funds on the on the saving deposits and hence this compensates the outflow and transaction and thus the overall effect on bank lending is actually not too bad or I mean that's relatively sad but now on the right hand side what do we have on the right hand side on the right hand side we have both and a change with respect to the payment vehicle and also with respect to the saving. So if we are here in the calibrated equilibrium and we go to the left that means that agents switch from liquid deposits to CBDC and from illiquid deposits to CBDC by the same percentage so there is an underlying equal preference shift such that the fraction of agent switching is equal. Now we see that there is also a decrease in the transaction deposits but as well there is a decrease now in saving because people also switch away from saving. Now one other thing that I want to mention is that you see that here on the right hand side the decrease in liquid deposits is not as big as the decrease that we see here. Why is that? Well here the banks cannot compensate the outflow and transaction deposits with saving deposits and hence they increase their interest rates by more and because they increase their interest rates by more we see that the outflow here in transaction deposit is not as big as it is here so that's one interesting observation but still because we have both an outflow in transaction deposits and in saving deposits the overall effect on bank lending as you see if you compare these two pictures will be bigger. If you want to have some numbers if we are at the calibrated equilibrium and we go to a situation where 10% of the agents switch from transaction deposit to CBDC then bank lending increases by about 1 to 2 percent and if we are at the calibrated equilibrium and we have an outflow of both or a crowding out of both from liquid deposits and from illiquid deposits to CBDC then bank lending decreases by about 3.0 percent so we see that these effects are stronger so policy makers might be interested in really thinking of how should we design the CBDC and whether depending on how we design it whether people will start holding it as a payment vehicle or also as a CBDC. Okay so let me conclude what did I do I built a general equilibrium model to analyze the effect of the CBDC on bank litigation I differentiate the effect depending on whether the CBDC is used only as a payment or also as a saving vehicle which is kind of my main contribution of the paper and I find that there is no influence of a CBDC on bank lending if banks thought voluntary reserves so if we are in an excess reserve environment however if banks do not hold excess reserves then we find negative effects on bank lending and the numbers I have just told you so yeah so here I want to conclude and thanks for your attention and I'm happy to take questions. Thank you very much Remo for very interesting paper and so maybe at this point we can open up for Q&A and I saw a few hands raised already maybe we can start with Mohammed. Okay thanks Remo so that's very interesting so I'll try to be quick so my question is more or maybe comment I don't know is about the terminology that you're using in terms of the store value so as long as I can as far as I can see in the model so there is no inter temporal decision in terms of saving so the two types of deposits are and correct me if I'm wrong maybe I'm wrong the two types of deposits are different from the perspective of the bank in terms of basically whether they are stable or flighty I would say right because one of them are kind of subject to reserve requirement one is not but basically both can both deposits are spent before you go to the next cm right so there is no inter temporal decision if I am right and in that sense so and why this is important because if there is inter inter temporal kind of substitution then that's different from the liquidity channel that we usually see in the new monetary models and I think you don't have that inter temporal decision so if that's correct maybe that would be better to kind of label these two deposits differently not and that would change your intro perhaps a little bit okay so I'll just stop but we can discuss offline if yeah should I answer now or you want to take all the questions yeah maybe you can answer it right away so there's not it's not the case that both are spent before the cm and I mean I generalize that now a lot because I just gave the intuition but so the consumers they're both they're finer to live and basically the early consumer works when he's in the cm and then wants to consume in the dm but he does not want to consume anymore in the next cm and he's fine at the list that's the og part and now the the late consumer he works as well when he's young in the cm but he only wants to consume in the next cm so he is basically idle in the in the dm and that's kind of so it is kind of there is the payment vehicle in the sense that this is spent in the dm and then only in the next cm is the late consumer that wants to use the the the saving vehicle and then basically from the dm to the cm the payment vehicles of the liquid deposits are going to be held by the seller who received them for producing the food but maybe I misunderstood the moment and I'm happy to discuss that afterwards okay so I'll I'll wait if there is time more time I'll follow but there was an offline okay thank you great okay we have a few hands maybe you can collect a round of questions from starting with Catherine yeah thank you so I really like the paper the question I have is I may have missed it so I assume that I understood that you assume that the cbdc is introduced and there is a fixed proportion of consumer switching but I would think that also the design of the cbdc would have an impact so whether it's interest bearing and what amount so how high the interest rate is and then the substitution or even if the the central bank sets a varying interest right on the cbdc this could be even more dynamic thanks thank you and charlie yes um I wanted to think a little bit about what excess reserves mean story because um it's it's good and very useful that you've used it on a case with on an error where there are no excess reserves but when you do have excess reserves you want to think about the question of what's the definition of an excess reserve there's going to be some right right now with lots of reserves out there you still wouldn't want to think about all of them as excess even if they're above the the the the official line because the banks have some interest in in in holding them for whatever reason so I'm wondering how does the effect of reserve interest interest rates and the rest translate into your story into whether reserve is excess or not and if that if that feedback is taken into a case does that mean that your statement that the that the case with excess reserve is extremely simple does that become less simple once you've got that once you've got that feedback in there thank you um russell yeah I want to be quick uh it is very light um very beautiful simple model and super elegant uh I like the insight so uh the insight it seems to be very very realistic in the sense that if they are sitting on a loss of excess reserve it will be like a buffer to absorb any negative consequence from cbdc in the deposit market so I guess the results the insight should be applied to more general environment I saw uh Dave very reputable ask the question okay how it depends on the market structure in the q&a so I think maybe we can handle it but it's a very very different question uh I'm just wondering uh like what Charlie mentioned so um in in in a system now we have floor system a loss of reserve is sitting around it seems to be that the dental bank can also cancel out the any negative effect of cbdc uh by running a loss of reserve in the system but go back to your original insight uh if uh if the bank they are holding a loss of reserve can cancel the effect of cbdc and in the end so this kind of cbdc have to be uh this kind of reserve have to be hold by someone in the system right at the bank so how is this channel if you leave them you back in in your model just one day thank you okay thank you all the panelists maybe uh Ramo you can um first respond to these questions and then we have a few questions in the q&a uh if we have time we can get to those questions yes um okay so first uh country's question so thanks a lot to all the questions first the country's question with respect to the design and whether it's interest sparing or not yeah what I do in the model is I think of the case where it is interest sparing and then yeah what we see is because we have this heterogeneous preferences that also depend on the interest rate differential what's basically happening is if we increase the interest rate on the on the cbdc and people will start to switch um to to the central bank money and that's basically a channel that I analyze in the papers whether that I haven't that I haven't showed you with respect to the excess reserves um yeah what I have in the model so basically it's yeah there's different ways how to model it right in some papers what they assume is reserves are basically exogenous and basically that's set by the central bank here it's rather kind of if you think about a deposit facility of the central bank where you can go and you can deposit you can deposit so what basically happens that's also related to ross's question how are reserves affected or how are they channeled what basically what happened is a consumer or whatever let's just call it any agent would go to the bank and say to the bank okay and I want to withdraw I want to withdraw deposits from uh I want to withdraw deposits from you so what the central bank does it reduces the reserves and it just generates new cash right it kind of gives the cash to the bank and the bank hands the cash to the or the cbdc to the agent so that's kind of how reserves will be channeled out of the system if there is a withdrawal of all of deposits and with respect to rates I also show uh in the appendix how the interest rates on reserves affect the whole story it's now a bit hard to summarize that in a sentence but I also look at that and what I can say you're right that it's a bit hard to say that there is just the fixed requirement that then everything above is excess reserves that's totally fair here I guess I just yeah it's just simplified to basically all the reserves are the same they all pay the same interest rate and that's yeah definitely simplified from from reality could I go to the q&a in the chat or I think Todd has his hand raised maybe yeah I just wanted to follow up on this excess reserves discussion I took I understood what you just described about you know the the consumer withdraws the bank loses deposits and it loses reserves and I think Charlie's point was that affects the bank's liquidity coverage ratio for example and and maybe some other things like resolution planning and so if you think about that that reserves are not just meeting requirements they're not just meeting traditional reserve requirements they're meeting new regulations that that's no longer so neutral that that that's going to tend to influence the bank's actions and so then you know I interpreted Charlie is asking should we think about your requirement as traditional reserve requirement or is also including these new things in which case maybe it's a little trickier to calibrate but but it might be interesting to think along those lines I once had a quick discussion oh sorry that was a very good question thank you okay no okay sorry then I misunderstood it sorry Charlie and thanks Todd for clarifying um I once had a quick discussion about that as someone else to remember but I yeah at the moment I have to admit that I rather think about classical minimum reserve requirements I know that in the US they also put it to zero percent but also if they put it to zero percent you could kind of argue that it's still relevant because it basically just means whether the set whether the banks do hold an optimal loan amount or not but so currently yeah I would rather think about classical minimum reserve requirements but it's totally fair point that I might want to think a bit more about this other newer liquidity requirements that we have okay um and then so I just want to bring you the attention in the chat box that Scott has shared some link to data on US holdings of $100 bills and and then in the chat box in the Q&A David Rappaport has raised question about both on the liability side the SSI of the balance sheet of banks first I think he asked uh on the liability side do banks right finance itself only with deposit but why not also with equity and other short-term wholesale funding and then on the SSI he asked about what what is the role of imperfect competition in the long market if you can you know spend maybe one minute or so quickly respond to it I can answer that quickly and so with respect to the first question yeah they can only finance itself with deposits so basically I kept the model simple I could add equity and short-term wholesale funding sources as well so yeah I haven't done that if people find it interesting I could do that as well so far I kind of wanted to bring more depth with respect to the two types of deposits and that's why abstract it from the other from the other parts to keep it simple basically but I could put out that in attendance so I said well what's the role of imperfection in the long market well I could mention that so the point is that I wanted to introduce some imperfect competition the point is because I have these two types of deposits and I wanted to solve the model analytically as well it kind of became hard to get the results or the results analytically when I introduced imperfect competition in the deposit market so that's why I for tractability I decided to focus on the imperfect competition on the long market because then I still have imperfect competition and I still have these effects that banks can adjust interest rates but I still could solve a lot analytically and I think the effects would actually be qualitatively very similar and quantitatively if I have perfect competition in both then then it might change a bit that the numerical thank you very much Remo for the answer and also your presentation now I think we have a few minutes for a short break before we go to the next paper at 1045 so let's have a few minutes short break next paper Remo I think it is a very are you here so can you hear me yes I hear you yeah I think it is elegant model so simple but sharp to the point but I don't know whether I am clear about the intuition so it seems to me the intuition and the insight is that as long as banks have excess reserve it can be used to buffer any negative consequence from from the deported outflow triggered by the cpdc right so and the insight seems to apply to a more general environment in general right so because you have some extra buffer so that the bank can tap so I I think and people might and then the next question that whether it's buffers really a buffer then it may be you it can connect you are maybe some of the reserve is not really above because maybe they have to satisfy some battle three we find my light talk mention maybe the maybe the lcr the liquidity corporate ratio something like that then maybe it is a good way to say okay even if we account for some of it so as the insight still there maybe so it's just maybe some of the reserve is not really maybe part of it is just for for satisfying you'll be fine maybe I think that the the the insight is applied more generally also with the the competition in the long market as well whether it's more have a competition in the deported market as long as you still have some buffer on the excess reserve so it seems that goes through yeah sorry yeah I think I yeah that goes back to the point of some of your smart yeah not for other reasons maybe I mean I mean more about it maybe all could generalize it in a way that yeah if you think more about the background in Spain I just thought that some fraction of the positive needs to be somehow fixed and you can't use it something else you know more about it maybe more generalize it in a way that yeah oh I hear myself um but yeah I agree with that that that there are these other I could also introduce all the constraints as well so in a way that I don't know where it helps in the baseline um but yeah I agree with that so Sam maybe we should uh uh let's try whether you can we can screen a little bit uh because we might be going too fast soon all right so so Sam maybe we should uh uh uh let's try whether you can because you're going to screen a little bit now because we might be gorgeous are you guys hearing an echo like not it not too much an echo but the same thing repeated a few seconds later yeah I think someone has the youtube video open in the background I think that's why but I'm not sure but I also hear an echo I say not too much an echo but the same thing so uh let's get started so our next presenter will be Sam so from yeah also Sam you will have 25 minutes and then we have 10 minutes Q&A it's all your all right so you guys hear me well all right so I want to first thank the organizers Jonathan Russell and Todd for inviting me to present my work here so my name is Sam and I'm a John market candidate from Yale economics and before I go I want to make a quick disclaimer you know before I talk about my paper that this is not really my quote-unquote John market paper per se so I'm you know very proud of this paper of the work that I've done here that I'm about to share with you today but I'm working on other projects and I'm working on other projects in digital currency space as well but I just wanted to ask you to keep an eye out for my John market paper that's going to be posted on my website next week which unfortunately is not related to the digital currency or the crypto space but I promise it's going to be you know equally interesting it's not worse so now that we got that out of the way let me just begin my talk so the motivation for this work comes from the fact that there have been many instances of distress in the cryptocurrency market such as you know the severe crash in May of 2021 due to the Chinese government's you know crackdown on the crypto market or many other instances of runs on stable coins that we saw a few months ago when investors ran on Terra and the market capitalization of more than 18 billion dollars you know crashed down to zero in a matter of days so this paper is more focused on studying the financial stability issues associated with you know stable coins of the second incident but the top two stable coins combined have market capitalization over 112 billion dollars as of a few days ago and not only that their growth has been really fast as their market capitalization as of January of 2021 so you know just a bit more than a year ago was only 25 billion dollars so given the significance of the stable coin market right now the research question that I ask in this paper are two rules first I ask if the cryptocurrency market is connected to the financial market as the title suggests and if distress in the cryptocurrency market can potentially spill over to the financial market second if they are through what channel are they connected and I'm going to argue that yes they are connected and distress in the cryptocurrency marketing can potentially spill over to the financial market through what I'm going to call reserve back stable coins so in the interest of time instead of just going through the entire literature if you want to mention this one contemporaneous paper by Bartholomew Gardner and Nguyen I ask a similar research question as mine but my contribution is to take advantage of the structure of the stable coin market for identifying the causal effect as I'll show you later in this presentation so before I go on to the main analysis I just want to give you a kind of a really quick high-level overview of the cryptocurrency market and introduce some key terminologies that I'm going to use throughout this presentation so cryptocurrency was introduced as an alternative form of money that could quote unquote decentralize finance away from the central government or control and it gained momentum especially after the global financial crisis when the government devalued government issued money by printing and disseminating an enormous amount of it through relief programs like quantity raising however cryptocurrency that are not backed by other assets or what I'm going to call in this paper fiat cryptocurrencies like bitcoin and ether are exposed to significant price volatility which makes it an inconvenient transactional medium so stable coin can emerge to remedy this situation so tether was established I think in 2014 so stable coins refers to digital currency that maintains its price stability by pegging themselves to a specific fiat currency such as the US dollar so I think a lot many of these stable coins pegged themselves to the US dollars and they act as kind of a safe asset in the cryptocurrency market or kind of a gateway to a fiat cryptocurrency market as approximately three quarters of trading on cryptocurrency platforms occur between a stable coin and other cryptocurrency and this is this is from like a chairman of the sec a few months ago so from a high level there are two types of stable coins based on their pegging mechanism first it's a reserve backed stable coins that keep a reserve of fiat currency denominated assets to maintain their price stability so the example includes uh you know the tether and usd coin which are kind of the major stable coins in the market right now the second type of algorithm stable coin and for this presentation at least I'm going to call stable coins that are not backed by fiat currency denominated asset algorithms so in particular they're either backed by other cryptocurrency like dye or they're doing kind of or they're kind of doing some open market operation type algorithm like tera to maintain their pet to the US dollar okay so this figure is taken from a paper by liao and kora michael and we can see that reserve backed stable coins in particular tether and usd coin are dominant in the markets if you see here like I think finance usd is also a reserve backed so like the reserve backed stable coins are like taking out more than 80 of the market capitalization and this here is showing you the left hand side or the assets are the tether's balance sheet as of the september of 2021 which is taken straight from their assurance report that was supposedly audited by an independent auditor so looking at this balance sheet you can see how reserve backed stable coins are pegging their stable coins to the US dollar stable coin issuances were effectively collateralized by the supposedly safe and liquid money market instrument we can also observe that a significant proportion of tether's balance sheet was specifically made up of commercial paper and this fact will be used of put it put to use for my main analysis of this paper okay but before doing that let's also look at how algorithmic stable coins are collateralized remember i'm going to call every stable coin that are not backed by yet currency denominated asset algorithmic stable point in this paper so we can see that um so i'm looking at the instance of dye and can see that dye is first over collateralized by other cryptocurrency and they're collateralized mainly by usd coin which is another dominant kind of the second largest reserve backed stable coins in the market so i'm going to utilize the structure of the stable coin market that i showed you uh and introduced until now to identify the causal effect of movement in the stable coin in the in the cryptocurrency market on the financial market so i showed you that i'm going to categorize a stable point market into two the reserve backed stable coins and algorithm stable point and i showed that at least until the third quarter of 2021 commercial paper took up a significant portion of reserve backed stable coin issuers balance sheets while other cryptocurrency like usd coin took up a significant portion of algorithmic stable point issuers balance sheets this means the issuance of reserve backed stable coins directly affects the commercial paper market while algorithmic stable points affect the commercial paper market only through their effect on their reserve backed stable point therefore i can use the issuance of algorithmic stable point as an instrument for the issuance of reserve backed stable point to investigate what the effect of reserve backed stable coins issuance is on the issuance quantity and yield of commercial paper as we saw that treasuries also made up a large portion of tether's balance sheet i also study what the effect of stable coin issuance was on the treasury yield curve okay so i'm going to argue that the issuance of algorithm stable coin satisfied the identification of assumptions needed for it to work as an instrument for the issuance of reserve backed stable point first i show that algorithmic stable points are backed by other cryptocurrency including the usd coin which is all which is a reserve backed stable coin furthermore it's been shown in the literature that stable coin investors don't really differentiate among different stable coins when they're making their investment decisions so when they're they're equally likely to hold a die as they are likely to hold um um usd coin for example and tether so these facts make it so that algorithmic stable coin issuances satisfy the relevance assumption furthermore i argue that algorithmic stable coin issuance also satisfies the exclusion restriction as the issuance of algorithmic stable coins do not require the issuer to claim dollar denominated asset as reserve which means the effect of algorithmic stable coin issuance on the commercial paper market works only through its effect on the reserve rec stable coin issuance so given that these two assumptions are satisfied hopefully i'm going to estimate the two-stage least square model as follows i'm going to see what the effect of the daily issuance of tether and usd coin is on the following days issuance quantity and yields of commercial paper in the us as well as the following days treasury yield curve by instrumenting tether and usd coin issuance quantity with die issuance quantity so in the first stage i'm going to request the combined tether and usd coin issuance on the die issuance from the previous day and use the predicted value from the first first stage to uh see what the effect on the dependent variables are which are logs of commercial paper issuance commercial paper yield and the treasury yield okay so the result of this estimation is shown in this figure here i show that a one standard deviation increase in the issuance of tether and usd coin which corresponds to about uh 330 million dollars results and 11 percent increase in commercial paper issuance quantity the following day and 18 basis point decrease in the commercial paper yield so this shows that the issuance of reserve back stable coins created an excess demand for stable on commercial paper which led to an increase in its issuance quantity an increase in its prices i also show that an increase in stable coin issuance or reserve back stable coin issuance led to a 15 basis point decrease in treasury yield which means it also kind of i don't have the quantity data so it's kind of hard to identify the demand effect but it kind of provides a suggestive evidence that there's a higher demand for treasuries due to the commercial paper issuance on a stable coin issuance as well but now i estimate the same model across commercial paper different of commercial paper maturities so i'm going to show you that the estimated coefficients with commercial paper issuance is a dependent variable with lines and lines above and below again as the estimate showing you the 95 percent confidence interval we can see that the positive effect of stable coin issuance and the commercial paper issuance is primarily driven by an increase in the issuance of the commercial paper with the shortest maturity of nine days or less for a commercial paper with maturity of 10 days or more the estimated coefficients are statistically insignificant and close to zero so this shows that kind of stable point issuer are really cognizant of cognizant of the need to manage liquidity of their reserves and thus affect the very short-term commercial paper market so now the result i've shown you until now includes samples only until the third quarter of 2021 and why did i limit my sample this is because starting around third quarter of 2021 both tether and usd coins started to move away from holding commercial paper as a reserve reserve all last year there were like rumors and stories about how these reserve stable coin issuers are holding toxic commercial paper from companies that were about to go bankrupt in the foreign country for example so this kind of was an effort to gain credibility among investors so here this is a tether's announcement from a couple weeks ago tether officially i think the date is october 13 2022 now tether officially announced that they'd no longer hold any commercial paper and we can see also in this news article here that they have been working on reducing their holding of commercial paper for a while since third and fourth quarter of 2021 and this shift in tether's asset allocation strategy can really be seen when we compare the balance sheet in the third quarter of 2021 and in the fourth quarter of 2021 in the third quarter we can see that the commercial paper and cd took up about 44 percent of the reserves while the treasury took up about 28 percent however in the bottom panel in the fourth quarter we can see that this kind of flipped commercial paper and cd's proportion was reduced to 31 percent while the treasury's proportion was increased to 44 percent so looking at the strategy shift it's natural to ask what its effect was on the commercial paper market and to do this i'm going to estimate this model where i include an interaction term with the dummy variable that equals one if the observation is from the fourth quarter after the fourth quarter of 2021 so the table shows that the estimated coefficients for beta one on the interaction term are negative which means after the fourth quarter of 2021 an increase in the issuance of reserve back stable point did not affect the issuance of commercial paper or actually had a negative effect on the issuance quantity of commercial paper estimating the same model but with the commercial paper yield as a dependent variable we can see the same thing the estimated beta one are either set to simply insignificant as in the first as in the aspect commercial paper case or even positive i'm not i'll be only 90 90 percent of confidence interval positive for various different types of on commercial paper so these results provide kind of a suggestive evidence that after fourth quarter of 2021 stable coin issuers were kind of indeed unloading commercial paper from their reserve so now you might be asking sure this kind of sounds cool and good but why didn't we see any of these effect last may when people ran on terra and more than 18 billion dollars worth of market capital days and got wiped out in a matter of couple days so i'm going to argue that this was because terra was also an algorithmic stable point that did not hold any dollar denominated asset and to show this i'm going to see what the effect of die issuance was on the commercial paper market instrumenting die issuance with terra issuance and since terra has come into a significance very recently the time series is a lot shorter and the first stage after statistics are a bit too low but considering all this we can see that the die had a muted effect on the commercial paper market so we can extrapolate from this result that terra meltdown last may did not affect the financial market that much because it was an algorithmic stable point that did not directly hold a reserve of commercial paper or treasury on their balance sheet to maintain their path to the us dollar okay finally i'm going to show you what the effect of cryptocurrency market is fiat cryptocurrency market is on the commercial paper market so fiat cryptocurrencies are those that are not packed to a specific currency like bitcoin or ether right so i show that a one standard deviation increase in the market capitalization change of the top three fiat cryptocurrency which is bitcoin ether and binance coin resulted in a 11.9 percent decrease in commercial paper issuance quantity 21 basis point increase in the commercial paper yield and 18 basis point increase in the treasury yield so i interpreted this result to mean that an increase in market capital evasion change which signifies a favorable fiat cryptocurrency market makes investors exchange stable coins for fiat cryptocurrency in order to increase their exposure to the fiat cryptocurrency market when the market is favorable this decreases the demand for stable coins which in turn decreases the demand for money like assets like commercial paper and treasuries so i'm going to use this in the last few minutes that i have to discuss kind of the policy implications of this paper so the financial crisis is an event when lenders there's a there's kind of a view that financial crisis is an event when lenders run on privately produced money like safe assets because it loses its role as money like we saw when investors ran on repo during the global financial crisis and there is an argument in the literature that argues that a reserve back stable points can be viewed as the newest form of private money so if you take this view stable coins are the newest form of money like safe assets or private money such as demand deposit or repo and stable coin issuance are the newest form of banks that hold very liquid safe safe and liquid form of assets on their balance sheets so from this perspective there's a potential for a run on stable coins so there's a long-running theory in the safe asset literature that when investors sorry when investors get anxious uh sorry when investors uh when investors say anxious about the collateral that is backing the safe asset and gain an incentive conduct when the costly information production on the collateral the asset turns information sensitive and loses the role of money so this is what we saw when you know for example investors ran on repo during the global financial crisis um because investors lost confidence on many of the nbs's that were backing the repo contract so in order for reserve back stable coins to be information insensitive and function as a valid form of money it needs to build confidence with the stable coin holders or investors that the reserves which is effectively the collateral for the stable coins is beyond reproach when it comes to safety and liquidity so i argue that tether and usd coins shift away from stable uh commercial paper to treasuries is on a attempt to build this confidence and retain information insensitive insensitivity of the stable coins because people were questioning uh and kind of clamoring for like the more specific kind of details about the balance sheet of these reserve stable coin issuers okay so let me summarize and conclude uh i showed you that the cryptocurrency market is connected to the financial market through what i call reserve back stable coins stable coin issuances increase in stable coin issuances led to the issuers to buy up the short-term money I see assets of this commercial paper and the treasury from the market which uh which exerted on the upward pressure on the demand for this kind of safe asset and it made that increase the issuance of this asset and lower the yields of these assets i also showed that when the market for the fiat cryptocurrency uh turn favorable investors trade stable coins for fiat cryptocurrency which lowers the demand for stable coins and we resulted in the opposite direction of the of the of the empirical relationships so finally we saw that the distress in the reserve back stable coin market can spread to the traditional financial market and ultimately to the real economy through commercial paper and the treasury market a run on stable coins means stable coins holders exchange stable coins for the us dollar on loss to honor this exchange request or momentum request the stable coin issuers need to sell off their um their assets including their commercial paper which is not really true right now because they don't hold commercial paper right now but this sell-off will put an extreme upward pressure on the commercial and treasury paper yields and ultimately affect the real economy by shooting up financing costs for every market participant in the commercial paper market and in the treasury market so that's kind of the channel through which i i i i imagine that the distress in the um cryptocurrency market kind of you can spill over to the financial and the real economy so let me conclude my talk here and i'll welcome any questions from you guys thank you so much our participant in the commercial paper market all right thank you Sen can you stop sharing screen and we are now uh it's very neat on the vacation strategy so we are now open for Q&A so if you want to ask questions please raise your head uh we will do for 10 minutes okay um maybe maybe mohammed first all right thank you very interesting presentation so just one question uh quantitatively uh it seems to me that the what you're saying is that like a one standard deviation which was you said you said around 330 million dollars uh lower the yield in terms of in the order of like 10 to 20 basis points that much so it seems a little bit large to me but my question is that do you like have you seen the literature uh um do they have comparable numbers in terms of like the literature that look at the kind of the treasury market and i i assume there should be some estimates already and do you have some kind of benchmark to as to compare your estimates with them quantitatively um yeah so i think the paper i mentioned in the introduction um um that they they do something similar to mine um and i think their numbers are a lot smaller than mine for sure so i was kind of surprised when i first came up with this number that you know the numbers that seems kind of kind of hard right kind of kind of large like in terms of like 11 percent decrease in the corner of the issuance so there might be something going on in terms of when i'm instrumenting the reserve stable points with um algorithmic stable points because i showed you that i mean there's kind of this um empirical result that there's these kind of stable point market market market capitalization have moved in lock steps so there might be some um i guess um not in 1080 but there might be some correlatedness within the market capitalization issuances among stable points that makes this estimate a little larger than it sounds but i think the i think the numbers quantitatively speaking numbers itself quantitatively speaking i think the other only the only other paper that conducts does a similar thing is the paper that i mentioned the introduction and they have smaller numbers as well but they don't have like iv strategy that i did so there might be you know and that's needed going on there too but yeah i'll look into it but in terms of like a quantitative the magnitude of the number itself yeah that's what i have right now so yeah any other paper that connect with crystal thanks for this really interesting presentation during your talk i was wondering about the the packing order from the viewpoint of an issuer uh you know uh you ask yourself the question whether you sell the more liquid as at first or which asset to sell and i was kind of wondering if you can say something about that and i guess that in practice also depends on the financial market condition which type of asset at which quantity you'd like to sell thanks yeah i mean i think that's really depends on the the specific market condition of for example of commercial i mean they don't have commercial papers i don't want to talk about commercial paper anymore but you know market for CDs versus market for treasuries right so they're going to sell off the treasuries the market is favorite for them so no is that the thing with this kind of a reserve stablecoin issuers is that you know there's been a lot of controversy in terms of like them not being able to they're they're kind of arguing that they're being audited but are they really being audited because you know they're based on this kind of Cayman islands for example and that there's no way for the authority to kind of really delve into kind of the liquidity or reserve management strategy that they're pursuing so and i don't have a deep insight on like the specific reserve management strategy that they're using but i think i agree with you christoph that you know there might be market favorably kind of thing might be going on there so for sure before the authority so before i move to jonathan so there's also a thumb question in q&a maybe related to what you have just discussed so maybe i can read right here so uh maybe i start from the second one so maybe a little bit more technical so uh there might be some that might be a unit root issue unit mean okay that's the classic trend in the level of use so is there anyway you can handle it in your estimation so that maybe and the second question is how the mechanism for stablecoin issuers are better than the commercial paper issue the mechanism you have in mind maybe related to what you talk to christoph also um i don't so i guess first for the econometric question i think i tested all the unit root issues with the the the the the commercial paper yields so uh the yield changes i tried yield changes and i think the results are pretty much the same so why i'm doing an issuance on changes changes on changes because now i cannot care about you know unit root issues in those kind of when i'm estimating the time series model so i think you can be sure that you know there's no really problem with that so in terms of their second question what is like the second question is why the mechanism the stablecoin issuer will affect the commercial paper i'm not really sure what you know you specifically mean by mechanism but what i'm saying is that when they have to issue more stablecoin they just need to buy up um um more you know safe assets from the market so they want to when they issue a dollar more of stablecoin they have to buy a dollar more of treasury from the market so that's you know how i'm thinking these two markets are connected so Jonathan what i'm saying is that yeah thank you zeng i think it's very topical question and very interesting result and just have to um two quick questions one is that as we know that um a die is partially you know backed by usdc i think since march uh 2020 right right so it's just a suggestion maybe it would be cleaner if we focus on you know the sample before that day so that we don't capture any indirectly frapped through usdc and then another question is how in terms of interpreting the result of the channel how can we be sure that it's not you know through the effect of a third common factor such as say policy rate change that may affect both the demand for die and the demand for uh uh commercial papers okay um i think i guess for your first question um yeah let me limiting the before limiting it to before um march 2020 might be good i'm not i'm sure like people who are you know engaging in empirical research in this space know that the time series is very limited right so every observation matter is very valuable and especially before like 2020 or 2019 like the the data is very noisy noisier because not many market participants were there in the stable point market it's been really been after the covid crisis where um the noise in the data kind of really subsided because there were a lot more market participants in the market so yeah i'll i'll be sure to know do a robust robustness for uh periods before they had the usd coin and for the second for your second questions um in terms of for example like no no for the policy rate changes potentially affecting commercial paper issuance and die issuance um yeah i mean there is i guess i have to admit that is a potential um endogenous problem but um what i'm trying to do is that they're affecting their the the effect of policy rate changes on the commercial paper market is much more direct than its effect on i think die issuance especially considering that um no no there's like especially considering that die is kind of over doing stable points so yeah i'll have to admit there is some problem with that but um my i'm using the um identification strategies to kind of clean up those kind of things as well so thank you so much for your comment on the commercial paper market so i was i was wondering maybe maybe i missed that part but but the issuance is uh and the investment in commercial paper is probably lumpy right so if somebody requests a thousand new stable coins then they will not immediately mint new stable coins or they will not immediately purchase commercial paper for that so i guess they have a their own reserve so maybe can you tell us a little bit more about the institutional details of right right how this process works so again like no like specific institutional details on how they exactly manage their reserve is like very opaque because no they really don't want people to know a lot about their reserve management strategy so like i'm really going down to the basics and seeing kind of i guess it's more of i guess an assumption that they're very active in this commercial money market and in terms of them you know trying to um manage a reserve liquidity because like when you see kind of the uh yeah i think so that's kind of what i'm kind of i guess it's more of an assumption that um yeah they're very active in the money market for sure yeah they're very active in this commercial right so uh we are in time uh and one minute earlier so maybe we i had better talk uh so we have that last one yeah yeah i think so that's kind of what i okay well thanks thanks song and thanks everybody so our final speaker today is luciano samosa from the swiss finance institute and atre se lezon luciano the floor is yours do you all see the screen looks good perfect so thank you thank you for being here thank you for organizing this event i'm very happy to be here and i start off with the anecdote um i have a cousin in midland who's a physiotherapist has zero interest in anything related with finance whatsoever and two years ago she had 20k to invest so she went to her bank and she asked for an investment portfolio and the financial advisor prepared a portfolio for her and before signing off she remembered she had a a cousin doing a PhD in finance so she reached out to me and said do you mind checking it out for me and said sure i opened it and she had 27.5 percent of her portfolio invested in cryptocurrencies she has no financial knowledge whatsoever that was the financial advisor uh prepared portfolio indirectly true at gfs right uh but still and then i realized okay maybe even if you don't believe in this thing you should really pay attention because it's pretty much uh everywhere and i hardly encourage all of you to go back home and check your relatives investment portfolios when you go home tonight so uh they're everywhere of course even like Fidelity introduced in the 401k's uh Goldman thinks is a good store of value uh truth is that the argument is often that okay maybe it's a diversification tool the fundamental value is not obvious therefore there can be an argument that since the fundamental value is not obviously correlated with the market might be diversification tool and then you have people like the man the CEO of man groups saying okay there is no inherent worth but it's good to trade which is very often uh a rationale for going into this market so of course in literature there've been some papers approaching uh cryptocurrencies i'll always talk about bitcoin because that's in my data set that's 80 90 percent of the trades right uh bitcoin say okay there might be a fundamental value which is linked with um transactions so some transactional value you can use this for transaction therefore there is some value and be some theory based on this this is definitely true if you look at this paper by folly amazing on estimating how much of the currency uh transactions are linked with uh illegal activities right and the answer is not that much so um it's not clear how you can explain price movements in bitcoin very extreme over the past few years using transactional values of fundamental empirically we know that there are network effects we know that um we know some patterns about it but we but we don't know today is why does it come together with the market which is my uh my key question what I look at so this is the uh correlation between s&p and bitcoin and you can see that was roughly uncorrelated until 2020 jumping around zero and then jumped up to almost 60 percent and he stayed there ever since this is s&p if I show you nasdaq this goes up to 80 90 percent so this begs the question what drives this crypto anti correlation why has bitcoin started behaving like a tech stock because especially when there are certain certain periods it really behaves like tesla or like game stop or some kind of tech stock right so this is my research question to give you a overview of the results uh my main takeaway is that retail investors are driving this correlation I explain you how and why uh I start off with this toy model to make the the argument a bit more more formal and get some insights uh I derive some some implication and test them in the data so I'll show you that retail investors trade cryptos and equities at the same time in the same direction uh I'll show you that these behaviors started in March 2020 really and that the stocks that these people like when these people are trading are the one exhibiting this correlation so it's not every stock it's just a specific subset so again it's a very nutshell toy model uh I show you this new data set that I'm using uh I go uh with this data set at the micro level at the investor level and then I go and look at global markets and show you patterns in the global market consistent with with my model I start with the kaal model is an extension of the kaal model so this is a very uh workforce model but maybe not everybody's familiar so just a couple of slides to uh for to be everything on the same on the same level so the kaal model is a model of how information enters the stock market and you have a informed and an informed and a market maker you have three types of agents the informed investor knows the true value of the asset they're trading the uninformed investor just trading at random and the market maker is observing the total order flow so they see the total order for try to understand whether the informed trader is trading or not and based on the prior the market maker has about the distributions and the fundamental value it adjusts the price based on the order flow it observes the equilibrium is defined by the order of the insider insider trader which is a function of the fundamental value and some level of aggressiveness and on the price adjustment by the by the market maker that is some prior and then some price impact of the order flow it observes now in this setting we add a second asset so we have one asset and we have a second asset with three key hypotheses so the first hypothesis is that the fundamental values of the two assets are uncorrelated now white might argue on why or how bitcoin might be correlated with the stock market but i take the view here that is fundamentally uncorrelated now bitcoin gives no right to cash flows it gives no right to real assets it's not obvious what the fundamental value is if you argue that is linked with transactions say sure transaction legal markets maybe but we take the view this is uncorrelated with the fundamental value of the stock market second hypothesis is that there are two market makers so the market is segmented which is the case today so today the main actors on the bitcoin market making is not the same as on equity markets this thing is changing a little bit in the recent months it will change probably in the future but today you have mostly specialized market makers you have all kind of algorithmic activity so it's very segmented market the market infrastructure is still very separated this hypothesis is key and i'll change this hypothesis at the end i'll show you what happens because the model seems trivial the one i'm going to show you but when you tweak this hypothesis it changed completed results so this is where we are now the third hypothesis is that then there are uninformed investors trading both asset classes with some kind of pattern right there is some correlation so whenever they are doing something on the market there is some correlation what they're doing on the other market i'm not saying it's positive or negative they are doing something so these are the three key hypothesis that bring inside the model so very quickly you see the first one translating zero correlation fundamental value the second the third one translating to this rock which is the correlation between the order flow of the uninformed investor across the two assets and the second one is the market maker all the observes one order flow so on on their own asset not on the other one this is the equilibrium which looks a lot like the kyle with the two different assets now it looks a lot like a kyle but then when we take the we look at the correlation between the prices right this is actually dependent on the correlation between order flows so this depends on rock and it has the same sign as rock so the first prediction of the model is since we observe positive correlation in between bitcoin and and the stock market therefore we have to observe positive correlation between uninformed traders order flows now in my setting uninformed traders are retail investors and they're trading a few thousand euros sorry francs in this case because they're in Switzerland that show you that asset and they have to kind of go in the same direction the second implication is that there must be something happening in March 2020 right so this behavior should not exist before March 2020 because there was no correlation before that date the third one is that this should be more this phenomena should be more important for when the retail investors are active so when they are actually trading because there are a lot of uninformed trade but these people are the one trading on both markets and the fourth one is that this phenomena should be more significant for stocks that they like so the one they they tend to trade so this is a data set i'm using it's a swiss bank which is called swiss quote now swiss quote is a bit of a special bank in the sense that it's a digital bank that's been around for over 20 years has been listed on stock market in swiss for over 20 years and it's also the the financial delivery provider of the post service in switzerland so it's very institutional and they offer both trading account on normal securities and cryptocurrencies but the cryptocurrency is not in direct trading it's not the tf you're talking about actual wallets so people can have a wallet with this quote they can transfer token in and out from other platforms and then they can trade on the on the bank uh crypto exchange so this is a special setting because there are very few banks worldwide with this kind of services and they were probably the first they started doing this in 2017 the reason is that switzerland regulation of cryptocurrency is very lenient compared to other countries so you can have bank offering this kind of services and there are the indisputed market leaders on this uh the bank is famous for online trading so people go there to trade so we observe how much money customers have with swiss quote but we don't know how much money they have in general right so uh customers tend to have real estate they tend to have um other savings for retirement and then and then they have maybe 20 30 000 francs to actively trade with uh with swiss quote so we have a rounder sample random sub-sample of customers of the bank of 77 000 uh we are holding between 2017 and 2020 and then of which what is 77 000 16 000 uh open cryptocurrency wallets so we have data before and after the open cryptocurrency wallet they tend to be more male they tend to be younger they tend to trade a lot more than the others and they tend to have high returns because they come from very well right so high return a much higher volatility in the portfolio so it's exactly what you would expect from an anecdotal point of view of what the what the cryptocurrency what um investor uh is this is just to show you zero is when they open cryptocurrency wallet and this is number of months before and after and this is an average number of um logins to the platform and what you see is that it goes from six to seven so once every three four days to up to 14 so once every two days so they look in much more often to the platform after the open cryptocurrency wallet this is just to say there is something significant happening to them um now from now on i'll use uh sorry for next two regressions i'll use a difference in difference uh stagger a different different setting clearly opening a crypto wallet is not an exogenous event so i cannot claim causality with it and just showing what happens to them after they open a crypto wallet the first thing i look at is trading activity because that's what uh they usually do with sysquad is the main objective in having an account with sysquad so you have fixed effects for investor and time so stagger definitive and then you have this variable which is crypto user uh sorry the dependent variable is stock turnover meaning what is the percentage of their equity position they traded in a given month so if i have on average 10 000 francs of equities and they insert order for 5000 francs this month this measure is going to be 50 percent okay so this measure can also be 500 percent if they trade a lot so it's all the orders divided by average holding of equities what we see is that after they open a cryptocurrency wallet they reduce their trading activities and equities by seven percent so they trade less equities as soon as they open a crypto wallet so since like they are moving their attention towards this new asset and we also saw that the trading in the two asset classes is correlated positively so they trade less crypto sorry less equities on average but when they do so they tend to do it at the same time as cryptos now what does this mean to me this is a strong indication i'm going to look into it an hypothesis that they think of the two asset classes as uh the same thing kind of the same thing right because they're substituting attention from one to the other and then they tend to trade them at the same time so one might think that the land of reason behind is that they think they're kind of the same asset so is there substitution well to see whether this is a story about attention i look at performance because we know that they tend to do horribly and then we know they have a lot of biases they tend to lose money on average compared to the market and there's a long literature of all the biases of retail investors so we look and say okay if they're paying less attention to equities they should be doing better on equities which is indeed what we find the dependent variable here is the sharp ratio of the equity portfolio and it increases after the open cryptocurrency wallet indicating strongly that they pay less attention to it because they tend to time the market wrongly they tend to do bad decision as we know from the literature so they do better and this can this goes in direction of saying okay they are substituting equities with cryptos and the training in the two asset classes correlates as i just showed you they tend to trade them both at the same time meaning that our assumption in the model is valid in data but the model makes a very strong prediction says okay but this correlation needs to be positive it's not enough time correlation needs to be positive now there can be two stories here right right because it's not obvious on one side you can say okay cryptocurrency are going well i sell some crypto and i buy some equities i rebalance a bit there are wealth effects here right and then we should observe a negative correlation in the order flows or you can say okay maybe on the other hand i'm very optimistic i buy both assets or i'm very pessimistic i sell both assets this is possible because they tend to have very large cash positions so they tend to have up to 50 60 percent of their overall holding with the bank in cash during periods where they're pessimistic and this can go down to 10 percent when they're very optimistic so a kind of a sentiment story i have one the following regression where i have is a dependent variable the net order flow inequities so buy order minus sell order and i have for the crypto i have buy order i have sell order and i have the net regardless of specification regardless of fixed effect this is always positive and highly significant which is consistent with prediction one so they tend to trade both asset classes at the same time and in the same direction this is the key finding of the of this first empirical part now from this as we said it's not enough this needs to start in march 2020 as i show you before on the graph and as a matter of fact it does so when we look at what happened here is that it goes from roughly zero this is the correlation between net flows in cryptos and inequities on this three-squot platform so this correlation is basically zero and then it jumps through the roof in march 2020 this is weighted by trading flows because here trading flows on crypto was close to zero so it was you know it was a bit noisy still very evident effect but a bit noisy here you clearly see because when you waited to buy trade you clearly see that all the effects comes really in march 2020 so what happens in march 2020 this is consistent with prediction two this is our interpretation we say okay these people had liquidity and attention shock now there is some literature showing that in the u.s. stimulus checks were used to invest in equities and in cryptocurrencies that's they had an impact on stock prices there's a paper by gringo de laritz and worker and also that a lot of people use this to buy bitcoin this paper by diva caroni that shows that the number of orders on the blockchain for 1200 which was the amount of the stimulus check spiked when on the dates people receive stimulus checks so a lot of people took the stimulus checks and invested in bitcoin now switzerland did not have stimulus checks but this still was a significant equity show because people were home they were receiving the full salary and they were not spending anything and switzerland is very expensive it's very very expensive so myself all of a sudden a lot more money on my hands and as a matter of fact i invested it with swiss quote uh at the time so this was a very common phenomenon and also keep in mind that at the time bitcoin was kind of up and coming so a lot of people kind of knew about it but they were not really familiar with it it's like i don't have money for a car but if i get a very good job on the job market i might buy a car meaning that i'm going to look into cars and gonna spend my money to buy a car it's kind of the same thing all of a sudden they had time and liquidity and they invested in this new asset class keep in mind that the spiking bitcoin price was not in march 2020 it was later on uh at the end of the year so here bitcoin is still around nine thousand ten thousand so the spiking price is subsequent um so now we are not we cannot be sure that the rest of the world works like swiss investors right so we have a sample that is localized in switzerland they tend to invest in us stocks uh the overwhelming majority of the trades are on us stocks so what do we do we look at global markets and uh we take the three thousand most traded us stocks and we rank them by the relative trade by swiss quote uh crypto investors so we just take the the trading the trading volume by swiss quote investor divided by trading volume on the stock and rank the stock so on the fifth quintile you have the stocks that are relatively most traded more traded by crypto investors and the first quintile your stocks are relatively less traded by crypto investors um keep in mind that crypto investors tend to trade on few stocks so the fifth quintile is where most of the trading is concentrated now this is the summary stats of what these quintiles look like so from one to five clearly is more tech is larger stocks and is growth it's growth stocks surprisingly also is healthcare stocks and this is well it's not surprising the fact is that retail investors tend to be attention driven so covid people heard about healthcare and they invested in it so it's very consistent and uh we run five regression so one on each sub-sample the dependent variable is the correlation between a given stock and bitcoin and the independent variable you have a few controls and the global volume on the bitcoin market what we find is that volume on the bitcoin market is positively associated with correlation between bitcoin and stocks and this effect is much stronger for stocks which are preferred by retail investors against those who are not preferred by retail investors now we are still looking at global bitcoin volumes here we're just looking at global markets so up there but there is a lot of volume on bitcoin which is not from retail investors so we add a second variable we had we had trading on swiss gold as a proxy for retail trading on on bitcoin clearly this measure is correlated with global volume but it's not multi-colinear because there are way more more things going on on the global market this is the result so as you can see the effect on bitcoin kind of fades away so it goes to zero and then it becomes even negative on the global market but then the retail measure picks up all of the effects and it becomes much more clean and much more linear across quintiles so this is just to help you visualize it this is the from first quintile to the fifth quintile this is the global bitcoin volume this is the retail bitcoin volume and this shows this is very consistent with the prediction of the model 3 and 4 so this correlation that we observe between stocks and crypto it's concentrated in periods where they are when retail investors are active and on the stocks they like which is our prediction 3 and 4 so to summarize this part cryptocurrency capture the tensions we set right so they kind of move the shift attention away from equities but they tend to trade both asset class at the same time in the same direction and this behavior really started in March 2020 now when you look at global markets what we find is that the stocks the retail investor who trade cryptocurrency like are the one who are the most correlated with bitcoin and are the one who are the most correlated with bitcoin when these people are trading so that this effect is very localized we look at the index okay smp is correlated with crypto but it's not really the case just certain stocks are and are these specific stocks now as i mentioned before this relies on hypothesis market segmentation in our model right in our model if we change this hypothesis i'm not going to go through everything again just show you the result which is the correlation becomes negative uh why because the market maker observes both order flows and is able to teleport is a more information is able to teleport the inform from uninform and is somehow trading against that inform and by doing this uh the correlation becomes negative so my point is this is an effect that we observe today given the market structure but in our model it's not necessarily the case tomorrow whether if it is going to be market integration so the takeaways uh three takeaways the first one retail investor divide the cryptocurrency correlation so if people think of bitcoin as a tech stock it will behave as such because of the market structure and because of uh what they show you so far right and uh i think that we have some compelling evidence that people do think of this asset as a tech stock because that's what they're substituting for in their trading activity so now clearly bitcoin is a bad diversification tool because it's highly correlated but during market tomorrow one should expect and sorry one should expect this to be even higher the correlation during market tomorrow so when retail investors are very active if you look at the bitcoin price over the past few months over the past year actually you see it was highly correlated in the spring where retail investors were panicking and slamming off and they uh and the correlation kind of flattened out during the summer until today because retail activity is extremely low uh now unfortunately all the data I showed you is 2017 to 2020 uh because we started paper uh before SwissQuote finished their annual report for 2021 so we had access to data to 2021 and 2022 we saw all the results hold and hopefully in a few months from now we'll be able to update the paper with the new data but still all the results I showed you is not just 2020s up to today and the final point is that for policy makers and I know there are a few in the audience this is a possible transmission channel so if you observe a meltdown on a crypto you should expect effects on whatever assets retail investors are holding because they panic they sell all the portfolio so this is a something that the thing is important to consider for for policy implications so thank you I'm the job market this year I've worked on digital money not just cryptos but a lot of central bank digital currencies and you can find my papers on the on my website and thank you for listening any question any email any comment really I'm more than happy and if you don't manage to to ask questions now please send me an email okay thanks thanks very much Luciano uh questions from the audience yeah so Russell go ahead yeah thanks uh yeah it is it is amazing data um so it one I have two questions one may be clarified question either maybe about the data so um so the story is a Kyle um market making story so uh it but one may be special about crypto it is uh there's no centralized market maker if you trade on chain so uh so I'm wondering so if people in your data if they choose to choose that in in your sweet bank so is there any premium uh they pay for the crypto compared to other the centralized market so that might be uh that might be I don't know whether it's evident or not to the to the Kyle story right if if there's a premium that may be a supportive to to the uh to the Kyle story so uh the other question is about uh the data so uh you uh so I I guess we observe the individual trade level data so every transaction every return every rebounding we can see so I'm just wondering whether there is a timing of of the of the trading so um maybe the timing they open a crypto account and door you know as well maybe II my portfolio uh crash so that's why I'm open a crypto account and move into the crypto space so then in this case it might change a little bit of a question you're running because at the time it and don't you think you should become a crypto guy thank you should I ask you directly or take all questions and go yeah let's try and answer directly okay yeah so the market making is true it's it's it's a very complicated market making on the crypto market and now for the Kyle if there is one market maker or a continuum of small uh risk neutral investors it's equivalent so these are things kind of save the day on the on the crypto side although there are uh it is it's it's developing for our for our setting Swiss quote is not doing market maker market making on equities right there's giving the trade to all to all their platforms they're doing some market making on on cryptos although we're not fully sure all the details on that side so they're not they're not the the kind of unified market maker that I have in model right so it's quite clearly isn't but again I think that there is some there is some good insight because there is additional information from the two flows and I don't think it's impossible in the future to think okay observe a certain flow on tesla but also several flow on bitcoin right so maybe this flow is uninformed I don't need to adjust my price as much um this is first on timing on trading Luiseno so one clarify so uh maybe maybe a related uh point is also ah do we see the price the people pay for the crypto on the on the exchange are very different from when they trade when they trade in some decentralized exchange so if there's some premium that might be signalling something about the information so they are trading on our they're trading with this quote on the on this quote platform right so we don't see what they would pay on all the platforms they in our that I said they can only trade with the bank uh their cryptos right and uh it's I I'm pricing they say they're the best in Switzerland I I don't know but they cannot use these tokens for the centralized applications that that for sure in this setting right they in order to use for the centralized application need to take the tokens take the amount of the platform and then maybe go on some other platform they can do it we cannot see it unfortunately in in this setting sorry I trouble unmuting Ramo yeah thanks for the great presentation so I think the story seems very legitimate to me I remember another presentation where they kind of showed that with the introduction of the futures market jicago futures bitcoin futures that this kind of was a leading indicator for maybe I was kind of the lead for the price that kind of troubles me a bit with with this story now I don't know whether you know something about that and you have something to say about that whether it's kind of disappeared again over time because I don't remember the paper properly but if kind of the jicago futures bitcoin market should be kind of leading indicator um that kind of will be a bit contradictory to to use story right or maybe yeah I'm not 100% sure I got the comment uh but I mean I think that here uh people think about bitcoin think about much more sophisticated actors usually right so you have you know the future etc etc etc here we're really talking about the average joe in switzerland who has no idea about anything we doesn't even know that probably the fees he's paying for trading on this market right and what we see is that all this correlation is really is really concentrated on this kind of investors um then of course I mean you have all of the effect that might affect this thing in the future but I think that at least so far the evidence is quite compelling that this this subset of investor is is really driving this correlation I think there is a Catherine Catherine yeah yeah so I would like to come back to your policy conclusions so because you're saying and I see uh how this comes up so the retail investors are driving the correlation and you say therefore it's policy relevant so typically central banks are interested about uh what banks are doing because they are their customers if they come into problems and finishes stability are threatened so with retail investors and at the beginning you said that these are guys that set her up away some 2000 francs and are trading on these 2000 francs so I see that they are driving a correlation but um they wouldn't threaten the financial system in my view so of course they they drive a correlation between crypto and um these tech stocks but as long as banks are not engaged in crypto so I don't really see why a central bank should be worried about this correlation I think it's a fair point my mention of policymaker is that I got policymaker reaching out and asking please can we use your data because it's very interesting and we're looking at impact on financial markets from these people so I agree that it's not maybe central in the sense right if Tesla collapses because crypto collapses maybe you don't it's not particularly relevant but I think that the overall mechanism and saying okay right now we understand there is this further channel here at least from conversation from people from Bank of England for instance they were very interested in this and might want to look more into this kind of mechanism I'm not arguing that this result can automatically lead some kind of regulation or anything peculiar but I think it's kind of helps the overall understanding of how these markets are linked because now there's a lot of focus on stablecoins rightly so right because this transmission channel there is self-evident they have all this backing in real markets here I say okay there's a further channel here which is if you have some kind of meltdown on crypto then you should expect at least probably some kind of heavy selling activity on certain but have you really established causality or are you looking at correlations I'm looking at I'm looking I'm trying to establish a reality in in a follow-up paper that I work on right after major market that's in my pipeline right now I'm trying to establish causality of I have some shock here I want to see what happens there with kind of cousin that that's part of my research agenda in the in the short term so maybe something related to the trading dynamic we talked before maybe related to catch report maybe one mechanism it may be the fire sale mechanism or maybe the the the household they they are crash in the crypto space and there it will be we'll trigger the holder we're also selling or thumbing the stuff in the in the traditional center so if we can see some this kind of trading dynamic maybe the one space in crypto can spill over to other stocks so that might be a one mechanism that you can see yeah I think this mechanism I see the mechanics so what I'm doubting is the size of the crypto market is causally driving a decline in the tech stocks that's that's that's my point um I think that's the impact of retail investors used to be very neglectable which is true but over the past three four years it did if you look at volumes by retailers went from 55 percent to 20 30 percent so and especially when they focus on certain stocks they can have a very heavy impact so because right now if I say okay retailers have an impact people say not convinced but if I tell you game stop do you think they had an impact on games so specifically most likely yes right so in certain periods when they have a focalized activity they can really have an impact and especially when you have this kind of huge equity impact like we had in covid then you should expect them to have an impact they're also growing literature showing that you have very inelastic institutional investors right so we have a very inelastic institutional investor and are faced with these flows from retailers maybe there are not many but they they can have an impact because everything else is quite rigid this paper by um van der ken Jonin showing this in 2020 with robin hood data so I think it's kind of a new phenomena and I don't think it's also all the time also again the the direction of this thing is not super clear because you can say okay maybe you know what happens on the stock market as an impact on crypto which is very important but not the other way around that's also possible they see Tesla collapsing they sell everything right so the causal link goes the other direction so that's why I focus the paper really on cryptos because I think that retail investors on cryptos they plausibly have a large impact and also everything you observe is very consistent with the price movements in crypto market or past two years very very much considered especially over the last year um while the other way around I'm not sure I get about the answer I hope I'll have a better answer in a few months with the with the follow-up paper so I had one general question uh about using the kyle framework to think about pricing bitcoin and in particular thinking about how things change in your results when you remove the market segmentation I mean I guess the question is who are the informed investors in bitcoin like when we talk about pricing stocks right I kind of know somebody has information about future cash flows absolutely I have difficulty translating that over but how do you think about that I think I can make a very concrete example um so bitcoin as we said the fundamental is derived at least in in theory we have so far on transactional values you can use to pay things right so when tesla for instance uh when tesla for instance said okay we accept bitcoin for uh buying cars this is a huge impact on bitcoin but tesla been building up the position in bitcoin before saying it and vice versa they sold their position then they said oh we don't accept it no longer and the price collapsed it was a clearly to me an informed trade right if I do the same trade nobody cares if Elon Musk does it or you know if you think of some black rock does it uh or jb Morgan does it that's the informed trade because by them buying it they automatically change the fundamental value of the thing because it it's more likely a general adoption of the of the currency if you have this kind of big players inside so they wasn't thinking of when I say informed so I agree that the kyle it's easy super easy on the stock market I think it's harder uh on the crypto market I think there are good arguments because again if you don't have a market maker but you have a continuum of a small literature it's the same informed I think that at least in this one example it's clear what it is and there are all degrees of these um I think it's I think overall it's a good model for it because it's very easy in the in the mechanism um and it really highlights the specific mechanics I wanted to I wanted to show in the paper okay I agree that's a very nice example I hadn't thought of that um okay is anyone else have any final questions they want to ask uh okay I guess we're right at the end time so we'll go ahead and wrap up so so thanks very much to the presenters thanks for everyone for attending and for providing feedback please feel free to reach out to any of the presenters about what they've presented or anything else and please take a careful look at their applications as they as they land in your inboxes uh finally let me just advertise our next session is on November 18th slightly earlier in the month because of the Thanksgiving holiday in the U.S. Tony Whitehead is presenting Will's central bank digital currency disintermediate banks I see um that there's a new version on SSRN that was just uploaded today and we've got Itamar Dreschler as the discussant so I think that's going to be a really interesting session okay thanks everybody and we'll see you again next time