Several friends and clients have asked for my thoughts on AI and the mania in both investment and stock valuations. While I have previously noted the shift from free-cashflow financing to debt financing, I held off writing a longer piece until I felt that I had something substantive to contribute. There is already a large amount of excellent commentary out there, from both top down and boots-on-the-ground perspectives1.

And then I stumbled across an interesting interview with Richard Sutton, the “father of reinforcement learning”2 and winner of the 2024 Turing award. Mr Sutton thinks Large Language Models (LLMs) will prove to be a dead-end: albeit in the very specific sense that we will eventually develop something better than LLMs. In his 2019 essay The Bitter Lesson, Sutton argued that architectures which do not scale cleanly with computing power have always lost out to those that do, over time.

If Sutton is correct, this would have profound implications for the current AI race. Current LLMs do not scale efficiently with compute3 - improvements thus far have followed power laws with diminishing returns, and there are other bottle necks such as memory, training data, and human feedback loops. Exponentially more resources are required for each incremental gain - hence the current bun-fight for GPUs, power, land, water and so on.

But while current AI leaders are competing with each other at ever increasing costs, they risk being outflanked by approaches that can overcome constraints on data-sets, human oversight, and so on. Through the lens of Sutton, the question is not will this happen, but when it happens.

In this article I walk through Sutton’s core position, why I agree with him, and the implications for how we should think about the AI mania.

Share

How LLMs Learn: Practice makes performance

Most current LLMs are trained in three stages.

1. Pre-training: Language patterns

2. Supervised Fine-Tuning: Following Instructions

3. Reinforcement Learning: Aligning with user preferences

In the first stage, the LLM is fed a huge amount of text and learns, through repetition, how to make reasonable guesses as to the next word in a sequence.

Consider the example: “The capital of France is”.

Depending on your age, the most likely next word is either Paris or “F”. But other words are possible: “to”, or “near”, or “considered”4. The LLM will have a measure of how likely it is that various words continue the fragment “The capital of France is” (or similar fragments) and, in the pre-training phase, the LLM will iteratively update its internal parameters until the distribution of its answers aligns with the data set. Not just for this phrase, but for all phrases in the training data.

The LLM will also learn to use the broader context in which the phrase was found to help guide it. And, like trying to sing along to a song where you half-know the words, this process gets better with more training and when given more context.

Thus, we hit our first non-compute bottle-neck for LLMs - once you have “drunk the internet” generating additional, high-quality training data becomes expensive. And this challenge has grown since LLMs became widely available, as more and more potential training data has been AI-generated. This is intrinsically less valuable than human-generated content and can potentially degrade the training process.

Supervised learning

After pre-training the LLM can respond to a prompt by continuing the conversation. In the next stage it learns to not just continue a string of text, but to respond - to answer a question, analyse a text, engage with a statement.

This is done by using a smaller, carefully (usually human) curated dataset, specifically based on the call-and-respond pattern seen in human conversations (as opposed to human-written text). In this sense, the learning steps are the same as those used in pre-training: the LLM predicts what it believes is the most probable answer, and this is compared to the actual answer in the training data. The LLM then updates its internal model in order to shift its answers towards the pre-determined answer - not for just one prompt/response pair, but in batches.

When thinking about the future of LLMs, this curated data set is another significant bottleneck. There are initiatives to ease it, by using AIs to train other AIs, but at the moment those outputs still need reviewing even if they are not human generated.

Reinforcement Learning from Feedback

In the final phase, the LLM begins to learn what humans like.

In essence this is the same as the preceding steps, except now instead of assessing its answer against a pre-set solution, it generates multiple different answers and human assessors rank their preferences5. Preferred answers are then fed back to the LLM, which adjusts its internal parameters in the usual way to increase the probability that it will generate answers that are perceived as more “helpful”. This is similar to the supervised learning stage, but instead of there being a single target answer for each prompt the feedback is based on a range of possible answers.

This stage, in particular, is where LLMs quickly became polarising once released to the general public - for many subjects there is no single hierarchy of preference. Politics is an obvious example and it’s easy to see how prompts such as “Which US president was the most corrupt?” are likely to divide a userbase, depending on the trainers’ political preferences.

But more fundamentally, this stage opens a window into some of the fundamental short-coming of LLMs - as discussed on the Sutton interview. For example, human interactions have no single, consistent value-maximisation function (goal) - the “best” answer to a question about US Presidents depends on who you are talking to, whether the conversation is public or private, and so on. As humans, we intuitively know that the “best” thing to say depends on who we are talking to, how they are feeling that day, what our desired outcome is, and so on. Humans are complex.

In these situations, LLMs do not solve for “truth” (nor could they) but for likeability - often framed as usefulness. The corporate goal behind many LLMs is to develop AIs that people choose in preference to competition. Some users base this choice on factual accuracy, or usefulness, but many users are likely to select for other characteristics. This, as we will discuss below, is a severe constraint on what LLMs will be able to do even as we scale compute to higher highs.

The Bitter Lessons (from Chess and GO)

The Bitter Lesson is a fantastic and, critically, short essay that simply states that over the history of AI/Machine Learning it has, thus far, always been the case that while different methods and strategies have yielded early success, in the end, the winner has always been the approach that best leveraged raw computing power.

In part, this reflects a historical “Moore’s Law” type effect that has seen computing power expand at an exponential rate. As such, methods that have sought to extend the cutting-edge of software intelligence through human-ingenuity have always been over-taken by the growth in computing power and by methods that better leverage it.

Two canonical examples are chess and (more recently) Go.

Early chess engines leveraged human-generated algorithms to evaluate positions. The first computer to beat a reigning world chess Champion (Garry Kasparov) combined human algorithms with “deep search” - playing out millions of possible future moves, to explore as much of the local probability space as possible and select the path that was most likely to win.

But AlphaGo Zero (Go) and AlphaZero (chess) learned not by being told which positions are better, but by playing games against “themselves” (another AI) and working this out themselves. Their training processes were not bound by the set of all games ever recorded (pre-training data), or by a special set of games that exemplified specific strategies (Supervised Fine Tuning), or even by the number of games it could play with humans (Reinforcement Learning from Human Feedback). Because the computer could learn by playing itself, the training data, and the training agents themselves, became themselves scalable with compute.

And because every element of the learning chain scaled, these systems were able to not only overtake human-centric strategies but ultimately to dominate these spaces.

Going down the wrong rabbit hole

Current LLMs do not appear to be capable of transformative applications that might justify the lofty valuations we see in AI stocks. But there is a belief, expressed by CEOs and stock market valuations, that we are close to a breakthrough. This belief justifies a sprint to cross the line first - to bridge the gap with vast amounts of training compute, even if current LLMs do not scale efficiently along this axis.

However, this is far from certain. Adding compute does allow models to be trained more quickly (faster training cycles) - but these gains have power-law relationships, so even the current, GDP-moving levels of investment will only generate incremental progress if past scaling relationships hold. Based on some rough estimates, the hundreds of billions of investments announced could increase training compute by 20-50x for next generation LLMs - but this would only increase performance modestly.

And if these modest improvements are not sufficient for LLMs to generate the required critical breakthrough (as yet undefined), then the perspectives laid in The Bitter Lesson are both deep and profound.

[Note that in this diagram the compute scale is logarithmic - each incremental step requires 10x more compute]

So, the question is: are the current generation of LLMs the equivalent of “Deep Blue”? Destined to be surpassed, but already good enough (or soon to be good enough) to reach a high level of adoption and value generation. Or do the problems that LLMs are trying to solve require an agentic equivalent of AlphaGo? Approaches based on fundamentally different methods: agents that don’t just apply human-derived models (as LLMs do) but derive their own estimates of value.

If we are solving a problem at least as complex as winning at Go, then LLMs are a dead-end. And the problem is likely even more complex than Go: unlike games, human conversations have no defined win-condition. When LLMs have been allowed to talk to each other without objective feedback, they devolve into nonsense because they lack an agenda or goal. It is telling that in areas where “correct” answers can be defined, for example in mathematics, there has been much stronger progress in recent years - a strong hint that the Bitter Lesson is again being learnt when it comes to the more humanistic applications of LLMs.

My position is that LLMs will not make the required break-through on the basis of brute-force investment in compute. If there is a breakthrough it will come from a genuinely new and innovative learning method that is not currently being widely used. Further, based on the Citrini report, I have changed my position somewhat on AI investments, in the sense that I think we may be slightly further into the investment bubble than I had thought. Debt is being issued (most recently Meta) to finance projects, many of which are already under construction.

If these projects fall flat, it will not just be a case of marking down future revenue growth for projects never undertaken (equity re-rating), but there will be real assets to be written down (capital destruction). And the write-downs could be brutal - in some cases Citrini identify that on-site generation turbines and other construction elements are being selected on the basis of what is readily available. This seems likely to (further) accelerate depreciation of these sites in the medium term.

Admittedly, from an equity perspective, the Sutton position tells us nothing about what AI stocks will do in the next six months. But it does put some pointed questions over those 30-year bonds - and that is good enough for me, as a credit guy.

Leave a comment

Money is a funny thing.

It’s so deeply embedded in most societies that the vast majority of us are barely able to explain it. Even within finance, the mechanics of money creation are frequently misunderstood - even central bankers frequently misspeak on the topic.

In part, this lack of understanding is due to how rarely the deep mechanisms of money matter: much as everyone reading this can use the internet without explaining how it operates, people can use money without needing to understand it. But one reliable trigger for these deeper conversations is when money markets1 start to move. Financial markets remain hyper-aware of movements in the “monetary plumbing” after a decade and a half of banking crises.

And so it has been this week. A few money markets have shown “strains” and the Fed has reacted by ending its Quantitative Tightening program. Cue heavy breathing and breathless reporting.

I disagree. My view is that what we are seeing is monetary policy traction as the Fed transitions away from a “floor system” for reserves (where the Fed maintains a huge excess of money in the banking system), to one where the Fed’s balance sheet is as small as possible, but no smaller. This requires a move higher in money market interest rates and some marginal borrowing of reserves. So the moves being observed were expected - they are a feature, not a bug.

In this article I provide a (very) short overview of how banks make money, how that process is used by central banks to generate monetary policy “traction”, and how this can explain recent movements in money markets.

How banks make money

We generally receive money in return for goods or labour, or perhaps as a gift from a friend or family. But where did they get it from: if we followed the trail of transactions all the way back in time, whence came the money that we hold?

Historically, most legal tender was physical currency2. This was an important type of money because it was, amongst other things, used to pay taxes and it came from the mint, or was printed by the treasury. But today, most money is digital - a number in your bank account that goes up and down when you send or receive currency. Where does that come from? Money, it turns out, is a surprisingly philosophical thing3.

In modern banking systems, which the US and Western Europe have had for at least 500 years, the answer is simple and somewhat unexpected. The money in our bank accounts comes from banks.

This fact creates a huge amount of confusion and misinformation, but banks are where almost all of our money comes from. Specifically, money is created through lending: whenever a bank originates a mortgage, or a credit card, or a company loan it creates that money. This is because, under double-entry bookkeeping, the loan that the bank makes is an asset and there must be an offsetting liability, which is the deposit that appears in your bank account.

To be specific: when banks make loans, they create bank deposits and these are a type of money - indeed bank deposits are the largest component of monetary aggregates4. Essentially all of the money you own as an individual is in the form of bank deposits, with the notable exception of any physical cash you may still carry.

The ability to make money is clearly a powerful one, so it is tightly controlled - for example, only regulated financial entities are allowed to make loans in this way. And broad policy tools are needed to avoid run-away lending, or “excessive credit creation” as central bankers call it. A system-level constraint on the total amount of money being created is one of the roles fulfilled by bank reserves.

M0 - the origin of money

Historically, one way that central banks would try to keep bank lending broadly under control was by requiring that banks retain a certain amount of cash “reserves” against the deposits of their customers. This was to help ensure that if customers tried to take their money out, for example to buy a house or a car, that the bank would have enough currency on hand to meet that demand.

This was often kept in large, elaborate, and expensive vaults

But reserves also acted as a constraint on the total amount a bank could lend. A bank with $1mn of cash on hand and a hypothetical 10% reserve requirement would only, legally, be allowed to make $10mn of loans. From this constraint was born the concept of the money multiplier - the relationship between the amount of reserves in a system and the total amount of credit in an economy5. Because most of these reserves were physical currency, it was possible to constrain the amount of coins minted and bills printed, and thus the total size of the banking system.

A critical point to emphasise here is that reserves are a bank asset - they sit on the same side of the balance sheet as loans but on the opposite side to deposits. This is the core of why reserves and deposits are fundamentally different types of money. Another special feature of central bank reserve currency (reserves) is that only banks can hold reserves6. The only way for anyone other than a bank to own central bank money is via physical coins and notes.

Even as money became increasingly digital, this remained one of the key ways that central banks managed the banking system - by controlling the amount of reserves in the system7. A bank that wanted to grow would need to borrow more reserves from the central bank or from another bank, because while banks can create deposits, they cannot create reserves8. A constraint on the total amount of the reserves in the banking system, along with the reserve requirement, limits total size of the banking system without requiring control of individual bank lending policies.

The lending of reserves to the banking system also creates a second lever for central bank policy - how much they charge banks to borrow reserves. Indeed, this is how central banks set interest rates. When the Fed, ECB or BoE cuts interest rates, what they actually do is lower the cost for banks to borrow reserves9. If reserves are made cheaper to borrow, then banks can pass that lower cost on through their lending rates. Thus, the central bank borrowing rates are expected to be transmitted to other lending rates - such as your mortgage. But that is far from guaranteed: the ECB struggled for almost a decade with a “broken transmission mechanism”, during which ECB policy rate cuts were not transmitted to borrowing rates for households.

Ceilings, floors and corridors

Pre-GFC reserve scarcity was, more or less, how the Fed, ECB, and BoE conducted monetary policy. The total amount of reserves in the banking system was kept at some level that constrained credit creation and banks would bid for additional reserves at weekly or monthly lending operations10. In parallel, banks lent reserves to each other on a daily basis, reflecting their individual needs.

Because every bank could obtain reserves from their central bank at the official rate, there was no reason to borrow from another bank at a higher rate. Hence, the policy rate would act as a ceiling on cost of interbank reserve lending. If the interbank rate moved higher than this, banks could borrow from the central bank instead.

But this system requires there to be a (small) deficiency in M0, i.e. bank reserves, such that at least some banks are bidding for reserves from the interbank market (or the central bank). When the Fed and BoE undertook Quantitative Easing, they injected vast amounts of central bank money (M0) into the bank systems. As a result, no bank needed to borrow more reserves - in fact they were trying to offload their excess reserves to each other and interbank rates collapsed.

In essence, QE threatened to kill the transmission mechanism from central bank policy rates to bank lending rates. In order to maintain some linkage between policy rates and market rates, the Fed introduced Interest on Excess Reserves (IOER), which allowed banks to put their excess reserves back to the central bank as a deposit - the ECB and BoE introduced similar deposit facilities. Overnight, banking systems and policy rates flipped from being a ceiling system to a floor system.

For as long as the system was flooded with excess central bank reserves, created by central bank asset purchases, interbank rates would be anchored by the policy floor. Indeed, several market rates traded below the central bank rate, when participants could not place excess liquidity at the central bank - this was particularly a problem in Europe where non-banks struggled to find a home for large deposits.

Back to the future

Since the pandemic, following the inflation surge in 2022/23, central banks have been trying unwind asset purchases and shrink their balance sheets to smallest possible size - but no smaller. This has involved gradual Quantitative Tightening (QT), where bonds on the balance sheet have been allowed to roll-off at maturity. Because this is the reverse of QE, QT reduces the amount of reserves in the banking system.

Initially the excess reserves in the system were so vast that removing some just made the excess reserves a bit less excessive. But it was always understood (by the Fed) that at some point reserve drains would get “traction”, in the sense that reserves would no longer be excessively abundant. And the effects would be quite clear:

1. Policy rates would no longer be nailed to the floor of the deposit rate

2. Banks would start to borrow again at central bank lending operations.

And that is exactly what is happening: market lending rates have been moving higher (relative to the Fed’s target rates) and US banks are starting to borrow small amounts from the Fed’s reserve lending operations.

The Fed does not want to fully return to a ceiling system, for various reasons that could fill an entire other article. But it is embracing a de facto corridor system, where the target rate (Fed Funds) sits between two policy rates - the lending rate and the borrowing rate. Market rates are expected to move within this corridor in relation to the relative scarcity of reserves, but none of this is a crisis warning.

Rather, this is how monetary policy operates. The movements are a feature, not a bug, and they don’t reflect any “brewing crisis”, but they do represent a transition from one monetary policy regime to another.

Beyond Fed Funds - SOFR and money markets

Adjacent to and intertwined with the conversation about Fed Funds, is a focus on the relationship between Fed Funds and SOFR - the Secured Overnight Funding Rate. To understand what is going on here we need to have an understanding of both and, critically, who participates in these markets.

Fed Funds: A legacy market

Pre-GFC, The Fed Funds (FF) market was where banks traded reserves between each other on a day-to-day basis, in order to optimise their own balance sheets. It was a hugely active market with massive volumes. But after the start of QE trading volumes collapsed - every bank had excess reserves, so hardly of them wanted to borrow.

The remaining volumes in Fed Funds are dominated by arbitrage flows between US Federal Home Loan Banks (FHLBs) and Foreign Banking Organisations (FBOs). Legally, FLHBs cannot earn interest on reserves held at the Fed, hence their reserves are lent out to generate interest income. But because no one actually needed reserves (for most of the last 15 years), the only borrowers were foreign banks (FBOs) who would borrow at rates below IOER/IORB11 and then deposit that money bank at the Fed to earn the spread.

Critically, because reserves can only be lent to other banks (and because of other constraints on the FHLBs) this money can’t go into other forms of lending. For these reasons the Fed Fund rates don’t reflect much other than the capacity of FBOs to arb the spread between Fed Funds and IOER/IORB.

SOFR: Heavily guided by MMFs

While banks no-longer trade much in the way of reserves with each other, there is still plenty of intraday financing that goes on that involves non-banks. As part of the post-GFC reforms, most of this activity moved to secured markets, which affect the pricing of SOFR. Much of the activity is in so-called repo financing: borrowing secured against US Treasuries (which are referred to as collateral in this context).

Critically, a huge amount of the lending in the SOFR market is done by non-banks, in particular Money Market Funds. And because, unlike the Fed Funds market, it is deposits (not reserves) being lent and borrowed, the lenders can and often do take their money somewhere else. In the case of US government only MMFs, the “somewhere else” is principally treasury bills. As such SOFR tends to be more closely related to T-bill yields than other short-rates - when US T-bill yields rise relative to SOFR, MMFs stop lending in repo markets and shift to T-bills (and vice-versa).

Bills, bills, everywhere

And so, we get to what is most likely the crux of the issue. Since the debt ceiling was lifted, the US Treasury has been replenishing its cash reserves by issuing T-bills. Lots and lots of T-bills.

This has two major impacts on money markets.

1. It drains reserves from the banking system12

2. It increases the amount of bills, thus causing them to “cheapen” (yields to rise)

As such, we would expect two things to result from the US Treasury issuing a lot of bills. First, that there would be some additional upward pressure on Fed Funds due to reserve drain - another reason why the Fed would have wanted to end QT, as the Treasury has in effect done a chunk of QT, independent to the FOMC.

Second, T-bills should cheapen as their supply increases. And we have seen this too. Further, because T-bills are a close substitute for repo and repo is what prices SOFR, we would expect SOFR rise as well. Given that the SOFR market has real price formation while Fed Funds is an arbitrage machine, you would also expect SOFR to underperform (rise more than) Fed Funds - and it has.

Share

There are too many T-bills and the Fed will have to buy some

Put in the proper context, the recent moves in money markets seem rather mundane. Yes, there are important changes taking place in money markets and Fed is alert to them, but only in the sense that it has been waiting for signs of reserve demand as the signal to halt QT. Exactly in line with this week’s announcement.

The Fed is trying to navigate its balance sheet to a size that is as small as possible, but no smaller. As a barometer, the FOMC intends to track the extent to which banks demand reserves by borrowing at its Standing Repo Facility (SRF) and when interbank lending rates (i.e. Fed funds) start to drift towards the upper end of the policy corridor. As such, the Fed is unlikely to be (or at least should not be) “alarmed”. But from here the Fed will be seeking to maintain the current level of policy traction, which will require it to deftly navigate the ebb and flow of demand for reserves within the banking system.

Given that banks continue to grow, that almost certainly means that the next step for the Fed will be to allow its balance sheet to start growing again - because if it did not then Fed Funds would move closer and closer to the SFR rate and the policy ceiling. As such, to some extent, the Fed will find itself accommodating the US government’s spending plans by absorbing at least some of the bill issuance that is ongoing - quite possibly within 2-3 quarters.

I look forward to posts about the Fed restarting QE in due course.

Leave a comment

Not wanting to let US credit markets have all the fun, in recent weeks Swiss courts ruled that the write-down of Credit Suisse Additional Tier 1 bonds (AT1s or CoCos) lacked legal basis. While Swiss legal rulings are not often high on my reading list, this one caught my eye because of its significance for the AT1 product, the upside for owners of the claims, and because it’s an opportunity to assure you, dear reader, that I was right… eventually.

“On 1 October 2025, the [Federal Administrative Court] passed a partial decision in one of the appeal cases. It confirmed the complainants’ right to appeal and revoked the decree of 19 March 2023. The FAC has not yet decided on the reversal request. 

This decision is subject to appeal to the Federal Supreme Court.”

Post-GFC bank regulation has generated a layer of bonds referred to as “Additional Tier 1 capital” for banks. These bonds can be written down or converted into equity outside of court bankruptcy proceedings. While the resolution toolkit has steadily evolved over 15 years, Credit Suisse was one of the first opportunities for regulators to play with their new toys and they were not shy about doing so.

In fact, they were so eager that they likely triggered the write-down of CS AT1s before they had legal capacity to do so.

That is the core of this case - did FINMA jump the gun? I have always been of the opinion that FINMA operated outside of their legal powers that existed prior to the emergency ordinance passed on 19 March 2023. More importantly, I’ve long argued that recapitalisation was the wrong tool for CS: it suffered a deposit run, not a capital hole. As such, the SNB was the body that should have been co-ordinating the rescue - but it appears that both FINMA and the SNB were overly cautious. This is likely a result of the post-GFC backlash against “bank bailouts”.

So how did we end up here, and what are the key lessons for AT bondholders and for bank regulators? In this article I lay out the historical contexts that led to the AT1 write-downs, why I believe that FINMA overstepped its powers, and how this court ruling could provide some necessary counter-balance to regulatory powers.

Bank AT1s: A (not so) little background

In my view, understanding the write-down of Credit Suisse’s AT1 starts with the historical backdrop of why European banks issue AT1 bonds at all. Like most of regulatory framework, this reaches back to the 2008/09 Global Financial Crisis.

How Old T1s failed

Subordinated bank debt is not a new idea: pre-GFC, banks issued capital securities in the form of Tier 1, Upper Tier 2, and Lower Tier 21 debt. These securities were, by and large, regular unsecured debt that was subordinated in bankruptcy - if a bank was liquidated by the courts, these bonds would only be paid after other creditors. Hence, they were at increased risk of losses if there were insufficient proceeds from asset sales2. Subordinated bonds also had features such as giving the issuer the option to suspend coupon payments or defer the final repayment date, creating additional risks to the timings of cashflows.

During the GFC, however, these capital instruments by and large failed to provide loss-absorbing capital to banks for a number of reasons. One of these, that should not be brushed over, is that the potential losses that banks were exposed to during the GFC were so wildly large that no amount of capital could have avoided wide-spread failures without government and central bank interventions.

A second issue was that even if there had been enough subordinated bond to absorb all the losses, pre-GFC capital instruments were designed to provide loss-absorption in a bankruptcy, i.e. after a bank had failed. Because widespread bank failures were, rightly, not allowed during the GFC, these bonds were mostly untouched bringing into question their real value as capital. Even where there were other levers for the banks, such as deferral of coupons and maturity payments, these were either too small to matter or came with poison-pills3 that dissuaded banks from using these options to their advantage - along with the associated reputational risk.

But the biggest issue was that governments had screwed up the rescue of banks. Because governments had been determined to prop-up banks before they failed, they had injected huge amounts of equity - in the UK this was the nationalisation of RBS, and similar actions were taken around the world. Governments inserted themselves at the bottom of capital structures.

This created a massive problem when banks needed to raise further capital in 2010-12, in response to the Eurozone crisis and the ratcheting up of capital requirements. Regulators wanted banks to raise capital, but European bank stocks were trading at large discounts to book value and governments often didn’t want to be further diluted.

One option was to focus on issuing AT1 bonds, but old-style instruments would be legally senior to the outstanding bank equity holdings. Governments had left themselves subordinated to bondholders and on the hook if there was any future crisis.

Legally senior, structurally subordinated

In light of all these issues, AT1s ended up with some extremely odd characteristics. Some of this was straightforward; features such as dividend stoppers and incentives to call such as coupon step-ups were phased out in order to make banks more willing to delay payments to bondholders if they needed to preserve capital.

Likewise, there was a strong push to make AT1s operate as “going concern” capital: to allow losses to be imposed on bondholders before banks go into bankruptcy. This is where the “contingent” convertible element of CoCos came from. Initially bonds had triggers (based on capital ratios) written into their documents, with some banks issuing both “high-trigger” and “low-trigger” securities depending on where the conversion ratio was set.

CoCos also came in two “flavours”: the original conversion structure which turns the bonds into equity and thereby recapitalises the banks directly; and a “write down” structure in which the bonds are written off, thereby indirectly recapitalising the banks4. In either case, the bonds would be written down before bank equity ratios fell to zero - making them de facto subordinated to equity, even if they structurally were supposed to be either pari-passu or senior.

An option to the regulator, not the issuer

This particular feature of AT1s, that they absorb losses before equity, was not an accident. It both served the regulator’s goal that AT1s should act as “going concern” capital - forcing bond holders to recapitalise banks without a full bankruptcy - and resolved the tension with government ownership, by creating a layer of bonds that would take losses before shareholders - i.e. governments.

Over time the regulatory framework continued to evolve and Europe finalised its bank resolution framework. This was a very large, very complex piece of legislation but of particular importance to AT1s was that it gradually superseded the trigger levels written into the bonds with “point of non-viability events” which allow regulators to write-down the bonds essentially at their discretion.

Because these powers kicked in a relatively high capital levels versus the 5-7% triggers written in bond documents, AT1s write-downs moved from being driven by bond documents to being options written from bondholders to the regulator. The capital ratios within bond documents are, in reality, an academic consideration as the bonds would de facto be triggered by regulators before banks formally recognised such extremely low capital levels.

Credit Suisse: Wrong for the right reasons…

In March 2023 Credit Suisse was in severe trouble.

The bank had been struggling with its long-term restructuring plan and had suffered extremely large outflows from its wealth management arm in Q4 2022. With the failure of Silicon Valley Bank (SVB) and Silvergate Bank due to deposit flight, the market became hyper-attuned to any bank at risk of a similar collapse. Suddenly the steady flow deposits out of Credit Suisse became an avalanche, and the bank was at risk of an imminent collapse.

The share price of CS collapsed in the days following SVB, and bonds traded to distressed prices while CDS spreads blew out.

I had a front-row seat of the chaos, and the good fortune to be working with some truly fantastic analysts at the time. It was a fascinating situation, not just because of the “will it/won’t it” question of whether CS would collapse, but because it was an extremely rich credit story where there were many different ways to trade potential outcomes given the number of bonds outstanding.

It was possible not only to take a view on the binary question of if CS would collapse, but also on how that would be done with the new bank resolution tools and where losses would fall. For example, investors could buy the senior bonds and sell subordinated bonds to express the view that if CS failed the loss-absorption would stop at the AT1s/LT2s5.

But one question that bothered me throughout was why bondholders needed to be bailed in at all? Even as analysts discussed on how much capital could be raised in various scenarios across AT1s, LT2s, and Senior HoldCo, no-one could tell me why capital was needed at all. Eventually I concluded that this was because CS didn’t need capital. Yes, CS was under severe stress, but this was a liquidity run not a capital hole.

This meant that CS didn’t need bailing in, it needed a lender of last resort. This is exactly why central banks were created - to backstop banks during a run. Indeed, there could hardly have been a more textbook example of a confidence run. As such it should have been the SNB, not FINMA, stepping in and not with resolution tools, but with central bank liquidity lines.6

To be clear, I was a minority voice - there was high conviction that CS would need rescuing and that this was likely to result in losses for AT1 bondholders. I was left with the strong impression that the regulator was overly keen to use its new bail in toolkit, even if that mean bringing a recapitalisation gun to a liquidity knife-fight. I didn’t like it, but I conceded the point: FINMA might have been solving the wrong problem, but that didn’t mean they wouldn’t do it anyway. After all, AT1s are options written to the regulator, and the regulator has the power to change the rules of the game.

On the weekend of 18 March 2023, CS was merged into UBS in a shotgun wedding and the details were remarkable. The SNB made available nearly CHF 200bn in liquidity, there was a first-loss protection for UBS, and c.CHF16bn of CS AT1s were written down. But equity holders were given CHF 3.2bn of UBS stock - confirming the inversion of the capital structure, as was always a latent feature of AT1s.

It also stood out that this was not done under the existing regulatory framework, but under an Emergency Ordinance. In the following days, several voices argued that FINMA had operated within its legal mandate. My counter was simple: the use of an Emergency Ordinance demonstrated, prima facie, that FINMA and UBS were unsure that the legal threshold to write-down the AT1s had been met.

Even at the time, CS AT1 claims, the ghost of bonds past, traded at a surprisingly high value indicating that a number of investors had a similar view. The merger of CS into UBS had been a hasty and not a result of capital inadequacy: that being so, the write-down of AT1s was legally challengeable. AT1s were designed to recapitalise an insolvent bank, not to triage a confidence run.

…right for the right reasons

This brings us, circuitously, back to the Swiss court ruling.

First, the court ruled that bonds could not be written down under their contractual provisions - in line with my own analysis at the time it found that CS was not under capitalised on the date of the conversion. The court also ruled that there were no statutory legal grounds for the conversion - i.e. that FINMA was not able to write down the CS AT1s under its pre-March 19 powers, aligned with the necessity of issuing an Emergency Ordinance.

In essence, this is a complete validation of my position on CS at the time - Credit Suisse was not insolvent and FINMA didn’t have legal powers to write down the AT1s under the resolution regime7. That said, the AT1s never fell far enough in price to look attractive and, as mentioned above, there were many much more interesting trades to put on at the time with far better convexity.

Finally, the court ruled that the Emergency Ordinance “proved unconstitutional in several respects” - which is critical because if the EO was unconstitutional and there was no other legal basis for writing down the CS AT1s (the first parts of the ruling), then there is a material risk of the bonds being written back up. Which is where all the interest around claims and upside for bondholders has come from.

Share

I hope the ruling stands

The failure of Credit Suisse shouldn’t have happened. A confidence run, triggered by the failure of two US regional banks, shouldn’t have been enough to bring down a large European institution. While the SNB did provide emergency liquidity and a joint statement with FINMA that CS was “in compliance” with capital and regulatory requirements, these statements proved to be too little, too late.

Part of this, in my view, is because regulators were anchored by the new resolution frameworks, which encouraged them to view the issue through the lens of solvency and to push for pre-emptive resolution of CS.

The issue is not the effectiveness of the resolution tools - two huge banking groups were merged over a weekend, with no losses for depositors or tax payers. Further, resolution of CS did avoid systemic contagion across the banking system and it also delineated that the breakpoints in bank capital are no longer theoretical. AT1s were written down, but LT2s were untouched along with senior unsecured. This is also part of why LT2s now trade so tight along with strong technicals, because the market is priced on the basis that if another European bank were to fail, there is a very high chance that losses would be contained to the AT1 bondholders8.

According to various reports and conversations I’ve had, since the court ruling the CS AT1 claims are trading at 20-30c in the dollar, up from 10-12c for much of the last few years. As such markets are putting a low, but material chance that these bonds are ultimately reinstated. I don’t own any claims and I doubt I could ever buy them, but there are reports that at least one retail fund may still own CS claims.

Personally, I believe it would be more than just a technical victory if courts determine that FINMA operated beyond its legal competencies. It would also draw a line in the sand regarding the ability of regulators to impose losses on a whim, and so provide some constraint on the exercise of regulatory options embedded in AT1s. Banks will fail and we need strong regulatory tools for dealing with those events, but their use should not verge into arbitrary and capricious - it is not the role of regulators to second guess if a bank might fail, only to judge that it has failed.

Leave a comment

Gold occupies an odd space in investment circles. Most of the time, sensible folks view the yellow metal as the last refuge of scoundrels and wearers of aluminium hats. It is expensive to trade, priced well above its value as an industrial metal, and generates no cashflows. Legendary investor Warren Buffett is quoted as saying “it won’t do anything… except look at you”.

But every so often, particularly when things really hit the fan, gold shines. Warren Buffett has also said that owning gold is a bit like going “long fear” and, in finance, fear is often synonymous with uncertainty.

It is also true that all assets have a certain quality when their price is going up rapidly - this was true of NFTs in 2021, and it’s true of gold today. So, when the investment equivalent of a conspiracy theory is talk of the finance town, something is most definitely up. Gold is making money, people are missing out, and there is a building narrative as to why it’s ok to jump into the pool even after a +50% rise this year.

Personally, I have always owned a little bit of gold in my investments as a sort of “everything” hedge. This is probably a hangover of starting my career during the GFC, which coincided with the last strong bull-run in gold, or perhaps because my first manager was from Egypt and I’m pretty sure he once mentioned his family having some offshore gold bars “just in case”1. And for 14 years, Mr Buffett was correct - the gold in my portfolio stared at me and did nothing.

But by April of this year, even my IFA had noticed that the dead-horse of investing had started to move. Like all good IFAs, he immediately advised me to consider selling. I considered it and then ignored him - I’d waited 15 years for gold to hedge me against something I hadn’t foreseen, and here it was protecting me against something that I had, namely that the US was EM-ifying.

But gold is up a further 28% since April, gold miners are joining in the fun, and the trade is now clearly both in fashion and attracting momentum inflows - as such I think gold is starting to look toppy.

Based on the analysis below, a run up of 2008/09 proportions would take gold to $5000 (+20%). Balancing this against a 50% year to date gain, it looks like a good time to start trimming into strength - unless you are all in on a near-term collapse of the USD.

Subscribe now

Strong fundamentals, and a dose of momentum

Aside from a general hedge-to-the-unknown, I have held onto gold in recent years for two “fundamental” reasons. First, the increased riskiness of holding UST as reserve assets if you happen to be a potential adversary of the US or Western Europe. Second, growing pressure on the USD as Trump pursues a series of policies that erode the institutional strength of the US.

With reserves like these, who needs enemies?

I think it was clear at the time, and has only become clearer since, that 2022 was a step-change for the world across many dimensions. A land war in Europe, the return of inflation after multiple decades, and a massive influx of immigration to the US and Western Europe as the world re-opened after COVID. In the midst of all the chaos, it was easy to overlook some of the more subtle second-order effects. In particular, I think it is often understated how dramatically the Rubicon was crossed when the Central Bank of Russia’s (mostly UST) reserves were sanctioned2.

This clearly signalled to the world that if there is any chance of being on the wrong side of the US, then US treasuries are not a reserve asset - other nations can and will withhold them. China, and likely several other countries, heard that message loud and clear and immediately started accumulating gold. Interestingly, the price of gold hardly moved at first, but the presence of such large buying flows would have tightened the market over time.

Gold did rise c.30% in 2024, as the market became more confident that the Fed was about to start a cutting cycle, but the price really started to accelerate after Trump’s election. This aligns well with my views on the direction of US politics and the desire of the White House to run the US economy (and inflation) hot. Gold prices were then further lifted over the summer when it was ruled that that US tariffs might apply to imports of gold from Switzerland. This layered on top of a broader depreciation of the USD in the first half of 2025, which also featured in my framework for US trade policy and a decline in the relative utility of the USD as a reserve currency.

More recently, though, it seems that there has been a rush of hot money flows into gold. Part of this reflects the rise of the debasement trade narrative (which is at least passingly similar to my US is an EM hypothesis). But there also appears to be some momentum chasing, as strong returns beget inflows that drive further returns.

These recent moves make me think that it is a good time to ask if price has gone (far) beyond the drivers that started the trend.

Share

Lies, damn lies, and historical analysis

As part of another side-project, I have been looking at how well past returns explain future returns for gold3. In a nutshell, they don’t. If you really torture the data, it could be argued that there is a negative correlation - but only in the sense that high past returns usually indicate poor future returns.

For example, the very large run up in returns in the 1970s, which peaked at 34% annualised in 1980, was followed by 5y annualised returns of -12%. This same pattern was seen in 1987, and 2011 In contrast, periods of low returns have often portended better future outcomes - for example 10y trailing returns troughed at -0.6%/y in October 2022 and prices are up +157% (35.8%/y) since then.

But if you try to beat the data into any sort of stable or quantitative relationship, it’s hard to generate anything defensible. That said, it does seem that qualitatively gold tends to oscillate, over very long time periods, around some sort of nominal “fair value” - which makes some intuitive sense. If I use 8% as a growth rate for “fair value” of gold, I get a chart like the one below4.

While this analysis is basic, it implies that the gold tends to spend long periods of time above or below some (unobservable) nominal trend. Accordingly, the periods in which it moves from below “fair value” to above, or vice versa, make for the best (worst) times to hold the yellow metal. Current values are, as a percent of the “fair value”, now approaching overbought levels similar to those reached in 2008/09.

The qualitative analysis is, I think, sufficient to make the point: gold returns tend to oscillate over long-term horizons - moving from high to low and back again. Given that gold has had a strong run up, approaching 2008/09 levels based on trailing returns, this implies the next move is likely a reversion to much lower returns over the next 5-10 years.

Near-term, a run of gold all the way to the 60-70% “overvaluation” reached in 2011 implies a price of $4800-5200, or about a further 20% from spot prices. Upside beyond that level would mean either aggressively playing short-term momentum, or looking for something akin to the Bretton-Woods de-pegging of 1971.

Sell when you can

Gold has done great and there were good, rational reasons to hold the asset since 2022, particularly if you view geo-political events through the sort of lens that I do.

But, even allowing for the roughness of the 8% return line, gold is now very likely above its historical trend and hence vulnerable to a reversion lower in the medium term. Perhaps more importantly, the narrative on gold has caught up to prices - flipping from the preserve of hard-money advocates and financial-system nihilists, to being touted as a core element of long-term asset allocation (one house is even calling for a 60-20-20 portfolio to replace the mainstay 60-40).

Holding gold from here looks much more speculative than rational.

How far could speculation go? Below I plug in a rise from $4200 to $5000 (c20%) over the next 6-9 months as one possible scenario. Running that forward in time, allowing for base-effects from 2020-2022, shows gold reaching 2008/09 levels of annualised performance. I then leave prices constant beyond 2026, to show how quickly returns could decline for anyone buying today even if prices continue to rise.

While I wouldn’t consider any of the analysis above to be convincing in itself, the mosaic of pieces all indicate that gold’s run up has already been historic and that $5000 looks like an upper-bound for near-term prices, unless you expect something more dramatic than either the GFC or the end of Bretton-Woods. With that in mind, and given that price inflation has bloated the size of my gold position to c.10%, I intend to start selling into any additional strength, while I can.

Leave a comment

European banks have, for many years, been a credit manager’s best friend. After the Global Financial Crisis (GFC) there was a blanket re-pricing of bank risk, justified by both the rolling crises of 2010-2023 and the root-and-branch overhaul of bank ratings and regulations. But, despite these ups and downs, the sector has long been favoured by active investors due to improving fundamentals, secondary-market liquidity, and the ability to leverage fundamental and macro views through the capital structure.

In keeping with this narrative, banks have outperformed since the failure of CS and SVB in 2023 - steadily tightening versus non-financial sectors. With the bank sector now only a couple of basis points wide to non-financials, the natural question is if this can continue (even) further.

As I wrote last week, I do not see much to worry about within credit1, but tight spreads compress the opportunity set away from generic, carry-driven returns2. For some investors this is sufficient - so long as the spread or yield on credit clears an investor’s hurdle rate it can make sense to hold onto high quality credit and collect the returns to maturity. But, for active managers it is necessary to either take a (timely) bearish view or to continue scraping the bottom of the barrel in search of Alpha.

In this context, European banks are likely to remain a popular long while the market is rallying, as they still have 10bp upside versus their 2021 relationships. However, this is highly directional - upside in a rally, but the sector is likely to see both fundamentals and technicals deteriorate when the market turns.

Subscribe now

Banks, Pre and Post

To frame the bull case on European banks, it is useful to understand the history of bank spreads in credit: how they went from the least to most risky sector, and their decade-long redemption arc.

Bank spreads traded at very tight levels pre-GFC, due to a combination of strong profitability, an assumed government backstop, and the spread-tightening impact of structured credit protection selling. Reflecting this, and the sectors’ outsized weight in Europe, the €-IG index traded below 50bp pre-GFC (vs 77bp today) and banks traded even tighter as many had AA ratings at the time.

Post-GFC many things changed - first and foremost the perception of bank riskiness rose massively. At the same time the implicit government support was stripped away when the Fed allowed Lehman Brothers to collapse. In Europe this was compounded by a sovereign credit crisis in which banks in Greece, Portugal, and Italy were buffeted by widening sovereign credit spreads - a contagion from the sovereigns to the banking system (versus the contagion from banks to the sovereign that occurred in Ireland, Iceland, and arguably Spain).

This deterioration in credit quality was crystallised into rating downgrades as many rating agencies constrain a bank’s rating relative to the rating of its sovereign and/or include sovereign support in the bank’s overall credit rating. Furthermore, as bank regulation was overhauled, rating agency methodologies were revised and updated resulting in mass downgrades of European banking systems.

Due to all of these factors, the beta of European banks fundamentally changed. Banks went from being the least-risky credit issuers pre-GFC to both the most liquid and often the highest beta credits post-GFC. This also made them popular vectors for expressing market views, either because it made fundamental sense; e.g. trading Greek banks to express a view on the Greece sovereign debt restructuring, or because they were viewed as a liquid and relatively high beta proxy, for example during multiple elections in Spain, Italy, and France.

The bull case: credit quality, profitability

By the middle of the last decade, the regulatory and credit rating methodology overhauls were more or less finalised, though it has taken almost a decade for Basel 3.0 to be fully implemented, and focus has already turned to Basel 4.

A key motivation behind the post-GFC reforms was to make banks more robust as independent entities, and therefore less likely to need bailouts in the future. While the stripping back of government support was relatively quick, amplified by the declining value of that support in peripheral Europe, the accumulation of stand-alone strength has been a slow process, phased over many years as banks built up capital.

The slowness of this process was compounded by central banks moving to zero (or in the case of Europe, negative) interest rates alongside massive QE. This stuffed the banking system with zero or negative yielding deposits that weighed heavily on their profitability, slowing internal capital generation. Indeed, capital generation (aka profitability) was so weak that European banks had to constrain the growth of their balance sheets for almost a decade. Capital accumulation only began to support larger balance sheets after 2020.

Low bank profitability also had an additional aggravating factor of keeping bank share prices supressed. This de facto reduced the equity cushion at the bottom of the capital stack, making their bonds more risky than they would have otherwise been.

Despite these headwinds, the banking system did transition successfully to the new regulatory model. It also adapted to the low interest-rate environment - so successfully that bank profitability exploded as interest rates rose, prompting new demands for “windfall” taxes on European banks. This has also triggered a sharp rise in bank share prices, which has mechanically improved their capital positions. The corollary is that banks have been using excess capital for shareholder distributions. Nonetheless, the benefits to bondholders in terms of greater protection from the equity layer are meaningful.

Reflecting these fundamentals, banks have seen a steady tide of rating upgrades in recent years as agencies have become comfortable with their new methodologies, including watershed moments such as Greek banks returning to IG. Indeed, the failure of CS showed that post-GFC capital structures provide ample protection to senior bond investors, even if regulators get trigger-happy with AT1s.

The increased “stand alone” strength of banks has had other benefits, including a fall in the beta of European banks to political volatility. It is notable that French bank spreads were less volatile than OAT-Bund spreads (i.e. the French sovereign) during last summer’s elections. While this may seem intuitive, given that the risk was French politics, in previous political crises the beta of Italian and UK banks was close to 1.

A further support for bank bonds is that issuance is unlikely to pick up materially even at current spread levels, as banks remain well funded by deposits at the system level and the unsecured paper that makes its way to bond markets is almost entirely a function of capital/regulatory needs3. Strong fundamentals, falling beta to exogenous risks such as politics, and steady supply is a powerful combination of factors in credit - arguing for banks to trade below the index average (i.e. below non-financials), as they did prior to Russia’s invasion of Ukraine.

Free leverage

Another factor that makes banks a popular sector for active credit managers, and supports bullish views on the space, is that their capital instruments - Lower Tier 2 and Additional Tier 1 (LT2/AT1) bonds - allow investors to take subordinated exposure to issuers. This effectively gives investors leveraged exposure to credit without funding, by moving down the capital structure.

For example, a portfolio manager who is very positive on XYZ bank can leverage that view by buying the LT2s or AT1s rather than the senior bonds. Currently LT2s trade very tight, reflecting both technicals (lack of issuance needs at this level) and also that LT2s were not affected in the restructuring of Credit Suisse. As such they don’t offer much value, but AT1s are a popular and liquid part of the bank capital structure for investors.

Share

The bear case: Second order everything

The bear case for banks is, and always has been, that they are second-order exposed to everything. This was true in the case of the European sovereign crisis, where exposure to European governments weighed on banks massively. It was also true when US shale oil imploded and analysts scoured the banking system for exposure to shale producers. And again during Brexit (and even the Scottish independence referendum).

Banks were exposed to the economic collapse during Covid, they were caught up in sanctions on Russia, there is exposure to Taiwan. US regionals were hammered by the rise in US treasury yields, and now banks are being scrutinised for their exposure to First Brands and other potentially bad credits.

In sum, whenever anything, anywhere goes bad it is likely that one or more banks is involved. It’s the nature of the beast that wherever there is credit lending, there is a bank. This means that when something goes wrong, from recession to reflation, banks are exposed at the second order.

Further the predictability of their issuance, with most of their unsecured supply being driven by regulatory considerations, is also a constraint on the flexibility of their financing. Most banks are pretty much required to refinance their unsecured debt regardless of market conditions. This means they have to keep issuing even in a weak market - which we saw clearly in 2022.

So when things do go wrong, the strengths of the sector tend to unwind on both legs - banks will be exposed (be that exposure small or large) and they will keep issuing into a weak market. Hence, it’s likely that banks will trade poorly in the next draw-down, even if they are not at the epicentre.

Make hay while the sun shines

For most of the last decade, I have been well aligned with fundamental colleagues in viewing banks as a long-term redemption story, as institutions have adapted to new regulatory frameworks and layered their capital buffers. Banks underperformed in 2022 due to various concerns and poor technicals (as the sector continued to issue while non-financial supply stopped), but I retained my view that the sector was under-rated - a view validated by the wave of upgrades over the last 2-3 years.

While the spread between banks and non-financials is now essentially flat, there is still room for compression. The sector traded 10bp through non-financials pre-Ukraine, and there is no fundamental reason why that couldn’t happen again. The main challenges are technical ones, with unsecured bank spreads bumping up against other markets such as covered bond spreads and EGB spreads. But the ability to take a leveraged view on the sector via AT1s is likely to win out over time - particularly if the sector continues to lose its beta to political risk, as it has done in recent years.

Yet, at its core, this is a beta-play. A view that as long as the market continues to rally, bank credit can continue to outperform by virtue of its beta and counter-cyclical issuance trends. There is no positive convexity here - as and when the cycle turns, the factors that have supported bank outperformance are very likely to operate in reverse, causing the sector to widen materially versus the overall benchmark. While bank beta to political risk has fallen, its beta to the credit market has not.

As such the sector is something of a “one way bet”.

Leave a comment

Credit spreads are back near their post-GFC tights, while equities make AI-hype highs during a US government shut-down. At the same time, one or two credits have blown up quite publicly, triggering the perennial question of whether credit markets are complacent and if we are on the precipice of the next down cycle.

Adding to the tension, non-credit people are waxing lyrical about how credit markets are delulu, while credit market participants are vocal that this time is different given strong fundamentals, yield-driven inflows, and absolute conviction that Private Credit is fine-actually-so-could-you-please-stop-asking.

In sum, it’s a strange time to talk about credit. Within credit the topics are rather dry, idiosyncratic, and the high-level conversations can be boiled down to two questions: how far can this go; and when will the music stop? In contrast, the outward facing conversations have been about arguing that it’s fine to keep allocating to credit (in particular whatever part of credit I happen to operate in).

So which camp is right, is credit a good investment or a pile of dry kindling? As I walk through my own risk framework, my sense is that both camps are looking in the wrong place - the major risks lie outside of credit. This makes the market even harder for credit investors to navigate, as risks are likely to emerge from blind spots.

Share

What goes wrong, when it goes wrong

Let’s start with why I don’t think credit is a risk to itself.

There are three key drivers of credit spreads: credit quality; risk aversion; and market technicals, including liquidity. Liquidity is mostly an internal (credit specific) factor, while risk aversion is largely external (a macro factor). Credit quality is a mixture of both micro- and macro- concerns, with their relative importance varying as we move through the ratings spectrum.

Each of these factors then has its own lower-level drivers, some of which are internal, but most are external and many of them overlap. For example, growth tends to push down on both credit risk, as corporate earnings improve, and risk aversion: hence credit spreads respond strongly to improving growth. Issuance tends to worsen the technicals in issuer’s curve but can either be positive (liquidity extending), or negative (leveraging) for credit quality.

Because of these broad concerns, spread cycles (as opposed to default cycles) tend to fall into one of three categories, which are again often overlapping:

1. A synchronised deterioration in credit quality

2. A broad increase in risk aversion

3. Liquidation of credit portfolios

For example, in 2022 the initial move wider was driven by a sharp rise in risk aversion, as Russia invaded Ukraine and Western Europe struggled with the sharpest tightening of monetary policy in decades.

As sanctions on Russia and attacks on physical infrastructure increased, European gas prices spiked causing huge mark-to-market losses for gas and electricity sellers who had sold-forward1 their generation. Many were technically insolvent, which triggered multi-billion loans from governments to avoid a liquidity crisis in the energy markets. At the same time, the Real Estate sector saw a massive erosion of credit quality due to rising interest costs that they were not able to pass through to tenants for various reasons. There were also fallouts for Russian government bonds and corporate credits with assets in Russia that became stranded due to sanctions (e.g. BP).

At its zenith, the rise in interest rates triggered the LDI crisis and a concentrated wave of forced selling in the gilts and sterling credit markets, due to margin calls. This liquidation flow also marked the wides of credit spreads in Europe.

This framework provides a lens through which to understand what might go wrong enough to justify being bearish on credit spreads.

Credit quality is deteriorating, idiosyncratically

Starting with credit quality, there is little to see here at the big picture level. Despite the current attention2, issuers going bankrupt is quite normal. Even in the good times, default rates for High Yield credits do not go to zero. Current default rates are neither historically high nor trending higher at the macro level.

The question is not can we find isolated examples of corporates rolling over, but whether there is any synchronised build up in either operating leverage or a deterioration in earnings. This is exactly what caused the European Real Estate sector to implode - a multi-year period of excess bond issuance followed by a sectoral spike in debt costs and a downturn in rents and property valuations. In that sense, the First Brands example is a touch more concerning, though the excellent FT coverage highlights that the founder has previously been accused of financing-related fraud.

Away from the two auto-finance blow-ups, the obvious place to look would be in anything AI-related. Thus far, most of the really ridiculous numbers being thrown around in this space have been either the aggregation of multi-year ambitions, or financed from equity and free-cash-flow. However, recent deals ($18bn from Oracle, $10.5bn from Meta) represent a shift from equity financing to debt financing and the start of a leveraging cycle.

To riff on some of the excellent equity analysis out there, the hyperscalers are shifting from being free-cashflow monsters to highly capital-intense beasts. As a result, they have started to take on debt as investment more than consumes their cashflows.

Turning back to Private Credit and First Brands, there are certainly also concerns over the re-emergence of pre-GFC type debt structures (I’m looking at you, SRT) but I view that as a somewhat separate, albeit not unrelated, issue. Corporates exited 2021 with balance sheets in surprisingly pristine condition - it seems unlikely to me that we have built up enough debt in the system to cause catastrophic fragility.

Of course, there is always the risk of fraud - but by definition that is idiosyncratic.

Demand will create issuance

What about the risk of a liquidity crisis or deteriorating technicals? As I have been arguing for years, higher yields have clearly led to higher demand for credit - even if the marginal source of demand in public bond markets has shifted away from institutional money and towards retail/household investors.

A factor that will further support this dynamic, and has already had some attention, is the impact of coupon reinvestment. With higher yields now holding for several years, the stock of low-coupon bonds issued in 2018-2021 is dwindling and being replaced with high-coupon bonds issued since 2023. As a result, the cashflows that need to be reinvested each month are rapidly increasing, creating follow-on demand for bonds from existing holders. In essence, credit no-longer needs inflows for demand to start outstripping supply - the market needs to grow c.3-4% per year (c.$250-350bn) just to keep up with reinvestment demand.

Institutional money, in contrast, is being aggressively wooed by Private Credit, but this still supports public markets indirectly, by competing for issuance. This is most easily observed in European High Yield, where the bond markets compete directly for issuance with loan markets (where CLOs have outsized influence). It is also becoming a factor in investment grade markets, as demonstrated by the £4.5bn deal between Apollo and EDF - the equivalent to around 1 month of issuance in £-IG.

Institutional money is also directly supporting credit, and fixed-income more broadly, through rebalancing mechanics as equity valuations inflate. For any sort of strategic asset allocation framework, the persistent outperformance of equities will leave funds over allocated to equities, triggering rebalancing flows out of stocks and into bonds. We used to refer to these as “happy rebalancing” and the implication is that the rise in stock markets is not just pushing down on risk aversion (positive for spreads), but also creating spill-over demand for credit. This is particularly the case for institutions such as US pension funds.

In sum, there is no reason to expect demand for credit to abate - more likely it will find an equilibrium by inducing more borrowing. That credit accelerator will, quite likely, support the current US and global expansion for longer than expected, even before we consider the US’s fiscal and monetary behaviour.

The US is hyper-stimulating its economy

If credit quality is holding up at the macro level and demand is outstripping supply, then the most likely trigger for a drawdown would be sharp uptick in risk aversion - and this would require an exogenous trigger, such as a growth slowdown.

Now, you are probably already familiar with my views on the current White House. But while I am not a fan of the current administration, I’ve avoided being excessively bearish on risk on the basis of my political distaste - in large part due to a core view that the Trump administration would seek to engineer a blow-off top in US assets, by hyper-stimulating the economy.

This seems self-evident to me when looking at both the fiscal policy and monetary preferences. The One Big Beautiful Bill is stimulatory (it increases the deficit): though it’s fair to argue that this is largely balanced out by tariffs (unless Trump follows through on his idea to cut everyone a tariff cheque). Likewise, we all know Trump’s position on the Fed in terms of both of its independence and where he believes policy rates should be3.

Thus far, my views have been largely validated - the Trump administration has found levers to pull to stimulate the economy fiscally and will likely get further support on the monetary front, despite inflation being on a steady upward trend. Responding to this inflationism, along with other stimuli, US equities are tracing through the blow-off top I always believed that Trump wanted to engineer.

Circling this back to our framework, VIX is at the lows and equities are at the highs. There is simply nothing to see here, for credit, for as long as this persists. Credit spreads are extremely tight, yes, but they can stay at the lows for some time and we have already had a number of head-fakes in this cycle.

How it unravels

Holding this analysis up as a mirror to the market generates a frustrating return on our analytical efforts.

First, there isn’t any clear, structural deterioration in credit quality or corporate balance sheets. If there is a bubble, it is in AI investment, and that space has only just started to see the sort of mega-issuance that could be problematic given the size of the balance sheets involved. Likewise, while there are some eyebrow-raising activities going on in credit structuring, or Private Credit as it has been rebranded, these re-leveraging activities are still relatively nascent: unlike the near decade of debt accumulation pre-GFC. We will probably end up looking back at these actions with our heads in our hands, but for now the music is still playing.

Instead, it’s more likely that credit deterioration would come from a slow-down on the revenue side of the income statement - the renormalisation of the cost of debt is well advanced and we are past the window in which rising interest rates are likely to cause problems to issuers who survived 2022/23. That means second guessing a US and global recession, but the White House is currently throwing everything it has at the US economy and might find yet more fuel for the engine if it follows through on mailing everyone stimulus cheques4.

Likewise, continued US hyperstimulation implies a further melt-up in stocks and persistently high bond yields that will keep demand for credit outpacing supply - particularly with the growing adoption of Private Credit as a pseudo-asset class. Yes, spreads are tight, but for the moment it is yields that matter and so demand will remain mechanically strong. There are no clear build-ups of positional leverage in public markets that could unexpectedly unwind.

Finally, it’s pretty clear to me that credit has been following equities through the bull market that started in 2023. For most of the post-GFC period credit led, as the Euro Area crisis, Brexit, the US HY shale crisis, COVID, and the Euro area energy crisis saw real deterioration in the credit quality and ratings of many corporates. But, as some unfortunate clients will have heard me saying for the last year or two, this time is different in the very narrow sense that, since 2023, whenever equities and credit have diverged equities were right and credit reverted to equities.

Equities consistently signalled that there would be no great economic crisis as the Fed was hiking and equity markets were right throughout. This leadership is likely to continue - the trigger for credit widening will be when stocks roll over. A sustained downturn in US equities would trigger a rise in risk aversion, and (likely) a rally in yields pushing many of the factors supporting the current market strength into reverse - we saw a small glimpse of this in April.

Of course, identifying the end of the equity bull-market is trivial in hindsight: the challenge is understanding if any specific pull-back is a dip to be bought or the start of a sustained draw-down. Given that the current equity up-draft is clearly retail led, the answers are more likely to be found outside of sell-side analysis. If I had to guess, it looks something like AI investment bots pushing retail investors into the same hype-stocks5, followed by a crash when the hot air escapes. Think 2001, rather than 2008.

Not the sharpest answer but, as I put at the top, I’ll know it when I see it.

Subscribe now

Leave a comment

What’s the return on holding a bond? The yield of course! But that just moves us on to the next question - what is a yield (or a spread)?

All bond metrics, be they yields, spreads or discount margins1, are short-hand for the expected return of the bond under certain assumptions, such as no-default and a specified maturity date. But how risky are those returns, not just in terms of mark-to-market2, but when estimating cash-on-cash returns all the way to maturity3?

I first raised the issue of reinvestment risk when talking about the 2061 gilt. To better understand the issues and to explore their implications, in this report I build a Monte Carlo simulation of long-end gilt yields, roll some dice (tens of thousands of dice), then explore the results and discuss their implications.

Subscribe now

The rules of the game

How should we think about returns from a bond? The simplest and most intuitive is the cashflows. Consider a 5-year bond bought at issuance, paying a 5% coupon. If you invest £10,000 then the cashflows look like this: £500; £500; £500; £500; £10,500. So as an investor you end up with cash returns of £12,500 from your £10,000 investment.

But for the mathematical readers, that is not what you expect from an investment that pays 5% per annum. Instead, you would have anticipated receiving (105%)^(5) x £10,000 = £12,762.82. So you might feel a bit swizzled by the missing £262.82. Where did it go?

The underlying issue is that the quoted yield on a debt security is the compounded rate of return on an asset4. That means that in order to get to the expected £12,762 return we needed to reinvest those £500 payments from years 1, 2, 3, and 4 - it is the interest payments on those reinvestments that explain the missing £262.82.

But what rate of interest will we get on those reinvestments in future years? We simply don’t know, unless they are contracted at the beginning of the investment5. If interest rates rise over time, then later payments will generate a higher return and the total payment at the end will be larger than anticipated. Vice-versa, if interest rates fall over time, then future coupon payments will be invested at a lower yield and the final return to the investor will be lower than expected.6

How large is this risk?

Despite my best efforts, I couldn’t find much written about this. Finance theory focuses heavily on very short-term interest rates, and constructing curves by bootstrapping short-rates. So, I decided to take a stab at doing it myself.

Share

Technical bits

To model the volatility of UK government bonds, I pulled observed daily yield changes for the benchmark, constant-maturity 10y gilt and scaled this to an approximate price change7. This was transformed into a probability density function, and re-centred on zero as yields were declining for most of the 40 years of data sampled.

This observed distribution of daily moves was sampled directly, to generate random movements in bond price - no assumptions about gaussian distributions or Brownian motion. Only very light-touch guard-rails were used, with the bond pulling to par over time and this effect being larger when yields were less than -2% or greater than 20%. The final bit of calibration was that random noise was scaled by the bond price (price/100) to avoid low-cash price bonds spiralling to negative values, and by the bond's tenor once it had less than 10 years remaining to maturity.

When the “volume” on the random noise is set to zero, bonds pull to par over their life with a steady yield.

When the volume is set to 1.0 (i.e historical observations), each simulation generates a random-walk in price that is converted into reinvestment yields. To calculate the cash returns, I bootstrapped the investment value by purchasing more of the bond at the prevailing price on each coupon payment day, from inception to maturity.

As a quick check, setting tax rates to zero and volatility to zero, we should get back the starting assumption of returns being equal to the compounded yield at the inception of the trade. In this case, a quick excel calculation suggests that the March 2035 gilt, a 9.444 year bond offering 4.689% yield to maturity, should pay us £15,415. A zero vol simulation estimates the returns at £15,490 (+0.5% vs our rough estimate) - so that looks pretty good given the vagaries of day-count conventions in bond math.

 SUMMARY STATISTICS 
------------------------------------------------------------
 Bond: 4.500% 03/2035 Gilt
 Clean price: £98.57
 Accrued amount: £0.261
 Dirty price: £98.83
 Yield to maturity: 4.689%
 Days to maturity: 3446

 ✅ Bond initialized

 SIMULATION RESULTS:
--------------------------------------------------
 Successful iterations: 1
 Initial investment: £10000.00
 Mean final value: £15490.10

The less often you roll the dice, the more certain the outcome

Model in hand, I ran simulations. Tens of thousands of simulations. Below I show a twenty sample walks for one bond: the 0 7/8% UKT 2046. As you can see, they are fairly widely distributed, but the weighted average change is zero, which is also the weighted average drift in yields.

To make life a little simpler, we can bucket these simulations by the final cash-return to an investor, to see the implied distribution of outcomes (cash-on-cash). Below are the results for 2000 simulations each of the high-coupon 4.50% UKT 2046 and the low coupon 0.625% UKT 2046, assuming a 40% income tax on coupons and historical volatility. A few things are immediately apparent8.

First, that clearly tax matters. There is a massive gap in expected returns between the low and high coupon 2046s, even though they have fairly similar starting yields (£26,650 vs £21,150). Second the distribution of possible outcomes is much broader for the high-coupon gilt; precisely because the larger the coupon cashflows are, the more influence they have on the returns that investors realise, and the more uncertainty there is over the return on those future (re)investments.

Interestingly, this only really becomes evident for 15y+ bonds. Even for high-coupon 10y bonds, there doesn’t appear to be enough randomness (and compounding of coupon returns) for the distribution of returns to really widen.

One final thing that was evident in all the simulations, across bonds and tax regimes, was that the distribution was right-tailed. What this means is that while the distribution of all possible outcomes had the right average, the median returns were lower - investors are 65-75% likely to get less than expected, even before transaction costs. But, because there are small number of scenarios in which yields spike to double digits, there are some outcomes where returns are much higher.

Put simply, there is a high probably you get slightly less than you expected for, offset by a smaller chance that you get significantly more - provided you keep reinvesting those coupons all the way down and back up. This is the sort of stuff they do not tell you about when you are investing in bonds.

Lies, damn lies and statistics

Aside from the benefits of polishing my Python skills, I learnt a fair bit working through these modelling exercises. For one, the distribution of outcomes is much broader and more negatively skewed (in terms of cash returns) than I had expected for higher-coupon instruments. This has helped me grasp a conversation I several years ago about insurance companies, zero-coupon bonds, and reinvestment risk. Not only are low-coupon bonds more tax efficient, they are logistically easier, and there is much less uncertainty over cash returns at the end of the investment period.

The other thing that struck me is that this approach to modelling risk doesn’t seem to be particularly common, at least not in publicly available material (maybe it is used in bank stress testing?). Academic finance seems to have focused on finding specific sets of assumptions that allow for closed-form pricing of securities, rather than modelling based probabilistic approaches. While that has advantages, including computational effort and clarity, it necessarily throws out interesting and important information about the range of possible outcomes.

For my next trick I might try repeating this exercise for US stocks, and then for a simple stock-bond portfolio. Or maybe Trump will do his thing.

[Edit] As a final thought, it’s also the case that reinvestment risk is much higher for institutions or for bond investments inside of tax wrappers, as the coupon income is higher when it is not taxed. This is also observable in the simulations as the distributions shift ot the left and become narrower as the tax rate increases.

Leave a comment

Thirty year bonds have had a fair amount of focus over the past few months, as yields continue to rise, curves steepen, and 2s30s quietly performs as one expression of my thesis on the current White House1. Some of this move has been interpreted in the context of the last Fed meeting and that absolutely wild dot. Some of it keeps feeding into the hullabaloo about the UK’s fiscal position - something I’ve been looking at in a more considered way with a colleague.

But the read-through to credit markets of these long-end gyrations is not intuitive and touches on a few adjacent topics, such as the current tightness of high-quality spreads, and the discussion of spreads versus yields. As a quick follow up to last week’s post, a few people let me know that I wasn’t explicit enough so let me try to clarify: the answer is yields2. But please do read the full note!

To help translate what is happening in the long-end of government curves to credit spreads, it’s important to understand who invests in long-dated credit and what drives their behaviour. Handily, due to a series of strange incidents3, I’ve spent a few hours on this topic in recent weeks - so I already had the crib notes to hand.

Who buys long-end credit and why?

Long-end (10y+, and usually more specifically 15y+) credit is a relatively specialist area of investment grade corporate bonds for a number of reasons. First, very few investors have the foresight to understand how a company will look in 20 or 30 years - consider the relative position of Microsoft today versus 1995 or, more strikingly, the examples of Enron and Lehman Brothers. Likewise, few companies have fundamental reasons for borrowing for such long periods, outside of instrastructure projects.

And, once issued, the large duration of these bonds makes them highly exposed to mark-to-market losses over the many years that they are outstanding. This gives them a “worst of” risk profile, as losses could be credit driven or due to a rise in interest rates (and resulting fall in prices). As such, even highly rated long-end bonds carry outsized exposures to both interest rates and credit risk.

Despite these headwinds, there is a significant amount of long-end debt out there: nearly $1trn in $-IG plus a significant slug of £-IG, relative to the overall size of that market. In contrast, there is hardly any long-dated €-IG debt. In each case, the presence or absence of a well populated long-end is directly tied to the demand (or lack of demand) from insurance and pension funds.

Pension funds

Historically, pension funds have had a strained relationship with IG credit. While in theory high-quality credit makes sense, as their liabilities are often discounted using long-term interest rates, the rise of portfolio construction and risk-factor investing has seen a significant decline in the support for holding IG credit over a combination of government bonds, illiquids, and equities. Instead, the major driver of fixed income investments by pension funds has been regulation and accounting.

For example, UK pensions funds were heavily invested in equities until the turn of the century. But in recent decades the investment decisions of large Defined Benefit (DB) funds were dominated by the sensitivity of their liabilities to falling interest rates and the need to mitigate the accounting risks for sponsors. As such these funds heavily re-allocated into bonds, including large slugs of long-dated IG debt.

This framework bias is particularly apparent if we compare legacy pension schemes with newer funds. Legacy (usually DB) funds which have an increasing number of members are entering retirement are heavily invested in fixed income. In contrast, new funds are often Defined Contribution (DC) and many, if not most, members are in the accumulation phase. As a result, DC funds are less constrained by interest-rate accounting and relatively few bonds are held by these types of pensions until very close to the termination date of the fund: underscoring how different the investment profile is with and without regulatory matching.

The accounting of legacy funds is powerful stuff and future liabilities (pensions) are valued on a discount measure, which means that the present value of those pensions (i.e. the cost to the fund) goes down as interest rates go up and vice versa. One way to think of this is that it becomes relatively easier for the pension to meet those future obligations through low-risk investments, such as bonds, and interest rates rise and hence they start to “de-risk”, i.e. buy bonds.

Related to pension funds are the UK Bulk Pension Annuity Aggregators. These are insurance companies that specialise in taking over the pension obligations of DB schemes, and hedging those obligations through their own investments. Under Solvency II there is a huge incentive, in the form of the so-called Matching Adjustment, for aggregators to hold bonds. There is also an incentive to generate as much yield as possible on their investments, while still retaining the Matching Adjustment. Accordingly, these investors are large buyers of long-dated corporate bonds, often at spreads well below what other investors would tolerate.

Insurance

Looking at insurers more broadly, we need to split their activity into two parts, starting with contracts where the insurer’s own capital is at risk.

In Europe and the UK, insurers are regulated under the Solvency II framework that lays down detailed guidelines on how to measure the risk of various asset classes. As a result, the risk-adjusted returns on capital for these investors tends to be well defined and only to change slowly. I have in the past showed, with a colleague, that these allocations have been remarkably stable over the last 10 years, despite the wild swings in both credit spreads and interest rates. As such, the direct impact of yields has very little implication how insurers engage with credit markets - the effects are indirect, and largely to do with the velocity at which they write new contracts.

That said, it is worth noting that traditional life-insurance is a long-dated product and historically it was hedged with long-dated fixed income. As a result, these investors were a major part of both sterling and dollar bond markets pre-GFC. And the absence of true life-insurance products in continental Europe is a large part of why a long-end corporate bond market has not developed there (yet).

“Savings products”

Insurers also tend to offer other types of products that are a lot more like “savings products”. I put “savings products” in quotes because they are usually structured to be insurance contracts in law, as this creates favourable tax treatment.

For example, the French “assurance vie” is a life insurance product in name, but it doesn’t only pay out on the death of the contract holder. It is de-facto a tax efficient savings product, with built in inheritance (death) planning. Likewise in Germany and other countries, there are combinations of savings accounts and returns guarantees (the insurance element) that are major vehicles for retirement savings. This creates an interesting liability profile, as the insurer is effectively short a total-return floor4 on the underlying assets.

Because the insurers are often also responsible for putting together the savings portfolio, the have strong incentives to de-risk the vehicle to increase the probability that they do not have to pay out on the insurance side. This means that these accounts tend to own a lot of bonds when yields are above the contractual floor. That said, it is also the case that European insurers historically run outsized reinvestment risk on what were effectively 30-40 contracts, by holding 10y government bonds. As a result, there was huge pressure on European insurers post-GFC, as contracts had been written at yields far higher than what government bonds could offer - driving a shift towards corporate bonds during the era of negative interest rates.

Of critical importance, products like these are also sold by insurance companies to savers in Taiwan - often based on the yields that can be achieved in the US market, as the TWD bond market is far too small for the scale of domestic demand5. Because these insurers have been much more fastidious about managing the duration risk, this has resulted in huge flows from these investors into long-end USD fixed income, and $-IG in particular. Like their European forebearers, their buying increases as yields rise to de-risk the policies that they have written.

Share

Many frameworks, but one implication

In my view, all of the aforementioned context is critical to understanding the nuances of long-end credit investing. But a significant simplifying factor in understanding the implications of rising long-end yields on credit spreads is that almost all of these drivers of demand for credit increase when yields rise… to a point.

In the case of pension funds, the biggest impact of higher yields is the improvement in solvency ratios, as almost all pension funds have liabilities that are more interest rate sensitive than their assets. When yields rise, the value of pension liabilities falls more than the value of the fund’s assets and solvency improves. This tends to trigger increased hedging (buying of bonds) and buy-in/buy-out activity - supporting an ongoig rise in UK BPA activity. And BPA activity creates its own demand for long-end corporate bonds due to the Solvency II Matching Adjustment.

For insurers, higher yields operate through different mechanisms. To a reasonable approximation, all insurance products are discount rates: the probability of something bad happening in the future, discounted back to a premium today. As interest rates rise, the present value of future insurance payouts falls, making policies look more affordable. As a result, policy writing picks up. Likewise, we have seen a sharp rise in annuity writing which is, again, priced off interest rates. More business, especially annuity writing, means more demand for fixed income.

This is also the case when insurance companies sell savings products - such as fixed maturity products in Europe or Taiwanese “life insurance products”. Particularly after a decade of ultra-low yields, the rise in interest rates allows insurers to build a portfolio or fixed income that yields more than the contract agreement. This derisks their back-book, while any additional spread they lock in is incremental return on capital. This creates a deep, yield-driven demand for corporate credit.

As a result, time and again we see yield-driven buying at the long-end as yields move up through certain levels and insurers seek to cover the guarantees they have written. This is particularly visible in the $-IG market because the €-IG long-end does not exist, and the £-IG market is too small (and rich versus gilt-ASW). As a result, every time US long-end yields rise you get a wave of long-end $-IG buying that is global in nature - the sum of all the liabilities written around the world and not just in the US.

This demand is also nearly spread agnostic, because the insurers that are buying the products are responding to the yields written into their “insurance contracts”. This is why $-IG long-end credit has led the charge to multi-decade tights.

As a final thought, it is always worth asking “what could go wrong” for long end credit spreads?

To me, the risk is that eventually this wave of back-book hedging will be exhausted. Particularly if long-end government bond yields reach levels at which they can cover their back-book liabilities, then the argument for taking high quality credit risk at tight spreads would start to look weak from a return on capital perspective. That said, this would be a whimper (lack of buyers) not a bang (force selling).

I’m also not convinced we are there yet and my Spidey-sense tells me there is a narrative shift in highly rated credit spreads underway6. But this effect has already played out in £-IG where the widening in 30y gilt-ASWs shifted the value in the long-end to gilts over corporate bonds - and BPAs adjusted their investments accordingly. That, plus the resurgence in structure credit products indicates that we are close to the floor in terms of liability-driven spread tightening.

Or maybe I’m just worried that my upside scenario of 75bp in €-IG for this year was not optimistic enough.

Leave a comment

Since 2023, the question of what matters more, credit spreads or credit yields, has been a surprisingly persistent one1. The volume of this question tends to rise as spreads fall and, accordingly, I was asked about this subject twice in the last month. So, if you are wondering why on earth I am writing this, blame those people.

They deserve it.

Share

What even is a spread?

I have long said, only half-joking, that the hardest question in credit is “what is a spread”?

In corpoate bond markets, a spread is the difference between the “yield”2 on a security and some other “yield” - usually that of government bonds, in which case it could be called the G-spread (or T-spread or bund-spread), spread to benchmark, or OAS (to treasury). But it can also be measured to the swaps curve, in which case it is likely to be called the I-spread, Z-spread (to swaps), OAS (to swaps), or ASW3.

There is also the spread between two bonds, which is strictly a spread-of-spreads and is usually referenced when talking about relative value; for example, the spread between triple-B and single-A credit. And none of these should be confused with swap-spreads which are to do with interest rates4.

But in very general terms, a credit spread is the return you expect to receive in excess of some other reference rate in return for taking credit risk. As such, it’s a barometer of credit valuations - how rich the asset class is when spreads are tight, or how cheap corporate bonds are when spreads are wide. For this reason, along with a significant dollop of inertia, almost everything within credit is done with reference to spreads. If yields happen to be high (or low) that is, at its core, a comment on the value (or lack of) in the underlying government bond market.

And yet, at the start of 2023, there was genuine concern that decade-high yields would be a negative for credit, as government bonds returned from their extended sojourn in ultra low (or even negative) yields. This conversation often involved a chart of spreads as a percentage of yields, arguing that the relative pick up was small enough to deter investors and asset allocators. For once being European was an advantage here - the same chart of €-IG has an extended period of divide-by-zero error that helps hammer home the point that this metric doesn’t actually correlate to anything.

When do spreads matter?

Despite worries that higher yields would cause money to flood into government bonds at the expense of corporate credit, over the last three years investment grade markets have seen strong demand on the back of steady inflows - it turns out that higher yields make credit (and fixed income in general) more attractive. As a result, $-IG spreads made 30y lows earlier this year and €-IG is hovering just above the post-GFC tights.

Surely this ends the debate: yields matter more than spreads. Demand for the asset class continues unabated. Long live the king!

Well not so fast. After all, if I have learnt anything from 15 years as a strategist it is the ability to argue both sides of the table and to make things more complex than they might initially seem.

First of all, spreads absolutely do matter when it comes to future returns. If nothing happens over its lifetime, the spread on a corporate bond very accurately represents the expected excess return for holding that bond versus the reference instrument. As long as the bond survives to maturity, it pays exactly what it says on the tin5.

And where spreads sit in their historical range also gives us some sense of the range of potential near-term returns - with spreads near their post-GFC tights, it’s unlikely that credit will generate materially higher returns than its carry6, but there is a high likelihood that spreads will be much wider at some point in the next 1-3 years. This creates a downward asymmetry in the distribution of returns for active investors, who do not just hold bonds to their final maturity.

So yes, yields are high and clearly people are happy to buy corporate bonds on that basis. But who are these people and why are they buying corporate bonds over government bonds when the marginal return is so low? Those are the real questions that we should be asking.

Who buys credit and versus what?

It would be very long, and rather dry, to discuss the broad range of investors who buy corporate bonds7, but a whistle-stop tour will help to start clarifying what is going on in credit, under the bonnet.

• Insurance companies (investing their own money): European insurers operate under the Solvency II framework, which lays down detailed guidelines on various asset classes. For corporate bonds, the key metric is usually spread to government bonds, as the incremental capital charge is relative to Govvies8. In my past life I showed that allocations to credit by insurers have been remarkably stable over the last 10 years, despite the wild swings in both credit spreads and interest rates, reflecting their rigid regulatory framework.

• Pension funds: Pension funds value their investments against pension liabilities that are discounted in various ways depending on the jurisdiction. In the UK this used to be the AA corporate yield curve, while in the Netherlands it was the Euro swap curve. However, all regions tend to have a higher demand for fixed income as yields rise (to a point), as this usually increases their accounting solvency and triggers bond buying to “de-risk” their investments.

• Households: A menagerie of retail savers, from High Net Worth Individuals to moms-and-pops. These savings will generally flow into one of two basic products: either actively managed corporate bond funds, or structured savings products that are backed by portfolios of bonds. Both products are generally marketed on a combination of past performance and/or their headline yields.

Further, it is important to bear in mind that the flows into credit are not just a function of valuations but also of the total amount of assets in the system - valuations drive allocations within a portfolio, but in recent years those portfolios have, themselves started to grow at a higher rate. For example, Euro Area insurance companies have seen their total assets grow by c.€1trn since Q4 2022, and pension funds by c.€0.5trn over the same period.

With equity valuations and coupon reinvestment flow surging, there is simply more money to put to work - everywhere. So while the marginal investment to credit might be lower at tight spreads, that is being offset as the overall pie continues to grow. Indeed, the rapid surge in equity valuations has almost certainly left many investors with an overallocation to stocks - which then needs to be rotated back into fixed income, including credit.

It’s the retail bid, stupid

Given the very rough outline above, it’s pretty clear to me that the demand for credit is overwhelmingly coming from households (via savings products) - albeit with further support from pension funds (especially in the UK) and insurers9. Indeed, anyone who has been active in credit markets over the last two years will have been bludgeoned into submission on the topic of Fixed-Maturity Products - savings products that promise a fixed return on a future date and have been enormously popular in France and Italy. Estimates are hard to come by, but inflows have been in the order of €60-80bn per year over the last three years - similar in magnitude to the buying power of the ECB’s Corporate Securities Purchase Programme (CSPP).

As the marginal buyer of credit, their activity is critical. So what is the outlook? On the one hand good: the spread for many of these investors is versus bank deposits, and bank term deposits currently yield a lot less than the c.3% on €-IG.

What is more tricky is the relative value of credit versus BTPs (in Italy) and OATs (in France). There have already been a few red flags raised, with €-IG now trading below the 10y OAT in yield. But, while this is a watershed moment of sorts, credit is a 5y product and the reference point should be 5y OATs where there is still daylight. Still, the point stands that the biggest retail buyers of credit have been in Italy and France, and if they could achieve the same or higher yield in EGBs, they would likely shift.

That said, there is an equal chance that the fund managers will shovel more corporate hybrids into their funds to pump up the headline yield.

Leave a comment

Leading into last weeks Non-Farm Payrolls, which were weaker than forecast as US immigration policy continues to bite, rising bond yields were pulling attention from equities. Long-end (30y) bonds drew particular focus, with the UK suffering yet more time in the spotlight as gilt yields touched multi-decade highs. The Chancellor might have breathed a sigh of relief as duration rallied post-payrolls (as did my personal gilt portfolio) but, taking a step back, I’m surprised by how little serious conversation I have seen about what the fair-value of rates really is.

For example, almost every comment on Fed policy describes it as restrictive - but on what basis? Not a single measure of US inflation has been sustainably at the 2% target since 2019, and prices have reaccelerated since April. The ECB and BoE, too, are generally described as holding rates at an elevated level - but again without any evidence that inflation is persistently undershooting.

Yes, yields are higher than they were for most of the post-GFC period, but that was because the global economy was in a liquidity trap1. Given that the US, and probably the entire global economy, most likely exited the liquidity trap at some point after 2020, “neutral” policy rates should be higher: likely at or above pre-GFC levels.

Through this lens, the current level of duration (long-term yields) is not screamingly cheap. A continued, messy range-trade in interest rates could be a great environment for traders, but the narrative of sustainably lower yields is only likely in a scenario of lower growth or a further loss of central bank independence - neither of which are bullish arguments for risk, prima facie.

Share

What would neutral look like?

One byproduct of being in research for a long time is that you say so many things that some of them turn out to be correct. As such, I have a catalogue of things that “I very definitely said” before they happened. One of those was in 2023 when I presented at several conferences a simple message: 2010-2020 had been a decade defined by low rates and monetary policy at or below the zero rate, but 2021-2030 would be a decade defined by higher yields and a greater focus on fiscal policy. A pretty bold call for a credit strategist, but it was already 2023 so I was only really taking a 7-year view. It’s almost 2026 and thus far I’m looking pretty good2.

Following the Global Financial Crisis Western economies had fallen into a liquidity trap in which both inflation and policy rates were stuck at very low or even negative levels. This was widely agreed. What was less agreed was how to exit a liquidity trap. My argument was that the truly titanic stimulus deployed during COVID, combined with a global supply shock, seemed like a good start. As inflation spiked in 2022, and economists floundered badly, it seemed increasingly clear that Western economies had exited the liquidity trap and hence the 2020s would look very different from the 2010s, at least in terms of interest rates.

Now, the thing I enjoy most about being a sell-side strategist is sitting across the table from genuinely world class investors, and telling them my random ideas. It’s a chance to be corrected (and learn) or to be agreed with (and learn)3. In Denmark one client asked me, as we discussed my theory, if central banks would ever admit that a decade of ultra-loose monetary policy, including QE, had been a mistake.

That was easy: no. Never.

There was simply too much literature and intellectual capital invested in post-GFC monetary policy for central banks to suddenly volte-face. Much more likely, I argued, was that we would see a narrative that COVID had pushed the neutral rate (which economists refer to as R*) higher. There was also a burgeoning discussion of “short-term R*”4. In any case, the argument would be that central banks had not made an error - they had been correct at the time, but that the global economy had changed.

Aiding this intellectual smoothing is the way that neutral-rates are determined. In simple terms, the neutral rate is the rate that neither accelerates nor decelerates economic activity and so neither pushes up nor down on inflation - hence it is neutral. But the neutral rate (R*) is not observable, it’s estimated using mathematical models that, in the broadest terms look at policy rates and then look at inflation outcomes on various lags5. From this data, the neutral rate is inferred.

During the 2010s, policy rates were set at or below zero, but inflation was consistently below target. If policy rates at or below zero are followed by a period of below-target inflation, then this implied that monetary policy was still restrictive. This, indeed, was the macro-economic orthodoxy of the post-GFC decade, underpinning the policy justification for Quantitative Easing.

But we can apply this same logic to the post-2022 period in which policy rates above 4% have, with a lag, generated inflation that remains above 2%. The longer this persists, the more forcefully it will imply that 4% rates (or 2% in the Euro Area) generate 2% inflation - and hence that current levels are not restrictive, but neutral. Indeed, a superficial reading of the current uptick inflation is that US policy was too loose 12-18 months ago, when Fed funds were still above 5%.

If this is neutral, where should rates be?

The recent data flows remain broadly aligned with my thesis, even if the calibration may eventually need a little bit of tweaking. First, there appears to be little to no chance of policy returning to the zero bound (or negative rates) any time soon. At least not without some material political intervention, or a massive recession.

Second, while orthodox economists still describe Fed policy as restrictive, and place ECB policy at the upper end of neutral, the disinflationary impulse since the end of 2022 has lost momentum. Inflation in Europe is sticky, in the US there is some hint of re-acceleration, and the UK is having a particularly poor time of things. While central banks have, after 4 years, found a new willingness to look through “transitory” inflation, their models will not be screaming for a cutting cycle. Instead, the longer that 4% (2%) rates generate 2% inflation, the more strongly that models will consider current rates to be at levels consistent with neutral.

Third, if policy rates are already close to neutral then the expectations component6 of rates, particularly for intermediate rates (7-10 years), should be centred near current levels. Near term rates, particularly in the US, will remain biased lower by political pressure and low employment growth, but with less pull on 10s and 30s than during QE. Much less. On top of this, term premium7 should be positive and likely in the 25-75bp region, which would put fair-value on the US 10y at 4.25-4.75%8 - albeit with a strong pull to the lower end of that range, given the pressure on the Fed and the overweighting of Tbill issuance by the US treasury9.

Conveniently for me, that is pretty much where yields traded until payrolls Friday. It’s certainly reasonable to argue that neutral might be as much as 50bp lower than where I’ve marked it, but that still leaves 10y bonds with little room to rally unless you also believe that term premium should return to the highly compressed levels seen during QE, despite continued QT and rising fiscal and policy uncertainty.

My gut sense is that we will eventually get back to the ranges that held since the start of 2024, as central banks fail to deliver more than a few cuts without a full recession. This has implications for credit demand, that I plan to comment on soon.

Leave a comment

You know you have hit a nerve when an article on the 2061 gilt is, by some distance, the most widely read report you have ever authored. At the end of that article was an interesting comment1 on constructing a ‘gilt-ladder’, rather than putting it all in a single bond. The idea is simple and attractive: it’s all very well having a lump of cash when you hit retirement age, but what about turning that into a pension?

Investing in a series, or ladder, of gilts that mature each year would result in a more equally spaced stream of income. Indeed, this is exactly what pension funds and Bulk Pension Annuity (BPA) providers consider when they undertake asset-liability matching (ALM) - constructing a series of investments that generate a schedule of payments, designed to match expected future pension payments. It is also an idea that is seeing more leverage in Defined Contribution pensions planning, in particular managing the post-retirement phase.

However, there are traps to avoid if you are an individual taxpayer, as opposed to an institutional investor. Particularly if you are leveraging the tax-free nature of capital gains on gilts, security selection can have a huge impact on investment returns.

Share

What’s the real return?

One of the major problems with fixed-income investing is that, in theory, it should be simple, but in practice it is complex. And if you don’t believe me, ask someone in the bond market to explain what a spread is. They may need more than one attempt.

In the case of gilts, it is critically important to understand that the stated yield on a bond represents the compound return on that security including reinvestment of coupons. So if an individual investor wanted to secure the quoted 5.47% return on the January 2046 gilt, they would have to reinvest 40 coupons, in the same bond, at the same yield as today2. This is pretty unlikely and, even if you are particularly diligent, the transactional costs on each coupon reinvestment would erode your returns. To make some of this a little more transparent, below I show the cash returns for the 2041 to 2016 gilts with and without reinvestment.

Source: https://www.dividenddata.co.uk/uk-gilts-prices-yields.py, as of 1 September 2025

The differences are meaningful. In cash terms, £10,000 invested in the 10 gilts shown above will return, without reinvestment, anywhere from £21,000 to £49,000. Given that these bonds all have very similar yields, the differences are largely due to the lengths of the holding (compounding) periods. But take note of the difference between reinvesting and not reinvesting: this can be can be as large as £20,400 (£36,000 vs £56,400 for the TG57), a huge loss to the saver when the bond matures. As a result, most individual investors will not fully realise the expected returns from holding gilts, due to the practicalities and costs of reinvesting many small coupons.

This is another reason for favouring low-coupon bonds over high-coupon ones, as reinvestment risk is naturally lower when the cashflows reinvested are smaller. I also think this is something that should be explicitly considered when making the argument for gilt-ladders - even under the best-case outcome, we can never truly be sure of the returns of a bond over its lifetime.

Post tax considerations are even more nuanced

Even bigger than reinvestment risk, which exists for all investors, is the complication caused by the different tax treatment of coupons (taxable as income) and capital gains (currently tax exempt for gilts). This tax difference does not apply inside of pensions, and so is not a concern for DC/DB planning, but it does matter for anyone looking to top up their savings by holding gilts outside of these tax wrappers (or an ISA).

To look into this, I built a little bit of Python code to turn spot gilt prices into estimated yields, net of tax for each income bracket3. Below I show the yield curve on the usual gross basis, and on a post-tax basis for the 2046 to 2073 gilts. Through the tax lens, the gilt curve is much more jagged and, arguably, downward sloping.

As a result, some gilts look more like snakes than ladders on a post-tax basis. Indeed, we can calculate the forward yield4 on each gilt relative to the 2046: this is the “reinvestment yield” that an investor can lock in now by building a gilt ladder, versus buying the 2046s and re-investing that money when the gilt matures. Viewed this way, it’s immediately clear that most of the potential ladder rungs are locking in less than 2% yields - and several of them lock in negative forward yields post tax!

It is hard to say what 10y gilts will yield in 2046, but anything below 3% looks unattractive to me. I would rather own the 2046s and decide how to build my pension annuity when the time comes. That leaves only the following gilts as reasonable rungs on the ladder: TG50; T51A; TG53; TR54; TG57; TG61.

So while I can find an argument for breaking up gilt investments into two or three lumps, the argument for building a full gilt ladder (and then reinvesting all of the coupons) is more theoretical than practical outside of a tax wrapper. That said, I will likely diversify into the TG50s as the implied forward (5.28%) looks pretty attractive!

Post script

I’ve really enjoyed working through this analysis, including building two yield calculators in Python and playing around with tax treatment and yield forwards. So a huge thank you again to for a great comment. I have also enjoyed reading the substack of , which has helped me climb the first few rungs of self-education on Tontines and Collective Defined Contribution funds.

This article is intended for informational and educational purposes only. It does not constitute financial, investment, legal, or tax advice. Readers should consult a qualified professional before making any financial decisions. The views expressed are those of the authors and do not necessarily reflect the opinions of any affiliated organizations

Leave a comment

Source: ChatGPT1

After declaring the US to be an EM and then following up with an article on how Bitcoin is part of the pay-to-play nature of this White House, I promised myself I would not slip into an endless cycle of half-baked US political commentary. If nothing else, I already have my hands full being a half-baked credit strategist.

But recent weeks have produced several new data points that support the arguments and (more importantly) give us further guidance on the likely implications. While the most prominent of these is the move to fire Fed Governor Cook ‘for cause’, the White House’s strategy of ‘flooding the zone’ has generated multiple stories that support and build out my narrative for the US.

One is the plan to take a 10% stake in Intel, which would set a precedent of US owned enterprises - national champions of the sort European and EM credit investors are familiar. Then there is the demand for a cut of NVidia’s revenues in order to ‘facilitate’ provision of export licenses to China, dovetailing with the DOJ’s updated guidance on the Foreign Corrupt Practices Act2. Finally, we have the already-forgotten Japan and EU trade deals that include promises of White House-directed foreign investments. These would be handy tools for MAGA as elections come back around.

All of these speak to the continued bending of the US economy and US corporates to the domestic and international political agenda of the current administration. And the implications for investors remain profound: this week, a commentator on Bloomberg TV noted that “the more time [NVidia’s CEO] spends at the White House, the more the share price rises”. I could not have said it better myself.

Share

USS Enterprise Intel

The announcement that the US will take a 10% stake in Intel was, as most US policy appears to be these days, announced over social media. Following a series of negative tweets about CEO Lip-Bu Tan by President Trump, the company agreed to sell a stake to the White House - bringing a reprieve for both the company and Mr Tan.

The natural question is: who is next? The Department of Defence took a stake in MP Materials, a Nevada-based company that mines rare earth metals, earlier this year. This weekend Mr. Lutnik suggested that defence contractors are being looked at. It seems to me that creating “US champions” in which the government has an equity stake is extremely aligned with both the goals and the execution strategy of the current administration. As such, we should expect more US backed enterprises.

Intel’s share price has not gathered much momentum after the initial announcement, but I think a US government stake raises a number of interesting questions for credit investors. To the extent that the US views Intel as a strategic asset, my gut reaction is that this is a credit positive. Intel was downgraded to BBB by S&P in December, so it’s been a deteriorating credit story. Now it’s also a political one3, with $9bn of equity and the potential for government support on an ongoing basis.

There are likely to be other equity stakes taken as the White House selects its champions. As such, there is an opportunity to both identify other beneficiaries of state investment, and also to learn from the history of SoEs in other regions.

Source: investing.com

Facilitation protocols

The reported 15% revenue share that Nvidia will pay the White House in return for China export licenses for their H2O chips is de facto a facilitation payment. In some ways, I appreciate the brazenness of the demand, as it removes any debate over the interpretation. This is rent-seeking behaviour by the US government that will erode the Nvidia’s (admittedly huge) margins. To give credit where it is due, during earnings Nvidia noted that they need a contract explaining how, legally, this is not a bribe, but I’m sure the DoJ will be able to draft something that the courts will uphold.

While not every example will be so cut-and-dried, this one helps to clarify what the macro-economic impact will be: lower margins for US corporates, albeit perhaps still better than the alternative of being shut-down completely. While the White House says there are legitimate policy reasons for taking these approaches, I note that these tariffs are justified on national security grounds - concerns that, apparently, can be balanced by a 15% revenue sharing agreement.

These actions also help explain how Japan and the EU both achieved lower import tariffs on cars than the 25% still being imposed on North America automakers4: because this isn’t really about national security, it’s about extracting concessions. With that in mind, I feel like Ford and GM should be working out how to give the President a win - maybe they can offer a 10% stake too?

Make it loan

One of the more interesting elements of the US trade agreements with the EU, Japan and South Korea was the inclusion of investment funds or commitments: $350bn from South Korea, $550bn from Japan, and $600bn from the EU. While the details remain vague and unsubstantiated, triggering understandable scepticism from investors, what stood out to me is that invesments made by the Japan and South Korean funds would be directed by the White House.

Setting aside how much money (if any) will be invested, and how much is incremental versus what would have happened anyway, this strikes me as similar to EU levelling up funds. Namely, these funds would allow the White House to reward compliance by directing investment into specific states and districts. If these are effectively state-directed lenders that can push investment into specific regions and industries, then this will be a extremely powerful tool for generating voter support. Think of the “this was funded by the EU” signs in Europe - but branded for President Trump.

Source: Bbc.co.uk

The ability to direct investment, even if it is not incrimental, would be a fantastic piece of apparatus both for generating compliance with White House policy and popular support for the incumbent government. It would come at the expense of mal-investment, credit losses, and fraud - but that is a problem for tomorrow.

Set FOMC to ‘Dependent’

Last but by no means least, we have Presidential firing of Fed Governor Cook, which is currently being litigated in the courts. On the one hand this shouldn’t be a surprise, Trump has been very clear that he is unhappy with the Fed and that he is willing to apply maximum pressure in his pursuit of lower interest rates - albeit stopping short of firing Jerome Powell thus far. But equally, I’m surprised that we haven’t seen a larger rise in risk premium for both the USD (lower) and 2s30s (higher).

Shortly after writing my first post, I was able to enjoy a pint or three with a friend and macro discretionary Hedge Fund manager where we discussed how to trade US policy. A push-back on his side was that the US could be viewed as a policy leader, with other DM’s likely to follow suit. This would make relative-degeneracy plays such as FX less convex. We settled on steepeners and USD/BTC as relatively cleaner expressions, both of which have performed well since the election5.

Source: investing.com

One perspective is to acknowledge that the White House will have their man in the Fed by May of next year, and likely with a majority of the appointed governors. Hence the firing Cook could be viewed as simply speeding up the inevitable. The FOMC is likely to comply with Trumps narrative sooner or later, even if data does not, so it’s appropriate to price cuts at the front of the curve. That, in turn means that any policy error premium should be expressed in the long end. As such, I think curve-steepening can keep on trending6.

The data has been… highly logical

A year ago, any one of these events would have been extraordinary. But from my, admittedly cynical, position they are all well aligned with my overarching thesis that the US economy and markets are increasingly driven out of the White House. As such, we should expect shareholder returns on investment to deteriorate, as more capital is deployed with political, rather than economic, returns in mind.

There will be select winners as more US state ownership takes root, supported by selective grants and loans. But always at the expense of lower growth and higher inflation. Moderate stagflation seems likely, with building tail risks.

This article is intended for informational and educational purposes only. It does not constitute financial, investment, legal, or tax advice. Readers should consult a qualified professional before making any financial decisions. The views expressed are those of the author and do not necessarily reflect the opinions of any affiliated organizations

Leave a comment

Chair Powell used his Jackson Hole address to tee-up a September cut by the Fed, though what comes after that remains data-dependent. In contrast, the market has moved away from pricing a cut at the next Bank of England meeting, after July inflation surprised to the upside and left CPI on track to move above 4% (CPI-H is already there). August’s services PMI also surprised to the upside, which is a somewhat more positive reason for the BoE to skip September.

Source: ONS

But while the Monetary Policy Committee1 is likely to leave policy rates unchanged, they will still need to decide what to do with their Quantitative Tightening2 program (QT). The BoE’s QT program is, to the best of my knowledge, unique in that it has an active element to it: the Bank does not merely let bonds mature, it has been selling them to the market since 2022. But with 30y UK yields at multi-decade highs, it seems like a good time for the BoE to reconsider. The BoE owns c.£53bn3 of gilts4 that will mature between October 2025 and December 2026. The marginal benefits of selling an additional c.£20bn of gilts are trifling from a monetary policy perspective, while there are benefits to undershooting market expectations of £75bn in QT. From both a monetary and fiscal perspective, a reverse-twist5 is more optimal than sticking to the current mix of tentative cuts and continued active gilt sales.

Source: Estimates based on Bank of England auction results

Share

Splitting monetary and fiscal

In August 2022 the Bank of England’s MPC voted to start unwinding its portfolio of gilts, bought under the Asset Purchase Facility (APF). Following the Fed’s decision to start passively reducing its own holdings of USTs in June 2022, the BoE was the second major bank to begin “Quantitative Tightening”. The ECB would start in March 2023, and the BoJ in July 2025.

But among the major central banks, the BoE is the only one that is actively selling bonds to the market - the other banks are merely letting their holdings mature. In some ways this is a practical issue: while many lament the amount of debt owed by the UK government, the gilts market is a fraction of the size of the other markets and is also unusually long-dated. Reflecting this, the BoE’s bond portfolio stretches out for decades and so the BoE is required to sell some gilts, or otherwise allow monetary policy to be constrained by the maturity profile of' the APF.

But active sales have a number of draw-backs, and two in particular now argue for the MPC to cease-and-desist in the active part of QT.

First, monetary policy conditions are already reasonably tight - particularly at longer tenors which is where QT has most effect. Short tenors remain anchored by the Bank Rate, so the only real argument for pursing QT is a view that specifically long-end rates need to be guided higher. But with the 30y yields at their highest since 1998 and above international peers that seems like a weak rationale.

Source: uk.investing.com

More subtly, one of the most powerful transmission mechanisms for monetary policy in the UK is via mortgage rates. However, mortgages price off the swaps-curve rather than gilt yields, and recently these have been diverging - with gilt yields rising, but mortgage rates falling. This further weakens the argument for pursuing QT, as if the BoE wants to tighten policy then swap-rates are more likely to respond to a higher Bank Rate while QT is instead likely to widen the gap between gilts and swaps.

I’m not the only person who sees the world this way. MPC member Catherine Mann raised the question of “equivalence” and of yield-curve effects in a speech she gave in June6. Specifically, she noted that “tightening effects of QT cannot be perfectly offset by cuts in Bank Rate”, implying that the Bank should be willing to slow QT if it is undermining the MPCs ability to set monetary policy in the desired configuration.

Second, HM Treasury’s indemnification of the APF has caused an undesirable de facto intertwining of monetary and fiscal policy. With central bank independence under torrid pressure internationally, it behoves the MPC to get out of the way on this one. Monetary policy is a relatively dry topic, so you know that something is up when UK media has repeatedly featured articles on how the Bank of England’s asset purchase program is influencing the Chancellor’s budget.

The broad rational for monetary policy tightening is that rising interest rates reduce consumption and investment, slowing activity and thereby inflation. But in the UK, there is growing evidence that the most interest rate sensitive sector is not corporates or households, but the government - rising rates are creating an increasing headache for the Chancellor. As such, QT is not slowing the economy through the monetary channel, but the fiscal one - which cannot be comfortable for the MPC. Moving to fully passive QT would help alleviate this headache.

What to do

In 2024/25 the APF cost the Treasury £23.3bn. Of that loss, £20bn was due to the mismatch between current interest rates (currently 4%) and the average yield on the portfolio at the time of gilt purchase (less than 1.5%). Against this, the crystallised losses due to active sales were £3.3bn. As such, the impact of active gilt sales is small, but not insignificant - and still higher than passive QT, even at the cost of skipping a policy cut (which would cost less than £1.5bn on the remaining £550bn APF portfolio).

But the indirect effects matter too. The DMO expects to issue £300bn of debt in 2025/26, so if continuing active QT pushed yields even 10bp higher it would cost the Treasury £300mn per year. And with an average tenor of c.13 years, that could rise to c.£4bn over the lifetime of the debt - on top of the up-front, crystalised losses.

The Treasury indemnification of the APF was an understandable effort to shield the BoE from accusations of debt monetisation in 2009. But it is now making the costs of QE explicit, while other central banks are able to sweep them under the rug. Moving QT to a fully passive process, or even transferring the remaining holdings to a new monetary policy portfolio (that will be required going forwards anyway) would provide both direct and indirect benefits for the economy. It would also help disentangle monetary policy from fiscal conversations, and shift focus to short-end rates where policy is more effective.

Source: Bank of England, uk.investing.com

If necessary, the BoE could offset slower this with a slower pace of policy cuts, though the BoE estimates that £300bn of QT has only increased term premium by 15-25bp so this might not be necessary. If the MPC did skip a cut on this basis, it would imply a reverse-twist, pushing up on short-term interest rates but taking pressure off the long end of the curve. This would be aligned, with the DMO7 which has reduced its supply of long-end debt, where domestic demand is in structural decline.

If the BoE wanted to be really radical it could lower the pace of QT to £50bn/year, and reinvest any excess maturities into T-bills. This would be aligned with the Fed and would allow for a very rapid reduction of the balance sheet if required in the future, as it would further shorten the residual holdings. Additionally, this would be closer to the sort of asset profile that the BoE will want in a steady state anyway.

Leave a comment

Source: Google Gemini

So farewell then, Ofwat. The UK government has announced that the regulator will be abolished and a “new, single, powerful” successor will be established for England. This was one of my more off-piste calls for 2025 - it’s not normal for a credit strategist to forecast the demise of a regulator, but it was part of my bullish thesis on UK water coming into the year. A reset of the regulatory framework had become a necessary step in rehabilitating the sector, and the flurry of UK water bond issuance since the government’s announcement is not a coincidence.

Why was a reset necessary? Because the regulatory framework was fundamentally broken: first by Ofwat and subsequently by the failure of Thames. By mid-2024, the sector was equally maligned by the public, investors, and ministers - the only possible outcome was a new regulatory framework, and a new regulator was, to my eye, the fastest way to wipe the slate clean. That doesn’t mean that I expect anyone to be held accountable; I imagine almost everyone at Ofwat will either be pensioned off or shuffled into the new regulator. But for investors (both credit and equity), the war is almost won: bill rises and higher investment are on the horizon.

Specificially for Thames, I expect a fudge. The remaining standoff is over “hundreds of millions” in potential fines, but if the government followed through on the threat to push Thames into SAR1 it would be on the hook for approximately £4bn. This is a bad trade for the Chancellor, so I expect creditors will stand their ground and negotiate an outcome in which Thames pays some fines but the costs are offset by bill increases.

Share

A hitchhiker’s guide to the regulator

My first look at Thames was in late 2023. The utilities analyst asked me what I thought the UK government would do if Thames got into trouble - knowing nothing of the situation, I nodded along in agreement that government support was probable and that senior bondholders would be spared from pain. With hindsight, I'd have done well to keep my uninformed opinions to myself.

By March 2024 the utilities analyst had changed and Thames was spiralling out of control. A widely expected equity injection failed to materialise as Ofwat tightened the screws in response to increasing media outcry over sewerage releases into rivers. The shit that had been hitting rivers was now starting to hit the fan.

The new utilities analyst didn’t ask me my opinion (!) but, given the importance of the sector to sterling credit investors, I listened carefully and started to piece together the issues. I quickly came to the conclusion that Thames, and the broader water-sector, was in an impossible situation - and it was almost entirely due to Ofwat.

For the unfamiliar, the regulatory framework for UK water companies is horrendously complicated: for each company the regulator generates spreadsheets with dozens of tabs and thousands of cells. But at its core, the key concepts are simple: UK water companies operate assets for water extraction, purification, and distribution along with sewerage collection and treatment. Every 5 years, Ofwat determines a level of investment (new assets) and an allowed return on those assets: from those numbers it determines how much bills are allowed to rise. The complexity comes from how Ofwat assess the relationship between bills and profitability, modelling the entire water business including convoluted debt and capital structures.

On paper this seems like a reasonable framework. The regulator sets targets for water companies to meet in terms of service satisfaction, pollution etc., agrees a level of investment to meet these targets, and ensures that the return on investment is not excessive. But in practise Ofwat had a very poor set of political incentives - namely, Ofwat’s only real performance metrics were how much bills rose, and how many companies were caught being naughty.

As a result, Ofwat was highly motivated to push down on bills while being relatively light on enforcement, as too many fines would bring into question how effective the regulator was at policing behaviour. And this is what happened: the cost of water in the UK remained low versus other European countries, but at the cost of chronic underinvestment and lax oversight.

https://smartwatermagazine.com/news/locken/water-ranking-europe-2020

Regulatory driven underinvestment

As Charlie Munger once said “Show me the incentive and I'll show you the outcome”. Ofwat was heavily incentivised to keep bills low, and the two levers it had to achieve this were (1) to push back against “excessive investment” by water companies and (2) to overestimate the return on equity for water companies.

The implications of (1) are pretty self-explanatory and on average Ofwat lowered the allowable Total Expenditure by 7-12% across the last six price reviews, spanning 30 years. In other words, Ofwat consistently told water companies to invest less, as a direct result of the regulator’s determination to keep bills low2.

Alongside this tendency to minimize investment, Ofwat was willing to allow very high levels of financial leverage (debt) at water companies as this pushed the returns on shareholder equity higher without raising bills. In the 30-year secular bull-market for interest rates that spanned most of the post-privatisation period, this worked. Water companies took on high levels of debt, but because interest rates were falling this was often at a lower cost than assumed by Ofwat at the start of each maintenance period. As a result, water companies made a slightly better return on equity, making the regulatory framework look sustainable while keeping bills low.

But when interest rates rose in 2022, the music stopped. The cost of debt, and there is a lot of debt, started rising and public outrage was building over waste water spillage. Ofwat was forced to recognise the massive underinvestment that had accumulated and large bill increases became inevitable. But the same public outcry that forced Ofwat to acknowledge the lack of historical investment also made it impossible to increase bills to pay for the necessary remedial work.

Source: BBC, https://www.bbc.co.uk/news/uk-england-berkshire-67088859

Ofwat was faced with two basic options. First it could admit that underinvestment over previous decades (and under enforcement) was a feature (not a bug) of regulation, and that the only way out of the hole that had been dug was to push up bills and start to accelerate investment. The second option was to blame shareholders and Thames.

Naturally the regulator tried to single-out Thames as a uniquely bad actor. But, as the same problems started appearing across the sector, it soon became clear that the regulatory framework was rotten. In part, markets forced the hand of both Ofwat and the UK government, as contagion across share prices and credit spreads indicated investor dissatisfaction with the regulator.

Source: https://uk.investing.com/equities

Many argue that the problems in UK water are a primarily result of private equity taking excessive payouts, but the shareholders that bought Thames in 2017 never took a dividend and ultimately wrote their stakes down to zero. So if financiers are to blame, the seeds were planted in 2010 and somehow we muddled through for 15 years before noticing what was going on. Personally, I believe that it is more reasonable to say that UK water companies were raided by Ofwat to the benefit of households, who enjoyed abnormally low water bills for decades - but at the cost of future billpayers. Yet another case of intergenerational transfers?

Ratings: Mostly guaranteed

As the situation deteriorated, I talked to more and more bondholders and was surprised to hear constant reference to how the regulatory regime required Thames to remain investment grade. There was genuine anger from bondholders towards Ofwat as Thames’ bonds were steadily downgraded, perhaps more so than the company.

It was true that the regulatory framework required water companies to maintain an IG rating or they would lose their license to operate3. But this had, apparently, been interpreted as an implicit guarantee that these issuers would to remain Investment Grade no matter what. This was a fundamental misrepresentation of the regulatory framework, possibly due to a deeper misunderstanding.

A regulator cannot guarantee your bond will always be IG rated - after all the rating agencies are independent and no-one, certainly not Ofwat, could guarantee the ratings on your bonds. What the regulation did was threaten to wipe out shareholders in the event that a water company lost its IG ratings, creating a massive, asymptotic incentive for shareholders to support the company’s debt ratings. But, critically, this incentive fell away as soon as the equity was written down.

By construction, the regulatory framework would have no traction if shareholders voluntarily wrote their stakes down to zero, which is exactly what happened in April 2024. At that point Ofwat had no leverage - you can’t threaten shareholders who have already walked away.

Initially the credit rating agencies gave some leniency, in the hope of a potential white-knight rescue. But the shareholder logic was indisputable: there was zero, indeed negative, value in Thames for shareholders without higher bills, and Ofwat was unable to deliver those bills because to do so required it to admit that it was at fault. Thus, a restructuring was inevitable which would, by definition, mean Thames bonds falling to sub-investment grade ratings.

The restructuring at the end of the framework

Based on my personal understanding of Thames balance sheet, the political economy of Ofwat, and the regulatory framework, I concluded that Thames was cooked - it would be junked and most likely restructure. Admittedly, by April of 2025 this was not a groundbreaking opinion - though many legacy bondholders still hoped that the regulatory framework might somehow kick in to save the day.

The more interesting question, to me, was how the debt would be restructured. While not quite as extreme as the 2061 gilts, Thames has a number of long-dated, low cash price bonds in its capital structure and the simplest trade was to buy bonds that were trading below analyst-estimated recovery prices of 70-75c.

The rationale behind that, at it’s core, is based on liquidation economics: where each bondholder is pari-passu and paid the same percentage of par (recovery). But, we know that this will not happen for Thames - the ultimate fallback remains government receivership (SAR). Hence, the resolution of Thames debt will be a restructuring and not a liquidation, dramatically changing the likely outcomes for various bondholders. For example, high coupon bonds are likely to take a haircut on the coupon itself, reducing interest costs. In contrast, low-coupon instruments are more likely to be written down and short-dated instruments to be extended. In general, the overall NPV4 loss is likely to be much more equitable across all of the (class A) bondholders than a simple cash waterfall would imply. Despite this, cash prices are still clustered around analyst estimates for recovery post-restructuring.

Source: LSE

Critically, given that none of this was going be resolved quickly, the March 2025s looked to me as if they were likely to be repaid (and they were). There were also some really interesting trades to be done in the index-linked bonds, for anyone willing to do the heavy lifting on corporate linkers (post RPI reform).

Don’t panic

By the end of 2024 Thames had been junked, shattering the illusion that the Ofwat would ensure IG ratings for the water sector. Public trust in the regulator was rock bottom, amidst a constant stream of scandals over sewerage discharges and evidence that some water companies had been actively misleading the regulator. Successive governments, first the Conservatives and then Labour, were unwilling to foot the bill on rescuing Thames, or to go for an even-more-aggressive debt write down of senior bondholders as contagion across the sector intensified.

Politically there was only one option: blame Ofwat.

A new regulatory regime was clearly required, that would give cover for swingeing water bill increases. As it happens, my late mother was a career civil servant: first in the Ministry of Agriculture; then later MAFF; and then, after the BSE crisis, at DEFRA. Having lived through a departmental crisis once before, I was of the view that Ofwat’s days were numbered (and that, like my mother, most of those involved would succeed to the new regulator).

Souce: Nature.

Ofwat tried to head-off this outcome by unexpectedly giving water companies much more of what they asked for in its December 2024 Final Determination, but it was too late. With the Cunliffe enquiry already underway, Ofwat’s days were numbered. In Q4 I was actively arguing for this outcome, as part of a “the worst is behind us” thesis for the sector.

Specifically for Thames, it appears that everyone has agreed that there is no reason to inject equity without bondholders taking a stiff write down. And bondholders have concluded that if they are going to take a write down, they might as well own the upside of holding the new equity. The transition of debt from bond investors to restructuring and recovery specialists in 2024 has facilitated this immensely.

The remaining sticking point appears to be fines. There is little reason to put equity into a company if the regulator will immediately extract it via fines. But the UK government is still beholden to the court of public opinion, and will struggle to let Thames Water walk away from past offences scot-free.

Given that Thames’ Final Determination has been referred to the CMA (and twice delayed), I will hazard a guess that the most likely way forward is that the CMA will further increase bills for Thames customers, which would allow the new holders of Thames to absorb potential future fines without taking a larger upfront haircut. This would be a win for the current government, as it would push yet more of the blame onto Ofwat while also allowing them to avoid footing the bill for cleaning up the mess. Both literally and figuratively.

There will, doubtless, be more protests in the court of public opinion, but the corpse of Ofwat will provide a convenient scapegoat for Ministers past, present, and future.

Leave a comment

Source: Google Gemini

August Non-Farm Payrolls were weak, the revisions were worse, and the reaction of President Trump was worse-r still. At +73k, the US economy added the fewest jobs since January 2021 (excluding a hurricane-afflicted October 2024). Alongside this, 258k jobs were revised away, as the Bureau of Labor Statistics (BLS) lowered its June and July estimates. These were the largest downward revisions since COVID.

Markets responded negatively to the data, though this appears to have been short-lived as spreads are back to the tights and equities regained their losses this week. But we should expect employment data to remain weak for as long as the US pursues the Trump administration’s tighter immigration policy.

Source: FRED

Subscribe now

Lies, damn lies, and statistics

President Trump firing the head of the BLS was disgraceful, but not out of character - after all, ‘The US is an EM’. But, if the President wants to blame someone for slowing employment growth, he should start with the head of Immigration and Customs Enforcement (ICE).

The Trump administration’s promises on closing the Southwestern land border and deporting criminals were very explicit, and have been put into effect with notable vigour. Given that these policies were a central pillar of Trump’s campaign, there a several pieces of analysis on the economic implications, for example here and here. Of course, these are just forecasts, and forecasting how migration will evolve under these policies is a huge challenge. But we are now starting to see some effects in the data: for example, encounters at the Southwestern border have fallen to extremely low levels. As such we can start replacing guesses with evidence.

Source: US Customs and Border Protection

Helpfully, the BLS splits its workforce survey between US and foreign born workers. The data doesn’t exactly align with NFPs, as it comes from the household survey, but it shows that employment of non-US born workers has turned deeply negative since March. This is not seasonally adjusted1, but according to the data the US economy has lost approximately 1.6 million non-US born workers in the last four months. Indeed, the loss of foreign-born workers has been the most severe since COVID.

Offsetting this, 2.6mn US born workers were added over the same period - but half of that was seasonal hiring in June2. On a seasonally adjusted basis, therefore, this data seems well aligned with the message from the August payrolls report - employment growth has stalled. Overwhelmingly, the key driver of the recent weakness in payrolls has been a collapse in non-US born workforce, a smoking gun that this is being driven by the new administration’s clamp down on “illegals”.

Source: FRED. Note the strong seasonality in the US born workforce data.

Share

Doing less with less

While it varies depending on the exact period you measure, slightly less than half of all US employment growth since June 2022 was foreign born workers. Making predictions is hard, particularly about the future, but if we assume that the foreign-born workforce flatlines for the remainder of the year, then payrolls would fall by roughly half. Given that NFPs were averaging around 250k/month until the start of this year, that would imply payrolls of around 125k going forward. Layer on a standard deviation of 170k, and further weak or even negative payroll reports are likely in the coming months. If the non-US worker count continues to fall (and not just flatline) then risks are further to the downside.

Given that employment weakness is arriving before tariffs pass through, the Fed are likely to deliver one or two cuts starting in September and this is probably what has allowed the market to recoup its initial losses. But I suspect that the FOMC will not manage more than a few insurance cuts before goods inflation starts to tick higher, on the delayed pass-through of tariffs. As such, volatility in short rates is likely to remain elevated, even more so as we approach the end of Powell’s tenor. Given that Trump is starting to run out of policy rabbits to pull from the hat, reducing credit/equity risk into strength feels prudent over the coming months.

Further out, lower workforce growth means lower US growth. President Trump may ease off on his migration stance over time, though perhaps not before banning the publication of immigration statistics or at least installing a new departmental head. In that case the US would benefit from a reacceleration in employment growth. But as long as US migration policy remains highly restrictive the economy will splutter.

To quote the Peterson Institution, “the US economy is significantly more dependent on foreign workers than on foreign imports." And with the bulk of tariff uncertainty (probably) behind us, focus should shift to the slowing of the US economy.

Leave a comment

It’s rare that a bond achieves infamy, or indeed even gets much attention at all. So when I saw multiple articles on the UK Treasury 2061 (TG61), it was almost obligatory for a fixed-income blog to comment. It’s a trade that has gone wrong for many, but also a teachable moment for everyone1.

My first encounter with the TG61s was in October 2022. I was seeing clients in Milan as the LDI crisis rapidly unfolded and, between meetings, I was glued to my phone, explaining to management in New York what an LDI was, and why they were selling £-IG, $-IG, and everything else they could get a bid on. Towards the end of the day, I caught up with a friend and colleague from Sales, who calmly asked me if 2061s were the trade of a lifetime. I pulled up Bloomberg to check TG61s first as a price, then as a yield, before giving my answer.

“Horrible trade, great investment.”

https://www.dividenddata.co.uk/uk-gilts-prices-yields.py

At the time my wife and I were saving for a house and it occurred to me that putting that deposit into the 2061s at 35c would provide a fantastic pension. Here was an opportunity to solve for retirement in one move - albeit at the cost of having to solve for the house later (or never). My colleague and friend was not interested: for him this was a “no brainer” trade for the short term. Gilts were, in his view, clearly oversold trading at distressed prices.

“What could go wrong?” He asked.

“They pay you par. In 2061.” I replied.

https://www.dividenddata.co.uk/uk-gilts-prices-yields.py

Subscribe now

Yields up (prices down)

Even in the context of my career, 2022 was a busy year for credit markets, as series of interconnected crises played out. Inflation had picked up over the course of 2021, but it took until December for markets to stop believing that it was transitory in nature, and instead come to the realisation that inflation forecasts were less than worthless and that central banks would have to tighten policy rates…a lot. The first real hiking cycle in over a decade was then super-charged by the Russian invasion of Ukraine, which triggered a spike in European energy prices. This set in motion a relentless grind higher in interest rates that, combined with the ill-fated Truss government, resulted in a crisis for UK LDI managers2.

One experience that was, I suspect, much more personal was the realisation that even extremely experienced bond investors were struggling to internalise the truism that bond prices fall as yields go up. On more than one occasion, I had found myself opposite an extremely excited High Yield/CLO manager asking why they shouldn’t buy some extremely low-price, long-dated GBP bonds such as BATSLN 2052s that were trading in the 50s.

Their argument was simple: these were large, blue chip companies, and therefore very, very unlikely to default - and if they did default the bonds would recover higher than their current prices anyway. So there was “no downside”. On the other hand, I knew that the low cash prices simply reflected that yields had risen and that these bonds had a very long duration. There was nothing surprising about them having low cash prices, it was just a mathematical artifact.

The gap between our perspectives largely came from HY being a cash-quoted market. High Yield managers simply were not used to seeing low cash bonds, unless the company was trading stressed/distressed. Through this lens, the potential to buy bonds from solidly investment grade companies at “distressed” (cash) prices looked like free money. Underlying this was a deeper, implicit assumption that the yield levels seen in late 2022 were a temporary blip. Transitory, if you will. After 10 years of zero yields, 5%+ returns were an aberration to be jumped on3.

So where was the downside? The downside was the risk that yields would stay at those Q4 2022 levels and that you got paid par at maturity, decades later. And that, more or less, is where we find ourselves in H2 2025. Still waiting for a “reversion” in interest rates that seems no closer than it was 3 years ago - even if policy rates have come down significantly from their peaks.

10y constant maturity benchmark yields. Source: FRED

Share

A great investment (to maturity)

All that said, I do think that low-price gilts are a fantastic opportunity if you happen to be either foreign, rich enough that your UK pension/ISA allowances have already been filled (!), or a cash saver. That is because gilts are capital-gains tax free, a situation that may not be unique but is certainly unusual across the government bond markets that I reviewed several months ago.

As a result, low cash price gilts look fantastic versus many other asset classes, offering (credit) risk free, guaranteed yields equivalent to 5-7% versus other, taxable asset classes. Provided you are able to hold them to maturity4.

Equivalent yield on a taxable savings account assuming 0% CGT, 45% income tax

The rub is that most people in the UK have ample capacity in the two basic tax wrappers available: pensions and ISAs. Investments inside these wrappers are already free of capital gains tax and so the tax benefit of low-cash price gilts is lost, and they are just bonds offering 3-5% (still beating most savings accounts). To benefit from the capital-gains tax efficiency of gilts you need to have a substantial amount of cash sat outside of your ISA. And even then, to the whole point of this article, you need to match the duration of the bond to your investment horizon.

I own both the 2046 (5.23%) and IL46s (2.20% Real Yield)5. When they mature, I will be about 62 which means they line up nicely with my (hopeful) retirement. I also own some of the 2026 and 2028s because they beat most savings accounts, even on a pre-tax basis (and definitely post-tax). Through this lens, the TG61s are a poor investment for anyone born before 1995, as many of my peers and colleagues have found out.

Leave a comment

Is this fair and, more importantly, was it predictable? Mostly, and yes.

Subscribe now

Tariffs matter, but largely from a political economy and distributive perspective. The US may raise as much as $300bn this year (c.1% of GDP) from tariffs and, despite President Trump’s assertions, they are almost never paid by the exporting country. Overwhelmingly they are paid by the importer through lower margins, by the customer in higher prices, or by the importer (again) through lower volumes. Hence, this is de-facto fiscal tightening. That said, fiscal policy is not a single line-item and we need to balance tariffs against the One Big Beautiful Bill. Looking holistically at US policy, there is a fiscal transfer from consumers of imported goods to wealthy households, and a transfer of power from congress (which sets taxes) to the White House (which sets tariffs). Indeed, Trump has floated the idea of making this trade-off explicit by sending households a cheque funded by tariff revenues.

https://bipartisanpolicy.org/explainer/tariff-tracker/

What about the macro-economic impacts? After the November elections, sell-side economists had the unenviable task of forecasting the impact on growth and inflation of various tariff scenarios. One house estimated a roughly c.0.7pp drag on EU growth in 2025, based on a scenario of 10% global tariffs, 30% tariffs on China, and retaliation of a similar magnitude. Tariffs look likely to settle at higher rates (closer to a weighted average of around 15-20%), albeit with plenty of complexity at the line-item level. And yet, despite this, economists have given up on calling for a sharp US recession as tariffs have been lowered from their “Liberation Day” levels. Net-net, tariffs are likely to land at higher levels that were estimated coming into 2025, but the economic impacts have been modelled away. How?

When it’s boring, fudge it

A point I discussed in every client meeting since the election was the composition of that forecast 0.7pp drag on growth. It consisted of two* elements: a ‘trade friction’ component, and a ‘confidence’ component. The first part was self explanatory and based a large amount of data-work to estimate marginal elasticity of trade once tariffs hit. The second part was a ‘vibes’ estimate of how much trade disruption would affect consumer confidence.

Critically, the frictional component contributed just 0.1pp of the 0.7pp drag. Almost the entire effect came from the confidence component. Or vibes. This aligns with an ECB estimate that trade frictions would lower Eurozone GDP by 0.2-0.4pp in the first year, made using Liberation Day tariff levels that were about twice what had been assumed in street models - so the estimates were well aligned on the frictional effects.

The fact that most of the impact on GDP from tariff modelling came from the ‘confidence channel’ (vibes) had four important implications:

1. The modelling was highly uncertain: For sure confidence was likely to be negatively affected, but determining the size of that effect was bordering on impossible. Heavy use was made of trade uncertainty measures - efforts that fell flat on their face when trade uncertainty indices promptly did this:

   

2. Confidence effects are non-linear: A technical but important issue is that the modelling assumed confidence effects are linear: that if you double tariffs then people are twice as upset. But this is highly unlikely. It is much more likely there is some low level at which you are indifferent to tariffs and some high level at which there is no trade. The transition between these states is likely to be over a fairly small range - in effect an S-curve. As a result of this issue, the initial yo-yo-ing of tariff rates that caused a constant restating of growth/inflation forecasts was largely meaningless and made calibration essentially guesswork.

3. Not a single-name story: Critically for credit/equities, the fact that trade frictions are modelled as being small explicitly told us that looking bottom-up for affected names was unlikely to yield much, away from acutely affected names (e.g. autos). The bearish macro call rested not on tariff effects, but on a consumer driven recession - as such the leverage to the theme was in recessionary plays rather than idiosyncratic supply chains issues.

4. Confidence effects are fast: Very fast and they don’t wait for change to happen, just the fear of change is sufficient. In contrast, trade frictions hit an economy quarters after the tariffs are implemented. This meant that street forecasts would be proven right (or wrong) very quickly… by Q2, in fact.

One reason I picked up on these points was that I had seen the exact same thing play out with Brexit. In 2016, economists scrambled to estimate the effect on trade from Brexit and came up with figures in the region of 0.1%. Embarrassed to publish reports saying “actually trade doesn’t matter for a services based economy like the UK”, the forecasts were fudged by assuming a consumer confidence shock**. It was too embarrassing to forecast that trade didn’t matter, so economists asserted that it did.

But the confidence shock failed to materialise, and the acute recession that was penciled in for H2 2016 never happened. Even later admitted falling victim to the “confidence-fairies” narrative of Brexit’s short term effects. To my eye, 2025 is a repeat of the same mistakes.

Share Macro Credit Thoughts

Trading the (actual) macro

Despite the cynicism, these points gave a clear set of investable insights.

First, forecasts of a moderate drag on growth from tariffs were always likely to be wrong - more likely the effects would either be vanishing, or they would trigger an immediate, sharp recession with little room for anything in between. Accordingly, the price action before, during, and after the April tariff shock was spot on - the market was testing whether Liberation Day was enough of a shock to trigger an immediate recession***. Now that it has not, tariffs are a rounding error for growth estimates.

Second, at least in credit, the best expression of tariff concerns was not by trading exposed supply chains, but by trading recessionary probabilities. Autos had already underperformed in Q4 after President Trump was elected to a second term. This premium then caused the sector to outperform in April, as other sectors were repriced by the increased recession risk and caught up (down) to autos. There were tradeable moves wider in a few auto names, but this largely reflected their beta****.

Right now, the market is correct in losing its beta to tariffs. Whether the final rate is 10% or 20% doesn’t matter, indeed it never mattered in the short run. All that mattered was if there would be enough of a “dead confidence-fairies” effect to trigger a US recession, and the macro data is clear that we have seen nothing of the sort. Trade is not going to knock the wind out of the market’s sails and was never likely to, once you dug into the modelling being done. If there is a bearish narrative for risk, the most likely one is that the Trump administration has fulfilled its policy agenda so effectively that it is running out of levers to pull to push markets even higher.

On a final note, I will flag that the commentators doing victory laps on the lack of US goods inflation thus far are getting inflation wrong again. We know that there was huge front-loading at pre-tariff prices and, while imports will resume with applied tariffs in the coming months, it’s likely that importers will feather the higher-cost inventory in with existing lower-cost inventory to smooth the price effects. So goods inflation is coming and will keep the Fed on hold, it’s just going to have a larger-than-expected lag. We saw the same impatience with goods inflation from markets in 2022-23, and with regard to monetary policy tightening in 2023.

Leave a comment

*Actually three. The trade friction component had an element for direct trade between the EU and US, a second for then trade redirection: but this was largely unimportant away from discussing the goods deflationary effect of China exports.

** In the UK, consumer confidence is closely correlated to the Services PMI, which is closely correlated to UK GDP. As a result, assuming a confidence shock is equivalent to assuming a recession. Running that assumption through econometric models was simply a case of Rubbish-in, Rubbish out.

***For example the risk of a total trade shut-down, or if the equity market sell-off became self-fulfilling.

****Even for the Autos sector, digging into analyst modelling showed it was less about the impact of tariffs on costs and more about concerns that a pass-through of costs would cause a recession-like collapse in unit sales. While this was a completely reasonable position, it runs counter to the general rule of thumb that unemployment (not prices) is the best predictor of recessionary behaviour. So again there was an assumption of recessionary collapses in auto sales underpinning the bearish narrative on auto credits.

With the GENIUS Act signed into law, the current US administration continues its “move fast and break things” bear-hug of cryptocurrencies. Whether stablecoins can single-handedly take down trillions of US debt deserves its own article, and some number-crunching on my side, but I will point out that the idea of stablecoins counter-balancing the fiscal profligacy of the One Big Bill only cropped up in the last few weeks. In contrast, the STABLE Act was drafted in February and stablecoin regulation has been in the works at least as far back as the de-pegging of Tether in 2020 (and the collapse of Terra in 2022).

Co-Pilot: 20/07/2025

There are many things I find fascinating about cryptocurrencies, which I have been loosely tracking since 2012, but it has long been clear that, by and large, they were a solution looking for a problem. As a result, the rationale for owning crypto has been in constant flux. Originally BTC was an computer-science novelty (solving the Two General’s problem) and a hedge to USD debasement via QE in 2010 (remember that?). Over time crypto has also been touted as the future of finance, a payment rail for the dark web, a meme (literally), a core allocation for pension funds (!?), and more.

In my mind, the chimeric arguments for crypto reflect two key narratives that have come into tension with one another:

1. By operating outside of the existing financial system, cryptocurrencies originally promised both a shelter from government laws and a shelter from government collapse. As a result adoption was rife in low-trust, low-law countries such as Russia, China, North Korea, and various other EM countries. In this sense, widespread support in the US was a surprise (or, perhaps, a warning).

2. Because price is a social construct, adoption is value. As cryptocurrecy adoption rose so did cryptocurrency valuations. But most people don’t want value that is trapped outside of the world in which they live. As a result, the wild success of crypto actively undermined its original value proposition: more and more of the cryptowealthy wanted to move their ‘money’ off the chain and began to push for legal frameworks and safe harbours for their digital wealth.

The institutional adoption of cryptocurrencies by the Trump administration is the convergence of declining governance in the US (because the US is an EM) and the increasing success of the cryptoverse in embedding into the world of “traditional finance” (or TradFi, as it the Defi community call it).

As such, even for the uninterested/uninvolved, I think there is value in keeping one eye on developments in the cryptocurrency space as part-and-parcel of tracking the US political economy. Or even owning some.

The three times I bought BTC (or at least tried)

I have attempted to buy BTC three times, only two of them successfully: 2013, 2017, and 2024. Each time the thesis was different and sadly (for me) in 2013 I was stymied by repeated requests for high-res scans of my passport from a Cayman Islands based FX platform that I found via Google. Eventually I gave up, figuring that identity theft was more risk that I was willing to wear.

My three conviction buys on BTC

https://charts.bitbo.io/price/

Thesis 1: Faith based investing

When Bitcoin first came on the scene (in computing circles) the cynicism was deep and well founded. I was also a sceptic, but what struck me was that the original BTC crowd were true believers - diamond-handed HODL'ers who were convinced that QE would trigger an age of hyper-inflation and currency debasement, leading BTC to replace all fiat currencies.

It is widely understood that faith is impervious to facts. Indeed, when presented with facts that contradict their beliefs, people overwhelmingly double-down on the side of beliefs. And that was my thesis: that there were enough people pointing out how BTC was nonsensical that the BTC believers, who were the only ones involved in the market such as it was, would double-down on the idea.

Unfortunately, unless you mined the stuff, it was remarkably hard to buy BTC in 2023, particularly from the UK.

Thesis 2: Bubbles make everyone look stupid

By the time 2017 rolled-around, crypto was making new highs. BTC topped $10,000 and I was doing the mental math on how I could have retired at 34 if I had just been more dogged in emailing my passport to the Cayman Islands (+10,000%).

Some people were getting rich in crypto and many other people were noticing. Indeed, even in my corner of investment-grade credit, clients were suddenly asking me what I “thought about Bitcoin” during meetings. Due to my earlier dabbling, I was probably one of the more qualified people to say anything on the matter, and even contributed on the technical side to a cross-asset report on BTC towards the end of 2017.

While the lead author argued that BTC had already reached max adoption and so max price (whoops!), I had a simpler take. Bubbles are equal-opportunity pain: you can look stupid on the way up, by not being involved; or you can look stupid on the way down when you lose your shirt. Only a few manage to get out (and stay out) near the top. This time around I chose to buy in because the pain of prices exploding higher a second time would have been too much for me to take: plus the stakes were £500. It was an emotional hedge.

Thesis 3: Corruption and safety rafts

My most recent purchase of BTC rests on third, but again simple, thesis. Over the course of 2018-2020 the crypto-ecosystem had re-oriented itself from being inward facing and confrontational, to being outward facing and compliant. With hindsight, I was both aware of this and ignored it - though partly because I had spoofed-myself into buying a sizable (for me) position in the Hedera alt coin in 2021, and I was stuck warehousing a significant draw-down*.

The crypto lobby spent years pushing for what it referred to as “on-ramps”, to make it easier for people to ‘invest’ in crypto. I referred to them as off-ramps. There was a sizable community of people who had huge paper wealth due to the run up in prices and were now less interested in HODL and more interested in cars and houses (see, for example, this FT article). These individuals wanted ways to get their wealth off the blockchain and into the real economy which meant that, instead of being antagonistic to regulation and regulators, crypto-exchanges started pushing for regulatory clarity to ensure the legality of wealth transfers. Wealth managers started pressing for ways to onboard newly minted millionaires as clients and so on. Eventually we even got bitcoin ETFs, which were a runaway success.

This multi-year effort to wire crypto in the existing financial system has sped ahead since the election of President Trump, who has repeatedly delivered for a community that unashamedly supported him last year. Alongside the GENIUS act there have been other rewards for crypto-maximalists such as the US Strategic Bitcoin Reserve, and the pardoning of both BitMEX and the man behind Silk Road. The President also got in on the action with his Trump coin (and then Melania coin).

So am I long corruption? A little bit.

But holding BTC is part of my broader thesis that the US is an EM. In this framework BTC serves three useful purposes for the Trump administration:

1. It has brought his campaign money and votes, as the cryptoverse was overwhelmingly pro-Trump in 2024. Promises made, promises kept.

2. It has brought him even more money. The de-facto legalisation of crypto while paired with light-touch regulation has created a wonderful channel for donations to President Trump (Legal Eagle has a strong take on this).

3. It’s a safety net for the rich and connected if Trumponomics goes wrong, harking back to the original fear that spawned Bitcoin.

My BTC purchase in November 2024 rested on (1) and (3). Populist Presidents love growth and a core view of mine coming into 2025 was that the US would continue to hyper-stimulate its economy with implications for rates (bearish), equities (bullish), and credit (bullish). This also weaves into an (even larger) thesis of mine, that Fiscal Dominance is back for this decade, but I digress. The One Big Beautiful Bill confirms that the US is likely to keep running a large, highly stimulative deficit for the rest of this presidency as I expected.

So far, so good. The US continues to spend, the economy powers on, and risk assets rise. What’s the problem? Well if life were so easy, governments around the world would not have turned to austerity post-GFC. But running a huge deficit in a pro-cyclical manner has historically had significant long-term costs, literally and figuratively. In particular, it tends to end in either a debt or currency crisis, or both - often via inflation, a bubble, and a crash.

With this in mind, embracing crypto is also is a hedge for the ultra-wealthy that surround the president and who cheered his tax cuts**. If the economy generates a blow-off top, then these people will benefit from their existing investments and/ from being close to the President. And if it all goes wrong, then BTC might even deliver its original value proposition from 2010 - as a hedge to a fiat dollar (though this is far from certain given that BTC has become highly correlated to the SPX).

So from an investment perspective, I still think that (even at these levels) owning BTC is an interesting way to own both tails. There is the potential for further gains if BTC sees continued adoption on the basis of past returns and greater legal clarity (or if there is a blow-off-top in US equities), but it may also outperform the USD if the administration hits exit velocity with regard to inflation, fiscal dominance and eventually dollar-debasement. Or if they fire Powell.

In some ways, BTC can be viewed as an inverse-UST: it is likely to do well in periods of high nominal growth and/or sovereign crisis. The downside risk is that neither of the above tails are realised: a final puff as the asset class is adopted and becomes a zero-yielding, low-volatility commodity, or if there is a regular-way US recession. But in that case USTs are probably rallying nicely.

Subscribe now

Share

*This eventually paid off in 2024 for about +200%, confirming that my HBAR position was giving me a material beta to BTC anyway.

** I refer to this as the looting phase.

*** If anyone is wondering why I wrote all this but I am not swimming in crypto wealth, those BTC trades in full are below. The bigger payoff has been the amount I have learnt from being involved in the conversations:

Image: Google Gemini

The need to explain yourself

Having been told in the past that I often talk around the point, I wanted to make sure not to bury the lede for my first post. So here is one of my most off-piste arguments: right now the US is, from an investor perspective, more EM than DM*.

Before I go any further, I need to be clear exactly what I mean when I say that the largest, most financialised, and most important economy in the world is an EM. While I have never covered Emerging Markets as my day job, aside from index-eligibility discussions when Greece was junked in 2010, it always struck me that the key difference in EM investing is the need to consider “rule of law” risk. This is why China, the second largest economy in the world, is an EM but Japan is not: I can buy a Japanese stock or bond and that investment will, by and large, stand or fall based on my analysis of the underlying issuer. But to invest in mainland China I would have to consider other risks, such as capital controls or changes in the political winds - as China HY credit investors found to their cost when the previously default-averse central government discovered an affinity for restructuring the off-shore bonds of property developers after 2021.

Another example here is Russia, where bondholders were left holding nothing but compliance headaches when US and EU sanctions were imposed in 2022. Of course, policy risk is always out there to some greater or lesser degree, but when it comes to EM investing I’d say that understanding the political economy of a country is at least as important as understanding their real economies.

So where does this leave the US?

As of today, the US is incredibly hard to invest in, in every sense of investing. Clean energy investments made under the Biden administration have been put at risk by the One Big Beautiful Bill, and investments under the OBB run at least some risk of being derailed by the next administration. More acutely, trade agreements are in a perpetual state of flux as tariff rates swing constantly from one headline to the next: a great environment for copper trading (BBG), but not for building efficient supply chains. Just a few weeks ago global investors had credible concerns that Section 899 might affect withholding tax on USTs (admittedly mostly due to misunderstanding). And on more than one occasion the President has been explicit about where and how he wants corporates to do business and to make their products.

At the same time, numerous commentators have picked up on eyebrow-raising trades in and around market-moving policy decisions (e.g. Here) and movements in US stocks such as TSLA have become increasingly correlated to the perceived relationship between CEOs and the President. This is all prima facia evidence that political economy is rising in importance, even when investing in US mega-caps: and because it is by Presidential design it is unlikely to abate**.

I am not the only person with this opinion: there are ETFs (such as GRFT) and investment vehicles based on the idea that being part of the Trump “inner circle” is a functional way to make excess returns. With Trump’s voting base adamant that the President is on their side, perhaps it makes sense that the baton of the US equity bull market has been passed to US retail, who may feel they have a better read on the President than Wall Street analysts.

How should a credit investor think about this?

While the preceding narrative helps me think about the current state of the world (accepting that it may be completely wrong), does it matter for credit investors if the US is a little (or a lot) more EM like? I think it matters for two reasons: first, because we need to change our investment process, with more emphasis on understanding the Trumpian political economy. The US president is quite happy to pick and choose winners and that has to matter for stock selectors. But there is also the fact that if investment decisions increasingly reflect a political factors over economic ones, then over-time that will impede efficient capital allocation, pushing down on growth and up on US inflation.

On the other hand, this all has frustratingly little bearing on credit spreads. Because $-IG is a multi-faceted asset-class, any build-up of risk-premium in US assets is more likely to be seen in a weaker USD (already down, but not for any of the above reasons**), wider UST asset-swaps (though on the other side there is SLR reform and stable-coin narrow banking**), or simply higher outright yields. Indeed, some credit investors have reached a similar landing point from an entirely different direction: arguing for a convergence in ‘credit quality’. between an Apple or a Microsoft and the US government as deficits remain untamed. i.e. That these issues are a US government credit problem, but not a US corporate credit problem.

There are other reasons to conclude that even if the US is becoming less investable, $-IG spreads are unlikely to bear the brunt of any repricing in the short, or medium term, related to the relative structure of $-IG and €-IG**. Nevertheless in a global IG portfolio, invested across multiple currencies, the tides that have benefited US credit since 2022 appear to now be reversing.

So I’ll go out on a limb and state that I view the second Trump administration as the swansong of US exceptionalism. Over time, the growing influence of political patronage is likely to distort and drag on the supply-side of the US economy, achieving what Mr Lutnick desires**: a weaker dollar and higher UST yields.

And I am comfortable saying this now, while market in general remain undecided on the matter.

* Ok not entirely off-piste as several articles with the same title are out there (e.g. This one), but equally not something that it would be easy to write down as a sell-side analyst.

**All topics deserving of their own posts in the future.

Post Script.

It turns out that after 16 years of writing research I have become used to, and soothed by, writing down my thoughts. I’m not sure how many of you will find value in reading these notes to myself, but please do comment below (or DM me) to agree, disagree, or suggest other topics that are of interest. All comments and opinions are my own.

Leave a comment

Subscribe now