|
Softpanorama |
May the source be with you, but remember the KISS principle ;-)
|
| ... The best known goals of the alchemists were the transmutation of common metals into gold or silver, and the creation of a "panacea", a remedy that supposedly would cure all diseases and prolong life indefinitely. Starting with the Middle Ages, European alchemists invested much effort on the search for the "philosopher's stone", a mythical substance that was believed to be an essential ingredient for either or both of those goals. |
This is the strategy recommended by mainstream press. Special calculators exist to provide you with the static allocation that supposly is suitable to your risk tolerance, investment horizon and behavior during crashes (asking naive question like "Do you tend to buy more or sell what you have ?"). One of the better calculator is CNN Money retirement calculator CNN Money retirement calculator
While it makes perfect sense to diversify assets into different classes of investments (stocks, bonds, precious metals, commodities like oil, etc) investment alhemism for some reason is concentrated in stocks only. Many such "perfect stock cocktails', especially those published in books, smells with "data mining" (attempt to fit the model to historical period used.) and look pretty silly after five or six years (you will never notice that if you buy only latest edition of the books ;-)
Investors, who adopted "stock diversification myth" and forgot about other classes of assets and first of all bonds which "in the long run" have returns not far detached from stocks outside very favorable for stocks period.
This is done usually with the
help of gurus, who published books, financial advisers or maintain mutual funds
web sites. Like alchemists in the past they all are essentially trying to find "perfect static allocation"
between four or more mostly stock mutual funds. Such a static allocation
of several (usually 4-6 but sometimes up to a dozen) mutual funds supposedly automatically guarantee the best
return while minimizing the risk while the owner of the "blessed" portfolio can
drink gin and tonic or do other more interesting things instead of dull job
of analyzing financial statistics to avoid blatant rip-off by Wall
Street professionals :-)
William Bernstein can be called the
leading proponent of complex static stock mutual funds cocktails with an additional
twist of blind rebalancing at the end of each year. In his book he claimed that
this primitive and completely unscientific strategy is the ultimate solution
suitable for intelligent investors. Of course no decent imitation modeling
of proposed portfolios was performed. Most of argumentation is faith-based.
The fact that he called this strategy "Intelligent Access Allocation"
represents a very interesting view on intelligence. In reality this is a close
analog of stock picking: finding the "winning mix" means that you can predict
the future development of economics. All authors mentioned above proved to be
notoriously bad at that as behavior of proposed in books published five or ten
years ago portfolios suggests. William
Bernstein is no exception and returns
from the portfolios proposed in his old book "Intelligent Access Allocation" are
far from being impressive in comparison with other investment strategies.
Blind annual rebalancing is suspect as it contradicts old Wall Street maxim "let
profits run". Another problem is that he relies on static
correlation, the assumption that is cargo cult in and by itself.
...The most recent Journal of Financial Planning shows that correlations among asset classes—the underpinning of any allocation strategy—are more changeable than we're led to believe. William J. CoakerII, senior investment officer for equities for the San Francisco City & County Retirement System, researched the correlations for 18 asset classes for the 1970-2004 period. U.S. stocks (the Standard & Poor's 500-stock index) and bonds (U.S. investment-grade issues) have a 0.23 correlation, which means that only 23% of the time are bonds and stocks acting the same way. But remember, that's an average, and the actual correlations range from 0.31 to 0.77. (A negative correlation means two assets move in opposite directions.)
What's more, Coaker showed in a previous article that correlations don't just bounce around but sometimes change their behavior. From 1970 through 1998, international stocks had 0.48 correlation with U.S. stocks. "But from 1998 to 2002, it rose to 0.83," he writes. These relationships are "inherently unstable."
If correlations are volatile, does it make sense to keep your allocations constant? Most advisers would say yes. But not all. Elliot Fineman, a senior vice-president at Compass Investors, a Kenilworth (Ill.) money manager, compares conventional asset allocation to driving at a constant speed, regardless of traffic and road conditions. "You'd have a serious accident," says Fineman. Compass issues asset allocation recommendations every five weeks based on a model that crunches economic and market indicators.
An even more radical approach: Put 85% to 95% of your money in an ultra-safe asset like U.S. Treasury bills or inflation-indexed bonds, and put the rest in speculative, leveraged bets. That's the sort of approach advocated by Boston University finance professor Zvi Bodie (BW—Sept. 10) and by hedge fund manager Nassim Nicholas Taleb in his best seller The Black Swan: The Impact of the Highly Improbable. It minimizes damage from an ugly downturn while maintaining some upside.
Does this mean asset allocation doesn't work? Not at all. But it may work better in some periods than others. If that seems like too much risk, you may want to consider one of the alternatives.
Still many multi-component strategies are demonstrably better then either pure index fund or pure bond fund (for example stable value fund) approach as long as total share or stock and total share of bonds are almost equal. In this sense they are more like benign exercise in creation of your own virtual stock and bond index slightly different from existing but without influencing results in a major way.
The factors that are unclear that are very superficially treated by advocates are:
Are multi-component allocations able statistically significantly
outperform two component allocation, for example 60% of diversified stock
and 40% of bond fund on various 10 year investment scenarios or differences
are statistically insignificant ? Of course in retrospect you
always can
create a perfect portfolio for any period. But it does not need to be
multi-component
:-).
How does return depend on properties of the index fund and
bond fund ? (fairy tails about low correlation are too simplistic answer
here; but high dividend stocks ETFs that the latest fashion and they might
make sense "in the long run" as historically stocks used to produce
dividends. A century ago nobody would sell stock that does not produce dividends
comparable to bonds :-).
It might be that value based funds are better then
generic indexes for such mixes but this is just a hypothesis.
Anyway
in the context of 401K investing this is a little bit theoretical thinking detached from reality: most 401K
plans has limited and pretty arbitrary selection of funds.
What minimum percentage of stock in static binary allocation
improves median and average returns without undue increase of risk
(optimum might be less then traditional 60% stocks; it might be close to
50:50 split, but in reality I just do not know; what I can tell it is
not 100% stock). BTW when interpreting result
of Monte Carlo simulations you should discard maximum returns for the period:
chances getting them are extremely slim as most returns in simulation of
binary stock-bond portfolio 10 years returns are clustered in the low third
of the possible range of returns, making the distribution of returns asymmetrical
and pretty far from normal curve (distribution that you logically can expect.).
Actually this looks more like Pareto distribution. So it might
be that your chances to get the return closer to minimum are much greater
that chances to get the return closer to maximum for any given period. That
also means that median return in imitation model runs is much better predictor
of actual return then an average return.
What level of improvement of median returns you should consider
to be adequate compensation for additional risk inherent in portfolios with more
then 50% of stocks (up to 50% adding stocks in simulations that I
performed actually improved
both minimal, medial and average returns, meaning that such portfolios
might be
less risky that 100% bond portfolio for 401K investors with active
investing period overt decade.
But if you take into account generally
low nominal annual return of most of 401K portfolios (if you use cost
averaging them for 10 years period starting from zero) the question arise,
what are we fighting for ? Here by "nominal annualized return" I means
return that is calculated using formula:
(final_sum/total_contributions)0.1
-1
For example is you contributed 130500 and got 191008 at the end of
your 10 year investment period then your ten years total return is 46%.
But you nominal annualized return is only 3.88%, not 4.6%. That
means that if you had put this sum (130500) into 10 Year CD with annual
interest rate 3.88. then at the end on ten year period you will get
then same amount of money ($191K). Of course you did not have all the sum
ten years ago, so "real returns" (return that you will get from a stable
value fund if you use the same cost averaging strategy) are higher then
nominal (I do not know how analytically calculate the difference but
my impression from simulations is that they are approximately 1.5 times higher:
real return is closer to 5.85% instead of 3.88% in this
particular case)
Paul Farrell is another leading financial alchemist., a proponent of various "magic" static mixes of mutual funds and he gives a pretty good exposure to basics of "financial alhemism" in his article They're lazy and they're boring, but they're winning portfolios.
Unfortunately multi-component stock-bond fund mixes proposed are all based on hearsay: none of the authors belonging to this category managed to demonstrate knowledge of elementary statistics on the level that ensure passing grade for the bachelor degree for some non-technical specialty. I think none heard about Monte Carlo simulations, despite the fact that some hold Ph. D degree :-).
Also none was ever able to demonstrate advantages of more complex multi-component mutual fund cocktails with the large doze of stocks in comparison with simpler two component stock index-bond fund mix based on equal split or some age related formula like classic (100-your_age). The latter becomes "equal split" of stocks and bonds when the person becomes 50 years old.
Nobody in this category is using basic statistical methods to prove that the selection they advocate can behave as they expect during various (different) historical periods. For example all mixes I studied behave quite differently during 1990-2000 then during 1996-2006. period. All-in-all the level of discussion is extremely, frustratingly primitive and is limited to calculating returns for the most recent one, three, five and ten years. They know nothing better. As Wikipedia explains this phenomenon:
Although some alchemists were indeed crackpots and charlatans, most were well-meaning and intelligent scholars; among their number can be counted such distinguished scientists as Sir Isaac Newton. These people in many ways served as innovators, and attempted to explore and investigate the nature of chemical substances and processes. They had to rely on experimentation, traditional know-how, rules of thumb — and speculative thought in their attempts to uncover the mysteries of the physical universe.
...Throughout the history of the discipline, alchemists struggled to understand the nature of these principles, and find some order and sense in the results of their chemical experiments — which were often undermined by impure or poorly characterized reagents, the lack of quantitative measurements, and confusing and inconsistent nomenclature.
... The best known goals of the alchemists were the transmutation of common metals into gold or silver, and the creation of a "panacea", a remedy that supposedly would cure all diseases and prolong life indefinitely. Starting with the Middle Ages, European alchemists invested much effort on the search for the "philosopher's stone", a mythical substance that was believed to be an essential ingredient for either or both of those goals.
Moreover in more complex mixes of mutual funds, individual funds
used often does not make a lot of sense as holdings overlap and behaviors strongly
correlate (and correlate more strongly when it is least desirable as was the
case during the recent recession of 2001-2003.)
I would understand adding cash, gold and REIT as independent (in case of REIT
semi-independent) classes of assets (but again that's not that easy to do with
static allocation -- allocation should be dynamic and take into account both
your age and some kind of generic stock valuation, for example P/E ratio).
As an example of financial alhemism here is one such recipe -- "lazy portfolio" from the article They're lazy and they're boring, but they're winning portfolios.
, five and ten years. But it is naive to assume that portfolio that demonstrated good results in all three time frames mentioned above will continue demonstrate the same behavior in the future.
The end of the period date of such simulations is also very important: it's easy to see that certain end dates are better then other. For example, fund cocktails created on the base of the best returns for periods that end on March 2000 have very little predictive power. They are just a unique historical phenomenon ( historical curiosity, if you wish). The same is true for any period ending with some kind of breaking one or several records up or down.
But alchemy is alchemy and the authors are stuck to mixing the components as the only viable strategy of converting le
If you think implicit bet (or hidden economic prediction) in this 66% stock and 33% bond portfolio is that economic power might drift to Asia in the coming years and that Vanguard managers will be able to exploit this opportunity better then anybody else. If we know for sure that Asia will go bust, and Europe is stagnate there is no much attractive in this mix.
Also combination of domestic and international stock indexes gives your own version of "global stock index" and the fact that this allocation uses 66% of this newly created index in the mix is just accidental -- you can use anything you wish. As recession of 2001-2003 had shown us the independence of domestic and international indexes should be taken with a grain of salt as the correlation is dynamic and increases dramatically when you need it less -- during the recession. It might be fools gold for another reason: many USA companies in Total Stock market index have huge international presence and vise versa, internationals like BASF, Toyota, Mitsubishi, Toshiba, Sony, Philips or Erikson have considerable US presence.
Also it's unclear, why use TIPs as dominant bond strategy. Historical returns for TIPs are not that impressive and due to the fact that they are long term bonds the assumption that they can help to beat inflation might be wrong. Bond funds are not bonds, they are more risky financial instruments: any significant raise in bond rates endangers your principal and TIPS funds are no exception here.
There are many even more complex variants of those mutual funds mixes. Tor example the portfolio below can be "reduced" to binary allocation by adding additional bond fund (for example Pimco Total return) and making all allocations equal to 25%. In this case it will satisfy the requirement of almost equal split between stocks in bonds.
Mixes that contains high doze of stocks usually behave less predictably in Monte Carlo modeling and judging about their risk from returns in most recent ten, five and one year(s) is "trained idiotism" or, if you wish, "Financial Lysenkoism".
Also change of risk with growth of stock component is highly non-linear for any significant, say, ten years period and cannot be approximated by calculating statistics on each single year. That fact for example was missed in a typical "asset allocation article Optimal Asset Allocation by Dr. Steve Sjuggerud who naively considers the safest mix to be 20% of stock (probably meaning large diversified index like S&P500) and 80% of bonds ( probably meaning some bond index) and the riskiest 80% of stocks and 20% of bonds with approximately 60% of stocks as optimal allocation:
- SAFE PORTFOLIO -- 20% stocks, 80% bonds. Throughout history, this portfolio has averaged 7.0% a year. Its WORST year was a loss of 10.1%. It lost money 17% of the years.
- BALANCED PORTFOLIO -- 50% stocks, 50% bonds. Throughout history, this portfolio has averaged 8.7% a year. Its WORST year was a loss of 22.5%. It lost money 22% of the years.
- RISKY PORTFOLIO -- 80% stocks, 20% bonds. Throughout history, this portfolio has averaged 10.0% a year. Its WORST year was a loss of 34.9%. It lost money 28% of the years.
That's definitly not true for my simulation modeling results for 10 years investment period starting between 1990-1996 (see below): here both 80-20 stock-bond and 20%-80% stock-bond allocation were inferior to approximately 50:50 static stock-bond allocation, if cost averaging with initial capital equal to zero is used (see table below). Actually in this case ten year investment period risk (understood as medial or average variation of returns) does not decrease after approximately 60% of bonds, only returns suffer. But risk is disproportionately high on the low end (with stock percentage over 75). It looks like a simple formula 100-your_age has much deeper wisdom in it then it is commonly assumed. Among other things it protects you from putting more then 80% of your funds in stocks...
To minimize risk he recommends to split each class into subclasses and that part might make some sense: for example as you age you might rotate bond portion of your portfolio into safer short term bonds or stable value fund depending on the form on yields curve and stock portion into more conservative value stocks. He recommends the make the cocktail more complex using multiple representative in all three major classes of investments (stocks, bonds and commodities (limited to precious metals in this case):
It unclear what is the difference between returns on this complex portfolio and simple 50:50 stock-bond split and even if there is difference if it is statistically significant. Also we might get a simpler portfolio with similar returns if we replace the first class with Vanguard Total Stock Index (assuming the many US firms have significant international presence) and the second with Total bond index or Pimco Total Return (please note that bond indexes generally are less good then well managed stock funds, and Pimco has notoriously high fees )
Actually Vanguard has prepackaged "funds mixes" with various percentages of stocks, bonds and cash, for example one such fund has 40% in stocks, 40% in bonds and 20% cash (see below).