The Sources of Value

It is one of the established 'facts' of finance that value outperforms growth over reasonable time horizons (see Chapters 22-34 of Behavioural Investing). However, which of the component sources of return generate this performance?

In a new paper Fama and French explore the composition of returns for value and growth stocks. They start by decomposing returns into dividends and capital gains. The chart below shows the results they uncovered using US data since 1963 with value and growth defined by price to book/size quintile intersections. It is worth noting that Fama and French perform their analysis in nominal terms (so the capital gains they show include the effects of inflation - around 4.5% over the their sample period). The chart shows that importance of dividend returns to value investors. The dividend yield on big value stocks is 50% higher than that of the market, and twice that seen by growth investors.

Fama and French then go onto to decompose the capital gain term into several sub-components. They show that the capital gain term can be broken down into an element due to growth in book value (effectively the investment carried out by the firm), and the change in the valuation ratio (price to book in this case).

They also observe that that the change in valuation can be decomposed into an element they call drift associated with the general upward trend in valuations over the sample, and an component called convergence which is due to a rise in profitability and a reversion to the mean in valuation. Drift is measured by comparing the price to book of the original portfolio with its new counterpart when the data are resorted each year. Convergence is measured as the price to book on the original portfolio at the formation date and the price to book on the same portfolio one year later.

The results of this further decomposition are shown in the chart below again for the period 1963-2006. The picture reveals show huge differences between value and growth stocks. Value stocks see hardly any growth in book value - not hugely surprising, they don't tend to invest large sums, in general they are more interested in cost cutting than investment. However, their is a very strong tendancy for convergence in price to book terms - that is to say their valuation rebound - although the decomposition is silent on whether this is the result of a bounce back in profitability or not.



The same can not be said of growth stocks. They see an enormous amount of growth in book value - as they engage in large cap ex and M&A. However, they convergence is negative, they witness declines in price to book as their profitability erodes and valuations return to 'normal' levels.

From a behavioural standpoint this is exactly what we would expect to see, if investors over pay for growth. Of course, Fama and French prefer a rational explanation which I find far from convincing, but the bottom line is that the return decomposition can't help us distinguish between the rational and behavioural explanations - we have observational equivalence.

In Chapter 43 of Behavioural investing I show a decomposition of returns for the US market. I argue that returns can be decomposed into the dividend yield, growth in real dividends, the change in valuation and inflation. I usually do my analysis in real terms so I can dispense with the last term. I have recently completed a similar exercise in terms of value and growth stocks. The results can be seen below.


The returns are lower than the Fama and French numbers because I remove the effects of inflation. The importance of the dividend yield is once again revealed, it contributes 53% of the real return to value stocks, real dividend growth accounts for a further 30% of the return to value stocks. It is noticeable than the real dividend growth of value stocks is faster than the equivalent rate for 'growth' stocks. So by buying value you get both value and growth.

Investors of all kinds ignore dividends at their peril.

Sector rotation: an investment dead end?

Two chapters in Behavioural Investing suggest that investors focusing on sectors rather than stocks are barking up the wrong tree. Chapter 32 outlines the evidence showing that value and momentum effects are much greater at the stock (and even country level) than they are at the sector level. It also cautions that sectors are rarely stable entities in terms of their investment characteristics. Pretty much every sector has been 'vale' and 'growth' or it's lifespan. So ruling out sectors because they are growth or value is a big mistake.

Chapter 19 also touches on sectors. This presents some of the work of Cremers and Petajisto who show that those fund managers with low active share, but high tracking error (those taking sector bets) manage to destroy value for clients (having a negative gross and net alpha). Such managers account for around 35% of the US market! Whilst this isn't proof that sector rotation strategies are hopeless (that would be to confuse the absence of evidence with evidence of the absence), it does at least make one stop and think about the role of sectors.


A new paper sheds further light on the fruitlessness of trying to rotate sectors. The paper provides evidence of the absence! Stangl, Jacobsen and Visaltanachoti explore the possibility of timing sector rotation across the stages of the business cycle. They identify five stages of the cycle shown in the diagram below.


They follow a rotation strategy that seems to me to capture the conventional wisedom regarding sector rotation, as set out below.

They investigate the US market from 1948 to 2006. In a heroic leap of faith, they assume perfect foresight on the part of investors. That is to say, they assume that investors know with absolute certainty which phase of the business cycle they are in.

Even assuming such prescient powers, the sector rotation strategy only outperforms by around 2% p.a. If one were to include transaction costs, and drop the perfect foresight assumption then this would quickly become a zero, or even negative, alpha.

When one examines the detail of the sector rotation strategy, some further issues are created. For instance, those sectors favoured by the conventional wisdom in early and middle stages of expansion actually have negative alpha over those phases!

Those sectors favoured in the late expansion did outperform, but were beaten by sectors whose attractions are usually assocaited with late stage contraction! In fact, it was only really in the late stage contraction where the conventional wisdom over sector selection was the best strategy.

Stangl et al show that even if you can forecast the business cycle with complete accuracy (see my earlier post on why we don't need economists for my thoughts on this) then following the conventional wisdom with regard to sector selection is an suboptimal investment decision. You would be better off following a simple market timing model which stayed long equities apart from during the early recession period.

So sector rotation (at least as represented by the conventional wisdom viewpoint) is not a good source of outperformance. This certainly calls into question the raison d'ete of many strategists!
In fact, all of the evidence mentioned in this post raises challenges to the way in which investment is done. Not only is sector rotation highly dubious, the fact that useful investment characteristics such as value and momentum are better defined at the stock level rather than the industry level brings the role of sector specialists into doubt. I have long argued that what we need is a few analysts with good investment skills, rather an armies of industry 'experts'.

The book is out

I'm delighted to say that Behavioural Investing was published on Friday. It is available from amazon at a £20 discount to the RRP. It is a slighter longer than I had envisioned, weighing in at 705 pages! However, the good news that it is designed to be dipped into, and not necessarily read sequentially.

I'm afraid those in the US will have to wait a little longer. The books are printed here in the UK, and sent over to the US. I believe it should be available in the US at the end of Nov.

Let me know what you think of it.

The myth of exogenous risk and the recent quant problems

Regular readers of my work will know that I am deeply skeptical over the idea of exogenous risk (like a gambler playing roulette, where the behaviour of other players is irrelevant). In Chapter 36 of Behavioural Investing I argue that many aspect of risk are endogenous ( like a gambler playing poker, where the actions of the other plays are integral to the game) to the way in which we invest. The problems experienced by the quant funds in August may help highlight some of these issues.

Now as my post on the 17 August detailed I suspect that leverage had much to do with the problems experienced by the quant funds. As one observation reader commented the hubristic use of excessive leverage is an all too human failing, and has nothing do to with the 'quant' process as such.

Andrew Lo (of MIT) and Amir Khandani have written a fascinating paper on the problems of August (available here ). They use indirect evidence to establish the scale of the issue and the kind of problems to be encountered. They use a very simple contrarian strategy of simply buying yesterdays losers, and selling yesterday's winners on a daily basis. They ignore all the issues surrounding transaction costs and turnover, as this is only an example.

They document several interesting features using this simple strategy. The first is alpha cannibalisation (or alpha decay as they call it). As more and more funds have set up in the quant arena so the return to this strategy has declined from an average daily return of 1.38% in 1995, to a mere 0.13% in 2007.

They show that in order to increase the return back to the level seen in 1998, leverage of around 8-times would have been required. As noted in my original post Goldmans Global Equity Opportunity fund was running 6 times! They also show that this strategy performed exceptionally badly in August. It witnessed several consecutive days in August (nearly 7% over 3 days, a 12 standard deviation event).

In a comparsion between 1998 and 2007, Lo and Khandani show that the big difference between 1998 and 2007 is the lack of spillover in the LTCM crisis. Whilst markets had a turblent time as LTCM problems became widely known in the wake of the Asian and Russian crises, simple quant strategies continued to work.

This hints at one possible (perhaps even probable) cause of August's events. A multi-strategy funds (or prop desk) took a big hit in the mortgage/CDO space, and as a result was forced to scale back on operations across the board, terminating their positions in equity space.

This in turn probably caused other quant funds to hit their limits, and unleashed a vicious spiral of selling, and further selling. The more crowded a trade is, the more likely this outcome becomes.

A survey for the the CFA publication Trends in Quantitative Finance (April 2006) showed that out of 21 firms using quant, 18 used them for return forecasting (with around $2 trillion invested in equities). The most popular factors uncovered were 'reversal' with 86% of those questioned citing it's use. In second place came 'momentum' with 81%, and in third was exogenous factors (such as valuations, earnings quality, capex ect) with 62% of respondents using such methods.

The recent so called 'quant problems' are a timely reminder of the endogenous nature of risk in our markets.

Are finance professors overconfident?

I wasn't going to post this week, as I have a couple of other things on. However, this paper arrived in my inbox and I couldn't resist the urge to flag it up to you. Professors Doran, Peterson and Wright survey the academic finance communities attitudes towards market efficiency and investing.

They manage to get around 700 usable responses (a high response rate for such a survey). Doran et al first asked the professors for their beliefs on market efficiency. In academic finance there are 3 levels of efficiency (weak, semi-strong and strong form efficiency). Each corresponds to the amount of information that is incorporated within prices. Weak form efficiency says you can't beat the market using just past prices, semi-strong form says that both past prices and public information are already reflected in the prices. Strong form efficiency says that all information both public and private is reflected in the current price.

The chart below shows the percentage of respondents agreeing (blue bars) or disagreeing (cream bars) with the statements that future returns can be forecast from (i) past returns (weak form), (ii) past returns and public information (semi-strong form) and (iii) past returns and public and private information (strong form).

Nearly 60% of the professors accept that the US market is weak form efficient (i.e. ignoring the enormous literature on momentum as a stock selection tool, incidentally something I am guilty of as well).

One third of the them accept that the US market is semi-strong efficient (i.e. that value strategies don't work since they involve publically available information such as earnings or book value).

Only when it comes to strong form efficiency do we finally see the academics reject the concept on efficiency. Some 57% said it was possible to predict future returns when using private information.



Two other questions of interest were raised in the survey. The first concerned whether the professors agreed that returns were a compensation for risk (i.e. the multi-factor world of Fama and French - a rejection of behavioural finance). Only a mere 27% of the professors disagreed with this statement. So we behaviouralists are definitely in a distinct minority in academic circles.

The other question concerned whether the professors thought that investment strategies that could consistently beat the market without taking above market risk actually existed. Only 17% of the respondents said yes such strategies existed. So academia seems to remain a bastion of market efficiency and passive investing. When asked, just 18% of the professors said their objective was to outperform the index.

There was one final finding from the survey which made me laugh (I know I'm sad). Doran et al found that finance professors decisions to actively invest were largely independent of their beliefs in market efficiency. Instead they reflected professors confidence in their abilities! As Doran et al note "A professor’s opinion on the general efficiency of US stock markets has little influence on his investment objectives relative to his confidence in his own abilities. Within confidence groupings, there is little dispersion in a professor’s investment goals as his opinion of market efficiency changes. However, within the opinion groupings , there is a substantial amount of near monotonic dispersion in a professor’s investment goals as his confidence in investing abilities changes. This suggests that a respondent’s opinion of market efficiency has little to do with his decision of whether to actively or passively invest. What matters is an investor’s confidence in his own abilities."

Sounds like finance profs are just like the rest of us - human after all!

Yet more evidence on the folly of forecasting, or why we don't need economists!

First a quick comment on the change of colour scheme. Multiple readers have told me that they struggle reading the white font on a background. To make life easier I have switched to this format. Let me know if this is easier on the eye - design never was by strong point!

One of the seven sins of fund management (section III of Behavioural Investing) concerns the folly of forecasting (Chapter 9). This is our obsession with trying to forecast the future. Yet there is an enormous amount of evidence to suggest that we simply can't forecast with any more accuracy than a coin toss (a frequently we perform worse than even chance!).

One of the papers that didn't make it into the Behavioural Investing book (but with hindsight perhaps should have been added in) was on the performance of economists in forecasting recession. In it I pointed that economists are simply hopeless when it comes to forecasting recessions (I could have stopped that sentance before the word recessions).

Their track record is truly appalling. The chart below shows that in recent history (1980 onwards) the consensus of economists has not managed to forecast either of the recessions that have occurred. The data for this charts from the Philly Fed Survey of Professional Forecasters.




In the past I have proposed that simple quant models often have the edge of human judgement (see Chapter 22 of Behavioural Investing). Some new research by the San Francisco Fed shows that my supposition that economists would be no different than many other fields in finding their subjective forecasts outperform by a simple model was correct.

In a new paper Glen Rudebusch and John Williams show that a simple model based on the slope of the yield curve has significantly outperformed economists in forecasting recessions. They show that even if we use the economists own probability of recession estimate (rather than their spot forecast), the simple model wins hands down.

The chart below shows the so called anxious index, which is the economists stated probability of recession over the next four quarters. As Rudebusch and Williams state "Even at a horizon of two quarters, and certainly at three and four quarters ahead, the probability forecasts appear to have little relationship with historical recessions."

Compare these economists probabilities with the probability of a recession from a two variable model (using the level of short rates and the slop of the yield curve). The economists say their is currently a 19% chance of a recession in the next 4 quarters, the simple model says it is closer to a 30% probability.



Of course, economists have been aware of such simple models for many years. But this begs the question why they don't use/follow them? My own answer is overconfidence. This seems to be supported by the Rudebusch and Williams paper which concludes "It is interesting to note that many times during the past twenty years forecasters have acknowledged the formidable past performance of the yield curve in predicting expansions and recessions but argued that this past performance did not apply in the current situation. That is, signals from the yield curve have often been dismissed because of supposed changes in the economy or special factors in‡uencing interest rates. This paper, however, shows that the relative predictive power of the yield curve does not appear to have diminished much, if at all. "

Yet another example of just how poor forecasting really is. We need to find a better way to invest than relying upon our disproven and discredited forecasting abilities.

Something the Boglehead wouldn't want you to know, or index investing isn't passive

Just for the record, Bogleheads are die-hard devotees of index investing. Occassionally someone will mistake my criticisms of much of the active management industry for support of the Bogleheads' position. However, this isn't the case. In fact I reject pretty much all the foundations that index investing is built upon (see Chapter 35 of Behavioural Investing). The only exception is that the Bogleheads are quite right to point out the importance of minimising costs.

I recently came across a paper which I thought deserved some attention. It goes by the title of Index Rebalancing and Long-Term Portfolio Performance by Cai and Houge. It focuses on one of the misnomers of investing, that index investing is passive. This simply isn't true. Many indices are relatively actively managed. In fact most indices are really momentum players effectively adding stocks that have done well and deleting stocks that have done badly. This raises the question as to whether this 'active' element adds or destroys value. That is to say would you be better off if you ignored the index changes made by the index setters?

Cai and Houge take the Russell 2000 index and examine its performance since 1979 and see if the index changes that have occurred managed to add value to the investor over various time horizons. The Russell 2000 index is a small cap index, and makes on average 457 index changes each year (around 10% of market cap).

Cai and Houge show that an average an investor would be 2.2% better off in year one if they ignored the index changes, this rises to 17% in year five! So a buy and hold strategy seems to generate substantially higher returns for investors (yet again evidence of patience being key to investors - see Chapters 30/31 of Behavioural Investing).

Effectively Cai and Houge show that deletions have better future long term returns that additions to the index. In fact they show that deletions outperform non-new issue additions by around 8.9% in year 1 and 28% over five years. If one includes new issues that are added to the index, the situation is even worse since they underperform the deletions by 40% over five years!

Given that they are considering the Russell 2000 (a small cap index remember) , some stocks will leave the index because they become too large. Indeed the returns on these stocks seem particularly important in generating some of the short term outperform of the buy and hold strategy.

However, the results that Cai and Houge uncover are not simply an artifact of the way of the index considered. Siegel and Schwartz (2006) show a similar picture for the S&p500 (where no stock is deleted for being too large!) They track the changes made to the S&P500 from 1957 onwards. Nearly 1000 index additions over the sample period, averaging around 20 a year.

Three portfolios are formed, allowing for different scenarios:

(I) The survivor portfolio consists only of shares of the original S&P 500 firms. Shares of other firms received through mergers are immediately sold and the proceeds invested in the remaining survivor firms in proportion to their market value.

(II)Direct Descendants’ Portfolio (DDP), which consists of the shares of firms in the survivors’ portfolio plus the shares issued by firms acquiring an original S&P 500 firm.

(III)Total Descendants’ Portfolio (TDP) and includes all firms in the DDP plus all the spinoffs and other stock distributions issued by the firms in the Direct Descendants’Portfolio. The only difference between the TDP and the DDP is that the TDP holds all the spin-offs rather than sell them and reinvest in the proceeds in the parent firm.

The returns to the various portfolios are shown below:

	Geometric return							SD	Sharpe
Survivors Portfolio	11.31%							15.72%	0.4343
Direct Descendants	11.35%							15.93%	0.4331
Total Descendants	11.40%							16.09%	0.4337
S&P 500	10.85%							17.02%	0.3871

All three of the constructed portfolios outperform the index with it's additions and deletions, and they do so with considerably less risk!

The bottom line appears to be that index investing is often very far from passive. The rules of index construction appear to destroy value. Of course, one of the best ways of avoiding this problem is to be a long-term investor (i.e. conduct time arbitrage).