I had one quick question for the smart AF members. At this time I am rebalancing my Roth and wish to diversify 10% into a REIT Fund (VNQ) and into a Commodities Fund (DJP). At this time my Roth is allocated thus far; 40% VTI - Total Stock Market Fund 20% IJR - Small Cap Blend 15% BJBGX - Total Bond Market Fund 15% VO - Mid Cap Blend 10% VEU - Total International Market Fund My plan is to back off some of the funds in IJR and BJBGX and place them towards the REIT and Commodity funds. While diversification is preached in every investing manual, I have also learned about diworseification which was coined by Warren Buffet. “However, there’s a fine line to be drawn. Too many investors suffer from what I call diworseification. Their money is spread amongst so many holdings that they’re no longer investing; they’re collecting. Their account statement reads like a history of the mutual fund. Every year’s favourite makes an appearance. Because their money is spread so widely, appreciation in any one holding barely impacts the value of the overall portfolio” WB; http://www.claritywealth.com/ca_05_text.html My question for you is would you see my holdings become too spread out among too many different ETF index products? While I’m careful to diversify, now I’m worried about over doing it. Thank you in advance for your insight!
VTI 35.10% VEU 19.20% IJR 12.90% BJBGX 8.30% DJP 8.20% VNQ 8.20% VO 8.10% After my contribution/rebalance portfolio weightings.
Are you just pulling those allocation numbers out of thin air? You may want to try a real mean variance optimizer. I spent a couple of days trying out a bunch of ones off of the internet and settled on one called “Visual MVO.” It rocks. I think it cost $50-$75. I’ll try to see if I can give you a copy of mine since you hooked me up. I"m traveling now so I can’t really check to see if it’s even copy-able. Last summer, I had a small stash of cash that I wanted to keep separate from my trading account. I decided to passively invest using mean variance optimizer. I decided just to hold my nose (deworsification) and do what the optimizer said. I put in historical returns for long bonds, cash, small caps, large caps, foreign large caps, foreign bonds, and a few other asset classes. I executed the plan it set forth, which, to my dislike included a huge portion of long bonds. At the time, I believe long bonds were yielding a solid 4%. Anyway, the portfolio cruised along for a while, not doing much, and the long bonds started rallying. I couldn’t sit on my hands. I had to sell those damn long bonds. TLT is what I was using as a proxy for long bonds. I sold them at $93. Most of the equities were in EFA (international large caps). EFA tanked, TLT continued to rally. Long story short… had I stuck with the original allocation and rebalanced after 6 months I’d be off only a few percent (as opposed to be down 25% like I am now in that account). The other thing I learned was that a lot of other asset classes don’t really improve your returns or lower your variance as much as you’d hope. Their effect seemed trivial. The long bond is the grand master portfolio tool and it performed stunningly well in this bear market. I still can’t stomach the idea of buying it now but I am respectful of it’s effect in a portfolio. In a way you have to just forget that buying a long bond now will guarantee you at most a 2.5% ytm. You have to just say that your return OVERALL will be insanely high IF (and ONLY if) you have the cajones to rebalance into equities when everybody else goes running for the long bond. Then… IT WORKS.
Pulling numbers out of thin air. Kinda. My Roth is ~$8k due to the market losing so much value. I have contributed since 2006 the full allowable amount. I’d be very interested in a portfolio optimization program. I was planning on investing the $5k this year as follows. $1.5k VTI $1.5k VEU $1k VNQ $1k DJP As you can see, I’m rather new to portfolio management. I graduated college in 2006 opened a Roth, and just started buying various index funds per the buy and hold strategy I was taught. I only tinker with the weightings on the buy ins which occurs annually. If you have a suggestion for lowering the variance of my portfolio I’d be very open to that.
i run MV optimizers everyday for institutional clients… and believe me… it’s all hokus pokus cr4p. your results are only as good as the inputs you put it: i.e., for beta level (asset-class-only studies), the most important thing is your input assumptions such as returns and std for each asset class and the timeframe for those projected assumptions (i.e., 5 yrs out? 10yrs? 20yrs? – most assumptions are for longer periods). correlation is boogus too if you ask me. it’s definitely a nonstationary element which is subjected to multiple revisions through time. everything is purely speculative, which exhibit little or no stability in the short- to mid-term periods. also, you’ll have to throw in all kinds of constraints in order for your results to looks somewhat reasonable and executable, else your optimizer will put like everything in alternatives (given that a MV-optimizer underestimates, or rather, completely avoids, skewness and kurtosis). even with that said, it is still nice to produce a frontier to look at. it kind of gives you a sense to guide you through the process of allocation and/or risk budgeting… or it can bias you to confirm your original beliefs of what’s optimal for your portfolio. nonetheless, i don’t believe much in diworsefication, especially when you don’t even have beta exposure to REITs nor commodities. how does that diworsify your portfolio? i say throw those two in there in a meaningful way (>5% each). just my midnight-taco-bell-craved + 2-penny rambling
Thank you. I’m thinking back to the Pertrac and Zephyr model I used to use at the firm I worked at. It was basically a plug for Excel to create portfolio optimization. While I don’t think my retirement portfolio allocation is bad thus far, I think I should get a more systematic approach than just throwing darts at the allocation levels. Now I’m trolling for free software on download.com to see if I can find a decent spreadsheet macro that will allow me to import security profiles and create a optimal portfolio on the efficient frontier.
well good night and good luck. i should start thinking about rebalancing my roth as well as my regular IRA.
If you convert a rollover IRA to a regular roth account, you will have to pay taxes at the end of the year for the amount rolled over? It is taxed at the Federal rate? What is going to happen 2010?
95% sure you pay it like its income in the year it is rolled. There was a limit on ability to convert (if you make over 100k you cant do it I think) that is going away either this or next year, so the window would be open. And yep, taxed just like its income to you…great thing to do if you have a year that you work PT at your local YMCA…
adalfu, good points. Here’s my take. A MV optimizer is a useful data point. It’s probably better than a subjective guestimate by itself, but probably you will want to use a bit of both together, because: 1) traditional MVO is just backwards looking… it more or less assumes the future will look like the past. If you think the future is going to be different (US declining, China rising, whatever), then you’ll need something more complex, maybe Black-Litterman or something. 2) regular MVO assumes that you know the “true” average return and the “true” volatilities of assets, when in fact these are estimates. Small changes in the estimates can lead to very large changes in portfolio composition, so that means you’ll want to do something like have a resampled efficient frontier (where you vary your estimates a bit and come up with an average efficient frontier to reduce the effect of estimation error). 3) I recall reading some studies that suggest that the optimal portfolio’s sharpe ratio isn’t hugely different than many equal weighted and rule-of-thumb allocations (though small differences can still compound over 20-30 years).
tvPM Wrote: ------------------------------------------------------- > 95% sure you pay it like its income in the year it > is rolled. There was a limit on ability to convert > (if you make over 100k you cant do it I think) > that is going away either this or next year, so > the window would be open. And yep, taxed just like > its income to you…great thing to do if you have > a year that you work PT at your local YMCA… I’ve spoken to this before but you do pay tax as if it were income in the year you roll- so now is a good time since the MV had dropped so much (saving you money). If the MV drops more over the course of the year you can "un-convert’ it and do it later to lower the tax burden. See some fine print here: http://www.fairmark.com/rothira/recon.htm I plan to convert when I take a year off to travel or do MBA when income is nil anyway.
If you don’t understand a MVO backwards and forwards (including all of its assumptions) then you probably shouldn’t be using one. They can output some dangerous stuff!
Level 3 is all about portfolio management correct? I may have to wait and see for myself when that time comes.
bchadwick, good points as well, made me think back about a project i worked on w/ the Michaud Optimizer bchadwick Wrote: ------------------------------------------------------- > adalfu, good points. Here’s my take. > > > A MV optimizer is a useful data point. It’s > probably better than a subjective guestimate by > itself, but probably you will want to use a bit of > both together, because: > > 1) traditional MVO is just backwards looking… it > more or less assumes the future will look like the > past. If you think the future is going to be > different (US declining, China rising, whatever), > then you’ll need something more complex, maybe > Black-Litterman or something. even black-litterman (as well as traditional optimizers – i.e., markowitz optimizers) suffers from point estimates. they all work the same way: taking a snapshot of the return and correlation matrix, which generates an estimate bias/risk. that is why i prefer your #2: > 2) regular MVO assumes that you know the “true” > average return and the “true” volatilities of > assets, when in fact these are estimates. Small > changes in the estimates can lead to very large > changes in portfolio composition, so that means > you’ll want to do something like have a resampled > efficient frontier (where you vary your estimates > a bit and come up with an average efficient > frontier to reduce the effect of estimation > error). this is exactly the Michaud Optimizer, which takes into account estimation risk by using a bootstrapping algorithm (or monte carlo simulation) to generate a large sample of frontiers. the “robust” frontier will simply be some methodology of averaging all the samples. it’s intuitive and plausible, but then again, i don’t recall any mathematical justification for this approach. couple of advantages for this approach: you’ll have 1) more diversified portfolios 2) smoothing effect for allocations; i.e., when you shift along the frontier, you won’t see a dramatic change in the allocation to each asset class. the traditional models will, for example, make a large jump from large cap to small cap even when you move couple of points along the frontier. very unstable and not very practical effect for an investor. 3) eliminates estimation risk of a “snapshot” frontier > 3) I recall reading some studies that suggest that > the optimal portfolio’s sharpe ratio isn’t hugely > different than many equal weighted and > rule-of-thumb allocations (though small > differences can still compound over 20-30 years). Lastly, i just want to comment on other possibilities such as instead of modeling through the traditional MVO, you can try using mean-VaR (or any form of it, mean-modified VaR or mean-conditional VaR) optimization. or a mean-expected shortfall frontiers…
To jack my own thread, how do many of you decide how to allocate your portfolios. This is scary that I need to ask, but it looks like I know what my next book at the library will need to be.
My portfolio is a bit small to use an optimizer because share sizes and odd lots will still affect it. The truth, I’m embarassed to say, is that I am using a lot of rule of thumb and seat of the pants allocation - I must change this this year. Part of this is because I am using a global macro approach, which is more of a trading approach than an investing approach. My Roth is more of a traditional 60-40 portfolio implemented with ETFs with a bit overweight in emerging markets. I have been looking at the Treynor-Black model lately. Although I supposedly learned it at L2, I’ve come to appreciate it more lately as a way to combine the mostly efficient market portfolio with some limited opportunities to spot alpha. I am trying to figure out how fundamental analysts do portfolio construction. Once you’ve decided that a stock is overpriced or underpriced, how much of the portfolio should it form? You can throw it into an optimizer, maybe. Or maybe you simply compute the total return assuming that the stock will converge to intrinsic value in a year and compute expected return based on that, and using historical volatilities and correlations. I’ve come to percieve that factor models of expected returns can’t really tell you if a stock is overpriced or not; they can just tell you whether your portfolio should have more or less of it than you have now. If you have estimates of future earnings and a terminal value, you can discount them at the factor model’s rate of return and find over/underpriced stuff that way.