Friday, 13 March 2015

An Example Case for Modelling Banks: The Rise and Fall of Market-Based Finance

In my last post I asked whether we needed to include banks in macroeconomic models.  My conclusion was that we might only need to do so if we wanted to explicitly model the way things like regulatory issues might shape the terms on which credit was extended.

As an example, I have looked here at the sort of developments that we might be able to represent and map out within a model.  This relates to shift between traditional banking and market-based finance that, in my view, played an important role in the financial crisis.  We could construct some balance sheets and behavioural rules and develop this into a model with a list of equations, but it's a lot easier just to look at some balance sheet movements.

The following draws on my distinction between traditional banking and market-based finance, as described in this post.

Shift to Market-Based Finance

The first development is an industry trend towards greater use of market-based finance.  The main driver within the business is the more attractive capital treatment.  At the same time, investors facing large exposures on unsecured bank risk are creating greater demand for collateral that can be used to back deposit substitutes.

The process takes time, but accelerates as deeper market makes assets more liquid and therefore more effective as collateral.

The diagram below shows the changes in the balance sheet of the financial sector.  Loans are securitised and used to back issues of secured funding, with deposits repaid in the process.  We can map this by refering to different capital costs and funding costs.

Balance Sheet Expansion

Liquid securitised loans can be held with a lower capital requirement, so the trend to Market-Based Finance leads to a reduction in capital usage.  However actual capital levels are unchanged, leaving excess capital in the industry.

Rather than repay capital, banks try to use it by expanding their loan base.  This requires relaxing credit criteria and reducing prime lending margins.  (Average margins on new business may go or down, because of the increased share of sub-prime loans).  Again, we can model this by making some assumptions about bank behaviour or deriving results from assumptions about profit maximisation with frictions in capital issuance.

A fall in prime lending margins on new business increase the book value of existing securitisations, as these are on mark-to-market valuation.  This enhances bank profitability and contributes further to bank capital.

New loans create new deposits and the sectoral balance sheet grows as shown below.

Liquidity Crisis in Market-Based Finance

As average credit quality and margins decline, the system becomes more fragile.  

If certain securitised portfolios cease to be acceptable as collateral, financial institutions dependent on secured funding may be forced to sell these assets.  This further reduces the liquidity of the assets, which further reduces their use as collateral.  It also reduces their market value.

Banks are forced to take assets back onto the traditional banking book incurring a higher capital charge.  At the same time mark-to-market losses reduce actual bank capital.

These developments are shown below.  Unsecured funding is repaid, expanding the stock of unsecured deposits.  The value of securitised portfolios is written down with a corresponding reduction in capital.

Banks now face the following:

- Reduced actual capital levels
- Higher required capital
- Increased funding costs due to increased reliance on unsecured funding

They respond to this combination of factors by cutting back on lending.

All of this can be neatly modelled in a stock-flow consistent modelling format, although the number of equations required is quite large, as there's actually quite a lot going on here.  This can be a useful exercise, as it helps form of sense of how these things play through and why they matter.  However, I think it would be extremely difficult to use this to make forecasts.  We can examine history and maybe also track and analyse current developments, but working out where it will go next is a different matter.


  1. Nick,

    For banks, I'd say that it's more than regulations because bankers' own animal spirits matter and then may not extend credit even if there's no shortage of any sorts.

    Independent of this, about the question on whether banks have to be introduced in the model, I'd say it is necessary if the private sector is deconsolidated into households and firms. A separate banking system also needs to be introduced. I think it's because of the paradox of profits. There's no paradox of course, but I think it leads to logical inconsistencies if there's no separate banking system.

    1. Certainly I agree that animal spirits matter, including those of bankers. The question I was trying to look at was whether it makes sense to model banks explicitly or just leave them implicit. When it comes to animal spirits, there is not much we can model. We just put in some exogenous factor and label it as that. Generally, providing we have financial assets in the model, I don't think it then makes any difference to the behaviour of the model if we have banks in explicitly or not. In other words, a model with banks, where bankers are drive by animal spirits in their lending, will give the same results as a model without banks where households are driven by animal spirits in their lending. We might include banks to show how it relates to the real world, but it doesn't change the results. That was what I was trying to get at.

      It was a response to people who say we need banks in models, but then don't have a good idea of why it might make a difference. So they often fall back on explanations based on the quantity theory.

      I think you can deconsolidate households and firms and lump banks (and other FIs) in with households. Theoretical banks are not involved in production at all, they just deal with redistribution of income. You can just look at interest paid by firms on loans - you don't necessarily then need to break that into interest on deposits and bank profit - although of course, you may want to, if you think it affects behaviour materially.

      I'm sorry - I'm not sure about the paradox of profits point - I'd have to look at that more.

    2. Nick,

      What I meant when I mentioned the paradox of profits is that while there's no paradox and the issue can be dismissed by simple accounting, it is difficult to do so without an explicit banking system.

      At this moment my point is a simple hypothesis and I will try to show it sometime.

      Wynne Godley in his book Monetary Economics (the second edition which has a second preface by him) says this:

      “However, to complete such a system of account, so that ‘everything comes from somewhere and everything goes somewhere’ , it is essential to additionally include all those flows of funds which show how each sector’s financial balance (the gap between its total income and expenditure) is disposed of. It will then be found that it is impossible to complete the implied matrix of all transactions in such a way that every column and every row sums to zero without calling into existence a banking system which provides the funds which firms need in order to finance investment, thereby simultaneously creating the credit money which households need to finance transactions and to store wealth”.

      I don't think it's proved anywhere in the book but an interesting point.

    3. Consider model SIM in Ch 7 of Monetary Economics. In this model M always equals L and rm always equals rl. I could therefore do away with M and just replace it everywhere with L, and likewise replace rm and rl everywhere with just r.

      Doing this would eliminate any explicit mention of banks - it would simply look as if households were lending directly to firms. But the results of simulations would be unchanged. So having banks (explicitly) in this model achieves nothing. You might say that it needs banks implicitly and I'd be inclined to agree, or that it makes it easier to relate to the real world, but neither means they need to feature in the model.

      The same point applies to model DIS, It is only when we come to model INSOUT, that we start to get assumptions about bank behaviour that matter for real results (through the loan rate charged to firms) and that would be difficult to replicate without explicitly modelling the banks.

      I'm not sure what is meant by that quote, but it doesn't sound correct to me.

    4. Nick.

      The initial models such as SIM have simplifying assumptions that there's no investment and no retained profits and so on. There's not even profit in SIM. It's not a realistic model of the world. (Of course not wrong but illustrative and a first step toward creating realistic model).

    5. OK, but I think I could extend SIM or DIS to include things like retained earnings, whilst still having a model where the inclusion of banks was not necessary - in that I could take banks out without changing the result (it would have to be a growth model to include retained earnings). It's only when you start to include portfolio preference for banks that differs in some essential way from the portfolio preference of households that their inclusion makes a difference.

      Maybe the clue is in that word "implied" in the quote, in that you don't actually need it explicitly to complete the matrix, but if you were to ask what payments lie underneath it all you would need to bring banks into your explanation. I have no problem with that.

    6. Nick, two arguments and a question, also with reference to your previous post.

      Fist, as people like Steve Waldman point out, deposits are generally owned by other people than those who initiated their creation by taking loans. And as deposits cannot be destroyed by people who did not take loans, this creates a “hot potato” effect, as these people can only get rid of these deposits by spending them, partly on other assets, which will lead to asset inflation. This is quite different from loans provided by non-banks, that do not increase the supply of deposits.

      Second, in their Growth model, G&L have introduced banks that operate in a growing economy in an uncertain world, that must manage their liquidity and capital adequacy, that are confronted with bad loans, that use retained earnings to cover loan losses and to maintain their capital adequacy, and that set interest rate on deposits and on loans to manage the amount of liquid assets in their portfolio ( in the Growth model: bills) and to be able to generate sufficient earnings to pay dividends to their owners, to cover these bad loans, and to maintain adequate capital. It seems to me that such banks have quite a (pro-cyclical) influence on the economy.

      What do you think about these arguments?


    7. Anton,

      Your first point is an important one and I think I'll cover it in a seperate post. But consider the situation where (non-bank) Z lends to X who then spends with Y. X and Y are in exactly the same position as if a bank had lent to X. So the issue is whether Z has to change his spending habits to make the loan.

      On G&L, their Growth model is certainly one where bank behaviour is modelled in an intricate way, that coukd not easily be merely implied. And the way that matters is through the terms on which credit is provided, i.e. the rate of interest on loans.

  2. I think you need banks in the model if the goal is to model an increase in the money component of an economy. In contrast, if the goal is to model only the transactions between economic players, a bank-free model that allows for a change in the velocity of money may suffice.

    Here is an example. In a gold rush, the goal is to increase the money supply but the first effect, before meaningful amounts of gold are found, will see the velocity of existing money supply rapidly increase due to the investment required by the development of a new mining complex. An initial money velocity increase, then, later, (perhaps) an actual increase in the (gold) money supply.

    I think the difficult thing about modeling banks is that two perceptions develop. First, for the bank, there is no increase in money supply with a loan. The bank, at all times, knows how much money is available in the economic system.

    The second perception is the perception of bank depositors. Bank original depositors know how much money the bank has. Then the bank may lend the entire deposits to a 'creditable' borrower who spends the money. Once spent, the money may return to the bank in the form of new deposits, in the amount of original deposits. This results in two claims on the single original deposit. How do we model that without branching in to a bifurcated bank model with two complete balance sheets, both correct but not equal?

    So, to repeat, I think the presence of a bank in a model is only necessary if there is a need to demonstrate an increased money supply. There should be no need for a bank in a model where a change in the velocity of money will account for observed or predicted changes.

    1. If you believe that the quantity of private bank issued money is an important causal variable, then you might well want to model banks. As you may have gathered from various of my posts, I do not believe it is an important causal variable. I see it as what you might call a dead-end variable. Once the model is rich enough to determine an endogenous money supply, it will already have largely determined the major things of interest, like output and prices, although there may be some small feedback effect. So I don't think having the model determine the money supply adds much. Obviously, most monetarists would disagree with me on this.

  3. Interesting post(s) and discussion. Thank you (all).

    Btw. - as a side note - the securitised loans was mostly used for credit expansion by the shadow banking sector, but which had often backstops from the regulated banking sector - and thus were eventually a problem to the central bank.

    1. Agreed. That's one of the reasons why I talked here about the sector and the industry, rather than dividing it into banks and non-banks. There is a lot of inter-linkage here.

  4. Why did you decide to put capital on the liabilities side. Capital, assets minus liabilities is derived from a balance sheet it doesn't appear on the balance sheet. If by capital you are thinking of deposits from shareholders well those would either net off with the banks debts to other banks or result in a transfer of reserves, and those reserve would be written on the asset side.

    1. It's pretty normal to put it on that side. Have a look at this, from the Bank of England:

    2. That document is completely wrong . It , incorrectly, calls liabilities a source of funds, it then states that liabilities are funded by assets and then shows that reduction in asset values wipes out liabilities, which, as someone who knows a bit about these things I'm sure you will agree is absurd.
      Liabilities are not a source of funding. What happens is that taking a deposit may result in a reserve being acquired, depending on the banks clearing with other banks, and if so , then that reserve then goes on the asset side. The reserves are the source of funding. That document is wrong.

    3. in a graphical representation of a balance sheet like you have there the asset rectangle would be slightly longer than the liabilities rectangle with the difference in length labelled capital, and the length of that difference, the capital, would vary as the assets change in value.

    4. I should have labelled the column "Liabilities and Capital", rather than just "Liabilities" and perhaps that was a bit misleading. That's my economist-speak coming out rather than my accounting-speak, which would have been better. Otherwise the diagram is supposed to reflect what you describe in your last comment there.

    5. Capital is a subset of the assets it doesn't have a different definition speaking in terms of economics as apposed to accounting .
      A liability from the dictionary definition is a thing whose presence puts one at a disadvantage Therefore to class capital as a liability puts a banks financial position in a growing disadvantage as its capital grows which is oxymoron.
      The correct treatment of Capital is the value of the assets over and above the liabilities, a situation that a banker is careful to maintain.

    6. "The correct treatment of Capital is the value of the assets over and above the liabilities"

      I don't really disagree with this. I had just used the term more generally. In sectoral balance sheets in national accounting, liabilities includes equity capital. But yes it would have been clearer to have labelled the column "Liabilities and Capital".

    7. You say that In sectroral balance sheets in national accounting liabilities includes capital. How can that be , what is going on there. How can an asset be treated as a liability, that is illogical. It doesn't matter what you label the right hand column, capital is a quantitative measurement of the assets, it belongs in the left hand column with the assets.

    8. In this post you talk of liabilities funding assets .That's not right and it would explain why capital in the post is in the wrong column. Banks obtain reserves by attracting deposits but the deposits themselves are not a source of funding. Deposits are an accounting entry on an idividual banks balance sheet, deposits don't move around the financial sector. By securitising some loans the nature of the corresponding liabilities change but they do not become secured funding. Assets fund liabilities, not the other way around.

    9. The facts here are just the accounting entries that are made. For a description of how accounting entries works for banks, see here:

      People talk about this by saying that assets are funded with liabilities (and capital). You can talk about assets funding liabilities if you want. Most people won't understand you, but as long as you agree with the entries (i.e. what's a debit and what's a credit), there's no fundamental disagreement.

    10. Well if anyone talks about liabilities funding assets they are not respecting the meaning of words and it must be that they incorrectly think that banks take on liabilaties in and lend them out , which is of course factually incorrect and impossible.

    11. And just to add

      The approach you are using there , with assets being funded by liabilities and capital requires an impossible logic that deposits are both simultaneously adding to a banks funding and reducing its capital at the same time. Which illustrates that that approach can't be used and that the T account has causality , a flow of funds from assets to liabilities that has to be adhered to.

    12. Payment by a bank of a dividend would add to deposits and reduce capital. The accounting entries involving adding to the deposit accounts of shareholders and subtracting from retained earnings.

    13. Exactly thats a case in point. A bank doesn't need anything but a pen and a ledger to create a deposit , it doesn't require agreement with , or expectaion of , an income from, a co party, which illustrates once again the absurdity of "liabilaties funding assets". A deposit is no less than, and in particular no more than , a record that the bank owes somebody something , and so the idea that the deposit records on a banks own account sheet fund any aspect of the bank's buisiness is a non starter.

  5. This comment has been removed by the author.