Banks, Intermediation, And Pass-Throughs

This is a topic that is of interest for my book on banking. It may overlap some existing texts written some time ago (which is creating a future editing problem). Note that I refer to “this book” which should be read as “previous articles scattered around on my Substack.”

A somewhat arcane point of debate is whether banks are “(financial) intermediaries” or not. The reason why this is supposed to matter is whether banks exist to match savers or borrowers, or whether they “create saving.” From my perspective, the problem is the term “intermediary” as it is too vague, and should be replaced by the somewhat less common term “pass-through entity.” This is yet another example of how heterodox/orthodox economic debates have drifted into terminology disputes over decades. I will first explain the debate as I see it, then touch on the debate as framed by others.

The problem with this debate is that “intermediary” is hard to define. In the extreme, I have seen one attempt to argue that banks are intermediaries because if they lose the confidence of their funding sources, they can no longer extend loans. Unfortunately for that argument, that would imply that every single entity that issues short-term debt and holds financial assets is a “financial intermediary.” Which is pretty much every major industrial corporation (other than a handful of tech firms with debt-free balance sheets) — as well as many households. Although one can argue that all firms tend to be “bank-like” in practice, we are not getting much information from a term that applies to almost all firms and households.

Pass-Throughs

For “financial intermediary” to be a useful term, we need to find entities that are definitely “intermediaries” and have distinctive behaviour. These are “pass-through entities,” which are legal structures that buy a mix of financial assets and then pass through the cash flows to the owners of the units of the legal structure.

As an example, take a mortgage-backed security. A sponsor might bundle up $100 million of mortgages, and then issue $100 million in notes to bond investors (who could not buy the individual mortgages themselves due to fund governance rules). The cash flows from the mortgages go to the holders of the notes, with the sponsor skimming fees off the top.

I will just make two tangential points about securitisations (like mortgage-backed securities) before continuing.

1. Bond investors all have fund mandates that set the spectrum of allowable investments. Generally speaking, they need to stick to holding pieces of large debt issues so that there is at least theoretically a market in the instruments they hold. If an investor buys a $1 million mortgage outright, it is too small to trade in markets, and nobody would offer a quote on it. Holders of such small pieces of debt have to value them on the basis of the whole book. This explains why asset-backed securities exist: bond investors can buy the chunkier notes issued by pass-through, and they get a theoretical proportional claim on the underlying small pieces of debt that they could not buy directly.

2. If animal spirits/stupidity is sufficiently high, the pool of mortgages will be split in tranches — French for “slice” — with junior tranches absorbing credit losses on the mortgages first. This creates (allegedly) safer senior tranches, which means the overall structure looks like the liability structure of a bank. As seen below, they are not completely like a bank, but we see convergence in the capital structure.

If we look at the financial flows, we see that the pass-through increased the gross amount of debt in the economy by $100 million. There was the original $100 million in mortgages, and then there is $100 million in notes issued by the pass through. Despite this, it was all financed by $100 million in bond investor cash. If we want to use economist jargon, there is a simple budget constraint on the pass-through’s purchases of mortgages: it can only buy mortgages that are funded on a one-to-one basis by its note issuance.

That is, the pass-through just repackages existing mortgages into a more convenient format, but the bond investors had the financial resources to buy the resources. If we are just looking at savings flows, we could delete the pass-through entity and not much changes — the investors in the notes issued by the pass-through could (theoretically) buy the assets that the pass-through bought. Alternatively, we could have a theoretical zero-cost liquidation of the pass-through and put the assets on the note holders’ balance sheets and their economic situation would not change.

Although “pass-through” is already defined in tax law and accounting, for the purposes of economic analysis, what defines one is that it could be deleted/liquidated and it would have no effect on overall funding flows and economic claims on cash flows. Admittedly, there is presumably an effect in the real world of making it easier for existing pools of capital with fund management constraints to buy mortgages, but if we want to do “supply/demand curves” for “saving” there is no effect.

Why Banks Might Not Qualify As “Intermediaries”

The argument that banks are not intermediaries is straightforward: they create cash as part of their traditional lending operations. They expand their balance sheets to make new loans, and they do not need a matching inflow. Yes, they need to watch liquidity, capital ratios, etc., but those are prudential norms, not funding. Going back to economist jargon, they have a more complex budget constraint — they can “self-fund” some of their balance sheet.

The difference with pass-through entities is straightforward: pass-throughs do not create new mortgages by giving the borrower a new note that is a claim on the asset pool. (Or at least they are not supposed to do that…)

The “deletion test” also fails for a second reason: banks are operating companies that are expected by investors to continuously make money via fees and cagey lending practices (as well as cross-selling other financial services). Investors in long-dated bank debt or preferred shares are doing so as they assume that the bank is a “going concern”: they would not be happy to buy subordinated debt (and all other debt is subordinated to deposits) if a liquidation is a realistic possibility.

What About Mutual Funds?

At first glance, a fund that invests in bonds or money market instruments looks similar to a pass-through entity: rather than deal with the complication of buying those instruments, people pass their savings to someone who allegedly knows how to best invest in the bond market. We could delete the mutual fund and not much happens: people could just buy the underlying assets. Meanwhile, mutual funds are operated by firms that extract fees from the assets — the funds themselves are not “going concerns” and the accounting of the fund implicitly assumes that the fund could be liquidated at any time, with the proceeds going to the unit holders. (The liquidation of a large fund in a smaller market might impair the market valuation of assets. From the perspective of a large fund, “market values” are somewhat of a philosophical question.)

Where things get complicated is when borrowers in the bond/money markets re-invest the proceeds into funds. A certain amount of money might end up reinvested with the original borrower. That is, some of the money borrowed ends up funding the borrowing, which starts to smell like traditional bank lending.

Although these effects are small for a single borrower, if we start looking at aggregates, we begin to realise that aggregate borrowing in the non-bank financial markets can also end up being at least partly self-financing, since all the small amounts of self-financing start to add up. In aggregate, this means that such entities start to look like banks.

Aside: Insurance Firms

Although insurance firms are typically lumped in with the rest of the financial sector, they definitely are not economic pass-through entities (as I define it). They are an operating business that take in cash inflows and make pay outs based upon contractually set terms. The viability of the firm is based on investing the firm’s capital and charging enough for their products to be able to meet the contractual payments.

Nevertheless, one might lump in the insurance sector with other non-bank financial firms in a macro model just to simplify analysis. One could pretend that other sectors self-insure. This will not capture various wiggles in the data that result from events in the insurance industry, but given the low quality of macro models, this is not a concern.

This section now returns to the “intermediation” debate.

Jakab/Kumhof Paper

The debate about bank intermediation was old, but it got kicked off again by Bank of England Staff Working paper by Zoltan Jakab and Michael Kumhof — “Banks are not intermediaries of loanable funds—facts, theory and evidence” (see references below for link).

The use a strict definition of intermediation — intermediation of loanable funds (ILF). They define what they call the baseline ILF model as: “the key mechanism for bank balance sheet growth is that banks need to collect deposits of physical commodities (loanable funds), from one group of nonbanks, savers, before they can make loans of physical commodities (intermediation) to another group of nonbanks, borrowers.”

The “physical commodities” looks strange, but imagine that they are gold coins. (The fact that they are described as physical commodities relates to the fact that this definition is used within a Dynamic Stochastic General Equilibrium (DSGE) model, and it is not entirely clear why non-physical money has value in exchange in such models.) Since I am not going to describe that model, I will note that we just need some instrument that cannot be created out of thin air by the intermediary — it could be an electronic claim on the central bank, etc.

What matters in the context of the paper is the budget constraint — banks cannot lend more than they take in, which makes them fall under my “(economic) pass-through” definition. That is, if we use the ILF model as representing banks, they are pass-through entities. As I noted already, real world banks are definitely not pass-through entities, and thus the ILF model does not fit. This explains the title of the Jakab and Kumhof paper — banks do not qualify as intermediaries under their definition of the term.

Selgin Response

George Selgin has a working paper “Banks Are Intermediaries of Loanable Funds” which allegedly rebuts the points made by Jakab and Kumof, as well as other heterodox banking authors. (Reference below. This working paper has been revised, and may be revised later, my quotes are from the version of March 18, 2024.)

He aims at people pushing “Thin Air Theory,” which in the limit, suggests that banks can create loans/deposits without any limitation. Since this book already discusses why extreme versions of that “theory” are incorrect, I cannot disagree with Selgin completely. Instead, the concern is whether Selgin is making a good faith representation of what people are arguing.

Selgin defines “intermediation” as “investment intermediation” as follows.

But in their discourse both those who subscribe to the Intermediation theory and that theory’s critics have a particular sort of intermediation in mind, namely, intermediation between savers and borrowers, or “investment intermediation.” (Recall, for example, Luis Angeles’s definition of an intermediary quoted earlier.) In any economy there are people who wish to spend or invest beyond their current earnings and others who wish to spend less. Persons in the first set may borrow to satisfy their wants; those in the second save by accumulating either debt based financial assets, including informal IOUs, promissory notes, and bonds, or equity-based assets such as common stock. Savings may be invested in these assets either directly or through middlemen or brokers, also known as investment intermediaries.

Unfortunately this is a terrible definition.

The first problem is that Selgin uses “persons” and “people” as wanting to save or borrow, which immediately throws out every single “intermediary” that is involved in borrowing or lending to firms. Presumably, this reflects the simplistic classical/neoclassical models where the balance sheet of the corporate sector is abstracted out of existence. Given that Selgin relies on stating how he thinks banks work in the real world, offering a definition that only makes sense in models where the business sector does not exist is somewhat puzzling. (Although the definition does refer to stocks and bonds, which are not normally issued by “persons,” so perhaps Selgin treats corporations as people.) For the rest of the discussion I effectively use modified version of the definition where we use “entities” instead of “people.” If one wants to insist that it only refers to households, it is not applicable to banks since they “intermediate” between firms as well, and that is the end of the discussion.

The second, more fundamental is that it is too vague. Any entity that issues financial liabilities and owns financial assets is an “investment intermediary” under this definition. You can find examples of that in every single sector of the economy. You need to find entities with either no financial assets or liabilities. About the only examples I can think of are young children, and the handful of firms and retirees with debt-free balance sheets.

To sum up, we can apply this definition to practically all entities in the economy, which adds no value to a model. This is unlike “pass-through entity” or the Jakab/Kumhof definition which imposes a property on the budget constraints of such entities.

Although Selgin claims that the two sides of the debate agree on the definition, he then states this about the Jakab/Kumhof definition.

But a moment’s reflection should make clear that this “physical savings” argument proves too much, for if dealing in “physical” commodities is what distinguishes genuine investment intermediaries from banks, then there have been few genuine investment intermediaries, bank or nonbank, since the days of the gold standard!

Which literally is the point of the Jakab/Kumhof paper — no banks in the real world meet their definition of “Intermediation of Loanable Funds,” and Selgin just admitted in this passage that this is the case.

Selgin had further text flailing around with the meaning of “physical savings,” which is reasonable as it is a strange phrase. However, it makes sense in the bizarre world of DSGE models, and those authors use “loanable funds” in parentheses behind physical savings. Loanable funds are economically equivalent to claims on physical commodities, even if they are in electronic form. In any event, admitting that no banks in the real world met their definition should have been a wake up call that they were not using the completely vague and useless definition Selgin favours.

I think the following passage captures most of the substantive point of Selgin’s critique of heterodox banking theories.

It’s of course true that Jane can make a check out for any amount, and that she might succeed in paying for something with it. To that extent, she may be said to exercise her own power to “create” money. But if she makes the check out for more than the value of her bank account balance, it will bounce. Likewise, if a bank writes up a loan without either having on hand, or arranging to acquire in short order, funds enough to back it up, it must fail at settlement.

This can be boiled down to: private banks need to avoid a liquidity crisis. (The central bank in a floating currency can get away with what it wants.) This is a topic that I discuss at length elsewhere in this book, so it should not be a surprise to the reader. However, I have not seen any academic paper that suggested as such — all I have seen is out-of-context quotes that others have misinterpreted. If we have simple models of the banking system that exclude liquidity leakages, private bank lending has very few constraints. However, such models are usually noted as being unrealistic in the articles I have read.

This observation of the real world appears to be introduced by Selgin to obscure that he studiously ignores that the critique of loanable funds is that bank lending changes the budget constraint. Yes, banks need deposits — and lending injects new deposits into the system. The fact that an individual bank cannot “blow out its balance sheet to infinity” has no effect on the argument about budget constraints.

Another issue that Selgin highlights is the possibility that some banks will hoard settlement balances at the central bank (“reserves”). (Since this comes up in a variety of contexts, I will not provide a quotation.) Although this is possible in systems where the level of “reserves” is set by convention, this does not occur in practice for systems where the amount of (excess) reserves is expected to hit zero. (For example, the pre-2020 Canadian system, or pre-QE American system, where excess reserves were not significant.) The “rules of the game” in such a system is that banks cannot hoard excess reserves, as otherwise, the central bank would have to intervene and inject excess reserves to the rest of the banking system to avoid a crisis. We see in the data that this was not a typical event (although it did happen infrequently).

One final point of outright misleading logic is his discussion of the need for the central bank to provide demanded reserves.

Like the Thin Air theory itself, the “endogenous” or “horizontal” theory of the supply of bank reserves is superficially appealing, yet fundamentally misleading. It is of course true that any central bank interest-rate target implies a reserve supply schedule that’s horizontal at the targeted rate. It’s also true that, as long as it remains committed to a particular target rate, a central bank must allow the stock of reserves to adjust passively with the quantity demanded at that rate.

But if monetary policy means anything in a rate-targeting regime, it means that the central bank routinely reconsiders its rate target, shifting the horizontal reserve supply schedule up or down whenever it sees fit to do so for the sake of meeting its macroeconomic objectives; and this ability leaves it no less in ultimate control of the outstanding quantity of reserves than it would be were it to instead target that quantity itself.

Classical economics is largely useless because it did not take into account the passage of time, and this is a great example of this. The central bank operations desk has to supply required reserves at the end of the day every single day or there will be a banking crisis because of inadequate reserves in the system (which amazingly enough, does not happen in practice). What the central bank does in the future, and the effects on the economy in the future, has exactly zero relevance to the requirement to provide those reserves right now.

In summary, one needs to be careful in interpreting simple models of the banking system — but any reader of this book has received that warning that many times. This does not imply that loanable funds models are valid, which is the theoretical bugbear aimed at by heterodox authors. Whether one wants to call banks “intermediaries” is a stylistic decision at this point — what matters is that they are not economically equivalent to pass-through entities (which qualify as an “intermediary” under any plausible definition).

Concluding Remarks

Although it is clear that pass-throughs act quite differently than banks, we need to be cautious about things like mutual funds — variants of which represent the bulk of the non-bank non-insurance financial sector. Although it seems safe to describe pass-throughs as “financial intermediaries,” it is unclear whether “financial intermediaries” is a meaningful phrase if it applies to banks.

References

• Jakab, Zoltan, and Michael Kumhof. “Banks are not intermediaries of loanable funds—facts, theory and evidence.” Bank of England Staff Working Paper #761. (2018). URL: https://www.bankofengland.co.uk/working-paper/2018/banks-are-not-intermediaries-of-loanable-funds-facts-theory-and-evidence

• Selgin, George. “Banks Are Intermediaries of Loanable Funds.” (2024). URL: https://www.jstor.org/stable/resrep58223