slice and dice tracking error study
slice and dice tracking error study
http://home.comcast.net/~rodec/finance/ ... _error.pdf
The above link is to a short study of tracking error using the Fama-French benchmark portfolios. Three pages plus figures.
Historically small value tilted portfolios have often had greater returns than TSM even when nominally they have been adjusted to have equal "risk" (admitting up front that the definitions of risk generally used are less than perfect, so please keep that in mind).
But to get that superior return an investor had to endure periods when performance lagged the returns of the total market, and many investors don't have the stomach to do that.
This study might help one judge if they will be able to endure the possible tracking error on the way to possible benefits.
Let me know what you think. Don't be shy. :lol:
The above link is to a short study of tracking error using the Fama-French benchmark portfolios. Three pages plus figures.
Historically small value tilted portfolios have often had greater returns than TSM even when nominally they have been adjusted to have equal "risk" (admitting up front that the definitions of risk generally used are less than perfect, so please keep that in mind).
But to get that superior return an investor had to endure periods when performance lagged the returns of the total market, and many investors don't have the stomach to do that.
This study might help one judge if they will be able to endure the possible tracking error on the way to possible benefits.
Let me know what you think. Don't be shy. :lol:
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.
Very interesting article! Thanks for writing this up so nicely.
However, I couldn't find where you got the numbers from,
ie. what is TSM or SV? Did you use the data from Fama's website?
It would be interesting to run these experiments with real funds as far as that is possible historically. This would factor in transaction costs and fees which can have a real impact especially for small and value tilting.
Thanks again,
-- Daniel
ps: I have to say though that I like your graphs from this post even more:
http://diehards.org/forum/viewtopic.php ... light=3x33
Perhaps you can combine those into the article? You can then become the John Norstad of slice-and-dice
However, I couldn't find where you got the numbers from,
ie. what is TSM or SV? Did you use the data from Fama's website?
It would be interesting to run these experiments with real funds as far as that is possible historically. This would factor in transaction costs and fees which can have a real impact especially for small and value tilting.
Thanks again,
-- Daniel
ps: I have to say though that I like your graphs from this post even more:
http://diehards.org/forum/viewtopic.php ... light=3x33
Perhaps you can combine those into the article? You can then become the John Norstad of slice-and-dice
Perhaps I should have defined things more clearly. I think the first time each such term is used I defined it, but did not draw attention with italics or anything and maybe I should have.daniel wrote:Very interesting article! Thanks for writing this up so nicely.
However, I couldn't find where you got the numbers from,
ie. what is TSM or SV? Did you use the data from Fama's website?
Yes the data are from the Fama website.
It would be interesting to run these experiments with real funds as far as that is possible historically. This would factor in transaction costs and fees which can have a real impact especially for small and value tilting.
Thanks again,
-- Daniel
ps: I have to say though that I like your graphs from this post even more:
http://diehards.org/forum/viewtopic.php ... light=3x33
Perhaps you can combine those into the article? You can then become the John Norstad of slice-and-dice
The problem with real funds, which would certainly be nice, is that histories are too short when you consider the cycles seem to be a decade or longer. That said, at some point I would like to compare some real funds to these benchmark portfolios as a reality check and write that up.
Transaction costs for Vanguard funds are not too much of an issue. I have modeled them, but since different people have different costs I don't generally include them. This is especially true over short periods like 5 years.
As to the other graphs, these were really meant as a counter point to those: they really are companion pieces and I should make that more clear. I hope to have time to put these all together in a web site. I like to keep these things short as that is what most people will read, but would like to make this a more clearly defined set of papers that people could read little by little.
Thanks for the comments.
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.
-
- Posts: 3665
- Joined: Thu Oct 18, 2007 11:42 am
I don't know, but I have always been underwhelmed by this "tracking error" effect. I think it's telling that we typically only consider UNDER-performance (vs. whatever) to be tracking error. When I do better than the "market", that's not an error, is it?
Anyway, my strongest experience with this sort of thing stems from sticking with an AA within equity of 50% foreign and 50% of that as EM -- for years and years. Sure there were times when I felt left out. But that was the idea in holding asset classes that wouldn't always march together.
If you rationally decide to split a TSM allocation into 50% TSM and 50% SCV, are you really going to sit at home nightly and feel bad that the market news on TV (i.e. TSM) seems better than your 50/50 portfolio? Can't you remember the good times to carry you through the "bad"?
Anyway, my strongest experience with this sort of thing stems from sticking with an AA within equity of 50% foreign and 50% of that as EM -- for years and years. Sure there were times when I felt left out. But that was the idea in holding asset classes that wouldn't always march together.
If you rationally decide to split a TSM allocation into 50% TSM and 50% SCV, are you really going to sit at home nightly and feel bad that the market news on TV (i.e. TSM) seems better than your 50/50 portfolio? Can't you remember the good times to carry you through the "bad"?
I am not a professional investment guy, but from everything I have read, the majority of investors deal very poorly with negative tracking error. People also often over-estimate their ability to deal with negative tracking error. This of course does not mean everyone.Tramper Al wrote:I don't know, but I have always been underwhelmed by this "tracking error" effect. I think it's telling that we typically only consider UNDER-performance (vs. whatever) to be tracking error. When I do better than the "market", that's not an error, is it?
Anyway, my strongest experience with this sort of thing stems from sticking with an AA within equity of 50% foreign and 50% of that as EM -- for years and years. Sure there were times when I felt left out. But that was the idea in holding asset classes that wouldn't always march together.
If you rationally decide to split a TSM allocation into 50% TSM and 50% SCV, are you really going to sit at home nightly and feel bad that the market news on TV (i.e. TSM) seems better than your 50/50 portfolio? Can't you remember the good times to carry you through the "bad"?
So, my hope is that this sort of thing will help someone make better decisions. Clearly some people will decide they can withstand negative tracking error, and some of those people will be correct. Others will come to the other decision. Either way is fine with me.
I should also add, maybe someone will read this and through gaining a better understanding find the will to deal with a period of negative tracking error. One can always hope.
Last edited by Rodc on Tue Feb 05, 2008 2:00 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.
-
- Posts: 3665
- Joined: Thu Oct 18, 2007 11:42 am
Yes. I guess what I'm saying is that I have always had a good idea of what my "track" was, and that I never expected to have my portfolio go along exactly as did the DJIA on TV. It's a test of one's commitment to the AA, no question, but all the portfolio can do is go along the track that you put it on, as opposed to someone else's track.Rodc wrote: I am not a professional investment guy, but from everything I have read, the majority of investors deal very poorly with negative tracking error.
If it was very important to me to go up and down with the DJIA, for example, then I'd design my AA accordingly. Then when EM inevitably had a huge year, I'd just have say, well that's not the track I chose.
Last edited by Tramper Al on Tue Feb 05, 2008 2:05 pm, edited 2 times in total.
I guess my hope is that this might help someone develop that level of understanding.Tramper Al wrote:Yes. I guess what I'm saying is that I have always had a good idea of what my "track" was, and that I never expected to have my portfolio go along exactly as did the DJIA on TV. It's a test of one's commitment to the AA, no question, but all the portfolio can do is go along the track that you put it on, as opposed to someone else's track.Rodc wrote: I am not a professional investment guy, but from everything I have read, the majority of investors deal very poorly with negative tracking error.
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.
-
- Posts: 3665
- Joined: Thu Oct 18, 2007 11:42 am
I'm sure that it will. Well done that.Rodc wrote: I guess my hope is that this might help someone develop that level of understanding.
I'll just add that I think it's a matter of reference. Ideally, I want to set up my AA and then have the Nightly Business Report and Wall Street Journal reflect it's weighted composition in their market news coverage. If I'm 10% TIPS, and that's the only thing in my portfolio that went up today, there should be a couple of positive stories about TIPS on NBR, right? The tracking error thing gets to me when the news is all about the DJIA and maybe a little NASDAQ thrown in. Wouldn't it be very different (and OK bizarre) if there were a small value feature on CNBC once a week, reminding viewers of the historic premium for these factors?
If the US equity markets are beating everything else out there, week after week or month after month, then I'm not going to keep up. The key is simply to remember why.
Costs matter!
This analysis is incomplete, and thus misleading. Not to pick on Rod, as he merely repeats the nonsense we have seen so often before.
In the final paragraph it is stated: "Expenses were not estimated or included; you should deduct the expenses associated with your own investment options." Costs matter! Current costs for tilting portfolios in such fashion are not relevant to such an analysis of historical returns.
The question is- what would it have cost to tilt a portfolio during the period of study? Commisions, bid-ask spreads, taxes have not been accounted for. Those costs were all much higher for tilted portfolios relative to Market portfolios than they are currently. Please also recall that tilting and rebalancing could not have taken place in tax protected accounts, as they did not exist thruout most of the study period.
Historical investors faced with such extra costs for tilting a portfolio would certainly have demanded a greater premium for doing so- an "expense premium" if you will. There is no reason that such a historical "expense premium" should still exist in an era of fundamentally lowered cost structure.
A woefully incomplete description of the past may lead to unpleasant surprises if used as a guide to the future!
Z.
In the final paragraph it is stated: "Expenses were not estimated or included; you should deduct the expenses associated with your own investment options." Costs matter! Current costs for tilting portfolios in such fashion are not relevant to such an analysis of historical returns.
The question is- what would it have cost to tilt a portfolio during the period of study? Commisions, bid-ask spreads, taxes have not been accounted for. Those costs were all much higher for tilted portfolios relative to Market portfolios than they are currently. Please also recall that tilting and rebalancing could not have taken place in tax protected accounts, as they did not exist thruout most of the study period.
Historical investors faced with such extra costs for tilting a portfolio would certainly have demanded a greater premium for doing so- an "expense premium" if you will. There is no reason that such a historical "expense premium" should still exist in an era of fundamentally lowered cost structure.
A woefully incomplete description of the past may lead to unpleasant surprises if used as a guide to the future!
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Now Z, tell me what you really think, don't be shy.
Those are good points. By the same token though, would not those same factors be true of all equity investing? After all there was no low expense TSM fund back very far either.
To the extent those concerns are valid in general, might one conclude that history then overstates returns, but perhaps not risks (however defined), and perhaps not relative strength of different asset classes (the subject here)?
Or does that mean history teaches us nothing at all about investing in stocks and we are left only with the hope that markets are like our simple models?
Seems to me the best course is to use both history and theory, with a dash of common sense and moderation.
Those are good points. By the same token though, would not those same factors be true of all equity investing? After all there was no low expense TSM fund back very far either.
To the extent those concerns are valid in general, might one conclude that history then overstates returns, but perhaps not risks (however defined), and perhaps not relative strength of different asset classes (the subject here)?
Or does that mean history teaches us nothing at all about investing in stocks and we are left only with the hope that markets are like our simple models?
And so too a woefully incomplete model lacking any data?A woefully incomplete description of the past may lead to unpleasant surprises if used as a guide to the future!
Seems to me the best course is to use both history and theory, with a dash of common sense and moderation.
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.
Costs matter- alot!
Not even close! Remember: bid-ask spreads on small stock much higher than on large, sheer volume of high commision trading required to re-constitute/re-balance much higher for tilted portfolios, and TAXES. Not only higher capital gains taxes (to rebalance/reconstitute) but killer (historical) dividend taxation which would especially hit value tilting.Rodc wrote: Those are good points. By the same token though, would not those same factors be true of all equity investing?
Of course not (on both questions). We are left with the hope that we can improve on our over-simple models. Perhaps the F-F Small and Value "Risks" are really just "Expenses"? Surely someone has looked into that(?).... does that mean history teaches us nothing at all about investing in stocks and we are left only with the hope that markets are like our simple models?
Yes, indeed!Rodc wrote:Seems to me the best course is to use both history and theory, with a dash of common sense and moderation.Zalzel wrote:A woefully incomplete description of the past may lead to unpleasant surprises if used as a guide to the future!
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Z,
I should be more explicit on the taxes question:
I have specifically not included taxes as I am thinking of tax advantaged accounts. I need to edit the paper to make that clear.
Taxes do change things hugely.
I should be more explicit on the taxes question:
I have specifically not included taxes as I am thinking of tax advantaged accounts. I need to edit the paper to make that clear.
Taxes do change things hugely.
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,
Thanks for the paper and charts.
To me it is important to highlight tracking error since many investors jump on the hot item and drop it when it goes cold (at a much lower price). To rationally tilt towards small and value means understanding why you are doing it along with the resolve to stick with your AA over the very long term.
To Zalzel's point, I think the costs for a small and value tilt are somewhat a function of how your assets are located. If you have a large tax deferred space for smallcap and value then you can avoid some of the costs of capital gain distributions for these inefficient equity funds.
I know in my own case I would have a greater tilt if my taxable/tax deferred ratio was not 50/50. As it is my bonds in my tax deferred take up much of the room there and I do not want to hold value in a taxable account - I use the Total Stock Market Index and various International Index funds.
Also, having invested for a long period of time I have unrealized capital gains baked into my taxable equity holdings that I am not keen to realize until I'm required to sell the assets to cover living expenses.
Norm
Thanks for the paper and charts.
To me it is important to highlight tracking error since many investors jump on the hot item and drop it when it goes cold (at a much lower price). To rationally tilt towards small and value means understanding why you are doing it along with the resolve to stick with your AA over the very long term.
To Zalzel's point, I think the costs for a small and value tilt are somewhat a function of how your assets are located. If you have a large tax deferred space for smallcap and value then you can avoid some of the costs of capital gain distributions for these inefficient equity funds.
I know in my own case I would have a greater tilt if my taxable/tax deferred ratio was not 50/50. As it is my bonds in my tax deferred take up much of the room there and I do not want to hold value in a taxable account - I use the Total Stock Market Index and various International Index funds.
Also, having invested for a long period of time I have unrealized capital gains baked into my taxable equity holdings that I am not keen to realize until I'm required to sell the assets to cover living expenses.
Norm
Costs (historical) vs. costs (current)
Hi Norm.AzRunner wrote:... I think the costs for a small and value tilt are somewhat a function of how your assets are located.
You are confusing the current cost of tilting with what it would have cost to tilt during the timeframe that F-F initially studied. An obvious question is: have the Small and Value premia been severely overestimated because of the failure to account for costs during the period under study? One would think that 15 years having elapsed since F-F published their landmark study, this question would have been asnwered. Perhaps it has.
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Oh yes... taxes...
Yes, they do, and they did! It would be just-plain-flatout-wrong to continue to ignore what expenses, including taxes, would have done to the theoretical historical portfolios of F-F.Rodc wrote: Taxes do change things hugely.
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Rod,
Great work buddy! You are never going to please everybody. Fact is, we have the data, and we should learn from it.
There is no way possible to evaluate every cost associated with making investment decisions 20, 30, or 50 years ago. Nor is it truely possible to incorporate the tremendous strides we have made in dealing with and overcoming fricitions in the marketplace over the last few decades. (by the way, apparently someone forgot to tell SG about the "cost/fee premium"? No doubt the highest fee/tax dimension of the market). The primary benefits of observing such historical behavior is to understand the tracking error risks and to help develop rational expectations. In the end, we are all free to assign the expected risk premiums we are most comfortable with.
That being said, its probably more than just a minor coincidence that the annualized 1993-2007 (longest period available) DFA style tilted US Equity portfolios (US Large Co., US Large Value, US Micro, and US Small Value) have exactly matched the FF/CRSP annualized returns from the 1926-1993 period, no*?
Somehow I doubt the conspiracy theorists will run out of ammo (blanks) any time soon! :lol:
sh
*literally, they are all within 0.1% of eachother...save the S&P 500 from 27-92 , which is about 0.3% ahead of DFLCX from 93-07
Great work buddy! You are never going to please everybody. Fact is, we have the data, and we should learn from it.
There is no way possible to evaluate every cost associated with making investment decisions 20, 30, or 50 years ago. Nor is it truely possible to incorporate the tremendous strides we have made in dealing with and overcoming fricitions in the marketplace over the last few decades. (by the way, apparently someone forgot to tell SG about the "cost/fee premium"? No doubt the highest fee/tax dimension of the market). The primary benefits of observing such historical behavior is to understand the tracking error risks and to help develop rational expectations. In the end, we are all free to assign the expected risk premiums we are most comfortable with.
That being said, its probably more than just a minor coincidence that the annualized 1993-2007 (longest period available) DFA style tilted US Equity portfolios (US Large Co., US Large Value, US Micro, and US Small Value) have exactly matched the FF/CRSP annualized returns from the 1926-1993 period, no*?
Somehow I doubt the conspiracy theorists will run out of ammo (blanks) any time soon! :lol:
sh
*literally, they are all within 0.1% of eachother...save the S&P 500 from 27-92 , which is about 0.3% ahead of DFLCX from 93-07
Potential conflict of interest? Conspiracy?
Your point? Past investors wouldn't have taken these into account? They don't matter for judging the import of historical returns? It is impossible to estimate which strategy they would have effected more?SmallHi wrote: There is no way possible to evaluate every cost associated with making investment decisions 20, 30, or 50 years ago.
Agreed. Some do, of course, have more of a vested interest in this than others. For example, one whose income is at least partly dependent on attracting and maintaining clients by advocating (some might say "selling") the benefits of tilting. I have no such potential conflict of interest. How about you SmallHi? Are you willing to disclose whether or not you receive compensation for assisting/advising individuals with their investments?In the end, we are all free to assign the expected risk premiums we are most comfortable with.
What conspiracy would that be?SmallHi wrote: Somehow I doubt the conspiracy theorists will run out of ammo (blanks) any time soon! :lol:
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
As far as the question of the quality of the Fama-French data: I’m not equipped to argue for or against it. I would note a few things: These guys are very highly regarded, their papers passed the peer review test, and the test of at least some time as the early papers have been out since 1992, through bear and bull markets, and through other academics trying to tear them down. A very successful investment company was founded on these and related data and these types of studies. If someone has actual data that refute the F-F data, by all means bring that forth; I will look at them with an open mind.
Despite the above I am always cognizant of the fact that sometimes highly regarded people make mistakes. It is possible in the twilight of my investing career I will have a conversation like this:
Z: Remember “Dogs of the DOW”! (laugh) How could people have been so stupid!
R: Yeah, and that F-F nonsense! Oh man, what was I thinking!?
But I may have a different conversation:
R: Remember “Dogs of the DOW”! (laugh) How could people have been so stupid!
Z: Yeah, and believing in silly little models that said TSM was the bomb! Hoo-boy!
In the end more than either of these scenarios, I believe that any rational highly diversified, low cost, low turn-over strategy held consistently over time will do well enough over time.
The point of this conversation was simply to look at one aspect of that “held consistently over time” part.
As to the question of fees going forward, this figure shows the effect of adding in fees and turn-over drag. For turn-over drag I used 1 basis point for each 1% of turn-over. I believe Mr Bogle has used this formula. Recently I saw someone else use 1.2 basis points for each 1% of turn-over, which is very close to the Bogle factor. I used Vanguard fees and turn-over for the TSM part (0.19%, 4%). Vanguard fees and turn-over for the SV part (0.23%, 25%), and again with DFA fees for the SV part (0.53%, 27%). These fees are per Yahoo. In each case I assumed you could buy T-Bills direct at no cost. Given the short period of compounding (5 years) it is no surprise that you can hardly tell the results one from another.
As to taxes, the consensus is that these sorts of portfolios belong in tax advantaged accounts in which case taxes are a non-issue. If one wants to do this in a taxable account, tax situations being so variable, they will need to address that with a tax expert.
Despite the above I am always cognizant of the fact that sometimes highly regarded people make mistakes. It is possible in the twilight of my investing career I will have a conversation like this:
Z: Remember “Dogs of the DOW”! (laugh) How could people have been so stupid!
R: Yeah, and that F-F nonsense! Oh man, what was I thinking!?
But I may have a different conversation:
R: Remember “Dogs of the DOW”! (laugh) How could people have been so stupid!
Z: Yeah, and believing in silly little models that said TSM was the bomb! Hoo-boy!
In the end more than either of these scenarios, I believe that any rational highly diversified, low cost, low turn-over strategy held consistently over time will do well enough over time.
The point of this conversation was simply to look at one aspect of that “held consistently over time” part.
As to the question of fees going forward, this figure shows the effect of adding in fees and turn-over drag. For turn-over drag I used 1 basis point for each 1% of turn-over. I believe Mr Bogle has used this formula. Recently I saw someone else use 1.2 basis points for each 1% of turn-over, which is very close to the Bogle factor. I used Vanguard fees and turn-over for the TSM part (0.19%, 4%). Vanguard fees and turn-over for the SV part (0.23%, 25%), and again with DFA fees for the SV part (0.53%, 27%). These fees are per Yahoo. In each case I assumed you could buy T-Bills direct at no cost. Given the short period of compounding (5 years) it is no surprise that you can hardly tell the results one from another.
As to taxes, the consensus is that these sorts of portfolios belong in tax advantaged accounts in which case taxes are a non-issue. If one wants to do this in a taxable account, tax situations being so variable, they will need to address that with a tax expert.
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.
-
- Posts: 3665
- Joined: Thu Oct 18, 2007 11:42 am
Really, is it the "consensus" that TSM belongs in tax managed accounts?Rodc wrote: As to taxes, the consensus is that these sorts of portfolios belong in tax advantaged accounts in which case taxes are a non-issue.
Maybe among those who have nothing but tax managed accounts or those who own nothing but TSM!
Yes, it might be better to turn that around. The point was just that you generally don't want to slice and dice in a taxable account. You could certainly put say TSM in taxable, and SV and bonds in tax advantaged, etc.Tramper Al wrote:Really, is it the "consensus" that TSM belongs in tax managed accounts?Rodc wrote: As to taxes, the consensus is that these sorts of portfolios belong in tax advantaged accounts in which case taxes are a non-issue.
Maybe among those who have nothing but tax managed accounts or those who own nothing but TSM!
At any rate you don't want to slice and dice in a way that you generate a lot of taxes, thus I don't consider taxes. If someone wants to slice and dice in a way that tax considerations are important, that is beyond what I want to deal with.
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.
Fiction piled onto fiction = fiction squared.
F-F did not attempt to account for the expenses a real world investor would have incurred in attempting to invest in anything approaching their idealized/simulated/imaginary portfolios/indexes. Thus, their model is incomplete. I doubt they would argue otherwise, as I doubt that they are fools. The model, at best, describes an idealized/non-existant past that may be viewed as a starting point to further assessing a more real past, and a rather more uncertain future.Rodc wrote:Despite the above I am always cognizant of the fact that sometimes highly regarded people make mistakes.
Overlaying current (incomplete, BTW) costs on top of data from a period in which such costs were an order of magnitude larger, in an attempt to demonstrate anything- anything whatsoever- demonstrates only a lack of seriousness (in my opinion only, of course).
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Z,
That particular graph was not directed towards any of your comments.
I am glad to see you edited your original post in this slot to be less rude. It is a start, if somewhat incomplete.
R
That particular graph was not directed towards any of your comments.
I am glad to see you edited your original post in this slot to be less rude. It is a start, if somewhat incomplete.
R
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.
Tracking error
.
Rodc,
Thanks for your efforts. It’s an important question to try to answer IMO.
FWIW - I made an attempt at the question in an earlier post on the M* forum. Here is what was posted earlier:
Here is the historical tracking error across a spectrum of US value titled portfolios with value loadings ranging from 0.2 to 0.8 with different size loadings ranging from 0 to 0.4. Tracking error is measured as the annual return of the value-size tilted portfolio minus the annual return of the total stock market. The table below lists the five largest annual tracking errors from 1929 to 2005 - and it can be significant - particularly for portfolios with large value tilts. For example the portfolio with a value loading of 0.8 with zero size tilt under-performed the total market by 26.6% in 1999 (+25.3 versus -1.3%), whereas a portfolio with half the value loading had half the tracking error. Interestingly the tracking error declines when adding a small cap tilt to a value portfolio. So the tables below may be a useful guide to matching value and size tilts to tracking error tolerance?
Again FWIW – my portfolio benchmark, with a US size and value loading of 0.2 and 0.4 has underperformed the total market on occasion by up to about 10 percent i.e. an annual return 10 percent lower that the return of the total US stock market. I should expect that this can happen again (or certainly not be surprised if it happens again).
Hope the tables are readable.
Robert
.
Rodc,
Thanks for your efforts. It’s an important question to try to answer IMO.
FWIW - I made an attempt at the question in an earlier post on the M* forum. Here is what was posted earlier:
Here is the historical tracking error across a spectrum of US value titled portfolios with value loadings ranging from 0.2 to 0.8 with different size loadings ranging from 0 to 0.4. Tracking error is measured as the annual return of the value-size tilted portfolio minus the annual return of the total stock market. The table below lists the five largest annual tracking errors from 1929 to 2005 - and it can be significant - particularly for portfolios with large value tilts. For example the portfolio with a value loading of 0.8 with zero size tilt under-performed the total market by 26.6% in 1999 (+25.3 versus -1.3%), whereas a portfolio with half the value loading had half the tracking error. Interestingly the tracking error declines when adding a small cap tilt to a value portfolio. So the tables below may be a useful guide to matching value and size tilts to tracking error tolerance?
Again FWIW – my portfolio benchmark, with a US size and value loading of 0.2 and 0.4 has underperformed the total market on occasion by up to about 10 percent i.e. an annual return 10 percent lower that the return of the total US stock market. I should expect that this can happen again (or certainly not be surprised if it happens again).
Code: Select all
Specturm of value portfolio returns minus total stock market return.
Year
.......Value loadings (with 0 size loading)
.........0.2......0.4......0.6.......0.8
1999....-6.6....-13.3....-19.9.....-26.6
1934....-5.9....-11.7....-17.6.....-23.5
1980....-4.9.....-9.8....-14.7.....-19.6
1939....-3.5.....-7.0....-10.4.....-13.9
1931....-3.2.....-6.5....-9.7......-13.0
.......Value loadings (with 0.2 size loading)
.........0.2......0.4......0.6......0.8
1999....-3.7....-10.3....-17.0....-23.6
1980....-3.8.....-8.7....-13.6....-18.5
1998....-6.9.....-8.7....-10.5....-12.3
1969....-5.0.....-7.0.....-9.1....-11.2
1990....-4.7.....-6.7.....-8.6....-10.6
.......Value loadings (with 0.4 size loading)
.........0.2......0.4......0.6......0.8
1998...-11.9....-13.7....-15.5....-17.3
1969....-7.8.....-9.9....-12.0....-14.1
1990....-7.5.....-9.5....-11.4....-13.3
1980....-2.6.....-7.5....-12.4....-17.3
1930....-4.8.....-7.5....-10.3....-13.0
1999....-0.8.....-7.4....-14.0....-20.7
Source: Derived from the Fama-French factor data on the Ken French website.
Robert
.
Question(s) to Robert
Hi Robert.
The numbers you elegantly provide, I doubt you would deny, are largely imaginary/simulated/stripped-of-real-world-costs.
The effect of:
- trading costs and bid-ask spreads an order of magnitude larger than those of today,
- absent (pre-1975) IRA accounts, and (pre-1981) 401k accounts,
- dividends subject to taxation (pre-2003) as ordinary income,
- and top tax rates of 63% and more (1932-1981)
no doubt entered into the calculations of investors.
All of these costs have impacted tilted-type portfolios moreso than Market-type portfolios. Thus, there can be no doubt that the F-F analysis overestimated the Small and Value premia that could have been obtained in such a real world. If you disagree with this then you need read no further.
Here is what I think is the interesting question: to what extent are the pre-(let's say) 1991 S/V premiums not Risk premiums but, rather, "cost of strategy" premiums? That is, would not investors have stayed away from Small and Value until the expected return compensated them for the increased cost of the strategy?
If the above is the case (I think it is), and the "cost of strategy" has come down markedly (of course it has), then I think the Small/Value premia going forward will necessarily reflect this.
The above does not specifically address the problems I see in the OP of this thread, nor in your addition to it, but hints at why I think both are in error and misleading.
Z.
The numbers you elegantly provide, I doubt you would deny, are largely imaginary/simulated/stripped-of-real-world-costs.
The effect of:
- trading costs and bid-ask spreads an order of magnitude larger than those of today,
- absent (pre-1975) IRA accounts, and (pre-1981) 401k accounts,
- dividends subject to taxation (pre-2003) as ordinary income,
- and top tax rates of 63% and more (1932-1981)
no doubt entered into the calculations of investors.
All of these costs have impacted tilted-type portfolios moreso than Market-type portfolios. Thus, there can be no doubt that the F-F analysis overestimated the Small and Value premia that could have been obtained in such a real world. If you disagree with this then you need read no further.
Here is what I think is the interesting question: to what extent are the pre-(let's say) 1991 S/V premiums not Risk premiums but, rather, "cost of strategy" premiums? That is, would not investors have stayed away from Small and Value until the expected return compensated them for the increased cost of the strategy?
If the above is the case (I think it is), and the "cost of strategy" has come down markedly (of course it has), then I think the Small/Value premia going forward will necessarily reflect this.
The above does not specifically address the problems I see in the OP of this thread, nor in your addition to it, but hints at why I think both are in error and misleading.
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Zalzel --
No one is denying investment costs were higher 30 or 40 years ago. Unfortunately, no one (excluding those just acting on a hunch) can prove that return premiums decline, or historical relationships are unattainable-- with real world strategies. Quite the opposite, as a matter of fact.All of these costs have impacted tilted-type portfolios moreso than Market-type portfolios. Thus, there can be no doubt that the F-F analysis overestimated the Small and Value premia that could have been obtained in such a real world. If you disagree with this then you need read no further.
What you fail to understand is, at the very minute we learn about a risk/return relationship, it is modeled and "real world" drawbacks are incorporated into the investment process. And the post-portfolio formation results indicate no disruption in results or return premiums. This has been the case for almost 4 decades!
A few examples should suffice:
Equities -- research into the behavior of markets indicates that broad based, diversified exposure to the entire market is a profitable long term strategy. Cap weighted index funds are developed at very minimal costs -- likely jeapordizing the future equity risk premium as more investors have direct access to it. Despite the overwhelming popularity, the equity risk premium from targeting the Vanguard S&P 500 fund has been 40% higher (net of fees) over the last 30 years than it had from the late 20s through the early 70s.
Small Stocks -- a historical premium available from buying small cap stocks is potentially absorbed by market impact. Therefore, a low cost passive strategy is developed in 1982 that targets microcap stocks without adhereing to a strict indexing approach -- allowing flexibility in the buy/sell decisions. Result: live net of fee returns since 1982 exceed the simulated results of the CRSP 9-10 Index.
Value stocks -- a historical premium potentially unattainable due to the difficulty in maintaining exposure to the lowest price firms marketwide. Strategies are developed to overcome these drawbacks through portfolio hold ranges and daily rebalancing with new cash flows. Result: Live returns since 1993 have exceeded the returns of the FF Indexes they were modeled after NET of fees from 26-92 -- successfully capturing the simulated/historical (and extremely lucrative) premiums out of sample.
On an after tax basis since 1993, a 50/50 Large Value + Small Value live fund allocation has matched the gross of fee and tax simulated returns of the FF xUtility Indexes they track over this period.
Oh, and, by the way...the relative diversification benefits of Small and Value (which were so helpful from 1966-1982 when S&P 500 lasted almost two decades with 0 real returns -- completely destroying any chance of a successful retirement for an entire generation) just happened to show up again from 2000 - 2007 as we endure yet another intermediate period where dedicated market exposure is unable to reward investors with any return above inflation. So the relative behavior (as well as the absolute behavior) can be harnessed in the real world as well.
Sadly, reading over your comments Zalzel, I am reminded of a recent Bill Bernstein quote I believe applies to you:
investing has always been, and will remain, an operation in which wealth is transferred from those without a working knowledge of financial history to those who have one
My advice to investors: focus on the big picture, let Gus Sauter, John Montgomery, and Robert Deere handle the details. They haven't let us down yet!
sh
Re: slice and dice tracking error study
Interesting paper... one thing I didn't understand was how you rebalanced between the portfolios and how you rebalanced for risk as well. It would be nice to clear that up along with any other assumptions made in the paper.Rodc wrote:Let me know what you think. Don't be shy. :lol:
On the first page you state: "Any such study comes with a host of caveats which are listed at the end of this article." Yet I didn't see a list of caveats at the end of the article.
Next, even though you can't track real funds you SHOULD track fees associated with those funds. For example, it's more expensive to maintain a SV fund than the TSM. I know Vanguard makes this cheap today, but that wasn't the case in the past.
Lastly, assuming that most investors keep FI in their retirement accounts and stocks in their taxable, studies like these should take into account taxes, which would kill all the returns of F-F funds IMHO.
Boris
EDIT: Wow, I must be tired, I didn't even notice all the replies and I guess some of the issues I posed have already been asked. Sorry
Short term moves in the market are like "a tale Told by an idiot, full of sound and fury, Signifying nothing." |
- John C. Bogle quoting Shakespeare
40 years (sic)
4 decades? The F-F analysis was published only 15 yrs ago. It took a while for people to catch on. The early adopters (I include myself) did well as the Small and Value premia were bid down. That's my view, and it allows for a prediction. People who expect Small and Value premia close to what F-F "demonstrated" are going to be very disappointed. Time will tell- not I, and not SmallHi.SmallHi wrote:
What you fail to understand is, at the very minute we learn about a risk/return relationship, it is modeled and "real world" drawbacks are incorporated into the investment process. And the post-portfolio formation results indicate no disruption in results or return premiums. This has been the case for almost 4 decades!
As for
...investing has always been, and will remain, an operation in which wealth is transferred from those without a working knowledge of financial history to those who have one
I do not for a second think that those who understand and adopt a Total Global Market Portfolio approach to investing are going to transfer their wealth to anyone. I asked on a previous occasion, and I continue to wonder about this, just who is it who is going to pay for all the Small/Value tilt yachts? It just seems too easy, too pat- doncha think?
One more time: time will tell!
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Re: slice and dice tracking error study
Not stated, but rebalanced yearly. For risk I took the entire 80 year dataset and added just enough T-Bills to get the standard deviation of yearly returns over the 80 years equal to 15%. Then the amount of T-Bills was held steady. For example 75%TSM/25%T-Bills was the "TSM" risk adjusted portfolio. The slice and dice were at about 40% T-Bills (differed plus or minus depending on 4x25, 3x33, etc.)Boris wrote:Interesting paper... one thing I didn't understand was how you rebalanced between the portfolios and how you rebalanced for risk as well. It would be nice to clear that up along with any other assumptions made in the paper.Rodc wrote:Let me know what you think. Don't be shy. :lol:
You sure?On the first page you state: "Any such study comes with a host of caveats which are listed at the end of this article." Yet I didn't see a list of caveats at the end of the article.
I did some of that in the follow-up graph. I can only indicate that current fees, the one you and I will pay going forward, would do. You can see it makes so little difference and everyone has different fees, so I left them out. If one stays with index funds the expense differences are fairly small, and only apply to part of the portfolio which makes the difference even smaller. But it is a judgment call. But the point is that fees going forward are not a problem.Next, even though you can't track real funds you SHOULD track fees associated with those funds. For example, it's more expensive to maintain a SV fund than the TSM. I know Vanguard makes this cheap today, but that wasn't the case in the past.
As you note below, I think this one has been taken care of.Lastly, assuming that most investors keep FI in their retirement accounts and stocks in their taxable, studies like these should take into account taxes, which would kill all the returns of F-F funds IMHO.
We've all been there at one time or another.EDIT: Wow, I must be tired, I didn't even notice all the replies and I guess some of the issues I posed have already been asked. Sorry
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.
I think that is an interesting question. FWIW: I expect TGMP buy and hold will get exactly their "fair share" so they will not be buying anyone any yachts.I do not for a second think that those who understand and adopt a Total Global Market Portfolio approach to investing are going to transfer their wealth to anyone. I asked on a previous occasion, and I continue to wonder about this, just who is it who is going to pay for all the Small/Value tilt yachts? It just seems too easy, too pat- doncha think?
The founder of a multi-billion dollar investment company called me up the other day after one of his clients pointed out one of my studies here. (Isn't the internet a marvelous place?) Apparently he paid a visit to DFA and asked the researchers there if I was on the right track and they did not disagree. Or so he told me.
He is interested in tilting, and does some for his clients all of who have at least $1 million invested. He is more interested in preservation of wealth than trying for extra returns. He however is properly concerned with how well we really know this tilting business is going to turn out well over the next 25 years which is the time frame of most of his clients (not at all sure was my reply and I sent him one of the tracking error plots, which is what prompted me to write this up as a follow up).
I commented about the problem that if we all tilted we'd all be above average! Despite his concern about tilting in general, he dismissed that out of hand: The big guys who hold the vast majority of the money have too much money to be able to invest in small cap stocks. I know Buffet has said much the same.
So the answer just might be the big guys with the yachts will spin off a percent or so of possible return to the little guys to whom it will look like a lot of money.
But you are absolutely right, only time will tell.
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.
Re: slice and dice tracking error study
Actually, that's the other thing that made me go "hmmmm" when I read the paper. I was surprised how much more FI was needed to keep the same SD with the F-F portfolios (especially the 4x25 one). All this talk of non-correlated assets and less volatility goes out the window.Rodc wrote:Not stated, but rebalanced yearly. For risk I took the entire 80 year dataset and added just enough T-Bills to get the standard deviation of yearly returns over the 80 years equal to 15%. Then the amount of T-Bills was held steady. For example 75%TSM/25%T-Bills was the "TSM" risk adjusted portfolio. The slice and dice were at about 40% T-Bills (differed plus or minus depending on 4x25, 3x33, etc.)Boris wrote:Interesting paper... one thing I didn't understand was how you rebalanced between the portfolios and how you rebalanced for risk as well. It would be nice to clear that up along with any other assumptions made in the paper.Rodc wrote:Let me know what you think. Don't be shy. :lol:
Bottom line is, your paper showed, just like others with Small/Value tilt that those particular segments of the market had higher returns. It's pretty much a given at this point. More risk, more reward. I don't think value is a free lunch deal, they've been doing well lately due to the tax-breaks and in fact I believe (as zalzel correctly pointed out), that a lot of the investment decisions are based on government regulations and taxes. Especially when you consider large institutions who always try to optimize their AMT/taxes/etc.
By the way, I'm not stating that there won't be a small/value premium going forward, I don't know. But I do forecast, and I can almost guarantee this, that much of the money will move around due to regulation/tax changes. For example, I learned today that about a decade ago treasuries were callable but today they're not. This will certainly impact how people invest in treasuries and their payout. Remember when bonds were yielding 15%+? Imagine buying a 30-year treasury paying that much and it not being callable. Insane.
Boris
Short term moves in the market are like "a tale Told by an idiot, full of sound and fury, Signifying nothing." |
- John C. Bogle quoting Shakespeare
.
Zalzel,
Surprisingly close (obviously can be interpreted a number of ways).
There seem to be three sets of views about the value (and size) premium:
As you say time will tell.
Robert
.
Zalzel,
A hypothesis would then be that the pre-1991 value premium would be larger to account for the ‘extra costs’ of the strategy in that period (so as to equate the after cost premiums with the post-1991 period).Here is what I think is the interesting question: to what extent are the pre-(let's say) 1991 S/V premiums not Risk premiums but, rather, "cost of strategy" premiums? That is, would not investors have stayed away from Small and Value until the expected return compensated them for the increased cost of the strategy?
- If we assume: risk premium pre-1991 = risk premium post 1991
An the hypothesis is: cost of strategy pre-1991 > cost of strategy post 1991
then:
(risk premium + cost of strategy pre-1991) > (risk premium + cost of strategy post-1991)
or written another way:
HmL pre-1991 > HmL post-1991
Code: Select all
Here is the FF data say:
Average HmL Annualized HmL
1926-1990 5.0% 4.1%
1991-2007 5.7% 4.3%
1926-2007 5.2% 4.2%
There seem to be three sets of views about the value (and size) premium:
- 1) It not a risk story but purely an implementation cost story, so after costs you simply get a market return.
2) Its both a risk and implementation cost story, and historically implementation costs were higher so the combined effect of a continued risk premium achieved at lower cost will reduce the future HmL premium (as tested in the above example).
3) Its purely a risk story so future premiums will likely be the same order of magnitude as the past.
As you say time will tell.
Robert
.
Yes to (a modified) "2"!
Hi Robert.
The following excludes, without any harm done, the 1992-present period. This is because this period may best be viewed as a transitional one of rapidly changing transaction/tax costs. At the outset, I will say that while I think that what follows is a vast improvement over the supposition that what F-F calculated are Risk Premiums that will persist into the future, I have for the sake of clarity left out one thing (which can be returned to later perhaps).
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Edit: I apologize to all for any and all confusion/consternation the following has caused. I retract it and ask that it be ignored. I leave it in place to remind me that I am fallible (and because attempting to rewrite history is never acceptable).
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Risk Premium pre-1991 = Risk Premium post-2007 = Risk Premium
Idealized, cost-free (Fama-French) Historical Returns - Historical Costs = Risk Premium
Forward Returns = Risk Premium - Forward Costs
Thus,
Forward Returns - (F-F) Historical Returns = - Forward Costs - Historical Costs
which is,
Forward Returns = (F-F) Historical Returns - (Historical Costs + Forward Costs)
This nicely shows that the greater were the (ignored) costs in the period F-F analyzed, the lower Forward Returns will appear compared to the (idealized) Returns F-F calculated! That may not, at first glance, be intuitively obvious, but it is correct. And please recall- that is assuming no change in the Risk Premiums between the time of F-F study and now!
Thanks,
Z.
This last need not be a "hypothesis" as it is easily verifiable.Robert T wrote:.
A hypothesis would then be that the pre-1991 value premium would be larger to account for the ‘extra costs’ of the strategy in that period (so as to equate the after cost premiums with the post-1991 period).
- If we assume: risk premium pre-1991 = risk premium post 1991
An the hypothesis is: cost of strategy pre-1991 > cost of strategy post 1991
You have muddled it here, as you have contradicted your assumption (above) of equal pre- and post- 1991 Risk Premiums.Robert T wrote:then:
(risk premium + cost of strategy pre-1991) > (risk premium + cost of strategy post-1991)
or written another way:
HmL pre-1991 > HmL post-1991
The following excludes, without any harm done, the 1992-present period. This is because this period may best be viewed as a transitional one of rapidly changing transaction/tax costs. At the outset, I will say that while I think that what follows is a vast improvement over the supposition that what F-F calculated are Risk Premiums that will persist into the future, I have for the sake of clarity left out one thing (which can be returned to later perhaps).
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Edit: I apologize to all for any and all confusion/consternation the following has caused. I retract it and ask that it be ignored. I leave it in place to remind me that I am fallible (and because attempting to rewrite history is never acceptable).
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Risk Premium pre-1991 = Risk Premium post-2007 = Risk Premium
Idealized, cost-free (Fama-French) Historical Returns - Historical Costs = Risk Premium
Forward Returns = Risk Premium - Forward Costs
Thus,
Forward Returns - (F-F) Historical Returns = - Forward Costs - Historical Costs
which is,
Forward Returns = (F-F) Historical Returns - (Historical Costs + Forward Costs)
This nicely shows that the greater were the (ignored) costs in the period F-F analyzed, the lower Forward Returns will appear compared to the (idealized) Returns F-F calculated! That may not, at first glance, be intuitively obvious, but it is correct. And please recall- that is assuming no change in the Risk Premiums between the time of F-F study and now!
Thanks,
Z.
Last edited by zalzel on Sat Feb 09, 2008 12:20 am, edited 1 time in total.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
The degree to how much extra FI is needed to normalize the standard deviation came as a surprise to me too. I expected that more would be needed for the tilted portfolios, but the magnitude was larger than I might have guessed. This is important because I’m not sure too many people who tilt increase FI over what they might otherwise hold. Of course most also diversify further than just US equity and Bonds and that must help some.Actually, that's the other thing that made me go "hmmmm" when I read the paper. I was surprised how much more FI was needed to keep the same SD with the F-F portfolios (especially the 4x25 one). All this talk of non-correlated assets and less volatility goes out the window.
Bottom line is, your paper showed, just like others with Small/Value tilt that those particular segments of the market had higher returns. It's pretty much a given at this point. More risk, more reward.
As to your second point, we always expect to see (even if it sometimes fails to show up, as in small growth which seems to be an anomaly to me) greater expected returns to go with greater expected volatility. This makes it hard to see whether small or value (or something else) has any return benefit beyond what would be expected due to the greater volatility. That is the whole reason to first normalize the standard deviation. Any difference thereafter is due to something other than simple differences in volatility.
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.
.
Zalzel,
Risk premium + cost of strategy = HmL (which you call the FF historical returns).
Forward returns = risk premium (or expected return = expected risk premium)
Or, using your notation:
Forward returns = cost-free (Fama-French) Returns – forward costs.
You seem to have subtracted costs twice, if I followed your sequence correctly.
Robert
.
Zalzel,
I don’t think so – at least by my understanding as:You have muddled it here, as you have contradicted your assumption (above) of equal pre- and post- 1991 Risk Premiums
Risk premium + cost of strategy = HmL (which you call the FF historical returns).
IMO this should read:Forward Returns = Risk Premium - Forward Costs
Forward returns = risk premium (or expected return = expected risk premium)
Or, using your notation:
Forward returns = cost-free (Fama-French) Returns – forward costs.
You seem to have subtracted costs twice, if I followed your sequence correctly.
Robert
.
Re: slice and dice tracking error study
This is essentially what Larry Swedroe does with his own portfolio, but in reverse. He keeps the return constant and uses the value tilting to lower the amount of equities required and allow a large increase in FI to drastically lower the SD.Boris wrote:Actually, that's the other thing that made me go "hmmmm" when I read the paper. I was surprised how much more FI was needed to keep the same SD with the F-F portfolios (especially the 4x25 one). All this talk of non-correlated assets and less volatility goes out the window.
Paul
Even Larry is only human, and must quiver before time.
Hi Paul.
Z.
Not to be pedantic, but it is important to keep tense correct here. Not even Larry Swedroe can do what you say. What he does do is: hope/attempt to maintain Expected Return with a lower dispersion of Returns. I have, in this and other threads, detailed why I think S/V tilters (including Larry) will be sorely disappointed. Only Time, in her terrible beauty and mercilessness, will tell.stratton/Paul wrote:This is essentially what Larry Swedroe does with his own portfolio, but in reverse. He keeps the return constant and uses the value tilting to lower the amount of equities required and allow a large increase in FI to drastically lower the SD.
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
.
Zalzel,
We seem to disagree on the following:
I believe that:..........................................Actual realizable return = HmL – Costs
You believe that (as I understand it):......Actual realizable return = (HmL – Costs) - Costs
Where HmL is the FF ‘value premium’ without costs, and ‘Costs’ = strategy implementation costs.
My question is why include costs twice?
Yes, HmL – Costs = the risk premium but why subtract costs again from the risk premium (the strategy has already been implemented)? The risk premium is the after cost reward investors demand for taking risk (they will only invest if they are satisfied that the after cost return is a sufficient reward for the risk taken). I also believe that costs are a small part of HmL, while he risk premium is a large part.
Just trying the clarify the area of disagreement.
Robert
PS: The above just uses HmL for simplicity. The full HmL is not easy to achieve – rather investors try to capture a share of HmL (e.g. value loading less than 1), which is easier to realize.
.
Zalzel,
We seem to disagree on the following:
I believe that:..........................................Actual realizable return = HmL – Costs
You believe that (as I understand it):......Actual realizable return = (HmL – Costs) - Costs
Where HmL is the FF ‘value premium’ without costs, and ‘Costs’ = strategy implementation costs.
My question is why include costs twice?
Yes, HmL – Costs = the risk premium but why subtract costs again from the risk premium (the strategy has already been implemented)? The risk premium is the after cost reward investors demand for taking risk (they will only invest if they are satisfied that the after cost return is a sufficient reward for the risk taken). I also believe that costs are a small part of HmL, while he risk premium is a large part.
Just trying the clarify the area of disagreement.
Robert
PS: The above just uses HmL for simplicity. The full HmL is not easy to achieve – rather investors try to capture a share of HmL (e.g. value loading less than 1), which is easier to realize.
.
Do-over
Hi Robert.
Evidently, I was taking down my earlier post at the same moment as you were putting up your latest. Sorry to you, and anyone else who got confused.
the historic (FF) Risk Premium is inflated due to the costs an investor would have had to overcome to capture that premium. Perhaps an example will make my admittedly jumbled thinking clearer?
Divide time into three eras: a-FF (ante-FF), transition, and p-FF (post-FF). I don't know the length of the transition period, but I wouldn't be surprised if it turns out to have been approximately 15 years (that is, roughly, 1992-2007).
Suppose that in a-FF, total cost for a .2/.4 S/V tilted portfolio was 3%, and for a TSM portfolio, 1% (I have plucked these numbers out of thin air- so please don't focus too much on them). Now suppose an investor wanted to tilt because they felt that there was a Risk to, and therefor a Premium for, doing so. Would such an investor tilt for a 1% premium? Of course not- not with it costing 2% to do so. He might bide his time until he felt the Risk Premium had increased to 6%. At that point, he might indeed decide to tilt at a 2% cost for a net 4% gain. That would represent a 200% return on his incremental investment.
Several years later, FF see that a tilted portfolio, such as the above investor's, returned a premium of 6% over a Market Portfolio. Do they therefor conclude that the Risk Premium was 6%? Sure. But what does that tell us about the Risk Premium for such a tilted portfolio in p-FF?
Suppose the costs have have come down by a factor of 10? That is, .1% for a Market portfolio, and .3% for a tilted portfolio? Would the Risk Premium remain at 6%? My first instinct would be that it should decrease to 4.2% so that by taking on the Risk the investor still clears 4% as a-FF.
But, should one believe that by adding .2% to portfolio cost the investor can now hope to attain a net 4% gain- that is, a 2000% return on the incremental investment as compared with the (only) 200% incremental gain of their forfather? One might infer that the p-FF Risk Premium, to satisfy arbitrageurs with similar value judgements as their forefathers, could be as low as .6%, thereby offering a net return of .4% on the incremental .2% investment (200%- same as during a-FF).
What do you think?
Z.
Evidently, I was taking down my earlier post at the same moment as you were putting up your latest. Sorry to you, and anyone else who got confused.
I was in error. I got confused from the following consideration:Robert T wrote: My question is why include costs twice?
the historic (FF) Risk Premium is inflated due to the costs an investor would have had to overcome to capture that premium. Perhaps an example will make my admittedly jumbled thinking clearer?
Divide time into three eras: a-FF (ante-FF), transition, and p-FF (post-FF). I don't know the length of the transition period, but I wouldn't be surprised if it turns out to have been approximately 15 years (that is, roughly, 1992-2007).
Suppose that in a-FF, total cost for a .2/.4 S/V tilted portfolio was 3%, and for a TSM portfolio, 1% (I have plucked these numbers out of thin air- so please don't focus too much on them). Now suppose an investor wanted to tilt because they felt that there was a Risk to, and therefor a Premium for, doing so. Would such an investor tilt for a 1% premium? Of course not- not with it costing 2% to do so. He might bide his time until he felt the Risk Premium had increased to 6%. At that point, he might indeed decide to tilt at a 2% cost for a net 4% gain. That would represent a 200% return on his incremental investment.
Several years later, FF see that a tilted portfolio, such as the above investor's, returned a premium of 6% over a Market Portfolio. Do they therefor conclude that the Risk Premium was 6%? Sure. But what does that tell us about the Risk Premium for such a tilted portfolio in p-FF?
Suppose the costs have have come down by a factor of 10? That is, .1% for a Market portfolio, and .3% for a tilted portfolio? Would the Risk Premium remain at 6%? My first instinct would be that it should decrease to 4.2% so that by taking on the Risk the investor still clears 4% as a-FF.
But, should one believe that by adding .2% to portfolio cost the investor can now hope to attain a net 4% gain- that is, a 2000% return on the incremental investment as compared with the (only) 200% incremental gain of their forfather? One might infer that the p-FF Risk Premium, to satisfy arbitrageurs with similar value judgements as their forefathers, could be as low as .6%, thereby offering a net return of .4% on the incremental .2% investment (200%- same as during a-FF).
What do you think?
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Surely it would have been better to have posted a correction as a new post. Now this thread makes no sense to someone just coming to this thread.Evidently, I was taking down my earlier post at the same moment as you were putting up your latest. Sorry to you, and anyone else who got confused.
Just joshing Z, we've all made posting mistakes, some of us even make the same mistakes.
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.
.
Zalzel,
No problem with the earlier post – we are all learning.
The after cost return on capital IMO should be the same across the two periods (for the same level of risk). The before-cost return on capital would likely be higher in the first period than the second period to attract the same level of investor demand.
Trying to use your example:
First period
Additional cost of a value tilted portfolio relative to the market = 2%
After cost return (premium) above the market return to investors = 4%
Before cost return = 6%
Second period
Additional cost of a value tilted portfolio relative to the market = 0.2%
After cost return (premium) above the market return to investors = 4%
Before cost return = 4.2%
In the above example, the after cost return on an investor’s incremental investment capital is the same across the two periods (4% not 2000% or 200%). If companies with the same risks as in the first period are to continue to attract capital from investors in the same way they will need to provide a similar after cost return to investors (4% not 0.4% or no one would invest in these businesses and they would probably not exist).
Just my take.
Robert
.
Zalzel,
No problem with the earlier post – we are all learning.
I think your first instinct was right in that investors (providers of capital) are compensated for risk taken (not for costs of a strategy). In your example you seem to have derived the estimated return (2000 vs 200) based on the cost of the strategy, and not on the capital provided by the investor. IMO it should be based on the latter not the former.What do you think?
The after cost return on capital IMO should be the same across the two periods (for the same level of risk). The before-cost return on capital would likely be higher in the first period than the second period to attract the same level of investor demand.
Trying to use your example:
First period
Additional cost of a value tilted portfolio relative to the market = 2%
After cost return (premium) above the market return to investors = 4%
Before cost return = 6%
Second period
Additional cost of a value tilted portfolio relative to the market = 0.2%
After cost return (premium) above the market return to investors = 4%
Before cost return = 4.2%
In the above example, the after cost return on an investor’s incremental investment capital is the same across the two periods (4% not 2000% or 200%). If companies with the same risks as in the first period are to continue to attract capital from investors in the same way they will need to provide a similar after cost return to investors (4% not 0.4% or no one would invest in these businesses and they would probably not exist).
Just my take.
Robert
.
Setting the record straight
To be clear: at the time I took down my post there were no responses to it on the board. Robert T's response appeared just as I went back to the Board to check things out. At that point, I couldn't put back my original post as I had not saved it. Every time I have made a change in a post that has been responded to, I have left the original intact and/or clearly delineate what changes have been made. I think that is the honest way to proceed. I could say more at this point, but I won't.Rodc wrote:Surely it would have been better to have posted a correction as a new post. Now this thread makes no sense to someone just coming to this thread.Evidently, I was taking down my earlier post at the same moment as you were putting up your latest. Sorry to you, and anyone else who got confused.
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
Thanks Robert.. Estimate of preFF costs?
Hi Robert.
Thank you.
BTW, do you have an off-the-cuff/what-the-heck/ballpark estimate of the pre-FF costs to achieve a reasonably diversified, sort of .2/.4 (S/V) tilted portfolio, for someone with a marginal tax rate of 50% (not by any means the highest rate during the preFF era)?
Z.
Thank you.
BTW, do you have an off-the-cuff/what-the-heck/ballpark estimate of the pre-FF costs to achieve a reasonably diversified, sort of .2/.4 (S/V) tilted portfolio, for someone with a marginal tax rate of 50% (not by any means the highest rate during the preFF era)?
Z.
"What we can't say we can't say, and we can't whistle it either." |
Frank P. Ramsey
.
Zalzel,
Here is the actual return of the Windsor fund (after fees, but not taxes) against the S&P500 index from 1964-1993. If I recall the value loading on the Windsor fund has historically been on average around 0.4 (but need to check). If we assume a 50% tax and a 20% turnover the after-tax return on the Windsor fund would have been closer to 12.5% (assuming I have done the calculation correctly). So if you add the expense ratio then the cost was possibly between 1.5 and 2%.
1964-1993
Vanguard Windsor Fund.........13.9%
S&P500..................................10.5%
Source: Tweedy Browne
FWIW - I believe that taxes will likely be the biggest cost (of any strategy). IMO taxes will likely only go up (may be wrong), and so I favor tax-managed funds over ETFs if they are available.
Robert
.
Zalzel,
Less than 3, but greater than 0.BTW, do you have an off-the-cuff/what-the-heck/ballpark estimate of the pre-FF costs to achieve a reasonably diversified, sort of .2/.4 (S/V) tilted portfolio, for someone with a marginal tax rate of 50% (not by any means the highest rate during the preFF era)?
Here is the actual return of the Windsor fund (after fees, but not taxes) against the S&P500 index from 1964-1993. If I recall the value loading on the Windsor fund has historically been on average around 0.4 (but need to check). If we assume a 50% tax and a 20% turnover the after-tax return on the Windsor fund would have been closer to 12.5% (assuming I have done the calculation correctly). So if you add the expense ratio then the cost was possibly between 1.5 and 2%.
1964-1993
Vanguard Windsor Fund.........13.9%
S&P500..................................10.5%
Source: Tweedy Browne
FWIW - I believe that taxes will likely be the biggest cost (of any strategy). IMO taxes will likely only go up (may be wrong), and so I favor tax-managed funds over ETFs if they are available.
Robert
.
TSM fund was started in 1987, not so long ago. The costs of that strategy would have also been more in the pre-computer era, along with those suspect Value stocks. I'm concerned that the historical cost of implementing a Small and Value tilt is being criticized without the consideration that all costs were much higher then.