Monday, 9 March 2015

Do We Need Banks in Models



The Bank of England's recent conference on its One Bank Reserach Agenda has prompted a few comments on the issue of the inclusion of banks in macroeconomic models.  Various people have observed that it is rather odd that the models used by central banks seem to completely ignore the role of banking.  So why might it matter whether we explicitly model banks or not?

Before we even think about including banks, we need to have a model with financial assets and liabilities, so at the very least we need some agents that are borrowers and some that are lenders.  Let's assume we have this.

Some people might think we need banks, as they can make loans without first needing to attract funding, when non-banks cannot.  It might seem that without banks in the model, we have no way for financial assets and liabilities to grow, they can only be passed around.   

However, in general this is not the case.  Most models with financial assets allow for the quantity of those assets to expand and contract in response to spending and saving decisions.  Although these are not generally modelled to the level of the individual payments, it is easy to add a payment system in and when you do, it makes no difference.  So, on this point, we can rely on an implicit banking system - we don't need to explicitly model it.

A more likely possible answer is that we need banks because some part of their liabilities constitute money.  Exactly what part depends on how we want to define the term.  

This would be the monetarist position - that banks matter because they influence the growth of the money supply.  Whether this means we needs banks in models would then come down to our view of the quantity theory - the question of whether money supply growth causes growth in nominal GDP.

For banks and money to matter in this way, we must assume people view holdings of money differently to holdings of other assets.  For example, we might assume that people want to hold money only in some proportion to their income, but that they have an infinite appetite for other financial assets.

The problem with this comes down to drawing the line between money and those other financial assets.  Balances which can be used in payment processes might be considered to be special, but these represent only a small part of bank balance sheets and they tend to respond to demand with a highly elastic supply.   Modelling bank behaviour, specifically lending behaviour, won't tell you much about these balances.

On the other hand, modelling bank lending might tell you about the overall size of bank liabilities.  But you then need to explain why these bank liabilities are so different from non-bank liabilities.  How much is it going to affect behaviour if someone has a Treasury repo, say, with a non-bank as opposed to with a bank?

So far none of this seems to me to suggest there is much to be gained by adding banks explicitly into models.  Where I think it does matter is when we come to think about the terms on which loans are made.

In the simplest model with borrowers and lenders, anyone who wants to borrow can borrow as much as they like at the same interest rate.  This is a model that ignores credit risk and for some purposes this is OK.  However, incorporating credit preferences into the model can help provide a lot of insight.  Once we include credit limits and credit spreads, the borrower's spending pattern increasingly depends on the preferences of the lender as well.

We still don't necessarily need banks to include credit preferences.  Savers will have their own preferences and if we believe banks' actions merely reflect the credit preferences of their depositors, then they would indeed be no more than intermediaries, whose existence could be safely ignored.

However, this is not how it works in the real world.  Bank lending decisions are considerably removed from the preferences of depositors and are shaped by other factors, including notably their regulatory structure.  Capital and liquidity regulations play an important role in determining how bank lending develops and these can only properly be captured in a model where banks appear explicitly.  This to me is the real reason why we might need to model banks.

In conclusion, I think that for many purposes, we don't need to include banks in models - we can just treat them as implicit.  We certainly don't need to include them simply because we believe in "endogenous money".  Indeed the whole point about endogenous money is that money is not an important causal factor as the quantity theorists would have it, but merely a by-product.  However, once we want to think about how credit limits and credit spreads might develop in a modern economy (and these are important things to think about), we need to start saying something about banks.

13 comments:

  1. "In conclusion, I think that for many purposes, we don't need to include banks in models - we can just treat them as implicit. We certainly don't need to include them simply because we believe in "endogenous money". Indeed the whole point about endogenous money is that money is not an important causal factor as the quantity theorists would have it, but merely a by-product."

    Are you sure? See a.o.:
    http://www.interfluidity.com/v2/4522.html
    http://monetaryrealism.com/krugman-and-tobin-on-banking/

    Furthermore, it seems to me that banks can and do play a very important role, as they did in the years before the crisis when they were actively pushing mortgage lending and driving up home prices in the US:
    http://fcic-static.law.stanford.edu/cdn_media/fcic-reports/fcic_final_report_full.pdf

    Anton

    ReplyDelete
    Replies
    1. I absolutely think that banks play an important role, particularly in relation to the crisis - but that comes down to their part in the availability and cost of credit, not in relation to the money supply. I think there is also a case that banks have not only played a part in how the demand for credit is met, but have actually influenced that demand. That aspect though is, I think, virtually impossible to include in a model.

      Delete
    2. Do you mean that banks can influence the amount of credit, but not the amount of bank money in circulation?

      Anton

      Delete
    3. No, I'm saying that distinguishing between a monetary measure that is broad enough to encompass most bank liabilities (and to therefore be closely related to lending volumes) and similar instruments issued by non-banks, is not particularly useful.

      Delete
  2. Considering that none of the models I have built have had banking systems, I'd agree you can do a lot without an explicit banking system.

    Where I think the difference comes in is within the payments system, as you note. Shadow banking operations implicitly require banking line backups as part of the expansion of credit; but the banking component disappears from the aggregates as "money" leaves the formal banking system for the shadow banking system. This sort of institutional detail is needed for understanding the transactions within the system, but it does seem unnecessary for a macro simulation that abstracts from the details.

    If we are monitoring real-world economies, you probably need to keep an eye on the details within the data, but it seems unlikely that we could fit a model to the extreme complexity of real world financial systems.

    ReplyDelete
    Replies
    1. I am inclined to agree. I think it is very difficult to include banks in a way that could help with any kind of forecasting. I think you can use banking models to help explain why things might happend or have in fact happened and also, as you suggest, to understand what might lie behind real world data.

      Delete
  3. Money should be limited to means of making payments including central bank notes in circulation among nonbanks and checking deposits of nonbanks kept in banks (this is basically the M1 money supply). The checking account level is driven by bank investment managers expanding aggregate bank assets, less the bank liability and equity managers seeking to convert checking deposits into other investments in the aggregate bank. The development of liabilities and equity from pre-existing checking accounts is necessary to keep interbank payments flowing on a small level of bank reserves, which banks try to minimize in a growth economy where the central bank does not pay interest on reserves. This means banks have a natural tendency to leverage the bank balance sheet on a small quantity of reserves which contributes to financial instability in a market crisis, so banks matter a great deal in economic models at the margin, where they contribute to inflation or deflation or require bailouts from governments. Before the financial crisis banks were making higher return on equity by charging fees to originate and distribute loans to nonbank financial firms, and this effectively created off-balance sheet liabilities for banks when those loans saw an increase in default rates. The interaction between the aggregate bank and nonbank sectors is crucial to understanding the stability of the economy. I developed a four sector model of the US economy from the flow of funds here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2458563. I think of the aggregate bank as a diode or check valve: it can satisfy the aggregate nonbank demand to convert money (transaction accounts) into investments, however, it cannot satisfy the demand of the aggregate nonbank to convert investments into money, because this fouls up the interbank payment mechanism and profit model.

    ReplyDelete
    Replies
    1. I think if I was to model banks from the perspective of saying that money matters, it would probably be something along these lines. However, my own sense is that both the demand and supply for this type of money are too elastic for it to be very useful in an empirical model.

      Thanks for the link to your paper - it looks interesting.

      Delete
  4. In frictionless equilibrium analysis banks can be taken into account as implicit. So most economic models are just fine without explicit banking sector (as per post). Yet if one tries to explain changes in money flows and frictions in price setting the explicit banking sector might be one way to go? Could you even say that this is one of the insights of Post-Keynesian economics?

    When credit expands the bank liabilities rise. The credit considerations are certainly needed but isn't there other channel as well? It is only a feel but I would say that there is something special in bank liabilities in general (without considering their structure). For example if the asset rich household sector is saturated with bank liabilities, how do they respond when the credit is still expanded?

    ReplyDelete
    Replies
    1. On your first point, I think most frictions do not need really need banks to be explicitly modelled. Adding them into the model might make it clearer what is going on, but it doesn't change the results. As far as PKE insights go, my view is that the most important one is that banks matter because they have particular portfolio preferences that do not simply reflect those of end investors. I see this as being in line with the reasons why I do think it might be useful to model banks.

      On the question of whether it makes a difference what assets households hold, I think it does make a difference, but on the whole the division between bank issued and other is not clear enough to make it useful enough to model. In my opinion.

      Delete
  5. Having thought about this, the difference may show up in an agent-based model, or similarly, an economy in a video game.

    In such a model, you are explicitly modelling the decisions of individual firms at the transaction level, and so the legally privileged position of banks should show up. That is, the lending decision rules should be different than for a bond investor.

    But if you are aggregating behaviour, and you are attempting to model the increase of debt within the capital goods sector (for example), you are not too concerned about the exact mechanism of how the debt increases.

    ReplyDelete
    Replies
    1. I agree that you often want to just assume lending increases and that trying to explain why won't add much, i.e. it will be no less arbitrary. I don't think you need to get to an agent-based level though to sometimes want to think about why bank lending might develop a certain way, but its difficult to turn this into anything useful for predicting what might happen in the future.

      Delete
    2. Episodic bank bailouts by governments are properly explained only by agent-based models incorporating endogenous creation of bank credit money and the nature of the interbank payment mechanism. Nonbanks must rollover investments in banks rather than decide to hold money or else the operation of the interbank payment mechanism coupled to investor psychology forces the economy into a bank balance sheet recession.

      Delete