Mental Models + Relationships
I think I promised I’d periodically talk about mental models and other things more loosely related to investing. Here’s my first crack at trying to explain a really general one, and also the way I try to keep it simple and organized in my head so that I remember to use it.
Mental Models
First of all, it’s worth paying some respect to and honouring the recent death of Charlie Munger who once said “You need to build a latticework of mental models on which to hang the facts.”
Munger’s method of trying to understand and organize the world (which is used — but often in their own, different ways — by everyone from many scientists to senior business executives and more) was to abstract things. He also tried to generalize mental models so that they could be applied in many different situations, meaning those models are more valuable than something specific. Munger’s view was that ultimately, there are only a small number of models that, in his words “carry 80% of the freight” in life.
Munger would learn from different fields and disciplines spanning law, engineering, politics, war, medicine, physics, architecture, etc., and would identify the “big ideas” from each. For example, the concept of the “factor of safety” in engineering is identical to investing’s “margin of safety” originally taught by Warren Buffett’s first mentor, Benjamin Graham. It’s that, in theory and in pure mathematics, you can work in precision and have exact solutions; yet in reality, there’s always noise, unforeseen outcomes, etc. Sources of variance. Therefore, you need to add a buffer to the analyses you do, the plans you make, and the actions you take. You’ve got to make room for error, and make sure you’re robust against really bad outcomes. If you try to get really specific in your views and planning, you’re going to get bowled over a lot.
I’m a lot farther behind than where Munger was in this journey, but I took a similar path because I saw it had practical value for me as I worked on it. After a decade or so of trying to learn and organize things, I’ve found there are now many concepts I understand more deeply than a lot of other people do (he said humbly), because I took the time to abstract and synthesize the more important ideas behind what’s going on.
I probably organize the ideas differently, though. I don’t know how Munger or anyone else’s mind really works internally, of course.
So here’s one idea I noticed came up over and over again everywhere.
Networks
This idea, this model, I don’t have a name for. For some things, my mnemonic to remember them is a picture. I mean, it looks like a network and it is. You’d be tempted to call it that, and you can. But that doesn’t encompass every place you can apply it.
What’s it represent?
Interconnected stuff. First, there’s individual things, which are the nodes. Second, there’s the relationships between things, which are the links between the nodes.
Before even using this, there’s one important thing.
A lot of people don’t put enough weight (especially Westerners because we’re individuality-oriented) on the fact that the links are an entirely different thing from the nodes. The relationships between things are a thing in and of themselves. They are not the nodes. They need the nodes to first exist, yes, but they can then act like a thing on their own, with completely different properties and behavior.
That’s important. It leads to “emergent” behavior. An example of emergence is how a colony of termites can make a mound without any individual termite understanding how to make a mound or even understanding the totality of what’s going on, even while the mound is being made. You can’t ask a termite how to make a mound, but you can watch a colony of them do it. It’s also like gravity, the relationship between massive bodies, forming a solar system or a galaxy, which is a thing on its own with its own properties separate from the stars and planets in it. It’s like a traffic jam, which emerges because of relationships between drivers, vehicles, and each other, the delays in each person’s reaction time, etc… Finally, emergence is how a stock market can price stocks reasonably correctly most of the time even though each individual market participant might not have an accurate view of the stock’s future cash flows.
(The stock market efficiency one is interesting — the easiest way to explain is that the individual errors cancel out as people take their own actions. A jellybean jar charity auction is a good way of seeing it, where people donate and enter a ballot with a guess of how many jellybeans are in the jar. Typically, the average of all the guesses is more accurate than any individual guess, because the participants don’t talk much to each other and they make their own independent guess, and therefore, if you average out the guesses, what is happening is that their positive and negative biases cancel out with each other, and you’re left with an average that’s very rational. Don’t even bother guessing. Just survey everyone.)
OK, how can you use the image?
Well, you could use it to organize and understand how a business works, for example. Every business is a series of activities, and then the activities relate to each other. One part of the business does a thing then passes it off to another part of the business and so on, eventually to the customer. And then there are ancillary activities other than manufacturing a thing, like advertising to get customers, or finance to count up what is being done and what the results are.
Zooming out of a business, the picture also applies to an industry value chain, like the (sort of old) way that computers were made and sold. There would be a bunch of different suppliers of different parts, like a graphics card (nVidia, ATI, etc) and a motherboard (ASUS, Gigabyte, etc) and a CPU (Intel, AMD, etc.), a hard drive (Western Digital, Seagate, Toshiba, etc.), some memory chips (Sandisk, Micron, SK Hynix, etc.) and then all those bits go to an assembler like Dell, IBM, Lenovo, etc, who puts it all together. They also do the marketing activities and sell through to the end-users either directly or via retailers like Best Buy (or some commercial distributor like a systems integrator like Accenture, Cognizant, or CGI), since those guys also own some of the consumer and commercial relationship. The assembler like Dell is also connected to the base software providers like Microsoft and Adobe and who originally paid Microsoft (and passed that through to the customer via the price of the finished device, along with the cost of all the other hardware they purchased from all those other suppliers).
And so you can put that whole drawing together really simply, and then you can see who is doing what. You can see both the flows of goods and products, and also the flows of money.
In those kinds of situations, you can also take the model one step deeper and ask some important questions if you’re analyzing any of those businesses:
Who pays who? Based on what? Who takes what risks?
What are each party’s incentives? Who aligns with who, or not? How would it change?
How do competing nodes like Dell and IBM compete with each other?
You could draw a picture of the automotive industry value chain to get a similar thing. Where you’ve got the parts suppliers, the assemblers/brands, like GM and Porsche, the dealers, the end-customers, the aftermarket and repair industry, etc.
Drawing out something like that is a really powerful way to abstract and organize it all.
Another way you could slice this idea is you draw out the pieces of a business, or the pieces of your investment and how they work, and look at what pieces rely on each other, and what’s more/less important for the whole thing to succeed.
Maybe you notice a bunch of activities or future things have to happen or parts of the investment have to work in series for the whole thing to work. That could be a point of issue, for example. If a business needs to do X and have it work well, then do Y and have it work well, then do Z and have it work well, and so on, then the odds of all those things need to be really high for the whole thing to work. Why? Well, they come one after another and you can’t do the next one without the one before (they are “in series”, as people in the electrical/computer world would say). Mathematically, the probabilities of each are multiplied together. If there’s a 90% chance X works but only a 50% chance each of Y and Z work, then the overall thing has just a 22.5% chance of working (90% * 50% * 50%). Not very good. If any node fails, that whole thing fails. You need to be really careful with investments or other things like that, because realistically you don’t accurately know the odds of each thing succeeding.
In complex systems with lots of nodes that dependent on — linked to — each other, this can get you something called “cascading failure”. You could look at the financial crisis through this mental model, and see that it was a cascading failure of interconnected actors in a complex system. You can see it was partly because of the incentives that each guy had in the system — the incentives at each node — just like we said you could do above for any industry value chain.
So in the financial crisis, you had, in a simplified way:
A person who wants a loan because they want to own a home. The “American Dream.”
A novel new product called the mortgage-backed security created a couple decades earlier by an investment bank. Those banks’ wanted to sell lots of those things because they got a fee to do so, but did not bear a cost if the loan defaulted. So they were OK with helping a lot of people get loans to own homes even if it wasn’t necessarily appropriate for that borrower to receive a loan, or appropriate for the investor they were selling the loans to (since they didn’t have any fiduciary duty to those investors).
Investors like pension funds who were not very wise to what risks they were really taking — because the product was new and because housing bubbles don’t happen often, so there’s not much history — bought into the idea that if you package one mortgage loan with a bunch of other mortgage loans, then your risk is “diversified” across many individual borrowers, only some of whom would default. They neglected to think about scenarios where many of the borrowers might end up in distress because of a circumstance that affected many of them all at the same time.
That whole thing put a cycle in motion as people bought in, home prices went steadily up, and everyone looked like a genius.
Once the borrowers become stressed as a group, not only do they stop continuing to borrow, but the ones who already borrowed start to default. Some thing starts the cascade, then the dominoes topple: first the borrowers default and stop borrowing, then the banks doing the securitizing business see less volume and their fees and such fall, and then the investors who bought the loan packages see losses, etc etc. The chain could keep going, too. You could consider the next domino or node as the slowing economy, or whatever.
Cascading failure is a common risk in systems that are interconnected with each other. It’s really common in finance, in sole-sourced supply chains, IT systems, and elsewhere. It happens when things rely on each other in order and there’s no redundancy (back-up options) in the system.
So I keep in my head that picture of the interconnected pieces to reinforce the idea that these things connect to each other and clearly influence each other. It’s more important what the chain looks like than what any one piece of the chain or network is.
Or maybe it’s the case that the linkages actually reinforce each other, like a company that is trying to get you into a multi-product ecosystem so that it’s hard for you to get out. A bank wants to sell you a loan. They want to sell you your checking/savings accounts (or treasury and cash management for a commercial client). They want to sell you foreign exchange services. Investment brokerage and advice. Merger advice. Etc etc. And if you’re doing seven things with them, then you’re linked up and integrated with that bank in many ways, and that stuff’s hard to break now. That network’s like a web you’ve found yourself caught in. It lets the bank price some of the products at below-market rates, like deposit rates being a lot lower than money market short-term interest rates.
Or, maybe the thing has linkages in parallel, or nodes that don’t really depend on each other, or nodes that can get a similar thing they need from more than one other node. This would be a system where there’s no single point of failure, which can create a robust system overall. That’s like a Brookfield or a Berkshire Hathaway. Both holding companies will do things like borrow money, but a lot of the money is borrowed at the operating company level. When Berkshire’s BNSF railroad borrows money, the debt is is held by BNSF and is not recourse to Berkshire: if the railroad ends up in default, the bondholders cannot go up to Berkshire and try to be made whole that way. They’re only entitled to BNSF and its assets. For this and other reasons, this network is durable since there are not many nodes (activities in the business and parts of the business) that can take down the whole thing. There’s redundancy and layers of defense. If BNSF disappeared tomorrow, you would still be left with all the other stuff at Berkshire, which is collectively most of Berkshire’s value. At Brookfield Corp, its Brookfield Energy Partners (BEP) business could disappear off the face of the earth tomorrow, and you would still have most of your investment’s value remaining (although in this case the asset management business would also be worth a little less because it is receiving management fees from BEP, so there is some linkage and dependency). In those two cases, the only activity — the only node — that impairs the whole network might be capital allocation, since there is only one or a few guys who are deciding what to do with the entire holding company’s profits from all the different businesses. They could easily destroy a lot of value by doing dumb things and misallocating the profits over time (like doing 1-3 really large, stupid acquisitions).
Ecologies and the Lindy effect
In the original picture is another idea, the idea of interconnected systems where the nodes have co-evolved in response to what the other nodes have been doing over time and how their relationships have changed — literally like how evolution works where species compete and the odds favor the ones that are best adapted. There are systems like that which have existed for a very, very long period of time. When you see one, you should approach it with humility rather than saying “aha, I know exactly how you could make it better”.
No, you don’t know.
Why?
That system has already been optimized as the nodes choose what to do iteratively in response to what' the others are doing. They have so many relationships, and they have all changed in ways to optimize across multiple relationships, not just the one thing you are looking at or are focused on.
In the book Seeing Like a State, Prussia and Saxony (now mostly part of Germany), a newer, modern system of forestry is created. One species of tree is chosen because it makes the “best” building material. Ecologically diverse, old growth forests are cut down and one species is planted in neat rows. What happens?
Many of these newly planted trees die in droves.
Why?
Because an ecology and habitat is a complex system with lots of nodes and relationships. Different animal species nest in certain trees more than others. The animals interact with other animals. Different animals, interacting with other animals, fertilize the soil with their dung or by eating and processing dung, dead animals, and dead plants. Mushrooms also do this work. Today we even know that trees and mushrooms and such “talk” to each other by the chemical composition in the soil and change a little bit of how they’re growing and such as a result.
You can’t just take that system apart, select one thing you think is “the best” for you because you need to build houses, and then expect there not to be consequences. The thing you chose is a part of a system and has adapted to be in the system, not to exist as a neatly-lined forest of only one kind of tree.
You took the thing out of its context and its network.
This is why when people try to invent something or change something or make a new thing, it frequently results in failure. Failure happens way more often than success because it turns out that the system was already pretty optimal, like an old growth German forest. This is also why the majority of genetic mutations do not succeed in nature — the existing animal species, as is, was already nearly optimal for its environment. So the odds of some “new” adaptation being even closer to optimal are low. The odds of it being worse are much higher, since there are just more ways for it to be suboptimal. Most mutations result in death without passing on the mutated genes.
This is also related to the Lindy effect.
That’s where the longer something has existed in the past, the longer it is likely to exist in the future (as long as it doesn’t already have a lifespan, like most animals do; these don’t count, although the species does).
That’s because that thing has already gone through a very large number of catastrophes or changes in its environment, and it survived. Hence, it’s durable. It passed the test of survival many times already. It is a successful part of the network or the system in which it exists.
This is the banking industry. Or the restaurant industry. Or the construction industry. Or the agriculture industry. Or the prostitution industry. These are really, really old things. You are crazy if you think you could rip them out of the world successfully. Your error and your hubris is that you’ve forgotten they’ve already survived a great number of attacks and changes in the system already — which implies they’ve got a lot of things going for them that make them so durable and valuable.
When you’re dealing with interconnected things or with Lindy effect stuff, you need to first approach it with humility and consider why the thing works the way it does in the first place, and the relationships and such that it has in the network. If you don’t do that, you’re likely to miss something. And that something might be the reason you can’t change the thing the way you think you can.
Chris