Book Review - Safe Haven: Investing for Financial Storms
You invest 100$ in an ETF today, and every year the ETF returns 10% on average, you are smart, so you know time in the market beats timing the market.
I said on average, let’s say that means the first year you lose 50% and the next to years you gain 40%:
Wait… did I just lose money? Yes, and the fund could have well been going down since forever yet had, on average, 10% average annual returns.
The arithmetic average was
1/10 = 1/3 * (-0.5 + 0.4 + 0.4)
The geometric average was
-1/50 ~= pow((1 + -0.5)*(1 + 0.4)*(1 + 0.4), 1/3) - 1
What’s the lesson? Don’t use average annual returns to judge an equity, use compounded growth rates over the period you’re investing in.
Wait, like, that’s all? Like, aren’t we already usually using that anyway?
i - What’s This Book Even Doing?
Since I’ve seen a bunch of people hyping up this book I feel like I need to start by bringing it down a bit.
I found this book to be rather boring, patronizing, and without an audience.
If it’s a book written for financial professionals, well, it sure does seem to me that it spends an awfully large amount of time going over 4th-grade probability theory. And yeah, I get that that’s important, but I really doubt 4th-grade probability theory is where all those people went wrong.
If this book is written for your average people that invest in various equities, then it does seem to miss three important points:
Nobody looks at those performance indicators anyway.
Over long periods of time, compounded returns and average returns usually average out to roughly the same number.
The math changes if you’re putting money in every month rather than playing with a fixed sum. Remember the first example, let’s recompute but assume we deposit an extra 100$ each year
“Average” case (1.1 returns each year)
The “volatile” case (0.5, 1.4, 1.4)
Wait… what, no!? But pirates! insurance? Niche’s demon! The sea of uncertainty!? Screams the book
Yeah, I’d pretty much ignore this book as an average save-for-retirement investor
Time in the market beats timing the market.
Money printers will go brrr.
Sync all money in VT, unless you’re bold, then calls SPX.
This book stinks of Talebian “everyone hasn’t figured it out because they are too dumb to grok my basic math” confidence that doesn’t quite cash into any revelations. It tries to tie in shitting on strawmen shaped like quants, some basic probability and cheesy quotes from pop philosophers into a narrative it take for granted as being among the biggest revelations of the century.
But, ahhh, Mark Spitznagel is an investor with a good track record, much better than mine, at any rate
Yes, I did like the book and found it rather inspiring. So with those disclaimers and complaints aside let me go on about what’s good in this book.
ii - The Big Picture
The first good thing about this book is that, if you’ve ever wondered what Taleb has to say about investing and didn’t want to go through 4 books worth of Taleb to find out, this book basically sums it all up in a 2 to 4 hours read.
The second good thing about this book is that it gives the kind of nuanced picture of individual finance I’d wish to have had as a kid.
You’re not living in a world where many possible paths average out, you’re living in a world where you take one of many branching paths, and even if there are a million alternatives, where they’d have taken you is unimportant, you are now on this path.
So don’t plan your future with the average in mind, because the 1% of best paths may raise the average enough to make you ignore the 99% of horrible paths, and suffering suck linearly, maybe exponentially, happiness on the other hand, tends to be logarithmic.
So think of the median path, or even better, think of a path that’s worst than 90% of all other paths. How does that seem? If the answer is not good, reconsider, ignore the 99th quantile and, above all, ignore the average.
More broadly, what this book seems to encourage the average not-hedge-fund-manager to do, is to essentially build a portfolio that will make one happy even in a fairly-bad-case scenario, which will incidentally allow us to be a lot more aggressive with the small percentage of our funds that we do venture with into the seas of speculation.
iii - Safe Havens
What this book misses, is that for your average Joe, most of their wealth already is in a safe haven. The issue here is that the author, understandably so, equates wealth with money.
But for most of us, most of our wealth is ourselves, and the ability to limit our personal spending, sell our services for money and make use of long-term high-value high-maintenance possessions like houses. As well as call for aid on a network of friends and family.
When the market falls by 50%, professional investors lose a lot of buying power, Joe-monthly-wage however, can now soak up a lot more free bargains. More importantly, most of his wealth, in the form of his house (which he doesn’t plan to sell anytime soon), cash in the bank, and a healthy body, is safe and sound. While not “increasing” due to the collapse, its buying power is proportionally increasing as capital is essentially devoured by the crash.
Barring very poor financial habits, most people work day jobs at mid and small-sized companies, i.e. most people, probably needn’t worry about financial safe havens much, if at all.
iv - The Paradigm Shift
Still, this book screams “I want to be part of a paradigm shift”, so let’s be charitable and call it part of the “probability without statistics” paradigm shift.
Statistics make sense in systems that are easily decoupled and are made up of similar parts.
But, to use the book’s own analogy, finance is no physics, companies aren’t particles. Finance is a cucu-clock, companies are its gears and frames and pulleys.
You can observe that physical objects behave probabilistically and use independent observations to figure out their regimes of operations, expressed as various probabilistic equations under a statistical umbrella.
You can’t do that do a cucu-clock. Sure, the components might behave probabilistically, maybe a gear’s irregularities follow a normal distribution, or a pulley's weak points in relation to force applied by that gear follow some sort of probabilistic differential equation.
But you can’t apply statistics to understand or predict the behavior of a couch clock. You can apply probabilities and say things like “If there’s a 10% chance that gear has irregularities beyond this point, then the mechanism will fall apart” or “if this pulley’s pattern of microfracture formation follows this or that distribution then this indicates a short-lived clock”. However, that sort of analysis is often impractical, and even if it weren’t, using it requires causal models.
v - Towards Better Use Of Data
To its great credit though, the book doesn’t conclude that probabilistic events are linked and then starts building quack theories and overiffting them.
Instead, it pulls out some rather advanced techniques (only if you are an econometrist) that should have been the foundation of understanding probabilistic systems (and are if you got your start in machine learning); Then it applies them to simplified real-world models and shows how they can be better understood.
By “pretty advance techniques” I mainly mean bootstrapping & cross-validation, but my feeling from reading economic papers is that, for the state of this field, this sort of approach might indeed constitute a paradigm shift.
Then again, one must ask, is the field stuck because of ignorance or is it stuck because using broken models more easily allows generating whatever conclusions the current political climate demands their eggheads back up their decisions with?
I guess I’m more in the “systematic issues are the desired mode of operation” camp than Mark. But alas, the way to fight fake experts serving power is to educate as many people as possible to spot basic flaws in their approaches. So as far as I’m concerned, this book might indeed provide a lot of value over time, by shifting the zeitgeist in a direction where people are more courageous about calling bullshit on economic models.
vi - Conclusion
I found this book a bit too popy, probing up simple and widely known mathematical concepts like it’s invented hot water.
I also found this book an enthralling read, with well-crafted stories that will forever be embedded in my brain.
It’s also served to reinforce my idea of splitting my investments 40/45/15:
40% Invested towards improving oneself, creating personal projects and enjoying life
45% Invested in “safe” vehicles like swiss government bonds and self-made indexes over stable companies and industries
15% Invested in wild shot startups, crazy ideas that will certainly flop and other zero-to-one vehicles
It may have played to my biases, hence why on the whole I’m feeling positive about it. Or I might have just used it to see the patterns I wanted to see and draw the conclusions I wanted to draw.
Either way, if you’ve got a few hours, I’d recommend giving it a read.
In case you didn’t already, I’d appreciate if you’d fill in the reader survey.