Asset allocation plus momentum , correlation

[rabidric]

Edited by Moderator: New discussion topic merits its own thread + title. Parent was this
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

i'll just leave this here as it seems to cover most sensible ideas on this subject.

1. go to the page, and download the 5-6 pdf's contained.
2. put on sunglasses to filter out the obvious marketing tint
3. read pdf's at your leisure and digest information contained within
4a. consider implementing some of the same stuff if you don't already do the same, or
4b. if liquid enough, perhaps just quit and invest with them?
5. whether you choose option a or b is immaterial, the unchosen option will be guaranteed to outperform

EDIT: (sorry i messed up the numbers of the exhibits first time, now corrected)
The AAA pdf is the main one. Personally I have serious queries as to whether the marginal benefit of exhibit 5 over exhibit 4(see the pdf itself to understand) necessarily justifies the boatload of extra complexity and introspective heartache introduced. But that just is, and probably will remain my stance throughout this thread.

Exhibit 6 uses a different time horizon at it's core and so whilst seems to be further improvement(or optimization) i consider it to be merely a separate anecdote.

The progression from exhibits 1-5 is certainly very interesting to see incrementally though.

[akhoury]

Thanks very much for sharing this...I enjoyed it and learned quite a bit from your thoughtful work......

[wguan]

I've done a similar research study and have found benefit into the framework of combining traditional allocation schemes from academia paired with momentum. Below are the results for momentum and risk parity on housing, us equity, foreign equity, bonds and commodities. (all total return data). Test period 1985 to 2012.

Observations: I am sure some will argue the drawbacks in the risk parity framework when we enter a higher interest rate environment but I argue that with a momentum overlay, such drawback are held at bay as we are not statically tied to any assets in the long run. I would test this strategy in the 1970s to 1980s for sensitivity analysis but I don't have the data.

[squaredQ]

i'll just leave this here as it seems to cover most sensible ideas on this subject.

1. go to the page, and download the 5-6 pdf's contained.
2. put on sunglasses to filter out the obvious marketing tint
3. read pdf's at your leisure and digest information contained within
4a. consider implementing some of the same stuff if you don't already do the same, or
4b. if liquid enough, perhaps just quit and invest with them?
5. whether you choose option a or b is immaterial, the unchosen option will be guaranteed to outperform

EDIT:
The AAA pdf is the main one. Personally I have serious queries as to whether the marginal benefit of exhibit 6 over exhibit 5(see the pdf itself to understand) necessarily justifies the boatload of extra complexity and introspective heartache introduced. But that just is, and probably will remain my stance throughout this thread.

Exhibit 7 uses a different time horizon at it's core and so whilst seems to be further improvement(or optimization) i consider it to be merely a separate anecdote.

The progression from exhibits 1-5 is certainly very interesting to see incrementally though.

I'm familiar with this work. If anyone finds a good long-term list of securities that could be used as a proxy for the 10 sectors he uses, please list them. I found some good long ones on a monthly basis, but not daily.

Along the same thinking, there's a good recent blog on this topic:
http://cssanalytics.wordpress.com/2012/ ... -variance/

[mojojojo]

I'm familiar with this work. If anyone finds a good long-term list of securities that could be used as a proxy for the 10 sectors he uses, please list them. I found some good long ones on a monthly basis, but not daily.

Along the same thinking, there's a good recent blog on this topic:
http://cssanalytics.wordpress.com/2012/ ... -variance/

There are ETFs that can be used for all 10 of those sectors. Some having a relatively long track record (for ETFs).

[squaredQ]

I'm familiar with this work. If anyone finds a good long-term list of securities that could be used as a proxy for the 10 sectors he uses, please list them. I found some good long ones on a monthly basis, but not daily.

Along the same thinking, there's a good recent blog on this topic:
http://cssanalytics.wordpress.com/2012/ ... -variance/

There are ETFs that can be used for all 10 of those sectors. Some having a relatively long track record (for ETFs).

edit:
@wguan
Thanks!
----------------------------------------------------------------------
Could you provide a list, please? thanks.

As far as I know, most of the 10 proxies will go to not much further than 90s. Either way, I'd like to have a list of 10 good candidate proxies (daily) if anyone wishes to share.

[wguan]

I used the underlying index of different ETFs as proxies.

GSG = SP GSCI Index (commodities)
VTI = MSCI US Index
VNQ = FTSE NAREIT (housing)
IEF = Barclays 5-7Yr Treasuries
SHY = Barclays 1-3Yr Treasuries Index
EFA = MSCI EAFE Index

A good site to get the etf for different asset class is http://www.etfdb.com. The Ivy Portfolio near the end has a huge list of diversified etf. If you are looking to get the possible mixture they used for backtest, a simple search on their websites will provide the possible asset classes they used.
http://www.macquarieprivatewealth.ca/d ... 012-06.pdf

[stopsareforwimps]

The progression from exhibits 1-5 is certainly very interesting to see incrementally though.

Yes very interesting reading. Interesting how Markowitz diversification is used to good effect here, when its results have often been underwhelming in the real world. Interesting also the short term horizon of this apparently successful strategy.

Some caveats and a disclosure...

Caveat 1. "All taxes, fees and costs are ignored for simplicity". It is a pity that life is not simple.

Caveat 2. Not all the algorithms used are disclosed. E.g., the "engineering improvements".

Caveat 3. Scenario 5 can have as few as two assets, so it may be possible that unpleasant surprises could happen that did not show up in the simulation.

Caveat 4. It is not clear how much curve-fitting or data mining has gone on here and how much the performance may suffer on new data.

Disclosure: I lost almost $40,000 on a Macquarie product a few years back (http://www.smh.com.au/business/macquari ... -1rls.html), via a financial planner. I may be biased against them.

[rabidric]

indeed. hence my "sunglasses advice".
re: caveat 3 especially, yes i agree. I think i said so in my original post, lot of extra complexity for questionable benefit, even the numbers they state show only a very marginal gain over exhibit 4.

[mojojojo]

I'm familiar with this work. If anyone finds a good long-term list of securities that could be used as a proxy for the 10 sectors he uses, please list them. I found some good long ones on a monthly basis, but not daily.

Along the same thinking, there's a good recent blog on this topic:
http://cssanalytics.wordpress.com/2012/ ... -variance/

There are ETFs that can be used for all 10 of those sectors. Some having a relatively long track record (for ETFs).

edit:
@wguan
Thanks!
----------------------------------------------------------------------
Could you provide a list, please? thanks.

As far as I know, most of the 10 proxies will go to not much further than 90s. Either way, I'd like to have a list of 10 good candidate proxies (daily) if anyone wishes to share.

You will have a problem finding ETFs going back prior to 2000 on most of those. ETFs didn't really start catching on until after that. I don't have anything handy but here are some places to start looking.

http://www.etfdb.com
http://www.xtf.com
http://www.indexuniverse.com
http://www.morningstar.com

Here's a new one that's in beta ... definitely has a way to go, but looks like it might be promising in future. http://www.etfg.com

[doubleR]

"Caveat 1. "All taxes, fees and costs are ignored for simplicity". It is a pity that life is not simple"

fees and cost indeed cannot be ignored however using monthly models these costs are reduced.

As per taxes, in UK for example, one can build 2 accounts for ISAs and SIPPs and trade those. GRANTED that one cannot put £1,000,000 there but if one is looking for a LTTF model to trade their savings tax free using ETFs...

I am sure in Australia there is something similar...