 Welcome, everyone, and welcome, Michael. This is Michael's second time at the Institute. He was here four years ago with his IMF hat. He worked in the IMF for almost 15 years. No, IMF 11, just under 11 years. And as you know, he's now with the Bank of England. Fascinating presentation today on the role of the banking system and the two different views of the banking system. So, he's going to talk for about 30 minutes and we have plenty of time for Q&A after that. So, Michael, can I hand over to you? Thank you. Thank you for the introduction. This is joint work with Soltan Jankaf, who is my home colleague at the International Monetary Fund. He has a body, your own working paper in May last year. So, here's the usual disclaimer that by Soltan, both the Bank of England and the IMF, this is pure research. Introduction. So, following the global financial crisis, all the central banks have been working on making up for lost time, essentially, by putting banks and financial markets back into models, especially for what is now known as microprudential analysis, the use of prudential tools for macroeconomic purposes. And I've been very involved with that at the IMF and now at the Bank of England. And what I will claim here today is that these recent models use what I would call the intermediation of the global funds theory of banking. This is a term I made up, but it has the two terms in it that you will find all over the country that banks have intermediaries and that what they intermediate is global funds that have been deposited with them. And I will argue that this misrepresents the way that credit and money is created in the real world. I will actually not talk about history very much, but we mentioned it over lunch and we can get back to it during question and answers. There is a very long intellectual history that makes the same point that I'm making here today and that current of thinking had actually won the debate by the early 1930s. And it lasted until about the early to mid 50s and then this intermediation of global funds theory took over sort of by stealth partly also because banking disappeared from the radar screen of the macroeconomics profession almost completely for many decades after that. And obviously we can no longer afford that. So the solution to what I call the problem with the intermediation of global funds theory is something that I call the financing through money creation theory. Again, I made that term up, but it essentially tells you what banks do. They finance new loans by creating new money. And this is a fundamentally different story and you will see that it has fundamentally different implications of how you think about financial shocks and how they are transmitted to the real economy. I will argue that this is consistent with the actual credit creation process. There is a very nice, there are two very nice Bank of England quarterly bulletin articles by my colleagues with Ray, Radia and Thomas that explain this really for the layman. You don't need even an economics degree or any kind of higher degree to understand what they're saying in those articles. But it is basically exactly the story that I'm going to be telling you here today, but in much more technical terms. And a little side story there, and I met John Vickers at the conference. He told me then when he teaches money and banking. He uses a standard textbook, but when it comes to the credit and money creation process, he asked the student to put the book aside for a bit and he uses these articles instead. That should tell you something. Before launching into what I think is wrong with the standard view, I will tell you what I actually do in my formal modelling. Let's not think about so much the formal modelling, but about what story about the real world that actually tells you. I grew up as a student of Guillermo Calvo of a very famous, grew up as an economist. As a student of Guillermo Calvo is very famous in international finance. And the class of models that we used in the classroom there were the standard models of the 80s and 90s where you would have money in the models. It would be exogenous government money and this money would be demanded by some representative household because it derives utility from it. Because there's a cash advance constraint or because it reduces his or her transactions cost. So that was very common. Then Michael Woodford et cetera took off and money was banished out of the models altogether. It was all just about prices including interest rates and not so much about quantities or not at all in many cases. The argument that I'm going to be making here essentially is that the main shortcoming of this old folk family of models is not that they use a reference. It is that they use a representative household, but rather that they look only at exogenous government money. Because the vast majority of money in today's economy that is actually relevant for the macroeconomic picture and for people being able to engage in transactions is created privately by the banking system. Not without the central bank having an influence, but that influence is indirect. You need to look at both what the central bank does and what the private banking system does. This is a little maybe too technical, but let's just focus on the intuition. In the loanable funds model you would have something like a saver and a borrower. The saver deposits something in the bank and then the bank takes it and lends it to the borrower. This is because typically the saver is patient, the borrower is impatient, something like that. These guys or these people would have budget constraints which say the change in my deposit is equal to my income minus my spending in very general terms. Let's just look at the saver for the borrower. It's basically the same story but in reverse because he's not saving his borrower. The saver accumulates deposit as a result of generating extra income for example by sacrificing his leisure and working or by spending less for example by just consuming it. This by nature because it's a physical process, it's slow. People would not just work 24 hours one week and zero hours the next because of something that happens in the economy. It's a slow process and therefore the change in deposits and the flip side of that on the other side of bank's bank sheet, the loans is therefore generally slow in this family of models. Plus if you look at how I've written this down, this is all entirely real. Income spending has nothing to do with money whereas what banks do is essentially monetary. In the family of models that I'm going to be presenting instead I put a representative household that interacts with the banking system and forget about this capital item that's not essential to this debate. The representative household would have income and spending but what is on the left hand side of his budget constraint is the change in deposits minus the change in loans. Meaning that if a representative household wants to have more deposits from the bank because by using deposits I can buy and sell machines, I can buy and sell cars and this is useful for my economic transaction. And if I need more I go to the bank and have the bank create this additional deposit for me and I do that by getting a loan let's say a loan of a million euros and the deposit of a million euros and that deposit is then useful for transactions. I can spend it but it'll also come back to me, it circulates through the economy. That has no longer anything to do necessarily with income and spending. I can change my deposits by getting an additional loan and nothing in my income or spending as a first approximation has to change. Which tells you that in models like that the changes in balance sheets can be very much more rapid and as I will show you in the data they are very rapid. And so this is an important feature. So more on some general insights about banking. So this is basically I just told you what am I going to do in the family of models that I'm going to use to and I'm going to just in the end produce one simulation that tells you what happens in a financial crash and how do these two model classes behave under those circumstances. But first some more general thoughts. Understanding banks. So two things that we need to get out of the way. First banks are not intermediaries of loanable funds so that is related to the intermediation story of banking that's out there. The other one is that the deposit multiplier is a myth because the deposit multiplier is another and distinct story out there about what banks do and it is very very you know I taught undergraduate money and banking for several years and it's all over the textbooks. Okay so the first one banks are not intermediaries of loanable funds and this is a little dense but it's on purpose because every sub-bullet here corresponds to a sub-bullet down here so that we can compare the two ways of thinking about banks. In the loanable funds model the postulated credit process is essentially that intermediation consists of the physical trading of real resources. We go to great lengths in our paper to show you that if you look at the book entries that you need to create in order to make sense of an intermediation model it cannot be anything to do with what we think of when it comes to banking like we deposit a check in the bank. Because when I deposit a check in a bank that's drawn on somebody my employer or something gave me a check that is drawn on Barclays Bank and I deposit it in my hundreds bank account in London this is not creating any new deposit. It is just moving a deposit from Barclays Bank to hundreds bank and hundreds bank also doesn't have any additional money to lend when it gets my check because it has already lent it out the moment it gets it because it has an accounts receivable on Barclays Bank. So there is no additional deposit being created here and the only way that you can think about intermediation where there would actually be additional deposits is what somebody would literally have to drive up to the bank with a truck and deposit some grain and tell the bank to intermediate this. Literally in terms of bookkeeping nothing else makes sense. So in this model banks collect a deposit of real resources from a saver and lend them to another agent the borrower deposits in this model are there for an input deposits and loans out. And the money in this model is essentially held as a store of value. We know about the hierarchy of the three functions of money store of value unit of account of medium of exchange it's a store of value story. And in order to explain rapid changes in credit in this class of models you would have to have people saying oh I don't I no longer want to hold my savings in the bank instead I directly want to hold bonds and stocks shares. And then the bank balance sheet could suddenly shrink because people use their bank deposits to buy shares from their own bank and then hold those shares and bonds directly. I will show you in the data that that's just not what happens that's just not the major thing behind real big changes in balance sheets instead it's what comes next. So in the financing model financing is the digital digital creation of monetary purchasing power that happens on a computer. There's no track involved. It's all just digits on a computer. So when when somebody came to work for Barclays Bank for five years so so if somebody came to me into the bank and it was a good business plan and I proved alone I would create an entry of a million pounds on the asset side of my bank's bank sheet. And at the same time create an entry in the name of the same person on the liability side which was the deposit that this person came to the bank for and he or she would then take that deposit and spend it on the actual purpose of the loan. So at the moment when a new loan is made there is no intermediation. It's just all between the bank and one person and then people would tell me yes but you're ignoring the second step. The second step is that then that person would go and take that money and spend it on something and then the depositor ends up being someone else. At which point I say yes but you are ignoring the third and the fourth and the fifth and the sixth step which is that once that money is then received by that other person it would then be spent on the goods that I have produced using that money in the first place. So let's think about an example. I'm a company when I come to the bank. I want to pay additional workers to put additional goods on my shelf. So I would get a loan from the bank. So as I said no intermediation. I have a deposit I have a loan. Now I spend the deposit by paying my workers. So now the workers are the depositors and I owe my workers. So what's the next thing that's going to happen at a very high abstract level? The workers are going to spend the money in order to buy the goods that I have just produced. Like macroeconomically that's what happens. And the money comes back to me. So it doesn't matter that at a particular snapshot moment in time the depositor and the borrower are not the same person. What matters is that when the new loan gets created there is no intermediation. There is money creation. Deposites in this model are there for an output. The loan is made in order to create the deposit. Loan in deposit out. Money in this model is held essentially as a medium of exchange. I do this whole transaction. Why would I ever go to the bank and get a loan of a million pounds and pay 5% interest on it and get a deposit that pays 2% or 0% interest. I would do this because the deposit is our generally accepted medium of exchange in the economy. That's why I do this. And then as we will see when we look at data when you think through what banks do through that lens then you can explain rapid changes in bank balance sheet by banks grossing up or down their balance sheet in response to macroeconomic conditions. So what you will see when you read a lot of academic papers and semi academic papers the standard story for banks is always there's a saver, there's a deposit, there's a loan and somebody else gets it. What I'm arguing is that the arrows actually go the other way around. The investor gets a loan in order to get a deposit. He or she then uses that deposit in order to do something with it. This by the way only holds for banks, not for non-bank financial institutions. Non-bank financial institutions are actually much closer to this up here. Except that what they intermediate is not good, what they intermediate is bank balances. The corollary, this saving does not finance investment, financing does. I think that's very important when you think about productivity of finance etc. There are two very nice papers by a German economist called Lindner who argues that aggregate financial saving is sort of a zero sum game. If I decide that I want to do additional saving in the bank, for example when my employer pays me, I'm going to save more than I used to. I'm going to keep that money in the bank. That does not increase the aggregate amount of deposits and loans in the economy. It just means that I keep a larger share of what already exists. The only way that the aggregate balance sheet of the financial sector can grow is by additional financing decisions, additional loans that create additional deposits. That's financial saving, real saving, and this goes back all the way to Keynes and probably earlier than that, is equal to investment as an accounting residual. Saving is equal to investment is not an equilibrium condition. You cannot say, oh, people should save more so that there should be more investment. The way you ought to think about this in my view is that there need to be conditions that are conducive to productive investment. One important condition for that is a banking system that finances productive investments and responds to market conditions to do so. You cannot say that saving finances investment in that sort of mechanical sense if I save more then there will be more investment. The logic is actually that if I finance additional investment, if the bank has good loan officers, they spot a good opportunity, the firm goes out and spends the additional money that has been created, then saving in the macroeconomic real sense is just a consequence. Saving is equal to investment. That saving is just an accounting residual and it has to be equal to investment by definition, not by equilibrium. It's a definition. So macroeconomic saving is equal to investment in a closed economy. Of course there are complications in an open economy. Saving is a consequence, not a cause of bank lending. The deposit multiplier is a myth. We all know about the deposit multiplier from the undergraded textbook. The Nobel Prize winners Kittle and Prescott showed in 1990 that actually the actual monetary transmission works in the opposite direction in the sense that broad money which is created by banks leads the cycle and narrow money which is provided by the central bank, very narrow money, lacks the cycle. So the banks decide first how much broad money they are going to create and the creation of reserves is a result. Under inflation targeting this is actually obvious. Almost all central banks nowadays do some version of inflation targeting. You're basically controlling in a market with supply and demand curfew, controlling the price, then you need to let the quantity adjust. And the quantity in this market is the quantity of reserves. So the quantity of reserves is therefore endogenous. So in just one sentence, this comes from a 1969 paper or speech by Alan Holmes, vice president at the time of the New York Fair reserves, reserve in the real world banks extend credit creating deposits in the process and look for the reserves later. So it goes from loans to deposits to reserves. And I will now show you that this matters when you think about macroeconomic simulations to especially credit shocks. So we build models, I don't want to go into too much technical details but I want to say some basic things about the models at a high level of presentation rather than going into equations and things like that. So I'm talking about bank balance sheets actually being explicitly modeled with loans on one side and deposits and equity on the other side. So what is the distinctive feature about bank assets? Well it's what we just talked about. Bank assets get created and funded, not by lending out pre-existing funds but by creating money. There are no loanable funds and this is something I say typically to provoke academics a little bit into thinking when it comes to the loanable funds theory as far as banks are concerned. Again, non-bank financial institutions are different. When it comes to banks there are no loanable funds. When I sat in my chair as a bank place lending manager I did not look over my shoulder or into my computer to say do we have some loanable funds? No, the funds existed entirely in my mind. I said if this is a good business proposition I'm going to create these funds. That's how that works. As far as liabilities are concerned I will basically say that they do something very, that they do something essential for the economy and that's where the link to the real economy is. They do this and create money and that money facilitates transactions. It makes it cheaper to buy and sell stuff and that provides a stimulus to the real economy when they do that. Again, in order to debunk a very common statement that you will find in literature that says banks collect deposits and then lend them out. In a macroeconomic sense that's not right. Banks do not collect deposits from non-banks. I talked to you earlier about that when I deposited my cheque drawn on Barclays Bank in my Handel's Bank account that is not a bank collecting a deposit from me. It's a bank collecting a deposit from another bank because as soon as I deposited it the value will be collected from Barclays Bank, not from me. So I do not put anything else into the banking system by depositing this cheque. In that sense banks do not collect deposits from non-banks. Bank equity will be modeled as subject to Basel regulation and I think I can jump over the rest. And then I will present models that are, the only point I want to make in this context is models that are identical except that one difference that I showed you earlier with those budget constraints that in one case the bank literally has to go to one set of agents collect savings and then lend them out to the other set of agents whereas in the model with the representative household there is one bank dealing with one agent who came to the bank in order to get additional money created for him or her. And that's the only difference. Everything else in the model is kept the same so that when we see differences we know where they come from. They can only come from one source. And so this is kind of the heart of the intuition that's going on and it looks a little messy but I think I wanted all the essential variables in one place to tell a complete story. So what we have here is a shock where the banker wakes up in the morning and all his borrowers have suddenly become a lot more risky. It could be some political event, whatever. Or in some country where there is a major storm and suddenly a lot of businesses are flattened or something like that. So that's sort of the story that we want to talk about here. So the borrowers have become more risky and this is not just in the mind of the banker in this model. This is actually real. And then we have here how does the macroeconomy respond and on the horizontal axis in each case we have 16 quarters i.e. four years. And this is highly illustrative. Don't look at the magnitudes too closely because it's not based on any kind of estimated model. It's an illustrative simulation. And in the right column here we have the financial variables in the economy and in the left column we have the real variables and I will tell the story starting with the financial variables. So this banker wakes up in the morning and says my borrowers are suddenly a lot more risky. I need to adjust. In which case he or she can do two things lend less or lend at a higher interest rate to compensate for the higher risk. The dash line in each case represents a loanable funds model where people have to save or dis-save in order for something to change with the deposits. The solid line is the financing model. With the financing model bank deposits change over time but very gradually we have to start dis-saving for the banker to be able to do something. So the banker who nevertheless sees elevated risk charges a much higher risky spread for his loans to his borrowers and so there is a big increase of 200 basis points in this particular case in the price of loans. And then also we have leveraged, the bank leveraged in this model is counter-cyclical meaning that leverage goes up in a crash. And what we know from recent empirical literature is that actually leverage is proscyclical. It tends to go up in a boom and down in a crash. Why does leverage go up in a crash in this economy? It's because there is a lot of action in bank networks. It declines. The bank makes losses here because there's been a bad shock to the quality of the loan portfolio. Bank takes a hit but not much happens with assets and liabilities at least on impact. It takes a long time for something to happen with assets and liabilities. That means because leverage is assets divided by a net worth and net worth goes down, leverage goes up. Okay, now as I will say later in the data leverage is actually proscyclical like in the financing model. What happens in the financing model is that the banker wakes up in the morning and again says I want to lend less at a higher interest rate. But in this case lending less is not a problem. Lending less is one of the things that he or she can easily do by calling in loans as they mature. And so here the bank balance sheet contracts by 5% in a single quarter. Sounds like a crazy magnitude but actually later we look at the data, it's not that crazy. And that means that the loan to value ratios of the remaining borrowers are now done because the banks are lending less to them which means that they don't need to increase the spread by quite as much. So more action on the quantity side, a little bit less action on the price side. Now if we go to the real variables now the key link between these two things the financial and the real variables is something that I call the effective price of investment and consumption. And what that measures is the ease of doing business as a result of having enough money circulating in the economy. When you suddenly withdraw a lot of money from the economy people can no longer do the purchases that they thought they would be able to do. Think of small and medium sized enterprises in the US for example after the crash of 2008. They were screaming they couldn't get credit therefore they couldn't get money therefore they couldn't purchase what they wanted to purchase. So in the financing model what happens in this crash is that the effective price of consumption investment goes up because there is so much less liquidity as you can see here bank deposits liquidity decreases a lot whereas in the other model that's actually not what's going on at least not on impact. And that's why you get a much bigger contraction in output, in consumption and in investment it's about twice as large in this particular simulation because the action is not just here through the price channel through the real interest rate. Although that action is also there but it is also the lack of money so to speak in the economy and that makes everything worse and that's an aspect that is completely ignored by the typical loanable funds model. So that was, there's a lot more in the paper but we have only so much time and remember some of what we discussed here because now we're going to look at data. So there are three interrelated predictions that we have seen in this simulation. First, credit and money can exhibit large and discontinuous jumps. Second, bank leverage is prosyclical and third, credit pressures have a large credit rationing component. There's evidence for all of these in the data. I don't really have much time, any time to talk about three, maybe a little bit about two, mostly about one. So here we have scatter plots for six national banking systems that plot on the horizontal axis the change in the log of assets, ie the percent change in assets against on this axis the percent change in equity which are the blue dots and the percent change in debt which are the red dots. And as in a famous paper by Hyun Shin and co-authors what you see is that the red dots are on something close to a 45 degree line. That means that changes in assets are matched more or less one for one by changes in debt. And this is not just for individual banks but for national banking systems. Now you can struggle mightily and some people have done so successfully using a loanable funds model in order to create this effect. It is possible but you have to struggle hard and our model is not even an issue. That's what happens naturally because when you increase loans and increase deposits at the same moment as the result of the same bookkeeping transaction. Also what you should look at is the magnitude and you might not be able to see that from the back of the room 4, 5, 6% in a given quarter is not that uncommon and this is national banking system balance sheets and this even holds after you take out valuation effects and it even holds after you take out interbank lending this also holds for just credit to the non-bank private sector. Let me skip over this because time is a little bit short. Now let's look at the magnitudes of bank financing in the United States and how they change and which model might line up with that. So what we have here is the change in millions of US dollars in various categories of credit between the early 90s and 2012 and the black line is overall credit and you see that there was a huge collapse in credit at the time of the financial crisis in the United States. On the previous slide we also showed that if you plot saving against this national saving from the national accounts it actually goes the opposite way which would tell you people save more there must have been more bank deposits doesn't make any sense. What happened is that banking systems compressed the assets and the liabilities were compressed together the loans were called and the deposits cancelled I mean I'm talking about loans and deposits in a very general sense in a banking system essentially. Now what some authors have claimed is that oh this must be because there has been a substitution to a different form of non-bank financing for example holding bonds directly which are these blue bars and the magnitudes are the same there was not very much of that there was a little bit here but no this was not a switching between different forms of financing it was a contraction in the overall amount of financing so our model has a very easy time explaining this whereas the standard model struggles here is also we add this time we just look at corporate financing which was one of the lines that declined rapidly on the previous slide and then we look at bond financing and here is equity financing because you might say ok maybe the firms didn't issue additional bonds to households they issued equity you see that that's also not part of the story it does not offset what happened to credit at the time of the crisis and finally we look at the contribution of households this is from the flow of funds you can look at the contribution of different sectors to this change in corporate financing where is the household sector involved in there that would have to be if you tell the story in terms of the loanable funds model where the household decides am I going to invest in the banking system am I going to invest in stocks and bonds and we see that no the change in the households position either equity or bonds vis-a-vis the corporate sector was small to negligible compared to the overall decline in the financing that happened in the financial system leverage is pro-cyclical banks lever up in a boom and down in a crisis because their assets and liabilities move even more than their equity and so Nwnion Thomas have a very well-known paper that looks at this and they basically look at the correlation of leverage with GDP and we do the same but we look at also at labs of GDP because banks do not always immediately react to a big credit event because they have things called pre-committed credit lines whereby borrowers in a crisis might actually draw on pre-committed credit lines even though the banks would rather wish they didn't because they would rather get out but legally that's not possible and it takes about a year or something like that until you can cut those credit lines and so if you allow for labs of output then you see that consistently across countries and we show that in the paper leverage is pro-cyclical so conclusions the key contribution of this paper was to just put a framework on the table by which we can think about the financing view of banking which as I indicated briefly at the beginning was very common in the 30s, 40s and 50s. This was an older professor at Meritus in London told me that when I grew up as an economist what you're telling us was conventional wisdom right? It was a long time ago but I'm basically trying to put a modern modelling tool on the table that can tell this story in a consistent way and what we found when we do that by quantitatively simulating such a model it predicts far larger and far faster changes in bank lending which arguably you see in the data it predicts much smaller changes in spreads whether you see that in the data it's sort of hard to we're actually doing cutting edge empirical work now not just looking at start-up aspects of these models. It also predicts larger effects on the real economy and the stylised facts we have just seen I'm involved at the Bank of England in sort of coming up with research projects that look at the interaction of banks with the real economy especially for micro-dentro policy analysis and virtually the entire literature uses some variant of the loanable funds model and what I'm concluding from this work is that much mileage can be gained from looking at the questions that we over the last 5, 6, 7, 8 years have already been studying again through the lens of this different model class and I think it will give us certainly quantitatively different answers in some cases perhaps also qualitatively different answers so there we go Thank you