I saw a comment by former Bank of Canada Governor Stephen Poloz that suggested that only population growth is keeping Canada from slipping into an “official” recession. (In the United States, the NBER is the standard recession-dating source, in Canada, the C.D. Howe institute aims to replicate that role.)
I finally have my kitchen back, and now can devote more time to writing and consulting. I am still pushing another project, so my output here will probably be limited. I have taken another look at my bank primer project, and realised that I have too much content — I will need to trim back the theoretical wrangling texts that I previously wrote. With today’s article, I think I have covered most of the content I want to be in the book, although I might stick in some cursory analysis of a few different banks’ balance sheets. For example, I might compare some teeny-tiny American bank versus larger European or Canadian banks as a way of indicating the dispersal of what “banks” are. This article has only been lightly edited.
If the economy was in a stable equilibrium dominated by agents forecasting their cash flows out to infinity, defaults would be a random process – defaults would occur, but without a pattern to them. Default risk would be an insurable risk (i.e., could be managed by actuarial calculations like life insurance). However, the existence and popularity of the term “business cycle” indicates that the flows of commerce are cycle – and defaults follow the business cycle. During an expansion, banks do face a persistent relatively low level of defaults and delinquent loans, which does accord with being a random, insurable risk. The problem is recessions – which see a spike in defaults. Although it is possible for there to be a recession without a default spike (as discussed below), the “interesting” recessions are the ones with default spikes. The “really interesting” recessions are the ones where the banking system itself joins in on the default trend.
Menzie Chinn just published a short note “Inversions, Bear Steepening Dis-Inversions, and Recessions.” He was responding to an article that argued that a bull steepening is a good sign for the economy, as it indicates less need for the Fed to cut in response to a recession. Chinn updated some recession probability indicators based on the yield curve.
Brian Adams1, Lara Loewenstein, Hugh Montag, and Randal J. Verbrugge just published the working paper “Disentangling Rent Index Differences: Data, Methods, and Scope.” URL: https://www.clevelandfed.org/publications/working-paper/wp-2238-disentangling-rent-index-differences. I did not have time to look over the paper, but the graphs are consistent with the story I have seen elsewhere: the rental data used in the CPI lags behind the latest changes, and the latest changes point to a lower growth rate in rent hikes.
This article continues the previous discussion of bank net interest margins. In it, I discussed how changes in the yield curve changed the net interest margin (NIM) for banks. This showed up historically — when bank balance sheets were shattered by the combination of holding long-dated mortgages with low fixed coupons versus having a sky-high short-term rate imposed by deranged Monetarists. In this article, I address a common macroeconomic story: yield curve inversions cause recessions by the alleged effect on NIM. As a spoiler, I do not think that story holds water in “modern” banking systems.
Since I am catching up with everything, I am unable to do a detailed analysis. If I do anything, I will postpone it until next week. I just want to add yet more words to the “is is a recession?” debate.
My view is straightforward: the NBER declares U.S. recessions, and I am not going to refer to a downturn in U.S. data as a “recession” unless the NBER has declared one. The thing to keep in mind that the declaration can arrive with a considerable lag, so I have no issues with texts arguing that one is currently in progress without the NBER committee declaring anything, so long as that condition is noted. i
I only have an informal version of the argument, and it runs as follows. The usual driver of a business expansion is private investment, which normally needs to financing by the firms/households that are investing. This implies a need for financial firms to be willing to extend credit. This willingness — the “animal spirits” of firms and financiers — thus drives the cycle. So far, this is pretty much what Keynes wrote. The wrinkle — which may or may not be original — is that the credit markets are in fact markets. If we could predict credit market trends, that largely implies the ability to time markets. Which implies that recession forecasting runs into some version of the Efficient Markets Hypothesis.
There is a lot of academic controversy about the Efficient Markets Hypothesis, which I am not interested in. I am referring to a simplistic version of it: if you think you can predict the direction of markets reliably, why are you publishing academic articles and not on a yacht somewhere? My belief is that this is close to the intent of post-Keynesian views on the business cycle, but they do not frame it this way due to being allergic “market efficiency,” and so they use more confusing terminology to explain the concepts I discuss here.
Although most historical recessions were associated with slowdowns in debt growth, it is possible for the economy to contract for other reasons — and such a possibility is a key risk at present. The figure above shows non-financial debt growth in the United States, along with recessions shading. We can see that credit growth historically coincided with slowing debt growth. In recent decades, the Savings and Loan Crisis was associated with the 1990 recession, the 2000 recession was associated with technology firms losing access to credit markets, and the 2008 recession was notoriously a debt crisis. These are the types of recessions I see as difficult to forecast.
However, we also can recessions driven by other causes — which force us to forecast those other drivers. We can have a policy-induced recession (e.g., Euro periphery), where we need to forecast policy. The recent recession was driven by a pandemic (and the response to the pandemic), which is a medical forecasting problem. We can also have downturns triggered by things like the loss of energy imports.
Currently, the most pressing problems seem to be of this type: the renewed lockdown in China, and the effects of constricted energy (and food) supplies. A credit contraction might be triggered by these events, but it is hard to see credit as the driver of any slowdown. As such, the comments here are not aimed at the current situation.
Credit risk being the constraint for debt growth is a problem for most macroeconomic models — agent-based models being the notable exception. It is hard for a mathematical model to capture the uncertainty associated with credit risk, and the tendency of risk appetites to change.
As a starting point, take the simplest macro model structure: stock-flow consistent (SFC) models. (I have a book on building SFC models in Python.) A SFC model has two categories of equations:
equations that define the accounting structure of the model (“accounting identities”), and
equations that determine the behaviour of sectors (that are constrained by the previous type of equations).
The arguments in my banking article can be re-phrased as follows: the accounting identities do not limit debt growth, rather it is driven entirely by behaviour.
We could imagine a model where there are some parameters in the model that act as a “credit growth” factor. That is, the model will settle in at a certain pace of private sector debt growth once other variables enter steady growth states. To then model a credit-driven recession, we could have the parameter(s) jump to new values that reduce debt growth, and this then has a knock-on effect on growth.
Although this technique will create a “teaching model” that helps explain why recessions occur, the problem is that it does not help forecasting. We need to forecast the parameter shift that leads to lower credit growth, which is no easier than forecasting a recession in the first place.
We could assume that the “credit growth parameters” are driven by variables within the model. We might then be able to forecast the parameter changes. However, that implies that we could use this technique to predict market developments (and therefore be on our way to yacht ownership).
DSGE models have notorious difficulties with credit. The entire premise of the models is that entities (households, firms) have a starting amount of tradeable goods and services, and then they trade these endowments amongst themselves to reach an optimal outcome.
“Money” in these models correspond to government money, created by policy, and so policy rules constrain its growth. Since those models are typically mainly used to discuss policy, this is not a problem.
The addition of private credit creates inherent problems with the optimisation-driven strategy of these models.
The models are built around forward prices of everything that are traded in forward markets. If firms can buy and sell everything forward, there is no fundamental uncertainty that can lead to defaults.
Since private credit “comes out of thin air,” there is no initial “endowment” for trading. Trying to impose a repayment constraint (“no Ponzi condition”) runs into the problem that if every firm can issue credit, there is no reason for nominal income growth to result in a finite limit if nominal interest rates are bounded. All the entities in the private sector can grow their balance sheets in concert, creating nominal income that allows for debt service.
The models rely upon representative agents that stand in for a sector, or part of a sector. If they are indeed representative, all the firms that are represented would default at the same time.
The methodology is not suited to handle credit risk, and so the best that can be hoped for is adding some heuristic behaviour that allegedly captures credit dynamics. We end up in the same position as SFC models: we need to forecast changes in model parameters.
Agent-based models are in a better position to capture credit dynamics, since they have entities that face the same information limitations that face lenders and borrowers in the real world. As such, it should be possible to create models that generate business cycles driven by credit.
The problem with such models is attempting to fit them to observed data, as they contain an extremely large number of state variables. Although there is research moving in that direction, the fitting problem is always going to be difficult to solve.
Mathematical economic models have a difficult time with modelling credit dynamics, since they are driven by herd psychology. They might be able to explain why the economy contracts if lending seizes up, but predicting such a seizure is just as difficult as predicting the direction of markets.
Market commentary I have been seeing has had an unusual split: warnings about the collapse of the bond market, as well as imminent recession. That combination would appear unlikely, given the usual reaction of central banks to recessions is to cut rates. However, it makes more sense if we accept that the United States and the rest of the developing world have diverging fortunes, as well as the reaction to the negative first quarter GDP print released last week.
(This is the first time I ran into anyone using “term spread” instead of “slope.” One could argue that it is a better term than slope — in other contexts, a slope is a rate of change, while a yield curve “slope” is a level difference between two points on the curve. Although it might catch on with the youngsters, I see having two words instead of one, with both of those words having multiple uses within fixed income, “slope” is more convenient for writing. In any event, either is infinitely superior to the practice of (mainly older) economists using “yield curve” to refer to a particular slope, when the yield curve is itself is the plot of yields versus the continuum of maturities.)
Matthew C. Klein also recently joined Substack, and wrote a lengthier piece on it. If the reader has not already gone through the details of the report, Matthew’s piece is a good place to start.
I will largely not comment on what he wrote, other than the big point he notes: there was a divergence between the Household and Establishment job growth numbers.
