Stuff I Read On The InternetI started thinking about this topic in response to the article "The Parable of the Fruit Trees" by Professor Nick Rowe. Unfortunately, I want to cover a lot of background material before I could get to his argument. His argument is not exactly the "excess demand for money" argument that I am straw-manning in this article. In his words,
It is not an excessive desire to accumulate assets that causes recessions; it is an excessive demand for one particular asset (the medium of exchange) relative to other assets. It's about the composition of their portfolios of assets, not about the total size of that portfolio.The distinction he draws is interesting, although my guess is that I would disagree. However, I think there is a more fundamental issue at stake.
Update: My initial interpretation of the parable was incorrect; there is a separate medium of exchange that was not one of the underlying commodities. However, this puts us back closer to the standard model setup within neoclassical models, with a "commodity-like" money instrument.
If we look at the parable and neoclassical models, monetary transactions appear indistinguishable from barter transactions. (One may note that this is also true for many simple stock-flow consistent models, which are favoured by heterodox economists.) The fact that we do not live in a barter economy is a standard sound bite critique of neoclassical models. The interesting question is: what distinguishes the real world from the "barter-like" conditions of those models?
The difference is that real-world industrial capitalism relies on cash, and cash has some very ugly theoretical properties that result from credit risk. (Note that I am using cash in the sense of bond market jargon, which are short maturity fixed income assets, which will show up in accounting as "cash and cash equivalents.")
I will get back to the cash/money distinction after some detours. However, for a more standard heterodox view, I would point the reader to "Saving, Investment, Loanable Funds, Paradox of Thrift" by Sri Thiruvadanthai.
Anything Can Cause a Recession...When I started thinking about Nick Rowe's arguments, I realised that we need to be careful about asserting that various factors "cause" recessions. (Since I want to write a book about business cycles and what causes recessions, this hits close to home.) The definition of "recession" is somewhat vague, and depends upon the country. (In the United States, the NBER is tasked with declaring "official recessions," other jurisdictions use a looser "2 quarters of declining real GDP" definition.) For our purposes here, we will assume that a recession refers to a period of declining real GDP.
Since the level of real GDP is an outcome of a complex system, it is very difficult to say that any particular factor "causes" recessions. Any factor that can result in changes to any aspect of economic activity could conceivably cause a recession.
For example, I often chortle at Real Business Cycle models that suggest that recessions are caused by a drop in productivity (or the more risible argument that workers simultaneously decide to take vacations). However, one can imagine scenarios where "dropping productivity" does cause a recession. For example, let us assume that some Bond villain organisation somehow manages to simultaneously sink the world's oil tanker fleets. The entire energy complex would be in chaos, and many jurisdictions would lose access to hydrocarbons. A great many activities would no longer make economic sense, and firms would either go bankrupt, or fire employees. The drop in real GDP could be legitimately interpreted as the result of the much lower productivity of a petroleum-free economy.
Obviously, such scenarios are not particularly common. However, one could try to explain some recessions as being the result of an inventory cycle, which is related to real variables. Conversely, the post-1990 U.S. recessions have been associated with financial crises, while the euro periphery was devastated as a result of policy choices. That pretty much covers all bases for types of causal explanations.
We also need to be careful of tautological statements. Since the usual result is lower output when you have less workers, one could say that mass layoffs "cause" recessions (if we use a real GDP definition). However, that is pretty much just a different way of saying the same thing, as we then need to know what causes the mass layoffs?
Returning to the "excess demand for money" story, one can see that GDP is based on (some of) the monetary transactions in the economy. If real GDP declines relative to the previous period, that means that some sectors are spending less money. If we want to call that an "increased demand for money," we can then see that recessions are associated with "an increased demand for money." However, that is just another way of saying the same thing.
Cash Versus MoneyIf we put aside the policy-induced recessions in the euro area, recessions in the post-1990 era in the developed economies were largely associated with financial crises of some sort (quite often housing market bubbles popping). For reasons I discussed in a previous article, financial crises almost inevitably start in the money markets, owing to market structure. The association of financial crises with money markets helps build the folklore around money and recessions.
Neoclassical economic models (and many stock-flow consistent models) effectively use government instruments as "money" in their models. In fact, the representative household "borrows" by having a short position in government-issued bills.In such a framework, "money" might as well be commodity money, and viewing it from a barter lens seems mathematically plausible.
Government liabilities are the bedrock of the settlement system, but they are not where private sector portfolios are allocated. Capitalism functions on the basis of cash.
There is a great deal of mysticism around banking and shadow banks, but practically all firms have financial and credit operations. However, this is no longer visible to casual inspection, as retail has been dumbed-down, and retail employees generally no longer have the authority to offer credit to customers.
Let us assume that customer A wants to purchase something from firm B. So long as customer A is "money good", firm B will sell the product to A on credit. The accounts receivable for B is a new financial asset issued by A, and can be traded for other financial assets, or easily financed. (The origins of the money markets are in the re-discounting of such short-term debts.) The private sector can create cash instruments to finance all desired transactions without any magic from fractional reserve lending or whatever -- under the assumption that there are no worries about credit.
The fact that it is so easy for even non-financial firms to create cash instruments is what causes the breakdown in barter models. From a mathematical perspective, since they net out for the private sector, we can fit this into mathematical models using government money, but we can no longer think about cash as a physical commodity.
Adding credit risk does not appear to be a problem from the perspective of neoclassical models. After all, we assume that markets are complete. Private debt with credit risk is just equal to Treasury bill plus credit default swap (CDS) protection (under the sorts of simplifying assumptions that economic models use). In case the reader is unfamiliar with CDS contracts, we will assume that they are the same thing as a credit guarantee -- a third party guarantee payment on the debt. All that happens is that we assume the accounts receivable are protected by CDS protection, and so the CDS spread will end up being incorporated into the buyer's budget constraint. (The usual assumption for the household budget constraint is that borrowing/lending rates are equal; this complication just introduces a wedge between the two rates. Whether such a model could be solved is an open question.)
The problem is the passage of time. Neoclassical thinking is based around solving optimisations in the current time period, with future activity being described by forward transactions in hypothetical complete markets. We can allegedly lay off credit risk forever using the CDS contracts going out to infinity.
This can break down when time actually passes. The credit risk associated with an entity at time t+1 is not necessary the discounted credit risk priced in at time t. (I explain this in the following example.) Furthermore, CDS contracts do not eliminate credit risk -- just redistribute it. Some entities have to be bearing 100% of the credit risk of the issuer -- and are exposed to potential credit losses.
For example, let us say that we want to close a deal on June 1st for $1 million. The customer is a respectable non-financial firm, and wants to pay on August 1st. The CDS quote is 50 basis points annualised, and we assume that we are operating in an arbitrage-free world (which is a good approximation of reality). We have a number of options.
- We keep the receivable on our balance sheet. Although it may not qualify as "cash", it is a short-term, relatively high quality receivable. From a planning perspective, we would lump it with other short--term assets, although possibly with a small loan loss provision.
- We keep it on the balance sheet, and buy CDS protection. The package looks like a Treasury bill that matures on August 1st.
- We rediscount the receivable with some third party. We get cash upfront (and some presumably low risk legal entanglements from the means of rediscounting).
- We borrow against the receivable. Assuming we have a relatively clean balance sheet, we might be able to borrow at an annualised spread of less than 50 basis points.
In summary, on June 1st, the receivable is a short-term asset that is readily comparable to money.
Unfortunately for us, on July 1st (happy Canada Day!), the borrower announces that they were um, lying to their auditors, and their financial statements were a fantasy. No one is even willing to quote a CDS price. That $1 million receivable has a market value somewhere around zero. Depending on what we did with it, some entity is eating a $1 million loss. If we passed the hot potato to some other entity, its solvency might come into question, as it may have been the only entity foolish enough to take on the borrower's credit risk.
Roughly speaking, neoclassical thinking is locked into the "June 1st world," where we believe that we can hedge out all possible outcomes, and the receivable is mathematically equivalent to any other money market instrument with a spread of 50 basis points. However, the transition to the "July 1st world" blows that assumption up: we now have credit losses that have a distributional effect on wealth and income. Plans need to be revised, based on the new liquidity realities of actors. If this cascades, we can easily generate enough economic disruption to cause a recession.
We do not need a full repeat of the 2008 fiasco in order to generate a recession. Fixed investment is typically debt-financed, and so rapidly growing sectors are the major source of new net borrowing. Once their credit quality is put in question, its growth will reverse. (The credit travails of the telecom industry helped scotch the 2000 tech boom.)
In summary, the possibility of transition from "cash" to "very risky asset" is where most neoclassical modelling approaches miss a key component of industrial capitalism.
Although it is easy to describe this effect -- and see examples in the real world -- modelling it is not straightforward. It is hard to see why this would happen in an aggregated model, since it is unlikely that an entire aggregate sector can default. Furthermore, it runs into the problem of model inconsistent beliefs -- why would investors be willing to lend at tight spreads to entities that default in the next model period? (Although one might point to empirical evidence that this is exactly what happens, such small-minded "empirical thinking" is not what gets your theoretical papers into the "top 5" economics journals.)
(An agent-based model would presumably be able to better cope with these issues than an aggregated model.)
Appendix: A Rant About the "Safe Asset Shortage"
I have often run into arguments that we should be concerned about a "shortage of safe assets." This is a ridiculous worry.
Going into 2008, the problem is that nobody wanted "safe assets." A deal structured by cows (allegedly) might have ended up trading at 20 basis points over Treasuries. If Treasurys were indeed in short supply, everything else would have been trading at wide spreads above them. (Historically, I believe that there issues with Treasury bill shortages during the brief U.S. federal surplus period, and spreads were disjointed as a result. I am not particularly motivated to try to track down those details in the market data I have available.)
If entity A has a boatload of toxic assets on their balance sheet, issuing a ton of Treasury bills to entity B is not going to help matters. The problem is that A would end up insolvent if the assets were marked to market (or default). There are only two ways to deal with this.
- Liquidationism. Let A deal with it.
- Industrial policy. Bail out A by buying the dodgy assets, or lending against them. In order to make this more palatable, we label this "lender-of-last-resort" operations. (If one were cynical, one might note that the most rabid anti-bailout faction in most countries are bankers, and yet we relabel the bailout of banks to avoid hurting free market fundamentalists' feelings.)
(c) Brian Romanchuk 2018