What We’re Reading (Week Ending 29 September 2024) - 29 Sep 2024
Reading helps us learn about the world and it is a really important aspect of investing. The late Charlie Munger even went so far as to say that “I don’t think you can get to be a really good investor over a broad range without doing a massive amount of reading.” We (the co-founders of Compounder Fund) read widely across a range of topics, including investing, business, technology, and the world in general. We want to regularly share the best articles we’ve come across recently. Here they are (for the week ending 29 September 2024):
1. Digging Into The Coal Industry – Matt Franz
There are two main types of coal: thermal and metallurgical (“met”) or coking coal.
Met coal has more carbon, less ash, moisture, and sulfur than thermal coal. It is rarer and commands a higher price.
Met coal is a key ingredient in steel. To make steel, met coal is first turned into coke by heating it to 1,000ºC in the absence of oxygen. Without oxygen, the coal does not burn. The coal swells and then releases its gaseous volatile matter. Nearly pure crystalline carbon is all the remains. Coke is to coal what charcoal is to wood. Caking is a coal’s ability to be converted into coke. Thermal coal has no caking ability, which is why it is much cheaper.
Coke is mixed with iron ore, flux (e.g. limestone), and hot air in a blast furnace to create iron. Iron is put into a basic oxygen furnace where oxygen reduces the metal’s carbon content. It is further refined to remove impurities and alloys are added to make steel.
The quality of met coal influences the quality of the coke, iron, and steel produced. A blast furnace fed with higher quality coke will require less of it, lowering production costs. Steel makers have an economic incentive to pay more for higher quality met coal.
It takes 0.8-1.05 tons of met coal to produce one ton of steel. (1.3-1.5 tons of met coal make one ton of coke. 0.6-0.7 tons of coke make one ton of steel.) That’s a lot! 70% of the world’s steel is made this way…
… The major met coal exporting nations are the US, Canada, and Australia. The major importers are countries with large steel industries relative to their domestic met coal supplies – China, India, Japan, and South Korea.
The US usually exports ~70% of its met coal. Export contract prices are tied to international benchmark indices. Domestic contracts tend to specify a fixed price and a fixed volume for one year…
…Thermal coal has a lower calorific value (CV) and a lower cost than met coal. It is primarily used to generate electricity. It is also used to make cement. It takes 0.1-0.12 tons of coal to make one ton of cement.
Thermal coal has been in decline in Europe since the 1980s and in the US since the 2000s. It continues to grow in Asia. Worldwide, coal demand reached its highest level ever in 2022 and again in 2023…
…Today, coal remains the world’s largest energy source for electricity generation. Coal may be losing share as an energy source, but it continues to grow in absolute tons.
China is the world’s largest producer and consumer of coal. In the early 2000s coal produced 75-80% of its electricity. Today it’s more like 60-65%. Coal lost share but grew in absolute terms. Chinese electricity production rose eightfold and its coal consumption rose sevenfold…
…Coal’s peak share of U.S. energy occurred around 2007-2008 at 50% of electricity. Today it’s 20-22%. That’s still a meaningful amount. The decline in US coal was driven as much by fracking and its byproduct, cheap natural gas, as environmental considerations. Should US natural gas get expensive, we could see a shift back towards coal…
…Coal’s share of energy production is falling slower than the increase in total energy demand. Jevon’s Paradox describes this situation. Demand for energy is elastic. As energy costs decrease, demand for energy increases even more. On balance, energy demand increases, even as energy consumption becomes more efficient…
…In Energy and Civilization: A History (2017), Vaclav Smil explains that energy transitions often take more than a century. The transition from biomass (wood) to coal in Western Europe took 200 years. The transition from coal to oil began in the late 19th century, but oil didn’t overtake coal as America’s dominant energy source until the 1940s, approximately 50-60 years later. The transition is still far from complete, and that’s despite crude being more energy dense and easier to transport (via pipeline).
One of the factors affecting the speed of the transition is infrastructure. The transition from coal to oil in America was slowed by the need to replace steam engines with diesel engines. The modern analog is the cost and complexity of switching a power plant from coal to natural gas.
Price is another factor. If there’s a new fuel that is much cheaper than the legacy fuel, there’s an economic incentive to rebuild the infrastructure faster. A wide disparity between coal and natural gas prices that is expected to continue will drive more US coal plants to switch to gas. That’s less likely to happen in Asia, where gas is less plentiful and LNG infrastructure is more expensive.
Once again, this suggests that the last ton of coal will be very expensive, not very cheap. Thermal coal may be a sunset industry, but it is going to be a beautiful sunset.
2. Fed up with Fed Talk? Factchecking Central Banking Fairy Tales! – Aswath Damodaran
As I drove to the grocery story on Fed Cut Wednesday, I had the radio on, and in the news at the top of the hour, I was told that the Fed had just cut interest rates, and that consumers would soon see lower rates on their mortgages and businesses on their loans. That delusion is not restricted to newscasters, since it seems to be widely held among politicians, economists and even market watchers. The truth, though, is that the Fed sets only one interest rate, the Fed Funds rate, and that none of the rates that we face in our lives, either as consumers (on mortgages, credit cards or fixed deposits) or businesses (business loans and bonds), are set by or even indexed to the Fed Funds Rate…
…While the Federal Open Market Committee controls the Fed Funds rate, there are a whole host of rates set by buyer and sellers in bond markets. These rates are dynamic and volatile, and you can see them play out in the movements of US treasury rates (with the 3-month and 10-year rates highlighted) and in corporate bond rates (with the Baa corporate bond rate shown)
There is a final set of rates, set by institutions, and sometimes indexed to market-set rates, and these are the rates that consumers are most likely to confront in their day-to-day lives. They include mortgage rates, set by lenders, credit card rates, specified by the credit card issuers, and fixed deposit rates on safety deposits at banks. They are not as dynamic as market-set rates, but they change more often than the Fed Funds rate…
…To test whether changes in the Fed Funds rate are a precursor for shifts in market interest rates, I ran a simple (perhaps even simplistic) test. I looked at the 249 quarters that compose the 1962- 2024 time period, breaking down each quarter into whether the effective Fed Funds rate increased, decreased or remained unchanged during the quarter. I followed up by looking at the change in the 3-month and 10-year US treasury rates in the following quarter:
Looking at the key distributional metrics (the first quartile, the median, the third quartile), it seems undeniable that the “Fed as leader” hypothesis falls apart. In fact, in the quarters after the Fed Funds rate increases, US treasury rates (short and long term) are more likely to decrease than increase, and the median change in rates is negative. In contrast, in the periods after the Fed Fund decreases, treasury rates are more likely to increase than decrease, and post small median increases…
…In the quarter after the Fed Funds rate increase, mortgage rates and fixed deposit rates are more likely to fall than rise, with the median change in the 15-year mortgage rate being -0.13% and the median change in the fixed deposit rate at -0.05%. In the quarter after the Fed Funds rate decreases, the mortgage rate does drop, but by less than it did during the Fed rate raising quarters. In short, those of us expecting our mortgage rates to decline in the next few months, just because the Fed lowered rates on Wednesday, are being set up for disappointment…
…How else can you explain why interest rates remained low for the last decades, other than the Fed? The answer is recognizing that market-set rates ultimately are composed of two elements: an expected inflation rate and an expected real interest rate, reflecting real economic growth…
…Interest rates were low in the last decade primarily because inflation stayed low (the lowest inflation decade in a century) and real growth was anemic. Interest rates rose in 2022, because inflation made a come back, and the Fed scrambled to catch up to markets, and most interesting, interest are down this year, because inflation is down and real growth has dropped…
…The Fed’s major signaling device remains the changes in the Fed Funds rate, and it is worth pondering what the signal the Fed is sending when it raises or lowers the Fed Funds rate. On the inflation front, an increase or decrease in the Fed Funds rate can be viewed as a signal that the Fed sees inflationary pressures picking up, with an increase, or declining, with a decrease. On the economic growth front, an increase or decrease in the Fed Funds rate, can be viewed as a signal that the Fed sees the economy growing too fast, with an increase, or slowing down too much, with a decrease…
…Viewed through this mix, you can see that there are two contrary reads of the Fed Funds rate cut of 50 basis points on Wednesdays. If you are an optimist, you could take the action to mean that the Fed is finally convinced that inflation has been vanquished, and that lower inflation is here to stay. If you are a pessimist, the fact that it was a fifty basis point decrease, rather than the expected twenty five basis points, can be construed as a sign that the Fed is seeing more worrying signs of an economic slowdown than have shown up in the public data on employment and growth…
…If you remove the Fed’s role in crisis, and focus on the effects of just its actions on the Fed Funds rate, the effect of the Fed on equity market becomes murkier…
…The S&P 500 did slightly better in quarters after the Fed Funds rate decreased than when the rate increased, but reserved its best performance for quarters after those where there was no change in the Fed Funds rate. At the risk of disagreeing with much of conventional wisdom, is it possible that the less activity there is on the part of the Fed, the better stocks do? I think so, and stock markets will be better served with fewer interviews and speeches from members of the FOMC and less political grandstanding (from senators, congresspeople and presidential candidates) on what the Federal Reserve should or should not do…
… The truth is that the Fed is acting in response to changes in markets rather than driving those actions, and it is thus more follower than leader. That said, there is the very real possibility that the Fed may start to believe its own hype, and that hubristic central bankers may decide that they set rates and drive stock markets, rather than the other way around…
…I believe that it is time for us to put the Fed delusion to rest. It has distracted us from talking about things that truly matter, which include growing government debt, inflation, growth and how globalization may be feeding into risk, and allowed us to believe that central bankers have the power to rescue us from whatever mistakes we may be making.
3. ‘There Are Real Issues in China Now,’ Ray Dalio Says (Transcript here) – Bloomberg Televsion
I think that there are real issues in China now, and they changed, really in the last four years, and that is that they need a restructuring. Individuals, 70% of their money was in real estate. Real estate has gone down. Stocks have gone down. Salaries have gone down. And so and as a result, they’re not spending and they’re concerned and they’re holding money in cash…
…At the same time, you have the government sector is a problem because most of the government spending – 83% of government spending – is spent by local governments. Those local governments got their money by selling land for real estate. Okay, there are no land sales and they borrowed a lot of money…
…It’s a situation that’s more challenging than Japan in 1990. It needs a restructuring in order to be able to do that. And then there’s also the question: Is the property ownership, is it respected? And Deng Xiaoping during his period said, “It’s glorious to be rich.” Is it still glorious to be rich?…
…Yes, there’s fantastic innovation in terms of technology, there’s nothing like it other than in the United States. Europe certainly isn’t a competitor in that. However, it’s very much government-directed. Can there still be entrepreneurship and that inventiveness? These are the big cosmic questions…
…I see investing in China as largely a very attractively-priced place that now has a lot of questions regarding the issues that I just referred to…
…There’s a small percentage of our portfolio which is in China, and we’ll stay in China, you know, through this process.
4. OpenAI’s New Model, How o1 Works, Scaling Inference – Ben Thompson
There are two important things happening: first, o1 is explicitly trained on how to solve problems, and second, o1 is designed to generate multiple problem-solving streams at inference time, choose the best one, and iterate through each step in the process when it realizes it made a mistake…
…There has been a lot of talk about the importance of scale in terms of LLM performance; for auto-regressive LLMs that has meant training scale. The more parameters you have, the larger the infrastructure you need, but the payoff is greater accuracy because the model is incorporating that much more information. That certainly still applies to o1, as the chart on the left indicates.
It’s the chart on the right that is the bigger deal: o1 gets more accurate the more time it spends on compute at inference time. This makes sense intuitively given what I laid out above: the more time spent on compute the more time o1 can spend spinning up multiple chains-of-thought, checking its answers, and iterating through different approaches and solutions.
It’s also a big departure from how we have thought about LLMs to date: one of the “benefits” of auto-regressive LLMs is that you’re only generating one answer in a serial manner. Yes, you can get that answer faster with beefier hardware, but that is another way of saying that the pay-off from more inference compute is getting the answer faster; the accuracy of the answer is a function of the underlying model, not the amount of compute brought to bear. Another way to think about it is that the more important question for inference is how much memory is available; the more memory there is, the larger the model, and therefore, the greater amount of accuracy.
In this o1 represents a new inference paradigm: yes, you need memory to load the model, but given the same model, answer quality does improve with more compute.
5. Learning from Richard Lawrence of Overlook – 14.3% CAGR for 30 years – Eugene Ng
The birth of Overlook’s Cap on Subscriptions originated when Richard Lawrence had lunch in 1992 in New York with Crosby Smith, a representative of the Dillon Family. Richard was asked why he would not just raise capital to generate fees like other investment managers. Crosby proposed that if Richard limited initial subscriptions into the fund at $30 million, the Dillon Family would invest $1 million. They shook hands, and Overlook had its first investor.
The Overlook Cap on Subscriptions was born in that spontaneous moment for Overlook. Richard Lawrence thought that the Cap on Subscriptions has proven to be the single most significant business decision in their 30-year history. In the early 1990s, Overlook decided to cap new subscriptions at 8-9% growth per year. This policy enabled the company to grow its AUM steadily…
…Effectively, the Cap on Subscriptions can smooth fund inflows, effectively lowering the cyclical volatility of AUM. One can then limit inflows of funds at the top of the market, have a ready queue of investors waiting to jump in during market declines, effectively making an investment fund more anti-fragile especially during market selloffs.
In addition, the Cap incentivizes investors to make a long-term commitment, which is aligned with a long-term investment horizon. Investors usually have to wait 6–12 months to gain access, so there are no short-term gains for investors trying to time the markets. The Cap effectively self-selects patient long-term investors.
What is the downside of having such a Cap? As you can imagine, such a Cap on Subscriptions is not for AUM/asset gatherers. The fund size grows much slower and takes longer to scale, and investment managers collect much lower fees…
…Time-weighted return (TWR): The TWR calculates the compound growth of a portfolio’s Net Asset Value on a per-share basis over a specified period of time. Fund managers most often disclose this number.
Capital-weighted return (CWR): CWR calculates the Internal Rate of Return (IRR) for an individual investor’s return and the return collectively earned by all investors in the fund. The CWR accounts for all cash flows into and out of the investor’s specific account and the fund since inception. Most fund managers do not report CWR, and CWRs typically underperform TWRs for most funds.
The Discount (the “Discount”) and the Premium: The discount is the difference between the TWR and the CWR for a specific fund. Discounts occur when CWRs are lower than and underperform TWRs. Peter Lynch was producing world-leading returns when he ran Magellan (high TWR), but the underlying investors performed far worse (low CWR).
Discounts typically happen for two reasons:
First, a fund manager can generate exceptional results as measured by TWRs at the fund’s inception when assets under management (AUM) are small. Then, the manager gets “discovered” and/or “promoted,” and an explosion of money enters the fund, to the great delight of the fund manager. However, with the larger asset base, the now-famous fund manager performs poorly, dragging down his TWR while crushing his CWR.
Second, CWRs are hurt when investments are poorly timed. Investors chase funds promoting hot themes, then bail out when markets turn down. This behavior inevitably decreases their CWRs. But even buying smartly and selling poorly, or buying poorly and selling smartly, can result in a Discount.
On average, the Discount increases when some of the following conditions prevail:
- Funds experience fast growth of AUM: the Discount tends to increase as the absolute value of a fund increases.
- Funds are invested in trendy asset classes.
- Funds are exposed to excessive valuation risk.
- Funds have excessive exposure to fund-of-funds’ investors.
- Funds are operated in higher volatility sectors…
…Richard realized that the elimination of Overlook’s Discount is overwhelmingly due to their legal Cap on Subscriptions. At first, he thought it was due to the success of their Investment Philosophy or the luck of their investors in timing their investments.
Overlook’s Investment Philosophy has helped them achieve outperformance of our TWR vs. the benchmark, but it did not impact CWR. The investors’ luck is not a factor either, as their investors have added funds consistently over time.
Instead, the answer lies exclusively in the legal Cap on Subscriptions because the Cap has allowed a limited amount of funds to enter Overlook steadily over the past 30 years. Control over the growth of AUM is the key to eliminating the Discount. The legal Cap on Subscriptions is the hero of Overlook’s story.
Disclaimer: None of the information or analysis presented is intended to form the basis for any offer or recommendation. We currently have no vested interest in any company mentioned. Holdings are subject to change at any time.
