Here is the Q2 2018 letter, formally announcing the launch of Poor Ash’s Almanack – a website with ~half a million words of content that is the best free mental models resource on the internet by a wide margin. I’ve been working on it in one manner or another since the launch of ACM, and work accelerated through Q1 and Q2.
My Q1 2018 letter, “The Asterisks,” looks beyond the surface of ACP’s stellar results since inception to identify areas for improvement.
Usually I’m not interested enough in “macro” topics to view them as worthy of much commentary; moreover, in most cases, I’m hardly qualified. I also usually don’t get on a high horse… but I’ll break from tradition for once.
Here’s something I’m getting tired of seeing all over the place: if you are an investor, and you write things, and those things you write occasionally concern the valuation level of the overall market, please stop referencing the Shiller P/E (also known as the cyclically-adjusted price to earnings ratio, or, cutely, “CAPE.”)
A few reasons: first, it just doesn’t work. As Blooomberg View writer Charles Lieberman put it in October 2017:
The only time the CAPE suggested stocks have not been overvalued in the last 25 years was in 2009, when it implied that stocks were fairly valued. Stocks have tripled since then.
Lieberman goes on to make the obvious conclusion (that something that’s been wrong for 25 years can’t be taken seriously as a useful indicator), and also makes the important observation that the CAPE:
“is inherently backward-looking, notably very far back, instead of forward-looking.”
Unfortunately, he doesn’t drill this point home. It’s obvious but overlooked. Plenty of investors lead with the CAPE, then proceed to hedge with statements that go, directionally “but interest rates” or “but tax cut” or something to that effect.
While these arguments all have their merits, they ignore the single biggest flaw with the CAPE that is extraordinarily rarely pointed out for reasons I can’t comprehend: it’s based on an incredibly flawed, cynical, and unrealistic premise, i.e. the idea that the way the world looked five or ten years ago should be the major driver of current valuation.
This makes any occasional successes of the Shiller P/E’s predictive value artifacts of luck rather than good methodology (similar to the famous “Super Bowl Indicator.”) The CAPE essentially rests on the assumption that the world is cyclical; the problem is that outside of certain industries, secular factors matter far more than cyclical ones.
To start with, over the past 10 years, the U.S. population has probably grown (in round numbers) from ~300 million to ~325 million; meanwhile, China’s GDP has more than doubled from ~$4.5 trillion to over ~$11 trillion today. Technology, meanwhile, has made everyone’s lives more efficient and productive, massively increasing wealth in real (lived) terms. Obviously, factors like that means more business for everyone.
Perhaps more importantly, though, the way the world looked in 2008 (smartphones merely a novelty, Amazon not yet dominant, “the cloud” let alone “apps” not something that many people were thinking about) is completely different than the way the world looks today. When you think about major index components like Google and Apple and look at a historical revenue chart, their revenues today are up 6-8x over 2008; in the case of Apple, EBITDA’s up over 12x.
On a bottom-up basis, using a “cyclically-adjusted” P/E to value these companies would make absolutely no sense – if you think that their go-forward earnings should somehow be modeled by an average of the past 10 years, you’re basically saying (in as many words) that we’re going to go back to a pre-mobile world and stay there, which I don’t think any reasonable analyst would view as a base case forecast. Whether or not you’re an Alphabet bull (I have no horse in the race), is there anyone who credibly expects search volume to fall back to 2012-2013 levels?
Similar stories could be constructed about plenty of other companies – for example, let’s take Fogo de Chao (FOGO), a restaurant stock which I’ve owned twice now (including currently). 10 years ago, FOGO had ~10 restaurants in the U.S.; today, it operates 38 with another four or five in the works. FOGO obviously isn’t in the S&P 500, but the CAPE should work as well in theory for other market valuation indicators, and similar stories could apply to many bigger restaurant stocks as well – are you telling me that FOGO or CMG or PNRA’s contribution to some market valuation should be “cyclically-adjusted” back to an average of 20 restaurants or something?
Obviously, there are puts and takes; I’m sure there are plenty of index components whose businesses today are structurally worse (bricks and mortar retail, newspapers?) than they were 10 years ago. Unfortunately, the CAPE wouldn’t work well for them either on a bottom-up basis – I pity anyone who tries to value a generic mall-based retailer based on its metrics from 2007 or 2012.
Therefore, using the CAPE at a high level is basically hand-waving and hoping that progress in company A is fully offset by declines in company B… which isn’t generally the case: since the Industrial Revolution, at least from a purely economic point of view, the world has been in a steady up and to the right trend.
Being aware of potential cyclical impacts is certainly important (as energy and mining investors learned), but most industries aren’t that cyclical. Perhaps fewer people would click on Google ads or buy Apple smartphones in a recession, but that number, and consequently the earnings of the respective company, is still going to be up a ton from 2008.
From a broader economic perspective, the long-term trend is something like X + some small sin(x)-looking component. Here is a Wolfram Alpha chart to show what I mean:
As you can see, when you zoom out, cycles are the little sin(x) fluctuation on the broader trend, but progress marches on notwithstanding – and this is what CAPE fails to capture. It, contrarily, assumes the world looks like this:
Obviously, that isn’t true, and if you think it is, market valuation multiples are probably besides the point (because you’d never be able to invest in anything for the long-term).
One of the things I’ve tried to do over the years is become more open-minded and less intellectually prescriptive; i.e., just because I hold a certain view or do things a certain way, I don’t automatically assume that it’s the always-and-everywhere optimal approach for everyone.
Spending a lot of time thinking about macroeconomics never has been and never will be part of my process, other than the really obvious bits, but I’m willing to at least entertain the thought that perhaps there are others smarter and more ambitious than me who could have the kinds of usable, actionable insights that I don’t.
However, seeing the Shiller P/E referenced so widely by otherwise-thoughtful investors, as if it’s some sort of useful or interesting data point, is annoying because it’s not only empirically indefensible, but conceptually worthless.
I certainly have no disagreement with the general conception that valuations are elevated based on my bottom-up view that attractive stocks are hard to find at scale and that, the way I value companies, far more companies are overvalued than undervalued, and those that are overvalued are typically overvalued by a much higher degree than those that are undervalued.
So if you want to make that argument, I’m totally willing to get on board with it. Just don’t use the Shiller P/E to do it, because while it may sound smart, it has no basis in or bearing on objective reality…
… thus ends my rare high-horse, soapbox-style rant. We now return to normal, understated, less-headdesk programming.
Askeladden’s Q4 2017 investor letter is now available here.
In which I restrain myself from making pointless football metaphors for the pure joy of it.