© 2016-2019 Fowler Drew Limited

  • Black Facebook Icon
  • Black Twitter Icon
  • Black LinkedIn Icon

Authorised and regulated by the

Financial Conduct Authority 

Search
  • Stuart Fowler

Have we seen the ultimate low for the S&P?


In The Long View in the weekend FT, columnist John Authers said US stocks ‘last year were never nearly as undervalued as they had been in the previous great bear market lows’. He thinks this is causing investors to fear that the ultimate low has not yet been seen.

Is he right?

In this increasingly popular view, the actions by governments to pump money into the stalling financial system are seen as giving stocks a reprieve before they reached the sort of valuation levels that properly discounted the post-credit crunch realities facing economies and businesses. Hence the anxiety now as the patient looks to be coming off life support.

John Authers’ preferred measures of value, using replacement value of assets (Tobins q) and a trailing 10-year average of earnings, supports the story that US equity investors, like the banks, got lucky and were bailed out before reality had sunk in to share prices.

I don’t think we can agree with that. Our own measure of value suggests that the last move down in the S&P at the start of 2009 took the market to an extreme level with only a few historical precedents in any major equity markets, all equally important as long-term buying opportunities.

Whilst we attach a great deal of importance to our measure of value as a basis for forecasting long holding-period return probabilities that can be used in asset allocation, it still does not mean there is a negligeable chance that the S&P will fall through the earlier low. Though it would be unprecedented for the US market, it happened in the UK in 1974 and in Japan in the 90s and again in 2003. What can happen in one place can happen in another.

There are also several historical examples where extremely high predicted value has persisted for a decade or more, Japan being one but Continental Europe in the late 70s and early 80s is a less familiar one, being disguised at the time by high inflation.

Though the FT piece was about the market most likely to influence the general direction of global equities, it is worth noting more parochially that on our measure the FTSE All Share Index is the least undervalued of all the major markets and did indeed recover well before reaching a historically exceptional valuation.

Lateral thinking about equity valuation

We agree with John Authers that Tobin’s q and trailing 10-year earnings will dampen the effects of extreme cyclicality in current fundamental inputs to valuation, particularly reported earnings. As an accompanying chart in his column showed, reported earnings in real terms reached a low point last seen in the 1930s although clearly this is not what has happened to ‘earnings power’ for American business. The q ratio dampens variance simply because replacement value of assets is very stable, so the significant information comes from price change. Averaging earnings over several years will also dampen the valuation variance but is transparently arbitrary and may also leave price variance as the important information. These may therefore be better than conventional price/earnings ratios when earnings are disturbed but they are still not rigorous indicators of extreme valuation, high or low.

As market historians, Chris Drew and I have since the late 1990s adopted a lateral approach to equity market valuation, formalised in a return-generating model called Lambda. We ignore fundamental inputs, because of the measurement problems, in favour of tracking the path of achieved real total returns from each market and comparing it with its own long-term trend, using ‘best fit’ regressions. The model has not been invalidated by subsequent return behaviour and has indeed handled several episodes of extreme valuation very well.

On this measure, the recent low for the S&P was as low as the previous extremes of the early 30s and 40s as well as 1974 which was revisited (in real but not money terms) in 1982.

Fig 1 shows the S&P continuously-compounded logarithmic returns with gross dividends reinvested and deflated by the CPI from 1925, as an index. The whole-history regression trend (colour) is 6.8% pa but the hindsight-free trend using data up to each point in the series (colour) was above or below this as a function of the particular period. In my book I showed a Wilshire series for US equities for nearly two centuries that revealed a similar trend in both centuries, in spite of massive contextual differences. Data for other major equity markets also suggests about 5-7% pa as a trend, with much narrower dispersion of the differences than economic context and culture might suggest. Global evidence suggests regression trends are not just a statistical accident but an economically sustainable quantity that is a core element of the equity return process. We believe there are strong theoretical underpinnings for this, based on both economics and risk aversion.

This idea of sustainable real equity returns over long holding periods is really important. It conflates into a single measure two crucial influences on returns: i) fundamentals (like earnings and dividends) and ii) investor behaviour changes (like risk aversion, sentiment and momentum following). Historically, even over long holding periods, the second set of uncertainty sources has usually dominated the first, which makes the traditional focus on fundamental inputs look odd. If jointly the two sources of uncertainty lead to bounded deviations from trend, as a further core element of the return process, we do not need to disaggregate them to model either returns or risk.

Fig 2 expresses the level of the real total returns index as a ratio of its own trend, or detrended, which we call the Market Value Ratio. We use this as a basis for adjusting upwards or downwards the mean expected returns extrapolated naively from the trend alone. Low price denotes high value and high future returns and vice versa. By making some assumption (itself based on historical analysis) of the time-dependent reversion to trend, we can estimate real return probabilities for every holding period as often as we gather new historical return data. The actual frequency is monthly.

What can we do with value?

Fig 2 makes the point of this post that the S&P low in March last year was extreme. This observation was sufficient to justify exceptionally high expected future returns, provided long holding periods could be assumed. For investors with naturally long horizons, this amounts to a lot. It is why we added with confidence to clients’ total equity exposures up to the low point, while others were selling, and reduced them either side of this recent year end. See Is your manager doing a good job? (22nd January).

Observing extremely low value by March 2009 may even have been enough to warrant assigning a very low probability to further downside risk over short horizons, which would also amount to a lot if, it could be relied on. We do not believe it can. We prefer to treat short holding-period returns as a ‘random walk’. That was not the basis of our decision to buy.

An aside for more numerate readers: short-run returns, whether deflated or not, also have the highly inconvenient statistical characteristic of fat tails, or extreme outliers that would not be predicted if assuming a normal distribution. Long holding-period real returns, on the other hand, appear to be normally distributed.

We sympathise with John Authers who is obliged (as a good journalist rather than a good story teller) to remind us that even a ‘true’ valuation of the market tells us very little about the chance of the market falling or rising or staying the same over the next few years. This might read as ‘sitting on the fence’ but it is more insightful to see that good valuation measures simply have restricted uses.

For us, as financial planners and even as portfolio managers, using them to project long holding-period real outcomes is actually more important than using them to try to time markets. Imagine yourself in 20 or 30 years time looking back and deciding you got the outcomes you wanted and planned for. Is it most likely you will credit the many market timing decisions that made up the path your money took or rather making the right long term decisions? We have no doubt and we don’t think our clients do either.

#markettiming #meanreversion #realterms #sp #valuation

0 views