With
equity markets surging from March lows, an increasing number of investors –
usually the one that are not yet particularly exposed to equities – have been questioning
current and forward-looking equity valuations. SPX trading at 20 (Price to earnings
ratio of 20), cheap or not?
It is an interesting aspect of
investors’ approach and psychology. In an effort to try to guess future equity returns,
investors use broad valuation metrics as a rational application of a sort of
normative investment guide. Is the market or this stock cheap, fair or expensive?
And this
is where things get complicated – or should I say – subjective. Indeed, cheap?
well maybe yes, but relative to what? versus historical average P/E ratio for instance?
but then, what historical average? 10 years? 20 years? and what about if the P/E
ratio is hovering around the historical average?
To add to
complexity historical valuation averages are not stable. A simple illustration
of this fact is historical standard deviation of the P/E ratio of the
US equity market. Standard deviation is a statistical measure of dispersion of a dataset. From 1950 to 2020, the P/E ratio exhibited a standard
deviation of 4.9 and an average of 16.7. i.e almost 30% variation around the
mean (!).
The last 60 years 10yrs z-score (number of standard deviations from the mean)
illustrates well the point: it often stays for long period of time at 2-3 SD
from the mean (!) see below chart.
Therefore,
it seems hazardous to deduct an historical average a sort of “reference valuation” of
what is fair value. Any average selected would therefore be context dépendent - macro context dependent.
But let’s say we agree on some sort of average that would represent an historical equilibrium and reckon P/E is below average so stock market might be cheap indeed. Are we sure this is a good guide of future returns?
Below the
trailing P/E ratio of the SP500 against SP500 5yrs forward returns (1952-2010). The orange
line is the 5 years forward total returns of SP500 and the blue line the trailing
12 months EPS of the SP500. At first glance, quite tricky to visually find a
stable relationship…
A simple
regression analysis confirms the weak relationship. With a r2 of 0.18 (r2 = conceptually
the higher the better as it means that the explanatory relationship between two
variables is high) the relation is very weak.
Same analysis run with 10yrs forward returns is bit better but still very weak with a r2 of 0.517. And if we strip out extreme the r2 goes down to 0.27...
We can safely conclude that trailing P/E have generally no predictions power whatsoever
below years. Imagine the same analysis on a 3 or 12 months horizon...
Even if
we use a more normalized ratio such as Shiller PE (the idea of this famous
measure is to average the last 10 years earnings so that we reduce the impact
of the business cycle on earnings – further adjusted for inflation) and SP500
10yrs forward returns.. The relation is still relatively weak with an r2= 0.65
Likewise,
if we strip out “extremes” the r2 goes down to 0.4..
The above
analysis leads us to safely conclude that valuations measure such as P/E have
no real prediction characteristics on future returns – baring some kind of mean
reverting properties at extremes.
The
reality is that valuation is a perception of value and ultimately the result of
the forces of supply and demand. There cannot be equilibrium in valuation as they
are the result of supply and demand dynamics. Therefore, valuations should be considered
most of the time a lagging indicators of supply/demand changes. Valuation I
guess provide only with a sense of comfort to investors…
Investors’
perception of value evolves over time. It is influenced by numerous context
dependent behavioral factors – the illusion of valuation equilibrium is essentially
a function of investors anchoring bias. As investors mood is bullish, a P/E of
20 or 25 might not be expensive. As investors mood is bearish the opposite
works. In 2007 investors justified a P/E of 20 as cheap – in early 2009, in the
depth of the bear market, investors estimated stocks trading at a P/E of 10 as
expensive.
Investor’s
anchoring bias is notably magnified and cemented by price trends (the higher
your investment goes, the better you will justify a high P/E for it). Forward
valuations are also a reflection of investors’ optimism or pessimism.
This does
not mean that valuations cannot be at-time a valuable tool for listed
instruments (non-listed instruments are another debate). As we saw earlier with
the lower r2 if we strip out extremes, at historical stretched levels (2-3 SD) valuations
seem to trigger some kind of mean reverting effect. At extremes, the audience
is usually larger: the message conveyed to investors has more echo and common
perception is easily shared.
Valuations
or fundamental analysis on single stock should benefit from higher predictive
power for long term investments than at an index level. Leaving aside all the index construction-induced
valuations bias, a long term stock picker that knows well a specific sector, company PnL
drivers and focus its analysis on free cash flow generation for instance, can have
a relatively well-educated guess about valuation (i.e guess the future earnings stream).
As some
of you might know I tend to focus most of my efforts on factual data and in
particular on assessing investors flows and positioning. My objective has
always been to assess the divergence between the rate of change of contextually
relevant macro data versus investors’
positioning. The larger the spread, the larger the asymmetry of the
opportunity, the larger the return will be.
I
therefore present below an interesting and probably less known way to try to
assess future equity returns based on estimation of the average investors’ allocation
to US equities. The fit is quite impressive. r2 is 0.87.
According
to this approach, the next 10yrs annualized return of US equities should be
around 2%..
Investors
flow/positioning are factual data, value is subjective. In my humble opinion, as
markets are increasingly mechanized (passive investment, smart beta flows, rules-based
strategies etc..) valuations will probably be even less predictive of future
returns as price discovery is further blurred by passive investments growth. The
r2 we talked about at the beginning of this article will probably be much lower
in the next decade. Fundamental investors…buckle up!
