Fundamental Indexing

[kyounge1956]

Good thread,
By "factor exposure" Bernstein means that certain factors (e.g. smallness or book-value ratio, a way of classifying something on a value-growth continuum) have been shown through repeated historical research to generate excess returns which most researchers agree accrue to the additional risks such companies face. (Small stocks are perceived to be riskier possibly because they can get cut off from credit during a credit crunch or value stocks are perceived to be risker because they are (often) companies whose stock prices are beat up because something has gone wrong with the business)

So what Bernstein is saying (and I agree) is that any outperformance by these fundamentally-ranked portfolios may only be the expected return for the additional risk they are taking on by having more small or value in their mix than the S&P

This may be just what you want, but it isn't a magic bullet way to earn a free lunch, extra return for no additional risk.

Also, as someone said in this thread, people have been putting similar portfolios together for years without Arnott's indexes by using small and value-tilts.

One trivia footnote --I worked for Arnott years ago (1982,83) when he was a portfolio manager at the Boston Company and I was a programmer pulling together the screens through the historical CRSP and Compustat data to look for things that would produce excess returns... I can assure you people have been looking for ways to mine that data to find a magic formula for a really long time. If anything gets found, the advantage tends to get traded away in a hurry. Made me into a lifelong slicer/dicer indexer...

I think the book is only claiming that a fundamentally weighted broad-market basket of stocks will significantly outperform a cap-weighted broad market basket. Using a "value" screen would eliminate the overpriced stocks from the mix, so it doesn't surprise me that the performance of a value index is similar to a fundamentally-weighted broad-market index. In fact it wouldn't surprise me if the value-screened basket outperforms the fundamentally-weighted stocks, because the value screen would remove overpriced bubble stocks from the portfolio altogether, while fundamental weighting only assures that they don't increase in proportion to the whole.

There are both mutual funds and ETFs based on fundamental weighting. The expense ratios I saw for a fundamentally weighted index of large US stocks, large international stocks and small/medium US were all lower than 1%, so the cost to implement fundamental indexing is less than the benefit when compared to a cap-weighed broad-market index, but I can easily imagine (though I haven't checked) that when compared to a cap-weighted, value-screened index fund, the cost might be greater than the benefit. Implementation of fundamental weighting requires inclusion of three characteristics in addition to the one the value index uses, and there would also be some additional trading required for reasons already mentioned earlier in the thread. I would guess that in the raw numbers, a fundamentally-weighted, value-screened basket of stocks would outperform the same basket with cap-weighting (but maybe not by a statistically significant amount), but that when the expense of implementing the fundamental weighting is included the cost would exceed the benefit.

 

[ESRBob]

<caveot>

as a topic aside, In my opinion Bob Clyatt's larger than 4 % SWR that he advocates in his book "Work Less, Live More" falls in the data mining (ie- unrealistic) category

MB -- this is a valid concern, and thanks for raising it.
Anyone using historical data to make investment decisions is at some level guilty of 'data mining'. But my own bias is that data mining is on its thinnest ice when it involves looking back through data on stocks to try to tease out factors that successful stocks shared, and then buying those stocks on the assumption that those relationships will continue to hold going forward.

The Work Less Live More approach (shared by lots of big institutional money managers, btw -- I didn't invent it but just adapted it to the needs of individual long-term retirees) is different in what might seem a subtle way but I think an important way. Rather than sifting the data for individual stocks, we sifted the data for baskets and proportions of asset classes which collectively delivered attractive characteristics over long periods -- not just in terms of market return but also in terms of volatility.

We sought not individual stocks or even funds or even asset classes but blends of asset classes which together deliver moderate volatility and reasonable returns over time. These weren't picky, either -- as long as the broad proportions of equities and fixed income were intact, there are lots of ways to assemble similar portfolios and get similar results. This is amply demonstrated by the three portfolios in the book -- a single-fund, an 8-fund and a 15-Asset Class portfolio all of which have about the same returns and volatility historically. (Agreed that the 15-asset class portfolio has special appeal to those looking to milk the last tenth of a percentage point of return, and may not deliver any better than the 8-fund over time, but the dream dies hard ;-)

Perhaps more to your point -- is the Safe Withdrawal Method, which indicates a 4%+ withdrawal rate from a diversified portfolio, a case of data mining? I think in this case this part of the research was just a matter of running lots of withdrawal scenarios and trying to get a sense of how and when withdrawals got to the point where they started to eat into portfolios in ways that weren't robustly recoverable. I don't think it is mining so much as looking at a natural process and seeing where and how it works, and where you want to be with it. Something like the analysis of how much to let commercial fishermen take out a fishing ground before the population collapses. Not an exact science but an organic process that can be monitored and understood, and optimized. Because we have lots of data and an ability to run lots of scenarios (given actual market results, what would have happened if I had withdrawn this percentage from that portfolio every year for 40 years, etc.) it is only human nature to want to have that information to base decisions on.

At the end of the day you take $ out of your portfolio to live on, and everybody likes to have a budget and know when they have enough to leave full-time work or whether they can go out for an additional steak dinner this month or whatever. So some numbers are needed -- the tenths-of-percentage points are not wrong (I believe) but they suggest more precision than is warranted given all the uncertainties in markets, and what our actual future results will be. Thanks for raising the issue -- a valid concern-- and I hope this sheds some light on how the number is derived and how it can responsibly be used to help long term retirees work less and live more.

 

[MasterBlaster]

ESRBob:

I don't doudbt your methodology. I just suspect that one can always find some combination that in hindsight always performs better.

So the question is... Is the combination that you found really better or just a fluke of the data? And more importantly... Is it repeatable with any certainty over the duration (my ER) that I am interested in ?

More specifically, Exactly how independent (statistically) are the asset classes you pick ?

 

[rayvt]

I think the book is only claiming that a fundamentally weighted broad-market basket of stocks will significantly outperform a cap-weighted broad market basket.

Problem is, there is a known definition of "cap-weighted" broad market basket. Well known and well understood.

But there is NOT any similar definition of "fundamentally weighted" broad-market basket.

Fundamentally Weighted Index says:
"Fundamentally-weighted indexes may be based on fundamental metrics such as revenue, dividend rates, earnings or book value. Proponents of these indexes claim......"

So any claim such as the book makes is bogus on its face. It simply begs the question. Since there can be an infinite number of metrics and weightings, the term "fundamentally weighted" has no concrete meaning.
Unlike "cap weighted", which has exactly one meaning.

 

[kyounge1956]

Problem is, there is a known definition of "cap-weighted" broad market basket. Well known and well understood.

But there is NOT any similar definition of "fundamentally weighted" broad-market basket.

Fundamentally Weighted Index says:
"Fundamentally-weighted indexes may be based on fundamental metrics such as revenue, dividend rates, earnings or book value. Proponents of these indexes claim......"

So any claim such as the book makes is bogus on its face. It simply begs the question. Since there can be an infinite number of metrics and weightings, the term "fundamentally weighted" has no concrete meaning.
Unlike "cap weighted", which has exactly one meaning.

"Fundamental weighting" is a somewhat nebulous term, but the book uses it to refer to one specific variety of fundamental weighting, a composite of four characteristics over five years, called the RAFI Fundamental Index, so claims can be made with relation to that specific method.

Claims could also be made about fundamental weightings in general. For example: "cap-weighted indexes have a built-in performance drag due to overweighting of overpriced stocks, and any variety of fundamental weighting is a potential solution because it breaks the link between price and weight". Does the data bear out this claim? Maybe or maybe not. If it works on raw data, can it be put into effect at a cost lower than the benefit it provides? Maybe or maybe not. But it is a testable claim and therefore not "bogus on its face", and which the data, on further investigation, will either support or disprove.

 

[HornedToad10]

Claims could also be made about fundamental weightings in general. For example: "cap-weighted indexes have a built-in performance drag due to overweighting of overpriced stocks, and any variety of fundamental weighting is a potential solution because it breaks the link between price and weight". Does the data bear out this claim? Maybe or maybe not. If it works on raw data, can it be put into effect at a cost lower than the benefit it provides? Maybe or maybe not. But it is a testable claim and therefore not "bogus on its face", and which the data, on further investigation, will either support or disprove.

Who decides what's "overpriced?" I thought Google was overpriced shortly after it IPO'd and it's possible that the fundamental index would have thought the same thing. However, as time went by, I'm glad the indices I owned held Google....

It's easy to say remove all the overpriced stocks. It's hard to decide which are overpriced and which are just growing rapidly.

It's easy to buy the underpriced/value stocks. It's hard to decide which are value plays and which are just losing all their business and plummeting...

 

[ESRBob]

ESRBob:

I don't doudbt your methodology. I just suspect that one can always find some combination that in hindsight always performs better.

So the question is... Is the combination that you found really better or just a fluke of the data? And more importantly... Is it repeatable with any certainty over the duration (my ER) that I am interested in ?

More specifically, Exactly how independent (statistically) are the asset classes you pick ?

MB,happy to restart this as a separate thread if there is more to dig into.
The combinations found are unlikely to be flukes as I avoided extreme combinations and just looked at ways to combine all (except metals - personal bias- lumped it in with commodities as it doesn't pay interest) of the major, credible asset classes in various constrained proportions (nothing less than 2% or greater than about 15 % of portfolio) to see if there were ways to get at low volatility, reasonable long-term return. There were plenty of better performing (return) yet extreme or highly skewed portfolios that didn't make the cut because they felt like 'flukes' to me, too.

The asset classes weren't randomly picked, if that is what you mean by independent. There are only about 15 and I looked at all of them. My bias is toward more diversification (again, to reduce downside risk) so I tended to want more asset classes rather than fewer. My sense from the data is that you could double or halve any of the target asset allocations and come out pretty close to the same performance (which might happen between re-balancings)-- these things are not that fragile. You can drop asset classes you don't like and come out about the same, too.

The point is to stay around 50/50 equities /bonds or 40-40-20 if you add the alternative asset classes like commercial real estate and commodities. Given that these are so very middle-of-the-road, designed to deliver average results, it doesn't feel like flukes at all. This is because I was looking for low-volatility portfolios with decent returns, not high-return portfolios. As a result, we ended up with a compromise portfolio that I and lots of long term and institutional investors feel we can rely on. We'd all love to live off guaranteed income streams, (4% Real returns TIPS anyone?) but in the absence of that I don't know anything better than this very diversified, conservative slice/dice portfolio for long run retirees. If you know of one or if I can find one it will certainly go in the next edition of the book!