Greater Fools In Credit Markets And Recessions

For a number of reasons, we want to be able to forecast recessions. Unfortunately, many recessions are the result of disruption in credit markets. As a result, forecasting recessions accurately is equivalently difficult to forecasting the direction of the credit market accurately. The lack of investors of achieving this feat suggests that recession forecasting is inherently difficult. This article offers an overview of this chain of logic.

(This article is an unedited first draft of a front section of my upcoming book on recessions. I am outlining the big picture of why we are interested in recessions in the first place. As a result, there are a lot of assertions in it. The remainder of the book -- which is in a partially finished state -- provides the details.)

Introduction

Recessions are of interest for both practical and theoretical reasons. From a practical perspective, calling for a recession when the consensus is for steady growth is a sure way to get attention. From the perspective of theory, recessions are where we can separate the usefulness of economic theories. During an expansion, economies are characterised by steady growth, and it is hard to distinguish any models that are able to generate trend-following behaviour. Recessions are when the interesting dynamics happens.

Being able to fit a story to a historical recession is interesting, but of limited usefulness. Any number of stories can be told about the data; one only needs to look at the ongoing debates between economic bulls and bears in financial market commentary to see that it possible to fit any number of theories against observed data. What we want to be able to do is forecast future recessions.

The first challenge to forecasting is that recessions are a generic economic event. For simplicity, let us use the simplest standard for defining a recession: two consecutive quarters of declining real gross domestic product (GDP) growth. Gross domestic product is an aggregate concept, and is the sum of a number of components. A decline in real GDP represents a decline in one or more of those components, and the reason why each component may decline can easily be different.

We can use an analogy from physics (since everyone loves using physics for analogies for economics). Let us assume that the event of interest is a body moving from point A to be point B. This can happen for any number of reasons.

• The body is self-propelled, like a car or a kitten.
• The body has metal components that are attracted by a magnet.
• The body has an electric charge, and is embedded in a magnetic field.
• The quantum wave function first collapsed at point A, then point B for inscrutable reasons.

If we put aside the quantum mechanics shenanigans, Newton’s law of motion will describe the dynamics of the move from point A, to point B, but the explanation why the movement occurred lies in some other fundamental physical laws.

Returning to recession forecasting, we do face a multiplicity of the number of potential causes for recessions; forecasting whether or not the economy will be in recession implies being able to forecast one or all of those potential factors.

For example, there is a broad consensus across economic schools of thought that argues that tightening fiscal policy sufficiently will cause a recession. (Since the term “tightening fiscal policy” is vague, it is the act of raising tax rates (or the volume of taxes raised) and/or cutting government spending; how we judge the relative importance of these effects will depend upon the theory used to analyse it.) Admittedly, there are some free-market economists who profess disagreement with that assessment, but I am unsure how serious those objections are. In any event, if we accept that tight fiscal policy can cause a recession, one aspect of recession forecasting is political forecasting: will the government enact such a policy on the forecast horizon? Although one would need economic models to assess the expected effects of the policy change, whether or not the policy will be enacted is outside the bounds of standard economic theory.

Nevertheless, we can narrow down most of our discussion to an interesting class: recessions caused by disruptions in debt dynamics. For example, the deep global recession associated with the Financial Crisis that peaked in 2008 was a sterling example of disruption in debt markets. Although we should keep in mind other oddball potential causes of recession, most of our recession forecasting efforts will be on debt dynamics.

With those preliminary comments out of the way, I will now lay out my main arguments about recession forecasting. I do not claim that they are in any sense original; however, I am unaware of other authors using identical language. Instead, the tendency was to use more flowery language that ends up being practically equivalent to my arguments. My concern is that the flowery language has ended up being hard to interpret, leaving a sense of mystery over the subject.

The Theoretical Thesis

The main points are as follows.

1. An interesting and important class of recessions is caused by the disruption of investment (or even consumption) in the private sector that coincides with a decline in borrowing flows.
2. In modern capitalist societies, lending decisions end up being de facto centralised because of the institutional structure.
3. Forecasting such recessions described in point (1) is an analytical exercise that is equivalent to being able to forecast accurately the direction of credit markets.
4. To the extent that the reader believes a loose version of the Efficient Markets Hypothesis – that market direction is impossible to forecast – quantitative recession forecasting is similarly impossible.

To make my stance perfectly clear – I am not saying that recession forecasting is impossible, rather that you would also have to be able to perform a far more lucrative analytical exercise, which is being an amazing credit market forecaster.

As a technical point, when I refer to “credit markets,” I am not referring to just public markets in bonds and money market instruments, as well as the myriad variations (such as asset-backed securities). I am including bank finance as part of the credit markets, even though bank lending decisions appear opaque and are not mark-to-market. There is a largely unexamined ideology in finance and economics that markets ought to be distributed, transparent, and mark-to-market, i.e., resemble futures or equity markets. The reality is that funding decisions happen in credit markets (as I define them), and those markets are opaque for very good reasons.

There is no good theoretical reason to believe that such forecasters cannot exist; the issue is finding whether any exist. Unfortunately, one cannot just point to the existence of investors with a strong historical investment performance. Based on my observations as a capital markets analyst, most investors make their money based on variations of relative value trades: buy cheap securities, sell/underweight expensive securities. The whole point of relative value investing is to avoid making directional calls on the market. If we look at the commentary in the financial press on directional trends by such investors, the commentary tends to be of very suspect quality, despite their exceptional investment track record.

Why are Credit Markets Hard to Forecast?

Why should credit markets be hard to forecast? We need to look at the structure of credit. Even if you believe the long-term prospects for a borrower stink, most borrowers have the liquidity to roll over a certain amount of their borrowings. So long as there is a “greater fool” willing to extend financing to the entity in the future, one outperforms by lending at an above-average spread (versus funding cost or benchmark) via instruments that mature while the liquidity buffer holds, or the projected supply of greater fools is non-zero. Furthermore, the concentrated nature of credit markets matters. There is an old banker’s adage: a rolling loan gathers no loss. As long as lenders roll over the borrower’s debts, the borrower can avoid default – no matter how uneconomic the borrower is. This can be justified on any number of reasons: long-shot odds of a turnaround by the borrower, a boom in economic conditions that bails out even the most incompetent borrowers, or just the need to delay the recognition of loan losses. The incentives to delay cutting off funding means that the supply of greater fools in the credit markets is much larger than one might suspect on theoretical grounds.

The theoretical claim is of considerable interest. Very simply, it suggests that mathematical economic modelling is a fundamentally doomed enterprise. Even if one can fit some model to the wiggles of the expansion phase (which has been done to varying degrees of success), the model will blow up when a recession hits – which is what we are most interested in! Sadly, I am unsure that I will be able to present arguments that will convince everyone of its truth. As a result, I have a more modest ambition: to set out a description of various economics theories that explain why credit market behaviour is of utmost importance. It will be up to the reader to decide whether my leap from that starting point to the final analytical result is plausible.

Fit the Model to Data (Sigh)!

As a final aside, I will explain why giving a “convincing” explanation of the thesis is impossible. The problem is straightforward: the thesis is fighting against a embedded cultural assumption that is almost impossible to break out of. The assumption is that economics is just like physics, and so we explain reality by fitting models against observed data. Although this is a charge often levelled against neoclassical (“mainstream”) economists, this is a culture-wide phenomenon. I have run across many physicists who have taken exactly that theoretical line.

The thesis I am pushing can be expressed alternatively: the prediction of the model is that economic outcomes cannot be accurately forecast at all times (with credit-based recessions being the primary example).

Then consider what the critics of the idea will inevitably argue: the model needs to be fitted against data. (This forecast is based on the previous responses of such critics to earlier heterodox economic arguments.) It tells us all we need to know about the modern academia that people can somehow qualify for a doctorate and yet they demand that models that predict economic outcomes cannot be forecast be fit to data.

It is very simple to prove my thesis wrong: construct a model that forecasts the economy accurately. Quantitative economic modelling started out with analog computers (e.g., A.W. Phillips) and the volume of models has exploded with the advent of digital computers. Researchers in academia, central banks, and finance have been throwing models at the macro data sets for decades, trying practically every numeric technique used in the applied sciences. (As an ex-control systems engineer, I was one of those people.) Despite this massive, directed attack on the set of possible forecasting models, humanity’s economic forecasting capabilities are lacklustre. Perhaps there is a fundamental reason for this continued analytical failure?

Concluding Remarks

Having laid out the theoretical roadmap of where we want to go, we still need to roll up our sleeves and do the theoretical work needed to get there. Since most of the points I raised are essentially consensus views within the heterodox community, the reader may be already familiar with the background. My objective is offer a guided tour of this theory for those less familiar with the subject.

(c) Brian Romanchuk 2019