Variable portfolio based on "Timing" the 4 asset classes.

[wizozz]

One of the basic assumptions of "permanent portfolio" is, you cannot predict the future and time the markets.

However, when I look at the long term historical analysis of the permanent portfolio funds, and 4 main assets, it is obvious that there is one very good performing asset each year, and at least one very poor performing asset. And generally, the best performing asset class is performing best several years in a row. It seems clear that global economy does not shift from one phase to other (recession, depression, inflation, prosperity) several times within a year or so.

So, why don't we choose a simple indicator, like 9-month simple moving average, and make monthly decisions about which of the 4 asset classes are doing best or worst, and take the 25% from the poorest performing asset, and invest %50 on the highest growing asset? Has anyone made back tests to compare the growth and volatility of such a portfolio? Or even a portfolio investing %100 on top earning asset at any given time based on 9month sma?

I know it is harder to calculate (you need monthly values for over 40 years, and need conditional formulas etc), but has anyone did any "Variable Growth Portfolio", based on such a simple criteria?

And what do you guys think about such a strategy?

[Pointedstick]

Interesting. Sounds a bit like a "lite" version of the Decision Moose strategy.

[wizozz]

What is "Decision Moose" strategy?

Edit: I have found "http://www.decisionmoose.com/Home_Page.html" online. But it does not seem to have a published criteria for selecting which fund is best.

Any such -and simple- criterias for selecting and timing the best ETF? I'd rather be able to make my own "timing" decisions.

[wizozz]

I have found some more information about timing the markets using SMA's.

http://www.gleasonreport.com/documents/ ... ingavg.htm

Here it is if you are interested.

I'm now considering a PP of %75 or my money and a timed "decision goose" kind of portfolio, only based on one of 4 assets (probably using ETF's), and simple moving averages timing.

[AdamA]

So, why don't we choose a simple indicator, like 9-month simple moving average, and make monthly decisions about which of the 4 asset classes are doing best or worst, and take the 25% from the poorest performing asset, and invest %50 on the highest growing asset? Has anyone made back tests to compare the growth and volatility of such a portfolio? Or even a portfolio investing %100 on top earning asset at any given time based on 9month sma?

It's debatable whether or not these moving average strategies really work.

Even if they do, they are difficult to adhere to psychologically because you can get whipsawed in and out of some of these positions many times before having any success with them.

[Thomas Hoog]

No one ?
I will start this VP, some kind of Ivy timing strategy, however with the HB assets (although a bit modified).
Assets will be Stocks (VT), Gold (IAU), REIT (in €), LT Bonds (€). Sell/Buy signals will be the 200 days SMA (simple moving average) on a montly base. Not invested money wil be parked in short Bonds.

comments ?

[rickb]

Faber has written about using Ivy timing for the HB PP, see http://www.mebanefaber.com/2010/10/28/t ... bernstein/ .

[Thomas Hoog]

Faber has written about using Ivy timing for the HB PP, see http://www.mebanefaber.com/2010/10/28/t ... bernstein/ .

Thanks, I have already found that post. That's one of the reason I don't use cash in the allocation.
I will post the results

[sophie]

There have been several useful threads on this topic.  I was strongly considering trying this out with a portion of VP, but I was dissuaded by two things.  First, keeping up with the required calculations is kind of a pain, and for most of us it violates Rule #1:  put your major effort into your career.  Second, read the prediction-system debunking chapters in Harry Browne's "Why the Best-Laid Investment Plans Usually Go Wrong".  He gives lots of examples of how you can get screwed by timing based on moving averages.

[Stefan]

I think the starting point for this strategy should be the Mebane Faber paper:
http://papers.ssrn.com/sol3/papers.cfm? ... _id=962461

The next improvement should be to look at a risk parity allocation as oposed to equal weights.

And the third should be controlling the portfolio volatility.

However, each of these key steps require some programming skill and a programming platform...

[melveyr]

Cool to see this thread pop up. I did my own little analysis on this a while back... The data is from 1978 to mid 2012.



This timer functions on each individual asset class. If gold, stocks, and bonds are all above there 12 month averages (inflation adjusted and including dividends/interest) than it is just like a vanilla PP. However, if one of the asset classes goes below its average, than cash is held instead of that asset.

The backtest generated higher returns as you can see above as well as fewer drawdowns and lower standard deviation.

Vanilla PP monthly stdev: 2.16%
Timer PP monthly stdev: 1.71%



I am not sure how much faith I would have in this going forward and I doubt I would do it but it is interesting.

[D1984]

Cool to see this thread pop up. I did my own little analysis on this a while back... The data is from 1978 to mid 2012.



This timer functions on each individual asset class. If gold, stocks, and bonds are all above there 12 month averages (inflation adjusted and including dividends/interest) than it is just like a vanilla PP. However, if one of the asset classes goes below its average, than cash is held instead of that asset.

The backtest generated higher returns as you can see above as well as fewer drawdowns and lower standard deviation.

Vanilla PP monthly stdev: 2.16%
Timer PP monthly stdev: 1.71%



I am not sure how much faith I would have in this going forward and I doubt I would do it but it is interesting.

I'm just curious and this may be a stupid question, but how come it only goes back to 1978....why doesn't it go back to 1968 or 1972? It would be interesting to see how this handled stocks during the '73-74 crash, long term bonds during the sharp rate rise from '73 to '75, or gold during the 1975 recession.

[melveyr]

I'm just curious and this may be a stupid question, but how come it only goes back to 1978....why doesn't it go back to 1968 or 1972? It would be interesting to see how this handled stocks during the '73-74 crash, long term bonds during the sharp rate rise from '73 to '75, or gold during the 1975 recession.

Not a dumb question at all!

I don't have intra year data for LTT from before 1978. If anyone finds it somewhere PLEASE let me know.

[D1984]

I'm just curious and this may be a stupid question, but how come it only goes back to 1978....why doesn't it go back to 1968 or 1972? It would be interesting to see how this handled stocks during the '73-74 crash, long term bonds during the sharp rate rise from '73 to '75, or gold during the 1975 recession.

Not a dumb question at all!

I don't have intra year data for LTT from before 1978. If anyone finds it somewhere PLEASE let me know.

Doesn't CRSP have daily 30-year (or whichever bond was closest to 30-year) returns (including accrued interest) going back to 1961 and monthly going back to 1941? Does your school have access to their data?

[Stefan]

Good to see some backtest results.
I think this could be improved by:

1. Allocating based on risk on not equal weights
2. Putting more cash to work, up to the targeted portfolio volatility, when the Risk is On  .

[melveyr]

Doesn't CRSP have daily 30-year (or whichever bond was closest to 30-year) returns (including accrued interest) going back to 1961 and monthly going back to 1941? Does your school have access to their data?

I don't believe I do have access to that database. Again, if anyone ever finds and wouldn't mind sharing granular (even monthly would be great) data going back that far I would really appreciate it.

[D1984]

Doesn't CRSP have daily 30-year (or whichever bond was closest to 30-year) returns (including accrued interest) going back to 1961 and monthly going back to 1941? Does your school have access to their data?

I don't believe I do have access to that database. Again, if anyone ever finds and wouldn't mind sharing granular (even monthly would be great) data going back that far I would really appreciate it.

Shoot...that sucks. The nearest place to me that has CRSP data access at no charge is about 60 miles away (and you have to be a student there to access it). I do wonder, though, if there is a way to use Excel to calculate daily returns--including accrued interest--based on yield. if so, would it be possible (just for a crude, quick'n'dirty approximation) to use the 20-year yields (Yahoo Finance has them back to 1-1-1962) but set the assumed maturity at 25 years. 25 year rates shouldn't have been that different from 20-year ones (heck, the 30-year rates aren't typically too far from the 20 year ones) and 25 years would be the rough average maturity one would have if one followed HB's plan of buying 30-year bonds and selling them when they reached 20 years.

Also, just AFAIK, from approximately 1947 or 1948 to early 1953 there wasn't a 30-year Treasury bond and not even really a noncallable 20-year one. The longest term Treasury security available between those dates was one that was issued during WWII, matured in 1972, and was callable in 1967. The US did issue a LT bond in the spring of '53 (think it was a 30 year callable in 25 years) and then issued a 40-year noncallable in 1955.

[MachineGhost]

Guys, if you're gonna prostitute market timing as better than the PP, then you better include mutual fund commissions and taxes in the results.  They were not insignificant for the time period in question.  Even nowadays, Fidelity still charges $75 each way for mutual fund trades!

I think you will find any alleged advantage is eliminated.  This is why moving averages "don't work".

[Thomas Hoog]

There have been several useful threads on this topic.  I was strongly considering trying this out with a portion of VP, but I was dissuaded by two things.  First, keeping up with the required calculations is kind of a pain, and for most of us it violates Rule #1:  put your major effort into your career.  Second, read the prediction-system debunking chapters in Harry Browne's "Why the Best-Laid Investment Plans Usually Go Wrong".  He gives lots of examples of how you can get screwed by timing based on moving averages.

I also thought there would be some treads but I coun't find them exactly
I don't understand "the required caculations" !. I will only use the SAM 200 and that's in yohoo finance. So it takes 10 minutes each month

Guys, if you're gonna prostitute market timing as better than the PP, then you better include mutual fund commissions and taxes in the results.  They were not insignificant for the time period in question.  Even nowadays, Fidelity still charges $75 each way for mutual fund trades!

I think you will find any alleged advantage is eliminated.  This is why moving averages "don't work".

You're right that's why I never tried a VP. However, costs have come down strongly. Based on 12 trades each year, it costs me 0,4 % extra. Other costs (taxes, commision) don't differ from a normal PP portfolio (for me).

[MachineGhost]

MG, Now that you can trade ETFs for free [or a few $], the commissions become kind of a non-issue. As far as the taxes, yes, you would pay them, but wouldn't that ultimately reduce the amount of tax you would ultimately pay later as you draw down the portfolio during retirement? Assuming you had no aspirations to leave an inheritance so the cost basis could be stepped-up, I would think the taxes would be a wash in the end.

I admit I'm intrigued by the allure of the free ETF's, but there will be a spread between the market price and the NAV, the bid and ask, as well as market impact.

But the market is not going to give anyone a free lunch just because trading costs are lower.  I suspect it will adapt and invalidate moving averages even more which did work wonderfully pre-70's.  It has done so by increasing volatility, choppiness and mean reversion at expense of trendiness.

Taxes are very detrimental because it kills the compounding vs non-trading.  It erodes a huge capital base at the end.  Plus most people are in a lower bracket at retirement.

Not saying it won't work, but playing victim to the hindsight bias is not how markets reward.  If you wind up with the same or lesser return as just buy and holding, then is it worth the stress and decision anxiety?  Look at the performance of GTAA (which uses Faber's simple moving averages) on an after tax basis.  Doesn't seem worth it.

[D1984]

I suspect it will adapt and invalidate moving averages even more which did work wonderfully pre-70's.

Since when did moving averages not work recently? If you had been using either 180-day SMA or 200-day SMA you would have been spared most of the 2000-02 stock market fall and almost all of the 2008-09 crash (and in both cases you would have only gotten back in a few months--perhaps one to three--after the market turned around and started going up again). Did either of these two methods do particularly badly in the 2011 market downturn over the debt limit?

As regards taxes and their drag on compounding: Agreed...but a moot point because that's what 401Ks, IRAs, Roths, variable annuities, and SVULs are for. One doesn't put an actively traded startegy in a taxable account except as a last resort.

On whether or not the market will adapt and invalidate timing strategies: if everyone or nearly everyone did them, cure...of course it would. But nearly everyone is NOT doing them. Probably a quarter of people don't have any savings and are in debt; a few more of those that aren't in debt are debt-free but don't have enough savings to last them through a few months of unemployment; about half of the remaining people have some savings in the bank and maybe a few thousand or few tens of thousands in their 401Ks but not much more (IIRc there was a study earlier this year saying the average retirment acount bbalance was less than $30K), and I would be willing to bet that of those that remain (the ones who do have significant amounts saved for retirement and/or have a decent chunk of money in taxable brokerage accounts) most don't even try to optimize by buying index funds instead of actively managed or by applying CAPM or by trying to deliberately hold assets that will do well in different economic circumstances. You have to remember that those of us on here (and Bogleheads, Fatwallet, Retire-Early, etc) are in a distinctly small minority when it comes to how much we think over and analyze our investment situations.

[Thomas Hoog]

You have to remember that those of us on here (and Bogleheads, Fatwallet, Retire-Early, etc) are in a distinctly small minority when it comes to how much we think over and analyze our investment situations.

Best guess: memberlist x 2 x 4 (boogleheads etc..)= 6000 = 0,0001% of the world population

With the pure timing model, you're moving the risk of a large drawdown into a risk that you underperform your relative benchmark, in this case the PP.  Take melveyr's graph and shift it to any random staring year.  Do that 100 times and plot it.

That's the transformation.  There's no free lunch.

I think you hit the weak point + trading cost
Taxes are equal in my country permanent versus timed

[Thomas Hoog]

Started the semi HB IVY timed VP Portfolio on april first
25 % Vanguard Total World Stock Index -ETF-
25 % iShares Euro Government Bond 15-30
25% iShares FTSE/EPRA European Prop Index
0 % iShares Gold Trust
so temporarily
25 % iShares Euro Treasury Bond 0-1

Performance after 4 days + initial cost (4 trades) included: -0,6 %

[blackomen]

Even if they do work in theory, the transaction costs and tax inefficiencies from the whipsaws will eat you alive.

So, why don't we choose a simple indicator, like 9-month simple moving average, and make monthly decisions about which of the 4 asset classes are doing best or worst, and take the 25% from the poorest performing asset, and invest %50 on the highest growing asset? Has anyone made back tests to compare the growth and volatility of such a portfolio? Or even a portfolio investing %100 on top earning asset at any given time based on 9month sma?

It's debatable whether or not these moving average strategies really work.

Even if they do, they are difficult to adhere to psychologically because you can get whipsawed in and out of some of these positions many times before having any success with them.

[Thomas Hoog]

Performance after 3 months: - 1,8 % (cost included)
Benchmarked against the European PP of - 7 %, it means an Alpha of 5,2% !
Number of trades: 7; 4 initial, now 3, only invested in VT at the moment
I missed the Gold crash completly.