Random Thoughts on the Inverted Yield Curve and Debt

There was much discussion in 2018 on a temporary inversion of the two and ten year treasury yields, and it almost certainly contributed to some panic as investors widely view inversions in the yield curve as leading indicators of recession. Surprisingly for those who take the position that market timing isn’t possible (it probably is to some extent, but it would likely be difficult and unprofitable to act upon), investors have good reason to believe yield inversions are meaningful, as they have preceded all nine U.S. recessions in the past 60 years with recessions on average occurring 18 months later, with a high level of predictive power according to the San Francisco Fed.

The knowledge and understanding of the relationship between the yield curve and economic growth seems to be quite low, even among academics. I don’t think there is a large body of research on the topic, and it is likely difficult to assess which are the independent variables responsible for change. Given the state of research and theory on the topic, I also have little understanding of it but wanted to speculate here as to what may be the more important causal factors at play.

The Fed of St. Louis recently published two articles discussing yield curve inversions and whether they merely predict or have the ability to cause recessions- 1, 2. The first discusses the theory of real interest rates and proposes that yields essentially predict future consumption. They note that every yield curve inversion has brought a future deceleration in the rate of consumption and oftentimes a decline in consumption. Their general summary is that real interest rates and therefore yields are representative of expected future changes in consumption and they believe fixed-income investors do a good job of assessing it, and therefore it is usually priced into yields. If a long-term rate is lower than a short-term rate, they feel that investors are pricing in shocks to demand and consumption, as real interest rates should be positively correlated with expected growth.

The second article questions whether yield inversions themselves can affect the economy and cause recessions. They note that banks borrow short and lend long, so therefore a yield curve inversion would likely lead to lower lending and loan growth, as bank NIM shrinks. As stated in the article, they surveyed banks and asked how a yield inversion would affect their lending practices, with most banks saying it would either “tighten lending standards or price terms on every major loan category”, as they would likely have to in order to remain as profitable as before. If a bank’s NIM shrinks and defaults remain stable, profits would decline. As theorized, the explanations that banks gave for their likely tightening of loan practices under inverted yields were that each loan would be less profitable, banks would become less risk tolerant, and on par with the first article, banks would view a yield curve inversion as a sign that future growth will decline. The authors then inserted this graph below which is pretty telling:


It shows that over the past 40 years, the yield spread and bank lending standards tend to move in concert, with bank lending standards generally increasing prior to inverted spreads (shaded periods are recessions). In their own words:

“The chart shows that the net percentage of banks tightening their lending standards on commercial and industrial loans began to rise around the time that the yield curve inverted in 2000 and 2006. Why is this important? Researchers have found that the economy tends to slow after banks tighten their lending standards, suggesting that an inversion of the yield curve could cause economic activity to slow by leading banks to reduce the supply of loans.”

This would conflict with traditional economic theory which assumes bank lending does not affect total demand or output. From this reasoning and data, we can’t say for sure whether yield curves affect lending or whether lending affect yield curves (if lending drops, future growth estimates can drop, causing an inversion). The answer is probably both to some extent, which causes markets to act cyclically-> lending and higher spreads lead to more lending, better GDP growth and therefore higher spreads, while lower lending leads to lower GDP growth and therefore yield inversions and inversions feed-back, causing lower lending. It may not matter which happens first, and they may just kind of happen at the same time without being predictable, but I am inclined to believe that growth in lending is the independent factor and first-mover here. The fixed-income market isn’t creating predictions as to future consumption and real interest rates ex-nihilo; the pricing depends on data and feedback from the economy. This means that the inverted yield curve would be a reaction from investors to economic data as well as current rates, and those data would depend on loan volumes.

The logic used by each fixed-income investor causing the yield inversion is probably a bit simpler. The market is willing to price bonds with a sloping yield as long as growth expectations are high, but when economic data becomes poor enough, they believe the Fed may have to cut rates to stimulate the economy, and therefore they all attempt to lock-in the current long-term rate. This flood of purchases for long-duration bonds would decrease the yield and lead to short-term rates being temporarily higher than long-term rates. Even if short-term rates are higher, if expectations of the future are bad enough and if the potential for future rate slashes are significant, the bond investor may be fine with locking in a 10 or 30 yr bond despite a slightly lower annual yield. Given that interest rates peak at the end of every cycle, they don’t want to be in the position of receiving a slightly higher yield on a shorter-duration bond but in 2 or 5 years having to reinvest it for much lower rates, they just want to punch in the peak rate for as long as possible. This, along with simple bond math, is a primary reason why long-duration bonds appreciate dramatically in recessions.

Every recession in the past 60 years has been preceded by a rise in interest rates, as you can see in the chart below.


The federal funds rate is determined by the Fed, and the FFR acts on all rates but most immediately on short-term rates. The Fed has less control over long-term rates which are controlled to a much larger extent by market forces. As stated, the primary cause of a yield curve inversion in my opinion, is likely when demand for long-duration bonds spikes relative to short-duration bonds as expectations decline quickly, and the inversion could affect lending patterns. Before this however, it could be that as interest rates rise, it increases the price of debt for consumers and businesses and eventually investors realize lending and GDP growth will be dropping in the short-term due to lower demand for debt, which causes the change in expectations as economic conditions deteriorate, leading to the inversion. Investors would therefore see lower growth in lending and lower GDP expectations, and price in a shock to demand sometime in the next few years, which would be seen via an inversion of the yield curve. However the downfall begins, the cycle starts over with the Fed slashing rates in an attempt to increase loan demand and liquidity, which eventually normalizes things.

An interesting thing I found when playing with the data and charting capacity the Fed provides is the growth in money supply in recessions, as well as what happens prior to them. In nearly all cases, money supply growth decelerates and then eventually the money supply drops in recessions, and the M1 supply usually takes a hit. The M1 supply is comprised primarily of currency, coins, and bank demand-deposits. In recessions, the supply of currency doesn’t change, but demand deposits (currently around 40% of M1) usually drop which leads to the drop in M1 and interestingly, the drops start prior to any noticeable changes in lending or yield inversions.


It may be difficult to see in the chart, so I’ll summarize. Prior to every recession shown, and generally years before a yield-curve inversion appears, demand deposits drop off to some extent from their peak and don’t recover until after a recession. From 1980-1981, deposits dropped 11%, with a recession occurring in mid 1981. From 1987-1989, deposits dropped 7%, with a recession occurring in mid 1990. From late 1996 to late 2000, deposits dropped 23%, with the yield curve inverting in early of 2000 and a recession occurring in early 2001. From mid-2005 to mid 2007, deposits dropped 7%, with the yield curve inverting in mid 2006 and a recession occurring in 2008. The drop in demand deposits tends to happen before a yield curve inversion, and the yield curve inversion tends to happen a few years prior to a recession.

Banks create deposits for their customers whenever a loan is requested and approved, as long as they have the necessary reserves (which they nearly always do via inter-bank transactions and short-term borrowings although banks prefer retail deposits which are nearly always much less costly and more stable). Both the loan and deposit are notes representative of the amounts owed to each party, which allows consumers and businesses to transact without cash and allows them to pull future consumption into the present. Banks create liquidity for exchange through their ability to issue these financial instruments. The easiest way to describe deposits would be tokens which people can trade around to buy things and do business and which can be accounted for and used as money but which isn’t currency. Banks allow for exchange without currency and so do not require currency from anyone else- deposits with banks obviously aren’t representative of currency they hold or have received, which is why it can be created and destroyed dependent on loan volumes. The loan and deposit volumes are dependent on both the decisions of the banks and borrowers- if banks tighten lending standards, or if loan demand drops, deposits will drop.

As for the drop in deposits being a leading indicator, my best guess here is that as interest rates rise (price of debt rises), loan demand and therefore liquidity and present consumption all drop, and so loan growth decelerates. As the Minneapolis Fed speculated in the late 90’s when deposits were dropping, another likely reason is savers moving funds from bank deposits to higher yielding assets such as bonds and stocks and this likely causes higher competition among banks for short-term funding, which reduces NIMs and loan growth. So as interest rates rise and the boom continues, loan demand drops as the cost of debt rises, and demand for deposits also drops relative to other assets, which can impact the loan supply via lower bank profitability per loan.

As loan growth decelerates and growth expectations drop, banks tighten lending standards which further reduces lending and growth. As suggested before, both booms and busts would be self-reinforcing through this manner. While everything may look fine if total loans outstanding continue to increase and charge-offs are in line with historical norms, the drop in demand deposits could be indicative of lower demand for present versus future consumption and therefore lower GDP growth in the short-term as loan demand eventually flat-lines. I would imagine that loan growth accelerates through to a peak, decelerates and then during the recession when we have moved from decelerating loan growth to declines in loan volumes, the rate of loan decreases will accelerate as lending standards tighten and soon after a sharp decline hit a trough and the cycle starts over with loan volumes increasing once more.

Like the yield curve inversion, there are no false positives for recession prediction after deposits drop 5%+, but using the prediction to take advantage of market prices likely wouldn’t work out well. In many cases you would profit enormously, such as in 1988 and 2006, but on average you probably would have been worse off by withdrawing funds from equities by leaving markets before prices collapsed (see 1996-2000).

As you can see in the period 2010 and on in the chart above, demand deposits and therefore also M1 have experienced historic increases. M1 has expanded around 120%, far more than in any other expansion and primarily due to a 300% increase in deposits. It likely has much to do with near-zero interest rates for over 8 years which may have pulled forward quite a lot of future consumption into the present, leading to much higher loan volume. Why wait to consume in the future, when you can consume the same amount now for only a very nominal fee? I suspect that the low cost of debt has skewed the balance of present versus future consumption quite heavily in favor of the former and it may have the potential to lead to unexpected (bad) outcomes - particularly because I seriously doubt that on average it has been used productively by either consumers or firms.

As you can see in the two charts below, both commercials loans and consumer loans have expanded rapidly with the historically low cost of debt over this boom.

industrial loans.PNG
consumer loans.PNG

Similar to the increase in M1 and demand deposits over this current expansion from 2010 on, private loans have expanded at a similarly rapid pace. The rise in debt in both absolute terms and relative to both GDP and the average equity of borrowing parties will almost certainly impact future consumption as it cannot keep up at this rate forever. With interest rates having risen and recently been stabilized (according to recent Powell statements), I view it as highly unlikely that deposits will pick back up in the next few years after their recent declines. I would expect historical outcomes to come into play, with deposits dropping in the short-run and with loan volumes decelerating to eventual declines.

Is there any way to profitably use this potential information, if it is indicative of the workings of business cycles? Probably not in equities markets, or at least I would say it wouldn’t increase returns proportionally to the risk you would be taking and with the potential for equity and real markets to disconnect, as you could miss out on quite a lot by trying to use predictions to time things. As of right now, assessing and following real-time data would likely have much more use in creating effective policy and better economic theory in order to understand outcomes. Unfortunately, the field of economics is short on observation, evidence, and data, and so its field of research is very limited in comparison to scientific fields in which studies can be carried out. The cause and effects of economic cycles is one particular topic in which academics and therefore the rest of us are highly uninformed. There need to be more academics and policy creators who are willing to create hypotheses based on data in an attempt to falsify much of the conventional thinking. It seems like the recent research of some regional Fed banks is a good step towards this outcome.