Will Trice wrote:
- quote -

Will - no, I don't use MVO, no more than I "use" the supply-demand
curves from Econ 101. I see it as an exercise that demonstrates one type
of optimization problem (one that would be more interesting if it had to
do with, say, finches and grub populations, instead of stocks & bonds).
I suppose if I was managing a pension fund I'd spend more time seeing if
I could apply it to come up with asset allocations, but I don't find it
particularly useful for individual-investor portfolios.


- quote -

My personal is that all this stuff is applicable only to baskets of
securities, where company/sector-specific effects won't predominate. It
shouldn't take too many but certainly one (or five, or ten) securities
can't be expected to fall in line with their 3-factor asset classes,
there's too much room for the narratives to drive the stocks'
performance. Was it a high book-to-market value that made Apple rocket
in 2004, or could it be, oh I don't know, THE IPOD? I'll put my buck on
the latter.

-Tad

Fetch wrote:

- quote -

To a certain extent you're right, I'm looking for a crystal ball. Or
more accurately, those who advocate MVOs as useful tools claim that they
are crystal balls. My question is, how cloudly are they?

MVOs are superb at telling us what the optimum mix of investments would
have been for a certain return or a certain risk - yesterday. Will that
mix hold for tomorrow? If it doesn't, then why are the results of an
MVO better than a random guess? For example, Beliavsky has shown us a
detailed analysis of the optimum mix of Vanguard investments (excluding
heathcare) for various risk/return goals over the period 1992 - 2001.
He has further suggested (I think wisely) that this type of analysis
could be even more useful if you apply constraints to the asset class
proportions. Let's say that he performed this analysis in order to
figure out how to allocate his assets in 2002 so that his portfolio's
volatility is minimized. The question is, if he did this, would he have
been successful (or close) in 2002?

Now that I've asked the question, I don't how I could devise a test to
check it. His data is based on quarterly results (which is a finer
grain than most other MVO data I've seen) so presumably an optimum
portfolio will minimize volatility on a quarterly basis. But I have
only four samples in 2002 (each quarter). I doubt I can get a
statistically valid standard deviation and expected return from four
samples. Does this mean that I would have to conduct an experiment
over, say, seven or eight years to test this? With maybe a rolling
window for my MVO over the previous ten years (because, as Tad pointed
out, the input data will change over time)? This gives rebalancing a
whole new meaning...

Apologies to Beliavsky, I haven't read the papers you suggested yet, nor
the book. Holidays, you know.

-Will

Will Trice wrote:

- quote -

MVO is a top-level decision tool - strictly mathematical. Apparently,
you are looking for a crystal ball.

Keep bets at the security level and use MVO to control overall risk
via allocation.

Will Trice wrote:
- quote -

I think Robert Haugen advocates MVO for individual stocks in his book
"The New Finance". Returns of stocks are estimated using a factor model
incorporating fundamental and technical (12-month monentum) variables.
One obstacle to its use in practice by mutual fund managers is that
they view risk as deviation from their benchmark, such as the S&P 500.
I think there has been research in the Journal of Finance showing that
MVO can give good results for stock portfolios, if constraints against
excessive concentration (no more than x% of the portfolio in any one
stock) and risk (no shorting) are imposed.

Without resorting to MVO, there are simple stock weighting rules, such
as equal weighting, that seem to outperform the cap weighting of the
S&P on average. I started a thread "S&P 500 not diversified enough"
about this in this newsgroup in August 2002, which you can find using
Google Groups. There now exists an ETF, symbol RSP, that tracks the
equal-weighted S&P 500.

A recent paper entitled "Fundamental Indexation", available at
http://papers.ssrn.com/sol3/papers.c...ract_id=604842 , finds that
weighting companies by fundamental measures, such as earnings, revenue,
or book value, has historically outperformed cap weighting. These are
diversified but value-oriented strategies.

Fetch wrote:
- quote -

Of course not. But does an MVO actually provide useful information?

- quote -

But are the MVO's predictions correct for the future? If they were
perfect, then I could use the MVO to make investment decisions. But
MVOs are obviously not perfect. How accurate are they at predicting the
future and under what conditions? If you say that I can use them as a
framework, then they must produce an output that is somewhat predictive,
right?

- quote -

I have yet to see a monkey throw darts. Feces, yes...

- quote -

What do you use for the MVO, then? Or do you use an MVO at all?


- quote -

I've read a lot of what's available from www.efficientfrontier.com and
it is great stuff. But it was this reading that lead me to my original
questions. In particular, Goetzmann's book (referenced on Bernstein's
www.efficientfrontier.com) indicates that asset allocation may not be
feasible for individual securities because of statistical estimation
errors, so it should only be used across asset classes. Yet Markowitz's
(the inventor of asset allocation theory) original study was on
individual securities, wasn't it (I haven't read it yet)? Even so,
Armstrong's book (also referenced by Bernstein) indicates that
correlations between asset classes varies widely depending on the time
period sampled. The risk-reward line even inverts occasionally (e.g.
during these times bonds may have higher return and risk than equities).
Armstrong suggests that a 20 year sample is the correct way to go, but
presents no evidence for this. He just says that 20 years is
sufficiently long-term.

This begs the question: Do MVO's provide a reasonable prediction of the
future? Especially with individual securities?

Just wondering,
-Will

- quote -

Absolutely!


- quote -

Surely you don't believe an optimization program coupled with a priori
probabilities is going to solve all of your investment decisions. Why
the confusion?

Most people actually double up on risk trying to diversify. The point
of MVO is to establish a frame work for choosing portfolios that
minimize risk for a given expected return - otherwise, it's simple
diversification.

- quote -

Go ahead and throw darts. Monkeys do it all the time.

Will Trice wrote:
- quote -

Will,
I think these are good points, and at the end of the day each investor
needs to form an opinion about this stuff. How much diversification is
enough, and how do you even choose the buckets to invest in? How do you
decided whether to put 20% or 0% or 5% in REITs?

Mine isn't actually too far from what you're alluding to - I don't look
to asset class data for MVO. The data analyses are interesting for
seeing how to dice up the market and yes, for seeing the ranges of
correlations between these different types of investments. But god
forbid we think there's some Pulse of the Capital Markets that we're
trying to measure. It fails common sense and that to me is when a theory
is most suspect.

On the flip side you have Beliavsky the self-described quant who
believes more strongly in the utility of MVO. And he's going to do just
fine, I imagine. The important thing, I think, may be the belief in what
you're doing and the perseverance with it. I suspect in 20 years both B
and I will have acceptable outcomes using methods that are at odds. Or
in English: there are a lot of ways to make a buck investing. The
important thing isn't whether your REIT allocation is 5.6% or 8%. Nobody
(well, nobody being honest) will say with a straight face that they know
with any degree of certainty which allocation is going to be better over
the next 20 years.


- quote -

No doubt with one of the inputs being something like 4% or 3.5%, for
inflation say. Reducing the precision to one or two significant figures,
but don't tell the software that. I tend to think in 5% blocks.


- quote -

These kinds of questions have actually been studied quite a bit. A good
start is the site www.efficientfrontier.com - it points as well to a lot
of good resources on the subject. Though I'm not sure it's necessary to
get into at that level for an individual investor, except perhaps to
"formulate an opinion" on the whole thing.

-Tad

beliavsky[at]aol.com wrote:
- quote -

I have read some of these studies as well. Intuitively it makes sense.
But interestingly, rebalancing doesn't require an optimal portfolio,
you just need any arbitrary asset allocation. Rebalancing then always
reduces exposure to assets that have appreciated and increases exposure
to assets that have lagged (i.e. buy low, sell high). This works for
more-ore-less the same reason that dollar cost averaging works, doesn't it?

- quote -

This is nice work. It would be interesting to know what the return and,
more importantly, the volatility of each portfolio was in the years
after 2001. Did the portfolios have close to the same volatility going
forward? Also, what do you think the effect would be of eliminating
assets that only represented a small portion of the portfolio. In other
words, at what point does an asset class become insignificant to the
results of the portfolio? At less than 3% of the portfolio value? Less
than 10%? Maybe you've already done this since I see that not all asset
classes (e.g. Energy) are represented in all the portfolios.

Will Trice wrote:
- quote -


I don't think there are clear answers to your pertinent questions. In
general, I think a quantitative approach based on reasonable
assumptions will outperform a purely heuristic approach. But then, I'm
a quant. It is possible to solve the mean variance optimization problem
under "reasonable" constraints that you impose. For example, you could
find the optimal portfolio under the constraint that

30% < U.S. stocks < 80%
5% < foreign stocks < 30%
20% < bonds < 60%
0% < REITS < 30%

Harold Evensky in the book "Wealth Management" describes how a
constrained optimal portfolio often has a Sharpe ratio almost as high
as the unconstrained optimal portfolio. I recommend his book to someone
interested in how MVO is used in practice. In another message I refer
to research that finds the MVO using high-frequency returns data.

Jonathan Clements of the Wall Street Journal, and other people, have
cited studies finding that periodic (say annual) rebalancing of a
portfolio does produce higher risk/reward ratios than a "drifting"
portfolio where assets weights change due to market returns. Since the
long term capital gains holding period is one year, one would have to
be careful about rebalancing more often, because rebalancing typically
involves selling the asset classes with capital gains. If you are
saving a regular amount each month, investing the new money in the
asset class that is currently most underweighted in your portfolio may
be sufficient to maintain a reasonable allocation.

If at least part of one's portfolio is in a taxable account, the
allocation to taxable bonds and REITS, whose returns are mostly
generated as taxable ordinary income, will be lower than in a
tax-deferred or tax-free account. Some academics have studied the
"asset location" problem of how to allocate assets in tax-deferred and
taxable accounts.

There are many index mutual funds that have almost identical
performance, and I think periodic, at least yearly, tax loss harvesting
should be considered. For example, if you have a substantial loss in an
S&P 500 index fund, you can sell it and replace it with a Russell 1000
index fund, which is very highly correlated.

Home equity is the most important asset of many households, and it can
be monetized through home equity loans, reverse mortgages, or a sale
followed be relocation to a less expensive area. There was an article
in the New York Times about retirees from places like New York and
California selling their expensive homes and relocating to cheaper
places. For some people, the volatility of house price changes and its
correlation to financial assets ought to be considered, but I don't
know of available free data.

A study I did in September 2002 of optimal portfolios of Vanguard
mutual funds is at http://members.aol.com/beliavsky/vgd_port.htm .

Tad Borek wrote:
- quote -

There is some academic research asserting that BECAUSE volatilities and
correlations are always changing, an asset allocation strategy using
current estimates of these quantities can outperform a static
allocation. The paper cited below can be downloaded from
http://papers.ssrn.com/sol3/papers.c...ract_id=304976 . The key
phrase is "a strategy based solely on volatility timing uniformly
outperforms market timing strategies, a model that assumes no
predictability and the market return in terms of certainty equivalent
gains and Sharpe ratios."

Sequential Optimal Portfolio Performance: Market and Volatility Timing

MICHAEL S. JOHANNES
Columbia University - Columbia Business School
NICK POLSON
University of Chicago - Graduate School of Business
JONATHAN R. STROUD
University of Pennsylvania - Statistics Department
Abstract:
This paper studies the economic benefits of return predictability by
analyzing the impact of market and volatility timing on the performance
of optimal portfolio rules. Using a model with time-varying expected
returns and volatility, we form optimal portfolios sequentially and
generate out-of-sample portfolio returns. We are careful to account for
estimation risk and parameter learning. Using S&P 500 index data from
1980-2000, we find that a strategy based solely on volatility timing
uniformly outperforms market timing strategies, a model that assumes no
predictability and the market return in terms of certainty equivalent
gains and Sharpe ratios. Market timing strategies perform poorly due
estimation risk, which is the substantial uncertainty present in
estimating and forecasting expected returns.

There are IShares ETFs for most asset classes, with daily prices (and
indirectly returns) available from Yahoo Finance. Someone who has some
knowledge of programming, econometrics, and optimization could probably
implement the strategies in the paper.

headless baby).
Flour
oil
onions
bell peppers
garlic salt, pepper, etc.
3 cups chicken stock
2 sticks butter
3 tablespoons oil

First stuff the heads, or make the patties (see index)
then fry or bake.
Set aside to drain on paper towels.
Make a roux with butter, oil and flour,
brown vegetables in the roux, then add chicken stock and
allow to simmer for 20 minutes.
Add the patties or stuffed heads, and some loose crawfish,
lobster, long piglet, or what have you.
Cook on low for 15 minutes, then allow it to set for at least
15 minutes more.
Serve over steamed rice; this dish is very impressive!



Stuffed Cabbage Rolls

Babies really can be found under a cabbage leaf -
or one can arrange for ground beef to be found there instead.

8 large cabbage leaves
1 lb. lean ground newborn human filets, or ground chuck
Onions
peppers
celery
garlic
soy sauce
salt pepper, etc
Olive oil
breadcrumbs
Tomato Gravy (see index)

Boil the cabbage leaves for 2 minutes to soften.
In skillet, brown the meat in a little olive oil,
then add onions, peppers, and celery (all chopped finely)
and season well.
Place in a large bowl and cool.
Add seasoned breadcrumbs and a little of the tomato gravy,
enough to make the mixture pliable.
Divide the stuffing among the cabbage leaves then roll.
Place seam down in a baking pan.
Ladle tomato gravy on top,
and bake at 325° for 30 - 45 minutes.



Umbilical Cordon Bleu

Nothing is so beautiful as the bond between mother and child,
so why not consume it?
Children or chicken breasts will work wonderfully also.

4 whole umbilical chords (or baby breasts, or chicken breasts)
4 thin slices of smoked ham, and Gruyere cheese
Flour
eggwash (milk and eggs)
seasoned bread crumbs
1 onion
minced
salt
pepper
butter
olive oil

Pound the breasts flat (parboil first if using umbilical
cords so they won

paprika

Remove the silverskin by loosening from the edges,
then stripping off.
Season generously, rubbing the mixture into the baby?s flesh.
Place 1 quart water in a baking pan, the meat on a wire rack.
Bake uncovered in 250° oven for 1½ hours.
When browned, remove and glaze,
return to oven and bake 20 minutes more to form a glaze.
Cut ribs into individual pieces and serve with extra sauce.



Fresh Sausage

If it becomes necessary to hide the fact that you are eating
human babies, this is the perfect solution.
But if you are still paranoid, you can substitute pork butt.

5 lb. lean chuck roast
3 lb. prime baby butt
2 tablespoons each:
salt
black, white and cayenne peppers
celery salt
garlic powder
parsley flakes
brown sugar
1 teaspoon sage
2 onions
6 cloves garlic
bunch green onions, chopped

Cut the children?s butts and the beef roast into pieces
that will fit in the grinder.
Run the meat through using a 3/16 grinding plate.
Add garlic, onions and seasoning then mix well.
Add just enough water for a smooth consistency, then mix again.
Form the sausage mixture into patties or stuff into natural casings.



Stillborn Stew

By definition, this meat cannot be had altogether fresh,
but have the lifeless unfortunate available immediately after delivery,
or use high quality beef or pork roasts (it is cheaper and better to
cut up a whole roast than to buy stew meat).

1 stillbirth, de-boned and cubed
¼ cup vegetable oil
2 large onions
bell pepper
celery
garlic
½ cup red wine
3 Irish potatoes
2 large carrots

This is a simple classic stew that makes natural gravy,
thus it does not have to be thickened.
Brown the meat quickly in very hot oil, remove and set aside.
Brown the onions, celery, pepper and garlic.
De-glaz

beef (buy steak or roast, do not pre-boil).

Pie crust (see index)
Whole fresh pre-mie; eviscerated, head, hands and feet removed
Onions, bell pepper, celery
½ cup wine
Root vegetables of choice (turnips, carrots, potatoes, etc) cubed

Make a crust from scratch - or go shamefully to the frozen food section
of your favorite grocery and select 2 high quality pie crusts (you
will need one for the top also).
Boil the prepared delicacy until the meat starts to come off the bones.
Remove, de-bone and cube; continue to reduce the broth.
Brown the onions, peppers and celery.
Add the meat then season, continue browning.
De-glaze with sherry, add the reduced broth.
Finally, put in the root vegetables and simmer for 15 minutes.
Allow to cool slightly.
Place the pie pan in 375 degree oven for a few minutes so bottom crust is not soggy,
reduce oven to 325.
Fill the pie with stew, place top crust and with a fork, seal the crusts together
then poke holes in top.
Return to oven and bake for 30 minutes, or until pie crust is golden brown.



Sudden Infant Death Soup

SIDS: delicious in winter, comparable to old fashioned Beef and Vegetable Soup.
Its free, you can sell the crib, baby clothes, toys, stroller... and so easy to
procure if such a lucky find is at hand (just pick him up from the crib and
he?

green onions, and parsley.
Place roast on top with fat side up.
Place uncovered in 500° oven for 20 minutes, reduce oven to 325°.
Bake till medium rare (150°) and let roast rest.
Pour stock over onions and drippings, carve the meat and
place the slices in the au jus.



Bisque à l?Enfant

Honor the memory of Grandma with this dish by utilizing her good
silver soup tureen and her great grandchildren (crawfish, crab or
lobster will work just as well, however this dish is classically
made with crawfish).

Stuffed infant heads, stuffed crawfish heads, stuffed crab or lobster shells;
make patties if shell or head is not available
(such as with packaged crawfish, crab, or headless baby).
Flour
oil
onions
bell peppers
garlic salt, pepper, etc.
3 cups chicken stock
2 sticks butter
3 tablespoons oil

First stuff the heads, or make the patties (see index)
then fry or bake.
Set aside to drain on paper towels.
Make a roux with butter, oil and flour,
brown vegetables in the roux, then add chicken stock and
allow to simmer for 20 minutes.
Add the patties or stuffed heads, and some loose crawfish,
lobs

- quote -

A good bond portfolio is essential. Bonds are truly a leveraged
investment.

- quote -

The SP500 has a risk grade of about 100. Normally, the SPX has a
historical standard deviation of about 15%-20%. The risk grade has
been normalized to 100 for the sake of simplicity (to make it easier
to compare to other investments). If xyz has a risk grade of 50, then
it is half as volatile as the index. Conversely, if xyz has a risk
grade of 150, then it is 1.5 times a volatile as the index.

- quote -

Keep in mind that the above references are general. Unless you can
replicate those portfolio's, the statistics are useless.

TB wrote:
- quote -

So this brings up an interesting question. Is asset allocation worth
doing at all? OK, don't kill me, I'm not saying that diversification is
not worthwhile. When I say "asset allocation" I mean the process of
saying "x% of my money should be in asset class A, y% in class B..."
Often this is done via an MVO. But the percentages that the MVO
generates are highly dependent on the inputs, and the inputs are
arguable. What should my sample size be? What should my sample
interval be? How coarse or fine do I break up my asset classes? What
benchmarks do I use for those asset classes? What happens if I invest
in individual securities and not in broad-based mutual funds?

If I go to ten different FPs, I'll get ten different answers on what my
asset allocation should be. All of them could be valid given their
input assumptions. For example, an FP worked up an investment plan for
me. In response to questions about my current financial situation, my
financial goals, and 5 questions assessing my risk tolerance, he was
able to generate a target asset allocation accurate to two decimal
places (e.g. I should have 30.91% of my assets in large-cap value
equities). Obviously, my buddy didn't do some kind of advanced analysis
to come up with this, this is just the raw output of the MVO.

But let's assume that I get the "right" input data and assumptions. I
can get an efficient frontier and find the tangent with my set of
risk/reward utility functions to determine my target asset allocation.
Will the assumptions this model reasonably hold into the future (if they
hold for asset classes, will they hold for individual securities?)? Do
I need to refigure my target asset allocation every time period (year,
six months, whatever) based on the latest data? Why should I believe
that a calculated asset allocation is better than the one that I have by
random chance (as long as I am reasonably diversified across asset classes)?

Thanks!
-Will

Fetch wrote:

- quote -

That's very possible. I am the rankest of beginners. I have recently read
another
book by Zvi Bodie ("Worry-Free Investing", coauthored with Michael J
Clowes) & found it very easy to read and understand. It's an interesting
idea the two authors put forward in this book--that the way arrive at
your
financial goal is likely to be U.S. Gov't inflation adjusted bonds--but
that's
another post--or will be if I feel like staying up late tonight.) I will
check at the
library and see if they have this other book.

- quote -

Could you say that again in words of one syllable? Specifically, what
does
it mean that the statistics are "scaled or normalized"?

- quote -

I am sure it is! However, by googling "correlation matrix" + "asset
class",
I did finally locate the sort of tables I was looking for, at a number of

different places, including

http//www.calpers.ca.gov/eip-docs/about/
board-cal-agenda/agendas/invest/200412/item04-03-02.pdf

and

http://www.efmoody.com/investments/correlation.html

I just wanted the data to "play with" & investigate some sample
portfolios
I've read about. Now to see if I can get my spreadsheet to cooperate.

Karen

Robin Mcleod wrote:
- quote -

Would you still want it if I told you that these numbers change
constantly, to the point where it's questionable whether cranking them
through software has much value?

-Tad

On Thu, 23 Dec 2004 05:50:09 CST, Karen Younge
<karenyounge[at]earthlink.net> wrote:

- quote -

Basically, you're receiving raw price data from yahoo - not
statistics. For example: http://finance.yahoo.com/q/hp?s=XTO

Repeat for each stock and then perform the required calc's. It's
pretty tedious and is something I don't normally do.

- quote -

I don't think you fully understand what you're asking for - I suggest
that you read Zvi Bodie "Investments" to gain a better understanding
of portfolio theory.

Riskgrades are JP Morgan's supposedly *improved* version of standard
deviation and related statistics. The riskgrade is basically a scaled
or *normalized* statistic and when combined with other assets, will
normally show a reduction in portfolio variance - hence, correlation.

It is easier to use riskgrades than to pull data from yahoo and do the
calc's yourself. However, you will have to read their methodology to
understand how to make riskgrades work for you.