Professor Shiller's numbers are persuasive; since anyone can remember, whenever the stock markets and/or the economy got into a train wreck, the bottom in the stock market that followed coincided when P/E ratios went down to about 7.
By that logic, plus the awful economic news (a little bit better is still awful), the S&P 500 is going down to 450, and the issue is not IF it's WHEN?
So here we are at S&P flirting with 900 and the doomsayers have been becoming increasingly agitated. First it was a "Dead-Cat-Bounce" then a "Sucker Rally" then a "Bear Rally" and now, well it's a "Conspiracy".
Meanwhile, some wag in the basement is playing that old Dr. Hook song "I Got Stoned And I Missed It".
Could 98% of SA contributors AND Professor Nouriel Roubini, be wrong?
Here's why they might be:
(1): It's earnings in the future that count:
What matters is earnings in the future, not the in past. The price of the market reflects market participants' opinions about what these will be - long term, not next week; sure earnings in the past are sometimes a guide to the future, and sometimes they are not. Likewise for earnings projections.
(2): Measurement Errors
There are different ways to work out earnings. A good example is the controversy over mark-to-market; depending on how you value your assets there can be huge differences (earnings are the change in assets net of liabilities over time, you generally know what your liabilities are; but if there are uncertainties over the value of the assets, well by definition there are uncertainties over earnings).
Case in point - the controversy over the Stress Tests. There was a difference of opinion about how much extra capital would be required, between what the Treasury team said and what the banks said of $50 billion or so. At a TCE of 4% that means a difference in opinion about the valuation of assets of $1.25 Trillion. So who was right? Well, we will find out soon enough.
What actually gets reported as earnings is a compromise between what the SEC, the regulator and the tax-man says, how close that is to reality can vary. But here is the rub, over time, if earnings are negative (regardless of what you report), you go bust, and that's not something you can argue about. So regardless of how they are measured, the idea is that in the end everything balances out.
That may sound pedantic, but look at it another way:
All the big banks reported stellar earnings up to the end of 2007. Well, either they lost all that money during Q1 and Q2 2008, or they hadn't noticed they had lost it (certainly their auditors who all signed them off a "going concerns" at the end of 2007 hadn't noticed).
Say you lend $100 to a deadbeat who has no intention of paying you back, three years later the penny finally drops. So when did you lose the $100, (a) when you handed it over or (b) when you figured out you weren't going to get it back?
Say the tax-man decides you can write off a new piece of equipment over ten years, but, in fact, it can run perfectly well and productively for twenty years (I've seen twenty-five year old bottling machines running just fine, they put stuff in bottles, it's not complicated they are just a rackety combination of pumps, mixers, conveyors and limit switches - why buy a new one when the old one ain't broke?).
But as far as the tax-man is concerned, your earnings were less in the first ten years (when you depreciated the machine) and much more in the next ten years (when the machine was basically free). But is that reality?
The value of all those potential earnings in the future is equal to the anticipated forward earnings, discounted by a yield to get today's Present Value where the yield is theoretically some sort of function of interest rates.
That's called income capitalization, which is one way you work out the "other than market value"1 of anything. (The other way is depreciated replacement cost which is related to book value, (but that's basically impossible to calculate for stocks mainly because you can't value brands or intellectual capital easily outside of mark-to market)).
So for the P/E "Rule of Thumb" to work it should account for yield, but it doesn't. That's like saying your plane flies so long as it stays on the ground; it's a denial of known reality.
What yield you use is of course anyone's guess, a good start is perhaps the 30 Year Treasury; although I guess the alternatives (short term rates) might also have an effect (mental note to look at that one day I can't think of anything better to do); plus how much more you want depends on the market, and that "risk premium" can vary depending on the business and perceptions of risk.
What just happened was that risk was absurdly mispriced; so perhaps the participants in the stock market also mispriced risk? Perhaps now they are overpricing risk?
If you are serious about doing valuation, you don't agonize very hard about "what should be", you just find out what the answer is.
The way you do that is that by definition when the price today is equal to the other-than-market-value, that means the market is not mispriced; so that's how you can check your input variables - long term each approach should on average give the same answer.
So what would happen if a yield was considered?
Well let's say (a) that the "correct" yield on a stock market earnings was long-term-interest-rates plus "X" where (b) "X" was a constant over the past 100 years and (c) the "normalized" earnings the way Professor Shiller works them out was a reasonable predictor of the future.
Working that out for the peaks and the troughs that gives the following chart for over pricing or under pricing of the S&P 500:
Well "rough numbers" (I just put in the peaks and the troughs), it looks as if perhaps either "X" went down over time or the way that earnings were reported went down, because the "best fit" regression of those points goes up over time.
How could that happen? Well it could be a "fundamental paradigm shift" which is what Allan Greenspan thought (I think he was wrong), it could be that governments in the USA have gotten a lot more aggressive about putting their hands in peoples' pockets (and smarter at it too), so the incentive to "hide" earnings increased (that's of course unless they are linked to your bonus scheme). Or it could perhaps be a progressive genetic imbalance in the Black Swan population caused by an "alien" recessive gene? Who knows, or more important, who cares?
Playing with that idea, how about re-adjusting the line so that on average over the past 100 years the market was correctly priced. To do that you just work out the degree of separation from the best-fit regression line:
So by that logic the market now is 55% under-priced, and historically it looks like it's about time for a bottom.
Comparing to market-long-wave analysis.
This is an approach that I have used to look at property prices for years; basically what I do is an "other-than-market" valuation of the market as a whole, then look at how the price deviates from the valuation. The "calibration" of that approach is simply that you "know" that over 100 years the market was in equilibrium on average, by definition.
Last January I tried that out on the stock market. I started off looking at earnings, but I rejected that for two reasons: (a) it didn't produce a result that fit with known reality (when you do valuations for a while you always check against reference points that are known reality, the more the better), and (b) I knew very well the problems of measuring earnings; I've spent half my life trying to persuade bankers and accountants that earnings are much bigger than what they think they are, and I don't think I'm the only one. It's like that tired old accountants joke "Profit Guv', what would you like it to be?"
In valuation you don't really care why things work, but you care very much about how clean the input data is, so I had a look at the cleanest data that exists, which is nominal GDP. That number is relatively easy to measure, and it has the advantage that (a) the way it's measured hasn't changed much over time and (b) it is routinely measured three ways (consumption, expenditure, capital formation), so it's got internal independent checks on the methodology.
It's the cleanest number there is (it doesn't mean it's always right, particularly in economies where tax is not paid - I'm always amazed that the basic "assumption" that all economists seem to make (or delusion depending how you look at it), is that people don't cheat).
So here is a notion, forget about what earnings are reported, how about if the reality is that over the long term the "real" earnings are a constant proportion of GDP, (after properly accounting for the $100 you gave to the deadbeat who you thought would pay you back but didn't, and the real profit from running a twenty-five year old bottling plant). In other words, the notion is that money churns around and on average people figure a way to transform a pretty constant proportion of that into earnings.
A notion...so how about dividing nominal GDP (a measure of real earnings) by LTIR (a measure of yield), and compare that to the Dow or the S&P 500?
Well, do that and there is a good correlation, much better than anything you can cook up using P/E ratios or P/E rations divided by LTIR.
There are two other things (a) the deviations from the "best-fit" coincide with known bubbles and busts, and (b) this is the really magic part, the amplitude of the under-pricing is almost uncannily equal to the amplitude of the preceding "wave" (after you adjust for the arithmetic that if you have a 50% increase from "X", you don't get back to "X" by a 50% decrease). That's called "efficient market hypothesis", regression to the mean or Farrell's Law #2, take your pick. The result of that analysis is shown here.
Plot that against the "adjusted" chart of P/E after accounting for a yield and you get:
So the two approaches (completely independent) give approximately the same answer. Which they should, given that they are both a methodology to compare an estimate of other-than-market-value compared to price using some sort of rough & ready income capitalization approach to valuation.
The Market-Long-Wave analysis says (a) stock markets are about 45% under priced and (b) they will not get any more under-priced.
Could it be that easy?
A criticism of this approach is I "back-adjusted". Well yes and no, dividing nominal GDP by LTIR and comparing that to the stock-market was just testing a notion, one string of clean data divided by another string of clean data is not monkey business.
The only "back adjustment" that was done was to say, that on average, regardless of how it happens, long-term in the past and likely in the future, markets will on average be correctly priced.
That may not be good economics, but from a valuation perspective (for example if you use International Valuation Standards), it's not just acceptable, it's mandatory. You are supposed to compare your valuation model with actual real-life market-derived data from the past, and only if you can show it works in the past, can you use it for the future.
Why that happens, I have no better opinion than anyone else. It's like you know that if you prod a bear with a stick, he will bite you. Why he does that, well pick a theory; that's not the point, you know with a high degree of certainty that if you prod you will be bitten. And that's all you need to know, that's what's called "the answer".
I've worked with a lot of economists, the joke we have is that they will tell you everything you could possibly imagine about a problem, except the answer.
The answer right now is that stock markets are not going any lower than the 9th March bottom, and if the economy picks up eventually, and interest rates don't go through the roof as they scratch around to raise all those trillions (and if they want to avoid that they have to STOP throwing money at the banks NOW), then there could be a sustained bull run as that 45% under-pricing is slowly eroded.
And then in about 2012 there will be a choice, the market will be correctly priced, and the choice will be...either (a) Do It To Me One More Time Baby, or (b) set up some sensible regulations so that idiot bankers don't borrow everyone's savings and go down and gamble them away in the casino using a formula that they get 20% if they win and you pay the whole lot if they lose. And then they get the lobbyists to persuade their cronies in the government to put their hands in everyone's pockets to pay off the debts.
In Saudi Arabia they cut off people's hands for doing stuff like that, now there's an idea, perhaps that might explain why they didn't have a credit crunch?
1Other-than-market-value is a concept used by International Valuation Standards; it is an estimate of what the market-value would be if the market was not in disequilibrium. It is not a concept used by eitherUS GAAP, FASB or IFRS (which is why if an accountant says your business is a going concern in USA or UK, take that as an "assumption", (and a wild one at that)).
Disclosure: Long Oil (effectively).