Updated figures and results on slice and dice vs TSM

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Updated figures and results on slice and dice vs TSM

Postby Rodc » Tue Dec 11, 2007 9:35 am

Last time I posted results like this I still had a slight bit of goofyness with large blend and small blend. Things turn out fine and hang together if you use F-F’s monthly weights between their benchmark portfolios, but if you use the yearly average weights things get a little off. Since the F-F large blend (mix of large growth, large core, and large value) averages about 92.7% of the total market, if you are not very careful TSM and LB can line up a little off of how you would expect them to line up. (Thank you Robert T in helping me work through that).

I also updated how I show risk adjusted returns. I think this makes things clearer and more useful at the intuitive level.

To recap: asset class benchmark portfolios (LG,LC,LV,SG,SC,SV) from F-F. Large blend (LB), small blend (SB) and Total Stock Market (TSM) from those benchmark portfolios using monthly market weights. All candidate slice and dice portfolios (blue dots in graphs) are made from all possible combinations using weights that are a multiple of 5% (a computer limitation on my part).

Four particular standard slice and dice portfolios are considered:
4x25: equal parts LB,LV,SB,SV
3x33: equal parts LB,LV,SV
TSM/SV: 60% TSM, 40% SV
LB/SB: 50% LB, 50% SB

Each has significant differences from the other portfolios, and each is similar to portfolio options often encountered in discussions here. The last is included because sometimes due to limitations within someone’s 401K value tilting is not an option, but they have LB and SB options.

One of the things I would like to be able to do is provide some guidance to a novice investor, someone who is not going to run off computing factor weights or want to get into what factor weights even mean. Note that to a naive investor each of the first three Slice and Dice portfolios achieves a similar apparent amount of loading, and as we have seen people often wonder which might be best. Very roughly and very naively:

4x25: 50% value, 50% small, “total tilt” = 100%
3x33: 67% value, 33% small, “total tilt” = 100%
TSM/SV: 40% value, 40% small, “total” tilt = 80% (in fact if one considers the value component of blend and small component of TSM the tilt is a little higher)
LB/SB: 0% value, 50% small, “total” tilt = 50%

The primary results are that:

1) Slice and dice has fairly consistently provided a risk/volatility adjusted benefit.
2) Any two vaguely similar slice and dice portfolios have about the same long term returns and volatility; how you get your tilt does not matter very much, and to the extent that it does, it is not easy to predict ahead of time which tilt is best.


Figure 1: 80 years of asset class returns on top of all possible portfolios formed from the F-F benchmark portfolios and 90-day T-Bills (more on how things change with different types of bonds shortly in another post). Note that averaged over long periods of time a combination of SV and T-Bills is right near the efficient frontier, and SG is about as bad as you can possibly find, and LG is not much better. Given that of all the asset class portfolios, SV has the greatest combined value and small tilt, this should be expected if one averages over enough time; similarly LG should be expected to do poorly. The last place finish for SG has often been noted, but seems a little odd to me relative to the F-F three factor model.



Image

Of course an investor rarely has the luxury of riding out 80 years of ups and down to achieve such results. Figure 2 shows the same information over the four independent 20-year sub-periods. In three of these 20-year periods you can only barely make out a difference between TSM and large growth. Note that TSM is averages about 7% small blend. Even in the period where you can clearly see a difference the difference is not very large. Note too that there are some things that are fairly consistent, such as the position of small value, and some that differ widely such as the position of large growth. The slice and dice portfolios consistently do well, if not quite optimally. The other important thing to note is just the large difference in volatility and return in these periods; you just can’t count on getting an average return even if you buy and hold for 20 years.

Image


However, these graphs do not show how the various standard portfolios compare on a risk adjusted basis. To show this I started with the notion that an investor sets a risk level they are comfortable with, and then they try to build the highest returning portfolio with that level of risk. Here for the sake of user ease I take risk to be volatility.

Suppose someone sets a risk level of standard deviation equal to 15%. Two questions arise: how do you achieve that with each portfolio, and having done that, what is the annual return? With perfect hindsight it is easy to see what level of T-Bills (or bonds) you need to get a standard deviation of 15% over some time period. In this case, over 80 years, you can use 75% TSM and 25% T-Bills, or with the more volatile slice and dice portfolios you need roughly 60% stocks and 40% T-Bills. Note these are fairly aggressive portfolios.



Code: Select all

   Percentages to maintain std = 15%   
                TSM      4x25      3x33      TSM/SV   LB/SB
stocks           75%      57%       60%         63%     62%
90-day T-Bills   25%      43%       40%         37%     38%



Figure 3 shows what happens with each portfolio as you dilute it with T-Bills, from zero percent in the upper right to 100% T-Bills in the lower left. The diamonds show the returns at a standard deviation of 15%. I think the most important points are that the value and small slice and dice portfolios run about 1% higher in annual returns if each is held to a standard deviation of 15% (i.e. equal “risk”) compared to the TSM/T-Bill portfolio, and that there is virtually no difference between the three standard value and small slice and dice portfolios on a risk/volatility adjusted basis. The small tilt only portfolio lies between the other slice and dice portfolios and TSM. Note that as an invest sets a lower volatility target the benefit of slice and dice decreases. Also note that costs are not included; if you have significant costs in obtaining the funds needed to tilt your portfolio those costs will reduce any gain in performance.

Image


How consistent is this behavior? Figure 4 shows the same information over the four independent 20-year periods. In one TSM/T-Bills is slightly better than the slice and dice portfolios, otherwise the slice and dice are clearly better. In period one it appears the benefit is all from small, as the LB/SB is as good as any of the value tilted portfolios. In general there is little difference between the three value and small tilted slice and dice portfolios.

Image


I looked at other measures of risk that might be useful to the novice investor, and may report on them later. Things like what happens if you look at 5-year periods, but in the end nothing changed the basic story.

rkbrashear
Posts: 104
Joined: Mon Oct 29, 2007 6:15 pm

Postby rkbrashear » Tue Dec 11, 2007 9:53 am

Rod - very compelling and well done. I love the graphs.

Did you allow for any cost/tax difference in the S&D portfolios?

Thanks,
Ryan

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Tue Dec 11, 2007 10:02 am

rkbrashear wrote:Rod - very compelling and well done. I love the graphs.

Did you allow for any cost/tax difference in the S&D portfolios?

Thanks,
Ryan


hi Ryan,

I have looked at cost differences, but not tax differences, but that is not included here. I struggle with how to cost things. Bogle once put out there that 100% turn over introduces an internal cost of about 1% in addition to expense ratios. I put in the code the option to use Vanguard investor class expense ratios and the current level of turnover. The total cost for TSM is 0.24% and for SV it is 0.48%, using the above turn-over factor. But for example I pay much less than investor class for some funds in my 401K, other people pay rather more than Vanguard fees. So you kind of have to look at your own fees and adjust for your own costs.

Rod

User avatar
Rick Ferri
Posts: 8200
Joined: Mon Feb 26, 2007 11:40 am
Location: Vagabond, USA
Contact:

Postby Rick Ferri » Tue Dec 11, 2007 10:03 am

Nice charts. When is the book coming out?

Your charts are consistent with the results of similar research I did for my All About Asset Allocation book back in 2005. I found a US allocation of 60% TSM, 20% SV, 10% REIT and 10% microcap worked well. That is very close to your 60% TSM and 40% small value blend.

Rick Ferri

Angst
Posts: 1533
Joined: Sat Jun 09, 2007 11:31 am
Location: St Louis, MO

Re: Updated figures and results on slice and dice vs TSM

Postby Angst » Tue Dec 11, 2007 10:22 am

Rodc wrote:One of the things I would like to be able to do is provide some guidance to a novice investor, someone who is not going to run off computing factor weights or want to get into what factor weights even mean. Note that to a naive investor each of the first three Slice and Dice portfolios achieves a similar apparent amount of loading, and as we have seen people often wonder which might be best. Very roughly and very naively:

4x25: equal parts LB,LV,SB,SV
3x33: equal parts LB,LV,SV
TSM/SV: 60% TSM, 40% SV
LB/SB: 50% LB, 50% SB

...

4x25: 50% value, 50% small, “total tilt” = 100%
3x33: 67% value, 33% small, “total tilt” = 100%
TSM/SV: 40% value, 40% small, “total” tilt = 80% (in fact if one considers the value component of blend and small component of TSM the tilt is a little higher)
LB/SB: 0% value, 50% small, “total” tilt = 50%


Very interesting information and wonderful presentation, but what proportions of "Value" and "Growth" would you say are comprised by "Blend"? If they are = approximately 50-50 Value-Growth, wouldn't the four portfolios net the following Value percentages: (I might be missing something.)

4x25: 75% Value
3x33: 83.16% Value
TSM/SV: 70% Value
LB/SB: 50% Value

Angie

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Updated figures and results on slice and dice vs TSM

Postby Rodc » Tue Dec 11, 2007 10:36 am

Angst wrote:
Rodc wrote:One of the things I would like to be able to do is provide some guidance to a novice investor, someone who is not going to run off computing factor weights or want to get into what factor weights even mean. Note that to a naive investor each of the first three Slice and Dice portfolios achieves a similar apparent amount of loading, and as we have seen people often wonder which might be best. Very roughly and very naively:

4x25: equal parts LB,LV,SB,SV
3x33: equal parts LB,LV,SV
TSM/SV: 60% TSM, 40% SV
LB/SB: 50% LB, 50% SB

...

4x25: 50% value, 50% small, “total tilt” = 100%
3x33: 67% value, 33% small, “total tilt” = 100%
TSM/SV: 40% value, 40% small, “total” tilt = 80% (in fact if one considers the value component of blend and small component of TSM the tilt is a little higher)
LB/SB: 0% value, 50% small, “total” tilt = 50%


Very interesting information and wonderful presentation, but what proportions of "Value" and "Growth" would you say are comprised by "Blend"? If they are = approximately 50-50 Value-Growth, wouldn't the four portfolios net the following Value percentages: (I might be missing something.)

4x25: 75% Value
3x33: 83.16% Value
TSM/SV: 70% Value
LB/SB: 50% Value

Angie


Good question.

Averaged over 80 years here are the weights relative to TSM are:

Code: Select all

SG       SC        SV        LG       LC       LV
2.40%   2.87%    2.06%    51.72%    30.97%    9.99%


You can of course use these to do LB and SB.

Yes one can be more careful. I was purposely trying not to be careful. The point I am try to make is you don't have to be careful in order to do well. I often see people agonize over decision of exactly how to tilt, and my point is just pick one then go off and do something more interesting with your life. (or course I don't always take my own advice. :))

User avatar
RaleighStClaire
Posts: 731
Joined: Mon Jun 18, 2007 2:08 pm
Location: Tilting

Postby RaleighStClaire » Tue Dec 11, 2007 10:36 am

Rick is right, the findings are very similar to those AAAA, but I liked these graphs even more. I think that showing the intersection of all portfolios helps show how much worse certain portfolios are compared to others. Great job.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Tue Dec 11, 2007 10:37 am

Rick Ferri wrote:Nice charts. When is the book coming out?

Your charts are consistent with the results of similar research I did for my All About Asset Allocation book back in 2005. I found a US allocation of 60% TSM, 20% SV, 10% REIT and 10% microcap worked well. That is very close to your 60% TSM and 40% small value blend.

Rick Ferri


I ought to read more, but reinventing the wheel is more fun. :)

User avatar
mas
Posts: 1461
Joined: Tue Feb 20, 2007 12:54 pm

Postby mas » Tue Dec 11, 2007 10:53 am

What software did you use to make these?

Angst
Posts: 1533
Joined: Sat Jun 09, 2007 11:31 am
Location: St Louis, MO

Re: Updated figures and results on slice and dice vs TSM

Postby Angst » Tue Dec 11, 2007 10:55 am

Rodc wrote:Averaged over 80 years here are the weights relative to TSM are:

Code: Select all

SG       SC        SV        LG       LC       LV
2.40%   2.87%    2.06%    51.72%    30.97%    9.99%


You can of course use these to do LB and SB.

Yes one can be more careful. I was purposely trying not to be careful. The point I am try to make is you don't have to be careful in order to do well. I often see people agonize over decision of exactly how to tilt, and my point is just pick one then go off and do something more interesting with your life. (or course I don't always take my own advice. :))


I do appreciate your point, and it's well taken. But please tell me, do the numbers above mean that only 12.05% of the TSM is comprised of Value stocks?

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Updated figures and results on slice and dice vs TSM

Postby Rodc » Tue Dec 11, 2007 11:21 am

Angst wrote:
Rodc wrote:Averaged over 80 years here are the weights relative to TSM are:

Code: Select all

SG       SC        SV        LG       LC       LV
2.40%   2.87%    2.06%    51.72%    30.97%    9.99%


You can of course use these to do LB and SB.

Yes one can be more careful. I was purposely trying not to be careful. The point I am try to make is you don't have to be careful in order to do well. I often see people agonize over decision of exactly how to tilt, and my point is just pick one then go off and do something more interesting with your life. (or course I don't always take my own advice. :))


I do appreciate your point, and it's well taken. But please tell me, do the numbers above mean that only 12.05% of the TSM is comprised of Value stocks?


By F-F's definition yes.

One of the difficulties in this endeavor is of course that different people have different notions of what is and what is not value, what is and what is not small, etc. If you want a nice long time history, it seems the only choice is F-F, at least that is the best I can find.

The exact details I wouldn't want to swear I fully understand in a court of law, but they define growth and value as the outer 30% (least and greatest) based on book equity to market equity, with core as the middle 40%. Note that the small distribution is closer to what one might expect, but the large is heavily growth oriented. This seems (per information on their web site giving average cap size) to reflect the fact that the large value companies as they see it tend to be much smaller than the large growth, so the 30% of the large companies that are value have much smaller market capitalization than the 30% that are growth. This true for small companies too, but to a much smaller extent (pun intended!)

it might also be worth mentioning their definition of large currently covers 811 companies, coming in between MCSI at 750 and Russell at 1000.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Tue Dec 11, 2007 11:23 am

mas wrote:What software did you use to make these?


Matlab

Angst
Posts: 1533
Joined: Sat Jun 09, 2007 11:31 am
Location: St Louis, MO

Re: Updated figures and results on slice and dice vs TSM

Postby Angst » Tue Dec 11, 2007 11:34 am

Rodc wrote:By F-F's definition yes.

One of the difficulties in this endeavor is of course that different people have different notions of what is and what is not value, what is and what is not small, etc. If you want a nice long time history, it seems the only choice is F-F, at least that is the best I can find.

The exact details I wouldn't want to swear I fully understand in a court of law, but they define growth and value as the outer 30% (least and greatest) based on book equity to market equity, with core as the middle 40%. Note that the small distribution is closer to what one might expect, but the large is heavily growth oriented. This seems (per information on their web site giving average cap size) to reflect the fact that the large value companies as they see it tend to be much smaller than the large growth, so the 30% of the large companies that are value have much smaller market capitalization than the 30% that are growth. This true for small companies too, but to a much smaller extent (pun intended!)

it might also be worth mentioning their definition of large currently covers 811 companies, coming in between MCSI at 750 and Russell at 1000.


Thank you! I learn so much simply lurking about this forum and only occasionally asking a question. I appreciate your helpfulness.

Angie

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Updated figures and results on slice and dice vs TSM

Postby Rodc » Tue Dec 11, 2007 11:46 am

Angst wrote:
Rodc wrote:By F-F's definition yes.

One of the difficulties in this endeavor is of course that different people have different notions of what is and what is not value, what is and what is not small, etc. If you want a nice long time history, it seems the only choice is F-F, at least that is the best I can find.

The exact details I wouldn't want to swear I fully understand in a court of law, but they define growth and value as the outer 30% (least and greatest) based on book equity to market equity, with core as the middle 40%. Note that the small distribution is closer to what one might expect, but the large is heavily growth oriented. This seems (per information on their web site giving average cap size) to reflect the fact that the large value companies as they see it tend to be much smaller than the large growth, so the 30% of the large companies that are value have much smaller market capitalization than the 30% that are growth. This true for small companies too, but to a much smaller extent (pun intended!)

it might also be worth mentioning their definition of large currently covers 811 companies, coming in between MCSI at 750 and Russell at 1000.


Thank you! I learn so much simply lurking about this forum and only occasionally asking a question. I appreciate your helpfulness.

Angie


My pleasure.

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

Drawdown risk

Postby RobertH » Tue Dec 11, 2007 12:37 pm

Rodc,

Very interesting stuff, nicely presented. Thank you for posting this. I've forwarded this link to some friends of mine who don't follow this stuff every day.

If you're thinking about building on this work in the future, I would be interested in an analysis that concentrated on drawdown risk of these portfolios. I believe many novice investors would also view that as an important - perhaps bottom-line - risk measure.

For myself, I view drawdown risk as having two components: maximum drawdown (how far does the portfolio fall from its previous high) and drawdown duration (how long does it take to get back to the previous high). I like to use inflation adjusted numbers to put the 1970s in proper perspective.

Robert

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Tue Dec 11, 2007 12:45 pm

If you're thinking about building on this work in the future, I would be interested in an analysis that concentrated on drawdown risk of these portfolios. I believe many novice investors would also view that as an important - perhaps bottom-line - risk measure.


Good suggestion.

That is primarily how I do my own risk analysis, although I have not updated that look personally in a couple of years. Might be a good idea to do so.

LFT_PFT
Posts: 265
Joined: Mon May 28, 2007 6:59 pm

Thanks

Postby LFT_PFT » Thu Dec 13, 2007 4:01 pm

Just wanted to offer a thanks for posting the graphs and analysis. Very interesting.

RodC wrote:

I often see people agonize over decision of exactly how to tilt, and my point is just pick one then go off and do something more interesting with your life.


The more you read and learn the more you come back to this key point....

User avatar
Diver
Posts: 102
Joined: Thu Nov 15, 2007 11:13 pm

Postby Diver » Wed Dec 19, 2007 11:37 am

Rod,

I have to agree with everyone on nice presentation and analysis. My GF works at MathWorks and some point
I got my hand on Matlab financial toolkit and tried to run some model portfolios but gave up after some struggle :)

I'm very curios to see where mid cap and mid cap value would fit on your graphs. There have been multiple
debates on this forum about mid caps and the data presented showed that mid caps have similar historical
returns and risks to slice and dice portfolios like 4x25 or 3x33.

I'm not sure if you have access to historical data for mid cap and possibly mid cap value but I think
many people would agree that it would be very interesting to see those asset classes on the graphs.

Thanks again for nice analysis !

User avatar
LH
Posts: 5490
Joined: Wed Mar 14, 2007 2:54 am

Postby LH » Wed Dec 19, 2007 12:12 pm

4x25: 50% value, 50% small, “total tilt” = 100%
3x33: 67% value, 33% small, “total tilt” = 100%
TSM/SV: 40% value, 40% small, “total” tilt = 80% (in fact if one considers the value component of blend and small component of TSM the tilt is a little higher)
LB/SB: 0% value, 50% small, “total” tilt = 50%


What does the ""total tilt" = 100%" mean? Is it a small value tilt I assume?
4x25: 50% value, 50% small, “total tilt” = 100%

If the 50% small was increased to 60% small, with 40% value, then the "total tilt" would be greater than 100 percent right?

Is there a definition of what 100 percent tilt means? Why did they pick that tilt to be 100 percent tilt if so? 100 percent SV tilt to me would be 100% in SV only. Or is it just a value tilt? if so then some of the small would still be growth right?

Thanks- great post.

LH

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Wed Dec 19, 2007 2:38 pm

LH wrote:
4x25: 50% value, 50% small, “total tilt” = 100%
3x33: 67% value, 33% small, “total tilt” = 100%
TSM/SV: 40% value, 40% small, “total” tilt = 80% (in fact if one considers the value component of blend and small component of TSM the tilt is a little higher)
LB/SB: 0% value, 50% small, “total” tilt = 50%


What does the ""total tilt" = 100%" mean? Is it a small value tilt I assume?
4x25: 50% value, 50% small, “total tilt” = 100%

If the 50% small was increased to 60% small, with 40% value, then the "total tilt" would be greater than 100 percent right?

Is there a definition of what 100 percent tilt means? Why did they pick that tilt to be 100 percent tilt if so? 100 percent SV tilt to me would be 100% in SV only. Or is it just a value tilt? if so then some of the small would still be growth right?

Thanks- great post.

LH


Good question.

My intention was to address how an unsophisticated investor might view these portfolios and how such a person might compare them. Done right you'd regress against the F-F factors, but my thought is is 99.44% (my favorite big percentage. :)) of investors are never going anywhere near such a calculation.

In this simply minded view of things you can tilt small, 100% could be 100% vanguard small cap. You could tilt value, 100% would be about 75% Large value and 25% small value (percentages depending on underlying indices), and total tilt is just the sum of the two tilts, so 100% small value gets a 200% total tilt. I could have done as you suggest so that 100% sv is maximum tilt = 100%, which makes sense, but it did not occur to me to do so, and would not change the message as they'd come out about equal at 50%.

This is not and was not intended to be rigorous (note that is completely misses the fact that a blend fund has some value in it), nor am I committed to this definition; I kind of like yours. It just seemed to fit my suppositions of how an unsophisticated investor might view things, and it just happened that such a view turned out to be useful in the sense that it predicted pretty well the long term results (which I was not expecting).

I did not pick the portfolios to come in at 100% or so, rather I started with the sorts of portfolios I see most often discussed, such as the 4x25, and LB/SB, and only later noticed that by this simple definition the "total tilt" was about equal. I did tweak the TSM/SV after I noticed this as I had started at 70/30 and moved to 60/40. So the 100% was just coincidental.

tan
Posts: 90
Joined: Sat Nov 10, 2007 1:13 pm

Postby tan » Wed Dec 19, 2007 2:49 pm

From an unsophisticated investor... thank you. This was wonderfully informative. It's nice that some people here remember us little folks. It's posts like this that help us step up the ladder to become more sophisticated investors.

tower
Posts: 155
Joined: Wed Mar 21, 2007 4:29 pm
Location: Durham, NC

Another calculation

Postby tower » Thu Dec 20, 2007 2:52 pm

Using the methodology described in the paper below (available by google search)

Do Vanguard’s Managed Funds Beat Its Index Funds? Looking for Prescient Stock and Style Selection

by Abel Rodriguez Edward Tower

I found that for the period January 1999-December 2006 the basket of Vanguard index funds which most closely mimics the style of the Total Stock Market Index outperforms it by 0.206% points per year. But it has a slightly higher standard deviation of monthly returns (.05% points per month). That basket of index funds is 75% VFINX (S&P500 index), 22% VIGRX (growth index) and 3% VISX (midcap index).

To the extent that there is transactions cost as stocks are shifted between indexes I would expect VTSMX to outperform. To the extent that rebalancing monthly (assumed in my calculations) to keep the fractions in each of the index funds means that one profits from regression to the mean, I would expect the TXMX to underperform. It appears that the latter effect dominates slightly.

A question: RodC, where did you get your data?

Ed

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Thu Dec 20, 2007 3:02 pm

A question: RodC, where did you get your data?


http://mba.tuck.dartmouth.edu/pages/fac ... brary.html

The stock returns are the standard Fama-French benchmark portfolio data, the 90-day T-Bills came from Global Financial Data. I'm used to using those as the "risk free" investment, although I could just as easily used the T-Bill data from F-F, which if memory serves are 30-day (but it makes no real difference either way). Indeed one can use just about any length bond funds and get very similar results (shown in another recent thread).

tower
Posts: 155
Joined: Wed Mar 21, 2007 4:29 pm
Location: Durham, NC

RocC Thanks

Postby tower » Sun Dec 23, 2007 8:53 am

I had not appreciated how much good stuff there is on the FF web site.

Ed

diasurfer
Posts: 1830
Joined: Fri Jul 06, 2007 8:33 pm
Location: miami-dade

Postby diasurfer » Thu Jul 10, 2008 2:42 pm

Rodc wrote:
A question: RodC, where did you get your data?


http://mba.tuck.dartmouth.edu/pages/fac ... brary.html

The stock returns are the standard Fama-French benchmark portfolio data, the 90-day T-Bills came from Global Financial Data. I'm used to using those as the "risk free" investment, although I could just as easily used the T-Bill data from F-F, which if memory serves are 30-day (but it makes no real difference either way). Indeed one can use just about any length bond funds and get very similar results (shown in another recent thread).


hi Rodc,
I see a lot of different datasets on French's website. Which are the "standard Fama-French benchmark portfolio data"? I'd like to write some backtesting code. BTW have you ever made you Matlab code publicly available? (hint hint)
thanks

Great thread

CrankyCube
Posts: 70
Joined: Mon Apr 07, 2008 10:26 am

Postby CrankyCube » Thu Jul 10, 2008 3:35 pm

very interesting thread. missed it the first time around.

thanks

:beer

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Thu Jul 10, 2008 3:44 pm

diasurfer wrote:
Rodc wrote:
A question: RodC, where did you get your data?


http://mba.tuck.dartmouth.edu/pages/fac ... brary.html

The stock returns are the standard Fama-French benchmark portfolio data, the 90-day T-Bills came from Global Financial Data. I'm used to using those as the "risk free" investment, although I could just as easily used the T-Bill data from F-F, which if memory serves are 30-day (but it makes no real difference either way). Indeed one can use just about any length bond funds and get very similar results (shown in another recent thread).


hi Rodc,
I see a lot of different datasets on French's website. Which are the "standard Fama-French benchmark portfolio data"? I'd like to write some backtesting code. BTW have you ever made you Matlab code publicly available? (hint hint)
thanks

Great thread


F-F_Benchmark_Portfolios_Annual.txt, down near the bottom of the page.
Listed as "Fama/French Benchmark Portfolios"

I'm afraid the code is hacked up as I keep using it for different things. However if you are familiar with Matlab this sort of stuff is not too hard to pull together.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

detifoss
Posts: 554
Joined: Sat Oct 13, 2007 10:38 pm

Postby detifoss » Mon Aug 18, 2008 1:49 pm

wow - thanks OP
I too missed this the first time around and have recently been mulling a bit of portfolio tweaking as I understand things a bit better.

This is one of the best posts I've read on here - thank you.

mike_slc
Posts: 966
Joined: Fri May 16, 2008 11:04 pm

Postby mike_slc » Mon Aug 18, 2008 2:33 pm

Perhaps this should be on the wiki?

Zelkha
Posts: 1
Joined: Thu Aug 21, 2008 11:10 am

Re: Drawdown risk

Postby Zelkha » Thu Aug 21, 2008 11:19 am

RobertH wrote:Rodc,

Very interesting stuff, nicely presented. Thank you for posting this. I've forwarded this link to some friends of mine who don't follow this stuff every day.

If you're thinking about building on this work in the future, I would be interested in an analysis that concentrated on drawdown risk of these portfolios. I believe many novice investors would also view that as an important - perhaps bottom-line - risk measure.

For myself, I view drawdown risk as having two components: maximum drawdown (how far does the portfolio fall from its previous high) and drawdown duration (how long does it take to get back to the previous high). I like to use inflation adjusted numbers to put the 1970s in proper perspective.

Robert



Rob,
As a novice to the asset allocation world, your note on "drawdown risk as having two components" strongly resonates with me. Please let me know if you do further develop this work. Thanks.

User avatar
Dan Moroboshi
Posts: 866
Joined: Sat Jul 07, 2007 7:10 pm

Postby Dan Moroboshi » Fri Aug 22, 2008 8:39 am

mike_slc wrote:Perhaps this should be on the wiki?


Another vote for posting this on the wiki.

I'd be interested in seeing the percentages of T-Bills required to adjust the various portfolios for different standard deviations - e.g. 10% and 20% as well.

(My compulsive side would like to see a range of portfolios adjusted for SDs of 5%, 7.5%, 10%, 12.5%, 15%, 17.5% and 20%, but I don't think there would be much additional information that couldn't be gleaned from the 10% and 20% numbers.)

Thanks for your time and effort, Rodc.

User avatar
PiperWarrior
Posts: 4068
Joined: Fri Dec 21, 2007 4:55 am
Location: right on course

Postby PiperWarrior » Fri Aug 22, 2008 9:18 am

Dan Moroboshi wrote:
mike_slc wrote:Perhaps this should be on the wiki?


Another vote for posting this on the wiki.

OK. This is on my radar. I'll do something unless Barry or others beat me to it. Currently, the wiki mention little about value tilt or small tilt.

veryanya
Posts: 55
Joined: Wed Mar 12, 2008 1:45 pm

Postby veryanya » Wed Jan 21, 2009 9:25 pm

this is very interesting..
where did you get the data to do this??

thanks,
anya

veryanya
Posts: 55
Joined: Wed Mar 12, 2008 1:45 pm

Postby veryanya » Wed Jan 21, 2009 9:27 pm

nevermind I see you mention it! thanks!

freedomfunds
Posts: 426
Joined: Thu Jan 15, 2009 4:17 am
Location: Los Angeles,CA

Postby freedomfunds » Wed Jan 21, 2009 9:49 pm


Your charts are consistent with the results of similar research I did for my All About Asset Allocation book back in 2005. I found a US allocation of 60% TSM, 20% SV, 10% REIT and 10% microcap worked well. That is very close to your 60% TSM and 40% small value blend.



Why do you ignore midcaps?

Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?
Index choice matters. | | Valuations matter even more.

User avatar
tetractys
Posts: 4496
Joined: Sat Mar 17, 2007 3:30 pm
Location: Along the Salish Sea

Postby tetractys » Wed Jan 21, 2009 10:58 pm

freedomfunds wrote:Why do you ignore midcaps?

That would reduce or eliminate the barbell effect, or the rebalancing benefit. The name of the game is putting together assets with lower correlations. -- Tet

On Approach
Posts: 437
Joined: Fri Aug 22, 2008 10:11 am

Postby On Approach » Thu Jan 22, 2009 12:40 am

99.44% (my favorite big percentage. )


That wouldn't have anything to do with Ivory soap, would it :wink: .

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Thu Jan 22, 2009 3:00 pm

freedomfunds wrote:

Your charts are consistent with the results of similar research I did for my All About Asset Allocation book back in 2005. I found a US allocation of 60% TSM, 20% SV, 10% REIT and 10% microcap worked well. That is very close to your 60% TSM and 40% small value blend.



Why do you ignore midcaps?

Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?


I'm not sure if that is directed to Rick who was quoted or to me.

FWIW: I "ignored" mid caps because there are no Fama-French mid-cap benchmark portfolios on the one hand, and on other hand, the number of possible portfolios blows up like number of assets factorial (n!) and ran into the limit of what I could run.

Tet also provides a reason why commonly mid caps are left out.
Last edited by Rodc on Thu Jan 22, 2009 3:02 pm, edited 1 time in total.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Thu Jan 22, 2009 3:01 pm

On Approach wrote:
99.44% (my favorite big percentage. )


That wouldn't have anything to do with Ivory soap, would it :wink: .


:)
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

GeekedOut
Posts: 248
Joined: Fri Jul 18, 2008 3:06 pm

Postby GeekedOut » Thu Jan 22, 2009 4:03 pm

This post is inspiring, and I just spend five minutes on "The Google" looking at pulling stock quotes for MATLAB. Apparently, this is now a built in function via the Datafeed Toolbox

The amazing things they come up with... I'll definately have to geek out a bit this weekend and see what I can throw together in order to spend way more time tracking minutia :P

Edit: I just noticed that you can pull historical data as well.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Thu Jan 22, 2009 4:22 pm

GeekedOut wrote:This post is inspiring, and I just spend five minutes on "The Google" looking at pulling stock quotes for MATLAB. Apparently, this is now a built in function via the Datafeed Toolbox

The amazing things they come up with... I'll definately have to geek out a bit this weekend and see what I can throw together in order to spend way more time tracking minutia :P

Edit: I just noticed that you can pull historical data as well.


I learn something new every day. :)
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

james22
Posts: 1091
Joined: Tue Aug 21, 2007 2:22 pm

Re: Updated figures and results on slice and dice vs TSM

Postby james22 » Fri Jan 23, 2009 1:45 am

Rodc wrote:To recap: asset class benchmark portfolios (LG,LC,LV,SG,SC,SV) from F-F.


Could we see the same using Vanguard fund data?

Rodc wrote:Suppose someone sets a risk level of standard deviation equal to 15%.


Could we see how a TIPS/SV (Swedroe-like) portfolio would compare?

Thanks.
Market Timer

YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Updated figures and results on slice and dice vs TSM

Postby YDNAL » Fri Jan 23, 2009 6:36 am

Rodc,

Was that Piper?

Great job with cool presentation, regardless of my sentiments about theoretical discussions of backtests. Wouldn't it be ironic if SV tanks for the next 80 years?... I'll be almost 137yo then. :)
Landy | Be yourself, everyone else is already taken -- Oscar Wilde

jimdigriztn
Posts: 403
Joined: Tue Feb 19, 2008 3:25 pm

Postby jimdigriztn » Fri Jan 23, 2009 9:00 am

freedomfunds wrote:Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?


80%

User avatar
deepdrive
Posts: 924
Joined: Sat Dec 20, 2008 5:03 pm

Postby deepdrive » Fri Jan 23, 2009 6:00 pm

jimdigriztn wrote:
freedomfunds wrote:Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?


80%


TSM is 72% large cap stocks.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Updated figures and results on slice and dice vs TSM

Postby Rodc » Fri Jan 23, 2009 6:04 pm

james22 wrote:
Rodc wrote:To recap: asset class benchmark portfolios (LG,LC,LV,SG,SC,SV) from F-F.


Could we see the same using Vanguard fund data?


Sufficient data do not exist.

Simba's back test sheet (search the site) will help you do something similar on what data do exist (not the graphs though).

Rodc wrote:Suppose someone sets a risk level of standard deviation equal to 15%.


Could we see how a TIPS/SV (Swedroe-like) portfolio would compare?

Thanks.


Same problem, TIPS are a new invention.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Updated figures and results on slice and dice vs TSM

Postby Rodc » Fri Jan 23, 2009 6:06 pm

YDNAL wrote:Rodc,

Was that Piper?

Great job with cool presentation, regardless of my sentiments about theoretical discussions of backtests. Wouldn't it be ironic if SV tanks for the next 80 years?... I'll be almost 137yo then. :)


Piper?

Ironic? Maybe, but it would not be the first time some really smart guys came up with a cool backtested theory that failed going forward.

DFA so far is doing pretty darn well.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Postby Rodc » Fri Jan 23, 2009 6:08 pm

deepdrive wrote:
jimdigriztn wrote:
freedomfunds wrote:Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?


80%


TSM is 72% large cap stocks.


Depends entire on how you define large caps: there is not one accepted definition. Depending on definition it also changes over time as cap weights change.

Last I looked though, the S&P 500 covered about 78% of TSM, but that was several months ago, certainly could be 72% today, or next year.

At the time I did the above, using the F-F definition, large cap was about 92%, so more like Russel 1000 than S&P 500 (as far as I know both are considered "large cap")

If memory serves Vanguard large cap uses the MSCI 750 which was about 86% of TSM at one point when I looked (but my memory is fuzzy)
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

jimdigriztn
Posts: 403
Joined: Tue Feb 19, 2008 3:25 pm

Postby jimdigriztn » Fri Jan 23, 2009 6:42 pm

deepdrive wrote:
jimdigriztn wrote:
freedomfunds wrote:Isnt TSM weighted 92 percent to the 500 stocks in S and P anyway?


80%


TSM is 72% large cap stocks.


Hi deepdrive,
I was using fund holdings info downloaded from Vanguard. I wrote a program that downloads the information and allows me to compare fund overlap. The stocks in VFINX make up 80.24% of VTSMX. On the other hand, the stocks in MGC (Vanguard's Mega Cap Index ETF) make up 73.75% of VTSMX, so that may be what you are thinking of in terms of Large Cap.

Jim

james22
Posts: 1091
Joined: Tue Aug 21, 2007 2:22 pm

Re: Updated figures and results on slice and dice vs TSM

Postby james22 » Sat Jan 24, 2009 1:15 am

Rodc wrote:Simba's back test sheet (search the site) will help you do something similar on what data do exist (not the graphs though).


Oh, I've spent a lot of time with Simba's tool; just looking for the pretty graphs.

Thanks.
Market Timer


Return to “Investing - Theory, News & General”

Who is online

Users browsing this forum: CurlyDave, rrppve and 33 guests