Gibson Asset Allocation 4th ed. Portfolio Results

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
stratton
Posts: 11080
Joined: Sun Mar 04, 2007 5:05 pm
Location: Puget Sound

Gibson Asset Allocation 4th ed. Portfolio Results

Post by stratton » Wed Feb 27, 2008 10:23 am

I decided to run Gibson's portfolio through Simba's Backtest Spreadsheet. I used the following assumptions:

Basic Assumptions

Short-Term Debt is 2 year short term Treasury

US Bonds are TIPS

Non-US Bonds are:
a. JP Morgan Global Gov Bond Unhedged 1987-1996
b. Pimco Global Bond unhedged PIGLX Instl ER 0.55% 1997-2007
c. Spreadsheet data at end of post.

US Stocks are Total Stock Market.

Non-US Stocks are Total Intl Stock

From Gibson's Asset Allocation 4th ed. borrowed from this thread

Image

Gibson Asset Allocation 4th ed. Portfolio Results

Code: Select all

1987-2007

Portfolio                               Sharpe   Sortino                     Max Draw   Max Draw
  ID        Total     CAGR    Std.Dev   Ratio    Ratio      C-US    C-Intl    Rebal     Unrebal
=================================================================================================
 2GR       $53,098     8.27     3.94     0.91     1.80      0.68     0.49      0.0%       0.8%
 2MOD       52,702     8.24     4.18     0.85     1.94      0.78     0.54      0.0        1.4
 2LTD       52,481     8.21     4.56     0.78     1.99      0.85     0.50      0.0        3.7
 2US        52,203     8.19     5.27     0.67     2.37      0.90     0.44      0.0        9.8
-------------------------------------------------------------------------------------------------
 3GR       $67,475     9.52     5.78     0.85     4.45      0.75     0.68      1.1%       4.3%
 3MOD       66,636     9.45     6.22     0.78     3.20      0.84     0.70      1.5        8.3
 3LTD       65,862     9.39     6.91     0.70     5.14      0.92     0.65      2.3       14.6
 3US        64,644     9.29     8.20     0.59     1.81      0.96     0.56      7.9       21.3
-------------------------------------------------------------------------------------------------
 4GR       $84,706    10.71     7.94     0.78     2.19      0.75     0.74      4.5%       7.5%
 4MOD       83,156    10.61     8.53     0.72     2.73      0.85     0.75      5.7       14.9
 4LTD       81,610    10.51     9.44     0.65     2.58      0.93     0.70     10.7       21.4
 4US        78,415    10.30    11.33     0.54     1.43      0.98     0.60     20.1       28.8


Conclusions

1. US only equities is bad.
2. The "greater diversification" of more equities can be less "risky" measured by Std.Dev than lesser equities of a US only portfolio.
3. Sortino calculations may be broke?
4. The more diversification the less the draw down. Commodities are probably above most peoples comfort level for greater diversification.

Info on Global Bonds

I've been looking at PIMCO's foreign bond fund vs. global bond fund both unhedged. The only difference between the two is the global fund is also allowed to invest in US bonds. By using US bonds the unhedged fund can "hedge" and this appears to capture most of the upside and less of the downside. The problem is the data here is spotty because PIMCO's foreign fund only has three years of history so I wouldn't give it much weight if any.

Comparing PIGLX to BEGBX does show returns for the global bond fund aren't as extreme. So the allowance of US bonds in the global bond fund moderates quite a bit of the downside and doesn't appear to hurt that much on the upside.

The global unhedged bond fund's strategy also appears to be similar to DFA's new international bond fund that engages in some hedging.

PIMCO Global Bond (Unhedged) Instl (PIGLX)
American Century International Bd Inv (BEGBX)
PIMCO Foreign Bond (Unhedged) I (PFUIX)
Vanguard Total Bond Market Index (VBMFX)

Code: Select all

Year    PFUIX   BEGBX   PIGLX   VBMFX
=====================================
2007    10.46    9.89    9.26    6.92
2006     6.64    8.25    5.85    4.27
2005    -9.06   -8.23   -6.36    2.40
2004            13.10   11.57    4.24
2003            19.91   16.60    3.97
2002            23.53   21.33    8.26
2001            -1.66    2.48    8.43
2000            -1.20    0.43   11.39
1999           -10.36   -4.28   -0.76
1998            17.87   12.44    8.58

I've added BEGBX since its been around at least 10 years instead of PFUIX. The difference in management might be a problem with the returns. At least with PIMCO you have a unified management strategy when comparing funds. Added VG's TBM for comparison

Global Unhedged Bonds 1987-2007

2.20 <--1987
6.80
14.00
8.60
15.50
4.60
12.30
1.30
19.30
4.40
-0.90
12.50
-4.29
0.43
2.48
21.33
16.59
11.56
-6.36
5.85
9.26 <--2007

User avatar
RobertH
Posts: 218
Joined: Thu Mar 01, 2007 11:35 pm
Location: Bay Area

Post by RobertH » Wed Feb 27, 2008 12:24 pm

Thanks for the analysis. Very interesting.

I'll update my version of Simba's spreadsheet with the historical return data you've provided for unhedged global bonds.

Does anybody have historical return info (real or synthetic) for unhedged global bonds prior to 1987? Preferably to 1972. Or know where I can find it?

Robert

P.S. I have had my own suspicions about the Sortino calculation in Simba's spreadsheet. If someone believes these numbers are correct, could they please attempt to explain them?

User avatar
Random Musings
Posts: 5209
Joined: Thu Feb 22, 2007 4:24 pm
Location: Pennsylvania

Post by Random Musings » Wed Feb 27, 2008 12:52 pm

My conclusions:

Running backtested data when current bond yields (and TIP fixed component rates) are low on the historical side (and relative to will provide you with results that mean less and less as the bond percentage of a portfolio goes up. Well, unless we pull a Japan in the longer run and end up with very, very low rates.

2-Yr Treasury Constant Maturity

6+ percent vs 2.5% is a big difference. Let alone the fact real returns are the ultimate truth-teller.

RM

User avatar
stratton
Posts: 11080
Joined: Sun Mar 04, 2007 5:05 pm
Location: Puget Sound

Post by stratton » Wed Feb 27, 2008 12:53 pm

RobertH wrote:I'll update my version of Simba's spreadsheet with the historical return data you've provided for unhedged global bonds.

Does anybody have historical return info (real or synthetic) for unhedged global bonds prior to 1987? Preferably to 1972. Or know where I can find it?

Robert

P.S. I have had my own suspicions about the Sortino calculation in Simba's spreadsheet. If someone believes these numbers are correct, could they please attempt to explain them?

Thanks for mentioning data going back to 1972 would be good. I didn't subtract the 0.55% ER from the JP Morgan data.

I've noticed a few other bugs:
1. putting 0.0 in the data set will generate a divide by zero error "sometimes". I had to put a 0.01 in some timberland data.
2. Only the first new asset class added on the data pages works. I fixed this in a previous version by "filling" to fullest extent of some tables.

Future data analysis
1. Craig Israelson has an article about replacing commodities with gold, real estate and energy. An approximation he came up with was 50% REITs, 30% energy, 20% gold to replace commodities. Since we have large amounts of real estate already 60% energy and 40% gold might be a better replacement. That would also be another excuse to lower the amount of commodities.
2. Rerun some tests wit gold, timberland, MLPs in various mixes
3. Break down broad US and foreign markets with SV tilting.
4. There was a set of TRP New Era (PRNEX) data back to 1972 I want to play with since that fund was originally designed to fight inflation.

Paul

User avatar
stratton
Posts: 11080
Joined: Sun Mar 04, 2007 5:05 pm
Location: Puget Sound

Post by stratton » Wed Feb 27, 2008 12:59 pm

Random Musings wrote:My conclusions:

Running backtested data when current bond yields (and TIP fixed component rates) are low on the historical side (and relative to will provide you with results that mean less and less as the bond percentage of a portfolio goes up. Well, unless we pull a Japan in the longer run and end up with very, very low rates.

2-Yr Treasury Constant Maturity

6+ percent vs 2.5% is a big difference. Let alone the fact real returns are the ultimate truth-teller.

I agree future returns won't necessarily match backtest results. It's probably more accurate to ignore the return numbers and only look at which combinations of assets have the higher returns vs. lower volitility. And then we run into the problem of interpreting "risk" as volatility. This is why it is not a good idea to have 100% EM or SV or Intl SV.

Paul

User avatar
DaveTH
Posts: 2504
Joined: Thu Apr 05, 2007 7:23 pm

Post by DaveTH » Wed Feb 27, 2008 6:51 pm

It's interesting that Gibson still does not include the International Real Estate asset class. Maybe he'll do that in his 5th edition.

User avatar
stratton
Posts: 11080
Joined: Sun Mar 04, 2007 5:05 pm
Location: Puget Sound

Post by stratton » Wed Feb 27, 2008 7:07 pm

DaveTH wrote:It's interesting that Gibson still does not include the International Real Estate asset class. Maybe he'll do that in his 5th edition.

I have a dataset on that back to 93 or 94. :) Splitting up real estate with Intl ought to smooth out the volatility a little bit more.

Paul

660ky612
Posts: 245
Joined: Thu Jun 14, 2007 11:32 am
Location: Hong Kong SAR

Post by 660ky612 » Sat Mar 01, 2008 3:22 pm

Dear stratton and forum,

Would you consider including more ex-US bonds and/or less commodities? (Rationale see below, and think of returns, SDs, and not just correlation.)

Thanks, 660ky612

----- cut ---------------------------------------
For the consistency of correlations, please read http://www.diehards.org/forum/viewtopic ... ht=#160923 of which I have just been able to understand something - tiny though.

--------- cut ---------------------------------
I am not sure if the following have ever been mentioned.
Consistency of correlations

http://www.fpanet.org/journal/articles/ ... 6354_1.pdf
Ignore table 1,2,3 but focus on table 4, especially p.62:

Useful background information:
. rho, the correlation coefficient takes values between -1 and 1.
. for this paper of Coaker, if 0.00 < rho < 0.49, we say the two asset classes are low-positive correlated.


Global Bonds is of consistently low-positive to negatively correlated to nearly every other asset classes, except Int. stocks, US bonds.
Image


Commodities CCF is of consistently low-positive to negatively correlated to nearly every other asset classes.
Image


US Reits is NOT of consistently low-positive to negatively correlated to every other asset classes.
US Reits is of consistently low-positive to negatively correlated to: Emerg. stocks, US bonds, Global bonds, Cash, commodities CCF, Long-short.
Image

! Of course, we will not be contented with the result from 1972 to 2004,
! we want a rolling analysis.
Oh! No! This is already a 3 year (36 months) rolling analysis. Hence it is done!

Thanks,
660ky612 from Hong Kong

Post Reply