Thanks for opening this board. After extensive research and reading I had about convinced myself of the validity of Strategic Global Asset Allocation. Now I think I will wait awhile before changing my portfolio. I would welcome any comments pro or con on MPT and the Efficient Frontier ideas. I am retired and have allocated my portfolio into a 30% MM and St Bond Fd and 70% in stock mutual funds. I have almost no foreign exposure. Lg Growth is the dominant portion. This idea has served well for the past 5 years, with the MM and St Bond Fd providing approx. 5 years of estimated income needs should there be a large market correction. At least I shouldnt have to sell depressed stocks when the correction comes. The idea of all equities US and foreign moving together seems to have some validity but I am sure they do not perfectly correlate. It is tough to decide on what is correct. All responses appreciated.Dawg on!
I'm skeptical about both MPT and Efficient Frontier. Both depend on unknowable assumptions about the future.Efficient Frontier assumes that correlations that were true in the past will continue to be true in the future. However, there are certain points of history where correlations have changed drastically (bonds vs. stocks for instance).Efficient Frontier says that it minimizes "risk"--but in what form? My portfolio is not static; as a young saver/investor I'm constantly adding money to it & buying stuff with dollar cost averaging. So I *want* assets that I'm buying to fluctuate, so that I have a good chance of driving down my average purchase price. I don't care at all about how my portfolio as a whole functions in the short term.For some people, Strategic Global Asset Allocation might meet their needs; for others Efficient Frontier is the ticket.Personally, I still haven't seen a compelling argument as to why I should be looking at risk-adjusted returns.
MPT is bunk.first off, MPT relies on linear mathematics and normal or Gaussian statistics. i don't care what aspect of the market you look at, market behavior does not fit simple Gaussian assumptions. MPT assumes that prices change in a random walk - that is also not true as the market is complex adaptive. on one extreme you have total order which is completly predictable. at the other extreme you have choas, which is completely unpredictable. in between lies complexity, and that is where the market is at - it is neither chaotic nor ordered - it is complex. like all complex systems it is partially predictable but never completely predictable and that is what turns market participants into mad hatters - the partial predictability. the efficient frontier - as defined by MPT - is not stable and likely to become more unstable as time marches on. bloomberg has had the hurst coefficient on their charts for several years now, the hurst coefficient is essentially an indication of connectiveness of price moves or the relation of one day's (or week or month or year) price move verses it's past price moves. if the movements are totally a random walk - then the hurst coefficient should be 0.5. but if you look at a lot of data you will see that the hurst coefficient- for either individual stocks or for indices is rarely even close to 0.5.MPT in my mind is akin to the Ptolemaic universe where everything revolved around the earth. it was an elegant theory and on the face of things, prior to telescopes, it seemed to "make sense". but once people started actually making measurements with telescopes and looking at the data as Copernicus did, it was pretty obvious that the sun did not revolve around the earth. MPT in the early 21st century is a similar case. the data just does not fit the model. however, there will always be holdouts - so i'm sure that there will be people talking about beta 100 hundred years from now.as far as the correlation between the direction (bull or bear) of foreign markets and our own, once again if you look at the data, really look at it, parse it up into different times frames etc, instead of just throwing it all into one pot and running statistical linear regression stats at it, you will see how the correlation has consistently become more positive with time the past 30 years. funny how that works - it wasn't until the 1960's that private US business investment began to go overseas in any meaningful way - and guess what? as that private business investment outside of the US grew and foreign investment into the US grew - the correlation between markets grew. not to sound glib but - duh.that's not to say that foreign investments are unsound, or do not pay enough to assume the currency risk - i think that we have a bad tendency in this country to assume that the US dollar will always remain the world's standard - so foreign investment is attractive if there is more growth overseas, more value to be attained, and as a hedge against a weaker dollar. but as some kind of alchemy mix to hedge standard deviation of return - that's useless imho.tr
solaris writes;first off, MPT relies on linear mathematics and normal or Gaussian statistics. i don't care what aspect of the market you look at, market behavior does not fit simple Gaussian assumptions. MPT assumes that prices change in a random walk - that is also not true as the market is complex adaptive. [snip]I couldn't agree more - it would be interesting to look at somesort of non-linear correlation coefficient where you could break down the correlation between markets - for example based on the size of price changes, where you'd probably find much higher correlations for large & sudden price movements (say >2%) than for small ones.I think that for the markets to get better analysis someone needs to teach the economists some more math (or maybe the reduced number of math PhD's not going to the NSA now will help somewhat). I think that until you have experience with working on (or modelling) complex/non-linear systems it can be hard to grasp how counter-intutive they can be.On the Berkshire board a couple of us were trying to explain a while ago to a poster why we thought that a couple of big catastrophies (which might cost the company $200-500m) would be the best thing that could happen to the company - simply because the insurance/reinsurance markets would adapt to the events, eliminating capital, rasing prices, etc. Not very linear, but perhaps not a bad model for the industry. The counter arguement was gut instinct :( Hopefully not by engineers :)peter xyz
peter xyz,have you read chris may's book Non-linear Pricing ?expensive but definitely worth the price imho.regarding super-cat, i agree with your general assumptions. past history indicates that large super-cat events improve pricing in the subsequent years. the problem with super-cat as an investor (fyi i am an owner of BRK class b shares) is that even with a longer term horizon, say 10, 20 years or even 30 years, there is a not insignificant probability that super-cat events can cluster within those time frames.interesting.as far as normalizing price structure volatility, in risk arbitrage i have been adpating some techniques developed in the currency markets by the folks at Olsen in Zurich - http://www.olsen.ch/library/research/oa_working.htmltheir focus of course is very short term, but i think that some of this stuff is very scaleable.tr
solasis,I havn't read Non-Linear Pricing, thanks for the tip - most of my non-linear experience was in an engineering/physics environment.As another Berkshire B owner, I'm not too sure that even a cluster of super-cat events would be such a bad thing if reasonable premiums were able to be calculated (big if i know) even after a substantial lag. These sort of events highlight reinsurer credit-risk and have the potential to pull a lot of capital out of the markets. That and there is quite a lot of fixed-demand in cat (rather than super-cat maybe) insurance - which if anything is likely to go up (interestingly Jack Byrne of GEICO fame is out of retirement at WTM - poss. worth a look if you follow insurers - and trading on insurers desire to get out of "ugly" businesses).peter xyz
Best Of |
Favorites & Replies |
Start a New Board |
My Fool |