Diversification Sweet Spot

[Nords]

So, it could be argued that a 4x25 portfolio, like in the Coffeehouse portfolio, actually weights value + growth, as well as large + small more equally than a TSM fund.

I remember a similar study claiming that holding equal amounts of the S&P500, S&P400, and S&P600 ETFs had a higher return than an equivalent amount invested in the S&P1500. And that excess return is after the "higher" trading costs plus the index fund's expense & taxes of buying/selling the additional stocks as the index changes.

 

[ats5g]

I remember a similar study claiming that holding equal amounts of the S&P500, S&P400, and S&P600 ETFs had a higher return than an equivalent amount invested in the S&P1500. And that excess return is after the "higher" trading costs plus the index fund's expense & taxes of buying/selling the additional stocks as the index changes.

And since the S&P1500 is market cap weighted, I'd bet that higher return is due solely to the fact that smaller stocks outperformed large stocks.

- Alec

 

[Nords]

And since the S&P1500 is market cap weighted, I'd bet that higher return is due solely to the fact that smaller stocks outperformed large stocks.

Exactly. It moves from cap weighting more toward equal weighting.

 

[ESRBob]

just to throw my gasoline on the fire --

Why should cap weighting be such a holy grail or standard? It is guaranteed to give you lots of big stocks. Instead you would want to hold different segments of the market in proportions that have historically generated good returns for the amount of risk (volatility) incurred, based on your own preferences for risk and return.

Once you think like this, you'll probably be 'overweight' smaller and vaue sectors, but only because they are so under-represented in the index. It doesn't make them 'bad' or 'less-desirable' asset classes. It would be like comparing the total value of Gold in the world to the total value of Stocks in the world in order to try to figure out how much of each to hold. Instead, hold the proportions in your portfolio that have historically generated the best risk/return performance for your taste.

 

[JohnEyles]

So, it could be argued that a 4x25 portfolio, like in the Coffeehouse portfolio, actually weights value + growth, as well as large + small more equally than a TSM fund.

I wonder why you call Coffeehouse a 4x25 portfolio ?

The versions I've seen have 10% each in large, large value, small, small value, and foreign.
Then another 10% in REITs, and 40% in fixed income. They seem to ignore commodity
altogether.

I wonder about ignoring mid-cap altogether, as most simple portfolios seem to do.

 

[Nords]

Why should cap weighting be such a holy grail or standard?

It's the mutual-fund equivalent of the human appendix-- an organ that aided digestion and was easily implemented but no longer serves a useful purpose as we've evolved.

I think cap weighting was the best way to deal with limited computing power and made more intuitive sense (better marketing & sales) than the DOW's price weighting. Today it's just waiting to be replaced with something better, although (like the appendix) it's expensive to remove yet can still kill you.

 

[justin]

The demise of the small-cap value run has been predicted for at least two of the last four years, but it just hasn't happened yet. Eventually there'll be a reversion to the mean... the trick is figuring out when.

I've tried to figure out why the small-cap value premium would ever end.

I get that small cap value may be "overvalued" today by, say, 25%. I've just assumed that is true, for the sake of argument.

But what is going to happen over the next 30 years? I've read the small cap value premium may be 2%. So the total market grows at 8%, and the SCV grows at 10%. 30 years out, the total market has gone up 906%, while the SCV has gone up 1645%. Even if the SCV drops in value (ie - becomes "fairly valued" over 30 years) by the 25% by which it is currently overvalued, you'd still get a 1209% return in SCV.

In other words, given a sufficiently long period of time to allow compound growth, the premium, if any, you pay for SCV today will be trumped by the higher rate of growth of SCV. Using the numbers I assumed, the breakeven point would be around 15.5 years.

Implicit in this analysis is that the SCV will grow at a rate greater than the total market (primarily the large caps). This premium makes intuitive sense, since it should be easier for a company with $100 million in revenue and a 1-2% market share in their industry to double their revenue and market share than a corporate giant with tens of billions in revenue and a 40% market share.

 

[ESRBob]

Interesting analysis, Justin.

One more point is that 'small cap' is what most of us would consider to be pretty large companies -- Vanguard's index has a median market cap of 1.6 Billion. It isn't as though owning small caps is like somehow betting on little venture-backed startups or the grocery store down on the corner. I think realizing the size and sophistication of what gets classified as a small cap helps people get comfortable owning more of them.

 

[wabmester]

I've read the small cap value premium may be 2%.

Your logic seems predicated on this assumption. What if the "premium" is simply an artifact of data mining? In that case, you might expect ScV to underperfom in the future as much as it has recently outperformed the total market.

If the premium really exists, and it hasn't been rationalized away by its wide publicity, then there should be no reason to hold any other asset class, right? 100% ScV for everybody! Hmm, what would that do to the Total Market?

 

[justin]

Your logic seems predicated on this assumption. What if the "premium" is simply an artifact of data mining? In that case, you might expect ScV to underperfom in the future as much as it has recently outperformed the total market.

If the premium really exists, and it hasn't been rationalized away by its wide publicity, then there should be no reason to hold any other asset class, right? 100% ScV for everybody! Hmm, what would that do to the Total Market?

My logic is predicated on the assumption that SCV will grow at a faster rate than total market over the long term. Whether that will hold true, I don't know.

What I'm trying to understand (and rebut) is the argument that "everyone knows about the small and value premium, so it's too late to exploit it". If the small and value segments of the market continue to grow at a faster rate than the market overall, then this continued outperformance year after year will make small and value a better proposition long term, even if the extra cost you pay today gets washed out in a correction to "fairly" value the small and value segments.

Small caps and value indexes are imperfectly correlated to the total market. Basic portfolio theory says that a mix of multiple imperfectly correlated asset classes will have lower volatility than what you would get by taking a simple weighted arithmetic average of the constituent asset classes. The goal of including a tilt towards small and value is to capture those small and value premiums, while at the same time incorporating multiple imperfectly correlated assets with the goal of reducing the overall portfolio volatility.

 

[saluki9]

Your logic seems predicated on this assumption. What if the "premium" is simply an artifact of data mining? In that case, you might expect ScV to underperfom in the future as much as it has recently outperformed the total market.

If the premium really exists, and it hasn't been rationalized away by its wide publicity, then there should be no reason to hold any other asset class, right? 100% ScV for everybody! Hmm, what would that do to the Total Market?

Does this mean we're going to have to start arguing the merit of the Fama/French 3 factor model? I'm not sure I have the energy today.

 

[wabmester]

Does this mean we're going to have to start arguing the merit of the Fama/French 3 factor model? I'm not sure I have the energy today.

Well, we can talk about the secret 4th factor they locked away in the closet: momentum!

Since Fama and French did their famous data mining, a lot of people have tried to rationalize the small and value factors as "risk" factors for which you get compensated, right? That never sat very well with me, but I think I've finally come up with a model that works.

ScV is really a cleverly disguised market timing model, and I kind of like it in this new light! Here's how it works:

1) Slice the market up, and identify "small" caps and "value." Both are obviously relative terms.

2) Publicize the small/value "risk" premium.

3) Everybody rushes in to buy ScV, which causes all the stocks in that slice to become mid-cap growth!

4) Nobody wants mid-cap growth, so it's time to reslice.

5) Now we have a new index with the next generation of ScV. In the meantime, the unloved mid-cap growth are getting cheaper since we took them out of the index.

6) Oh look, all the mid-cap growth got so cheap that they've become ScV again. Reslice, reindex, and repeat forever!

You basically capture the growth of stocks going from small cap to mid cap and from value to growth on a continuous basis. Pretty brilliant, isn't it? Forget the "risk premium" nonsense.

 

[lswswein]

3) Everybody rushes in to buy ScV, which causes all the stocks in that slice to become mid-cap growth!

4) Nobody wants mid-cap growth, so it's time to reslice.

5) Now we have a new index with the next generation of ScV. In the meantime, the unloved mid-cap growth are getting cheaper since we took them out of the index.

6) Oh look, all the mid-cap growth got so cheap that they've become ScV again. Reslice, reindex, and repeat forever!

You basically capture the growth of stocks going from small cap to mid cap and from value to growth on a continuous basis. Pretty brilliant, isn't it? Forget the "risk premium" nonsense.

Nice conspiracy theory - have you been watching too many movies lately sir!
Anyway DFA already knows and understand that ScV is ovebought. They "own" huge % of a lot of small stocks and the universe they are playing is rapikdly shrinking. So instead of investing in ScV - they have new core portfolios which includes small, Midcap & large.
Aslo they don't trade out of their small cap stocks when they are no longer in the index - they have their own indices. They have buffers around their ports.
Btw I am not a DFA shill or anything - I just keep an eye on them because they do a good job. Also its not an ScV premium, its small & Value premiums - ScV is just the steriod version. Which might not always work out.
I think the biggest reason, the whole world does not load up on ScV is all the active investing around you. Let the whole world first start using indexing then will we think about ScV
-h
p.s: btw I am not dissing anyone

 

[Nords]

Your logic seems predicated on this assumption. What if the "premium" is simply an artifact of data mining? In that case, you might expect ScV to underperfom in the future as much as it has recently outperformed the total market.

Well, Dimson & Marsh worked on a century's worth of data to come up with their numbers. That pretty well exceeds any time horizon I'd be able to expect.

If the premium really exists, and it hasn't been rationalized away by its wide publicity, then there should be no reason to hold any other asset class, right? 100% ScV for everybody! Hmm, what would that do to the Total Market?

Sssssshhhhh, you fool! You'll spoil it for the rest of us!!

I think people who've come across the ScV premium have also passed due to its volatility. Or else they keep it a small portion of their portfolio. And then there's the ScV mutual funds who really really try to avoid fund bloat but inevitably fall on their swords.

My logic is predicated on the assumption that SCV will grow at a faster rate than total market over the long term. Whether that will hold true, I don't know.
What I'm trying to understand (and rebut) is the argument that "everyone knows about the small and value premium, so it's too late to exploit it". If the small and value segments of the market continue to grow at a faster rate than the market overall, then this continued outperformance year after year will make small and value a better proposition long term, even if the extra cost you pay today gets washed out in a correction to "fairly" value the small and value segments.
Small caps and value indexes are imperfectly correlated to the total market. Basic portfolio theory says that a mix of multiple imperfectly correlated asset classes will have lower volatility than what you would get by taking a simple weighted arithmetic average of the constituent asset classes. The goal of including a tilt towards small and value is to capture those small and value premiums, while at the same time incorporating multiple imperfectly correlated assets with the goal of reducing the overall portfolio volatility.

Yup. Or else going for large helpings of ScV and ignoring the volatility.

 

[Hydroman]

Interesting analysis, Justin.

It isn't as though owning small caps is like somehow betting on little venture-backed startups or the grocery store down on the corner.

I work for a "little venture-backed start-up" with part of my compensation being paper equity and stock options That is more then enough micro/small cap exposure for me. I limit my investment portfolio to a healthy dose of large cap and fixed income.

 

[rmark]

The advantage of cap weighting is that the relative weight of stocks in the portfolio adjusts automatically, with no added trading costs.

 

[rmark]

Equal weighting was tried with the original index fund back in 1976, but trading costs to rebalance were high.

 

[heyyou]

From my reading, rebalancing does not boost returns, but does dampen volitility. I thought that would help returns but I'm not literate in statistics.
At the Diehards, post 54500, look at the tables and TrevH's short remarks below each table. The time period is 1927 to 2005. Callan Periodic Table led me to equal slices of several asset classes. Turns out that a barbell of Large Blend or Market with some amount of Small Value has less StDev than 4 x 25 slices. I don't understand how 2 slices have less volitility than 4 slices. Note that equal weighting of FF deciles has the most StDev of all the combos in Table 2. That shot a hole in my thinking.
Those tables do confirm why Bernstein and others advocate barbells and not Sm, Mid, Large. Something else of interest is the tightness of the spread of StDev--for about the same volitility, some allocations produce double or triple the returns over the 78 year period. Someone young who wills money to future grandkids could easily have that time horizon.
My new opinion is broad diversification seems less important than the low correlation of what you diversify into. Is that Buffet instead of Slice and Dice?

 

[Rich_by_the_Bay]

From my reading, rebalancing does not boost returns, but does dampen volitility. I thought that would help returns but I'm not literate in statistics.

Yes, that seems to be the consensus -- no rebalancing bonus. However, all the simulations and spreadsheets I have seen on this assume the money remains invested. Intuitively, rebalancing leaves that money in the portfolio to gain or lose value with all the other investments . Over time it comes out in the wash.

But suppose you actually withdraw funds to rebalance as many retirees might do. In other words, given a $1mm portfolio you remove $40k annually, and that money comes entirely from rebalancing, or at least to the extent that returns allow. Now that money is off the table and -- in a relative sense -- you have sold high. Compare that to just selling 4% of each holding, pro rata.

I once emailed Gummy about that, figuring he would be someone who knows how to do this. Guess he wasn't interested since I got no reply. I still think it's an interesting question.

 

[wabmester]

you have sold high

The problem is that selling "high" is only a Good Thing if the asset class doesn't continue to keep going higher. In the case of stocks vs bonds, stocks will generally continue to outpace bonds, which is why there is no rebalancing bonus (even in the case of a withdrawal).

Rebalancing for a "bonus" is simply thinly-disguised market timing. That's not to say that market timing is always bad, but to expect a random periodic rebalance to pay any bonus is nonsensical.

 

[Rich_by_the_Bay]

The problem is that selling "high" is only a Good Thing if the asset class doesn't continue to keep going higher. In the case of stocks vs bonds, stocks will generally continue to outpace bonds, which is why there is no rebalancing bonus (even in the case of a withdrawal).

Rebalancing for a "bonus" is simply thinly-disguised market timing. That's not to say that market timing is always bad, but to expect a random periodic rebalance to pay any bonus is nonsensical.

Agree, but you have to take the money out somehow if you are withdrawing periodically. So you can sell the winners mostly, or you can sell 4% of everything. Maybe it's a different calculation in that case.

 

[wabmester]

Agree, but you have to take the money out somehow if you are withdrawing periodically. So you can sell the winners mostly, or you can sell 4% of everything. Maybe it's a different calculation in that case.

How about selling a gain/loss-weighted amount of winners and losers. Then you get a bonus by paying less in cap gains taxes.

 

[Rich_by_the_Bay]

How about selling an equal amount of winners and losers. Then you get a bonus by paying less in cap gains taxes.

I like the way you think.

Alas, mine is all tax deferred, all ordinary income to me.

 

[lswswein]

Rebalancing bonus is quite small according to Bernstien but exists.

http://www.efficientfrontier.com/ef/100/rebal100.htm
http://www.efficientfrontier.com/ef/996/rebal.htm

This is one where is write when it does not work.
http://www.efficientfrontier.com/ef/197/rebal197.htm

I think this captures the concept well:
"Under what circumstances does it pay or not pay to rebalance?" Rebalancing works best with volatile, uncorrelated assets whose returns are roughly similar. Transaction costs will result in further downward adjustment, and taxability may completely eliminate any rebalancing benefit at all.

 

[wabmester]

Rebalancing bonus is quite small according to Bernstien but exists.

The last paper you listed from Bernstein is the only one you should read. That's where he realizes the mistake he made in his earlier papers and says:

In fact, however, this "excess return" is illusory.