[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference nyoss1::market_investing

Title:Market Investing
Moderator:2155::michaud
Created:Thu Jan 23 1992
Last Modified:Thu Jun 05 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:1060
Total number of notes:10477

474.0. "Asset Allocation/Portfolio Management" by NOVA::FINNERTY (Sell high, buy low) Wed May 19 1993 14:44

    
    There have been a number of notes asking about how to create a
    portfolio and allocate the assets between k different funds.  For the
    quantitatively inclined, there's a mathematical solution to this
    problem that was developed by Harry Markowitz (in 1952) that forms the
    basis of modern portfolio theory.
    
    The end result of this theory surprises many people.  Some people
    reason as follows: 
    
    	I'm in my 30's/40's, I have a long time to go before retirement,
    	so I might as well put all my money into growth funds.  I'll buy
    	and hold, and by the time I retire I'll get the maximum return.
    
    The theory says that this reasoning is flawed.  Putting 100% of your
    portfolio into a growth fund has some expected return E and some
    standard deviation (risk) s.  It is highly likely that there exists
    a more diversified portfolio with the same return E but with standard
    deviation s' < s.  There are also portfolios with the same s but with
    higher expected return (which may involve borrowing or short selling).
    
    The inputs to the algorithm are the estimated returns on each asset
    class (typically stocks, bonds, and bills) along with standard
    deviations and correlations.  An additional input is a personal risk
    aversion factor that determines how much of the riskless asset (e.g. T
    Bills) to mix with the optimal risky portfolio (e.g. stocks and bonds).
    
    It is curious that this algorithm is not more widely known among
    individual investors; my guess would be that it is used by 100% of 
    mutual fund managers or their staff.
    
    A good reference is "Investments" by Bodie and Marcus.
T.RTitleUserPersonal
Name
DateLines
474.1I'm not so sureKYOA::BOYLEDirty Jobs Done Dirt CheapWed May 19 1993 15:0314
    I'm not sure that very many people still pay much credance to Modern
    Portfolio theory.  It requires some information about the future
    performance of assets and their inherent betas (return variability as
    referenced to some index).  This information is NOT easily obtainable.
    
    Secondly, much of the theory is based on short selling assets, again
    based on their relationship to some index.  Meaning that you ANTICIPATE
    the movement of indices like the S&P 500, etc. which no one can do
    accurately.  
    
    When all is said, for the long term invest in stocks and stock based
    securities. 
    
    Jack Boyle
474.2The FunnelTLE::JBISHOPWed May 19 1993 15:2334
    My personal objection is formed by my own experience in the
    inflation of the '70s: T-bills are _not_ riskless in real terms
    even if they are riskless in nominal terms.
    
    Can your model be modified to include inflation?  I'd like to use
    it if it can.
    
    I've been influenced by a much less sophisticated model, the "funnel"
    model: you think of a portfolio as filling a conical funnel with money.
    As you fill it, you move from less-risky kinds of investments to more
    risky ones.  Each level of risk takes more money to fill up, and you
    don't start investing in a riskier level until this one is full (this
    is a rule I don't entirely agree with, by the way).  You can adjust 
    the angle and the scale to fit personal risk preferences (on an intuitive
    basis?), e.g. for a "typical" person:
    
    \          /  Aggressive Growth, Global funds (four years' income)
     \        /   Growth Funds (two years' income)
      \      /    Balanced or Index or Bond Funds (one year's income)
       \    /     Money Markets (six months' income)
        \  /      Checking (one month's income)
         \/       Cash (one week's income)
    
    What goes "above" this depends on how much people want to do individual
    investing.  There aren't many institutional funds with very high betas,
    and I think hedge funds (which would fit, risk-wise) have much higher 
    minimum requirements than most typical people can meet, while managed
    commodity future funds and the like have poor records.
    
    Personally, I like the idea of adding a little bit of the higher-risk
    stuff even at lower levels; I also tend to leave out the many levels
    of bonds which most explications of the "funnel" include.
    
    		-John Bishop
474.3moreNOVA::FINNERTYSell high, buy lowWed May 19 1993 15:5241
    
    re: .1
    
    >> I'm not sure that very many people still pay much credance to Modern
    >> Portfolio theory.  It requires some information about the future
    >> performance of assets and their inherent betas (return variability
    >> referenced to some index).  This information is NOT easily
    >> obtainable.
    
    Betas may be easily estimated from past price movements if the fund or
    stock, etc. has been around for a few years.
    
    Sure, using the model requires estimating future returns, but then so
    does _not using_ the model.  If you truly had no information about
    future returns, then your estimate of expected returns of all assets
    would be identical.  In this case, you'd have no incentive to invest in
    any security that had any risk (since virtually riskless/low yield 
    investments are available).
    
    Whether your estimates of future returns are explicit or not, your
    portfolio selection makes sense only insofar as it is consistent with
    your expectations of the future and of your estimates of the
    correlations between the returns of the items in your portfolio.
    
    >> much of the theory is based on short selling assets
    
    this isn't true in general.  it is true that if you set your desired
    return high enough, the only way to do it will be by borrowing money
    and/or short selling... but this is a function of the individual's 
    aversion to risk and not a necessary part of any portfolio.
    
    >> When all is said, for the long term invest in stocks and stock based
    >> securities.
    
    the model makes no comment about the investment time horizon, and includes
    both traders and long term investors.
    
    as for sticking exclusively with stocks (i.e. not diversifying your
    portfolio into bonds, international stocks, real estate, or even cash),
    then I think you're on shaky ground.
    
474.4VMSDEV::HALLYBFish have no concept of fireWed May 19 1993 17:2923
>    When all is said, for the long term invest in stocks and stock based
>    securities. 

    Spoken like a true mutual fund salesman.

    Just how long is the long term?  The Roman Empire is dust, shares in
    ancient products and services are worthless, only gold coins carry
    any objective value today.  Collectibles from the era would be worth
    a lot of money were there a liquid market which there isn't.

    The same story can be said for millions of companies over thousands of
    years; eventually all companies go bankrupt, eventually all empires
    fall and their assets are taken over by conquerers and creditors.
    Buying stock is a guaranteed way of going broke over "the long term".
    I say this on a day the DOW closes at an all-time high, 3500.
    
    Portfolio theory requires one to work with insufficient information,
    while a "buy and hold" approach requires less thinking.  Sometimes the
    simplest ideas are the best, to be sure.  But a simple-minded approach
    does not eliminate risk, it simply ignores it.  If you ignore risk,
    risk will come to you.  Portfolio theory attempts to manage this risk.
        
      John
474.5let pi be the rate of change of beta...TLE::JBISHOPWed May 19 1993 17:5333
    I think the offical objection to alphas and betas is that they aren't
    constant over time, so any beta is only a guess.  While it's better
    than nothing and we have to use something, as noted, I suspect that
    replacing "2.01506" with "2 plus or minus 1" makes a big difference
    in how well the algorithm works.  Here the time horizon makes a
    difference: we can probably assume that betas, etc. change over time,
    so that the historical "2.01506" is within .01 for this week, within
    .1 for this year, within 1.0 for the next decade and so on.
    
    I know of a large portfolio operated on an interesting principle which
    I wish was available to me without doing my own stock picking: they
    have several "funds", some aggressive, some less so.  When a
    corporation stops being an aggressive growth candidate, stock in it
    is transfered (free of any taxable events) into the "growth" pool;
    similiarly stocks are moved from "growth" to "income".  Only if a 
    stock is judged undesirable for any pool is it sold.   This reduces
    turnover and expenses.
    
    Alas, in my case, if 20th Century Growth sells something, I have a
    taxable event and can't somehow have that stock moved to my 20th
    Century Select account!
    
    Fidelity has a nice feature in their directed dividends--you can have
    your money-market fund use its dividends to buy shares in an equity 
    fund and vice versa.  I use this in my IRA account there to do easy
    dollar-cost averaging.  Given a "funnel" it's easy to imagine having
    the income from level "x" used to invest in the level one higher. 
    Other than a purely-Fidelity solution, I don't know of any way to do
    this easily (I'm a big believer in automatic investment).
    
    I feel another topic coming, on investment styles and life experiences!
    
    		-John Bishop
474.6NOVA::FINNERTYSell high, buy lowWed May 19 1993 18:0714
    
    re: stability of alpha and beta
    
    alphas don't appear to be either very stable or very statistically
    significant to begin with.
    
    betas, on the other hand, seem to be "reasonably" stable, at least
    according to one extensive study.  Stocks were divided into 10
    categories according to the beta calculated during each time period,
    and the category boundaries were defined in such a way that each
    category had approximately the same number of stocks.  40-70% of betas
    were stable +/- one category for the subsequent 5 year period.  The
    study looked at stocks over a 30 year period.
     
474.7As stable as home addresses, roughlyTLE::JBISHOPWed May 19 1993 20:319
    re .6
    
    This means that 60-30% moved from one decile to another in five
    years, which doesn't sound "very stable", particularly if we are
    trying to set up a portfolio for the next several decades.
    
    How sensitive is your algorithm to such variation?
    
    		-John Bishop
474.8rebalancingNOVA::FINNERTYSell high, buy lowThu May 20 1993 10:0327
    
    re: -.1
    
        the algorithm I have is just the standard Markowitz algorithm... I
        didn't even write it myself.
    
    re: sensitivity to betas
    
        well, that depends on what the assets are.  If you're investing in
        a stock index fund and, say, a U.S. bond fund and U.S. T-Bills,
        then there's no problem.  The beta of the index fund is supposed to
        be managed to maintain beta = 1.0 at all times.  In this case the
        stability of betas is not an issue.
    
        if the assets are individual stocks, then you'd probably want to
        rebalance your portfolio periodically (quarterly, say).  If you
        _don't_ rebalance, then your portfolio will become increasingly
        overweighted in whatever stocks happened to do well in the early
        periods, and the portfolio beta will likely change.  So even if
        you're a buy-and-hold'er, you can still adjust the amount of your
        holdings to maintain balance.  obviously there's a tradeoff here
        between transaction costs and the value of risk redution, so you
        can take this too far.  Quarterly rebalancing is probably 
        reasonable unless your transaction costs are very high.
    
    /jim