Very bad idea: using index values in portfolio backtesting

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Very bad idea: using index values in portfolio backtesting

Post by fluffyistaken » Wed Jan 06, 2010 12:17 pm

This is going to be a bit rambling, but I hope this post will convey my reasons for taking many back-tested portfolios discussed here with a huge grain of salt... The basic problem as I see it is the use of various indices for older data, as opposed to actual fund returns.

I assume that Simba's spreadsheet contains the best available data sources (Trev H and others -- if you have other data sources, please post here). Looking at Simba's "data sources" tab, for some commonly backtested equity sectors, the following are the earliest years for which the results are based on actual fund returns and not on some index:

TSM: 1993
LCB: 1977
LCV: 1993
LCG: 1993
MCB: 1999
MCV: 1996
MCG: 1996
SCB: 1992
SCV: 1999
SCG: 1999
Microcap: 1998
Extended Market: 1988
REIT: 1997
Int'l Developed : 2001
Int'l Emerging: 1995 ? (not clear if 1995-2008 data is from index or actual fund)
Int'l Value: 1997

So we can see that with the exception of LCB (and probably safe to assume TSM since it's very close to LCB) the actual fund data goes back only to 1990s, sometimes late 1990s.

Why is this a problem?

Looking on page 33 of "Modern Investments and Security Analysis" by Fuller and Farrell, 1987 we find the following table summarizing trading costs, based on a study done in 1983 (excuse the poor image quality -- I can try to get a sharper photo if people have trouble reading the numbers):

Image

Using the data in another table on the same page (you can see some of that table at the top of the photo -- I can take a full photo of this table as well, if people are interested), we can replace the leftmost column of the table with the percentage of US TSM at the time as follows:

Securities with market cap of $0-10M: 0.36% of the market
$10-25M: 0.89%
$25-50M: 1.59%
$50-75M: 1.60%
$75-100M: 1.27%
$100-500M: 15.65%
$500-1000M: 12.29%
$1000-1500M: 8.87%
$1500+M: 57.48%

Please note the huge difference in a round-trip trade costs due to spread and commissions for smaller issues compared to larger issues! Comparing (roughly) apples to apples, consider a round-trip trading cost of a $25,000 block of a security with $25-50M market cap and a $2,500,000 block of a large cap security (over $1500M market cap): 7.6% vs. 2.8%! Of course not every security would go through a round trip every year, but even now Vanguard's SCV fund has 30% annual turnover. We're still talking around 2% annual underperformance due to higher trading costs for small caps compared to large caps.

And the above is 1983 data for US equities. It's probably safe to assume that the spreads were even larger prior to 1983 and larger still for foreign securities. Yet here on bogleheads we routinely see backtested portfolio charts that include SCV and foreign equities from 1970s!

Anyway I'll stop rambling now. My point is that it's disingenious at best to present index-derived data, especially for small and/or foreign securities, as "what actually happened". It would be a good idea for anyone posting backtested portfolio data to either use only actual fund returns (which limits us to only 15 years or so of data for most asset classes) or include a huge asterisk noting that data may well be overstating actual investor returns by several percentage points annually.

I suspect the same problem holds for historic bond returns, but probably to a much lesser degree.

All comments welcome.

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Re: Very bad idea: using index values in portfolio backtesti

Post by ddb » Wed Jan 06, 2010 12:29 pm

There are obvious problems with backtesting, but I'm not too concerned about the issue you raise. "Good" index fund providers somehow figure out a way to overcome trading costs. Observe the Vanguard Tax-Managed International Fund (VTMGX), which is an EAFE Index Fund. We have the following data:

<table class="tableizer-table">
<tr class="tableizer-firstrow"><th>Year</th><th>EAFE Index</th><th>VTMGX</th><th>Tracking Error</th></tr> <tr><td>2000</td><td>-14.17%</td><td>-14.29%</td><td>-0.12%</td></tr> <tr><td>2001</td><td>-21.44%</td><td>-21.94%</td><td>-0.50%</td></tr> <tr><td>2002</td><td>-15.94%</td><td>-15.62%</td><td>0.32%</td></tr> <tr><td>2003</td><td>38.59%</td><td>38.67%</td><td>0.08%</td></tr> <tr><td>2004</td><td>20.25%</td><td>20.25%</td><td>0.00%</td></tr> <tr><td>2005</td><td>13.54%</td><td>13.60%</td><td>0.06%</td></tr> <tr><td>2006</td><td>26.34%</td><td>26.27%</td><td>-0.07%</td></tr> <tr><td>2007</td><td>11.17%</td><td>11.15%</td><td>-0.02%</td></tr> <tr><td>2008</td><td>-43.38%</td><td>-41.27%</td><td>2.11%</td></tr> <tr><td>CAGR</td><td>-1.75%</td><td>-1.39%</td><td>--</td></tr> <tr><td>Average</td><td>--</td><td>--</td><td>0.21%</td></tr></table>


So, the average tracking error has been roughly equal to the fund's average expense ratio over the time period. I have no idea how they do this, given other estimates I see of trading costs, but they are pretty consistent across different asset classes in matching the index return minus expense ratio.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 12:36 pm

ddb, these results are for a large cap fund from the past decade. Do we have any evidence that the same was true for, say, small caps in 1980s? I don't see how any fund manager can overcome market spread and commission costs. Even now Vanguard's SCV fund has 30% annual turnover, I doubt it would've been less in 1970s or 1980s.

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Post by richard » Wed Jan 06, 2010 12:40 pm

If I understand correctly, the issue is that backtesting SV is misleading because of the high trading costs that existed more than a decade or two ago. If nothing else, actual investors would not have been able to achieve the reported index returns.

One might suppose SV companies had a higher cost of capital in those days (compared to both today and to large cap companies) due to these trading and liquidity issues. You'd need a higher expected return to overcome these costs. In today's market, costs are lower and returns might be expected to be lower.

This would add to all the usual problems with backtesting - lack of adequate amounts of data, lack of reliable data, changing conditions, statistical significance, the market taking prior data into account, etc.

mikep
Posts: 3698
Joined: Wed Apr 22, 2009 9:27 pm

Post by mikep » Wed Jan 06, 2010 12:44 pm

I have seen Trev include expense ratios or estimates (such as 5bps per month or similar) in his charts when using only the indexes.

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 12:48 pm

richard wrote:If I understand correctly, the issue is that backtesting SV is misleading because of the high trading costs that existed more than a decade or two ago. If nothing else, actual investors would not have been able to achieve the reported index returns.
Yes. But what struck me is the magnitude of trading costs for smaller issues. We're not talking a fraction of a percent here -- it's a cost of 5-10% or even more for a round-trip trade for a small cap trade in 1983. What practical use is FF data from 1940s for small caps given that even in 1983 the relatively higher trading costs come close to or exceed the expected SV premium?

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 12:52 pm

mikep wrote:I have seen Trev include expense ratios or estimates (such as 5bps per month or similar) in his charts when using only the indexes.
That would be a large underestimate, according to the tables I'm looking at. Spread/midprice (i.e. (ask - bid)/((bid + ask) / 2) ):

Securities with market cap of $0-10M: 6.55%
$10-25M: 4.07%
$25-50M: 3.03%
$50-75M: 1.86%
$75-100M: 1.46%
$100-500M: 1.13%
$500-1000M: 0.76%
$1000-1500M: 0.65%
$1500+M: 0.52%

This is not even considering the commission costs or the effects of "moving the market" with a large trade, just the spread, as of 1983.
Last edited by fluffyistaken on Wed Jan 06, 2010 1:26 pm, edited 2 times in total.

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Post by ddb » Wed Jan 06, 2010 12:56 pm

fluffyistaken wrote:ddb, these results are for a large cap fund from the past decade. Do we have any evidence that the same was true for, say, small caps in 1980s? I don't see how any fund manager can overcome market spread and commission costs. Even now Vanguard's SCV fund has 30% annual turnover, I doubt it would've been less in 1970s or 1980s.
I realize it's a large cap fund; I was responding to your claim that this issue is important "especially for small and/or foreign securities" by showing a foreign securities fund.

I have no idea what effect trading costs would have had on a small cap index fund in the 80s, but I also don't really care since I'm not making portfolio decisions based on how a non-existent fund performed in the 80s.

In other words, I wouldn't use backtesting at all when making go-forward portfolio decisions. I think investors should form risk-return expectations for various asset classes under consideration and allocate as appropriate.

The other issue is that it's hard to even evaluate with current funds the impact of trading costs. Take the WisdomTree International SmallCap Dividend Fund (DLS) - inception-to-date returns of -0.31%/year, but the index has had a return over the same time frame of 0.33%/year, so the fund lagged the index by 64bps per year. But, the expense ratio is 58bps, so there's only 6bps of unexplained tracking error there. So, did trading costs have no impact, or did they have an impact but the fund was just lucky that the sampling strategy led to positive tracking error which offset the transaction costs? Who knows? But, for the funds that I personally use, I'm comfortable that they will deliver the underlying index returns minus expense ratio over the long-run, because I haven't seen evidence which convinces me otherwise.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Post by richard » Wed Jan 06, 2010 1:00 pm

fluffyistaken wrote:Yes. But what struck me is the magnitude of trading costs for smaller issues. We're not talking a fraction of a percent here -- it's a cost of 5-10% or even more for a round-trip trade for a small cap trade in 1983. What practical use is FF data from 1940s for small caps given that even in 1983 the relatively higher trading costs come close to or exceed the expected SV premium?
Agreed. This is very significant data.

On first glance one may think the problem is solved due to today's lower trading costs and therefore the historic index spread between SV and large indexes should continue. It's important to realize that one reason SV had to offer such large returns in the past was to overcome costs. Therefore, today's lower costs does not mean we should expect the historic premiums to continue.

dumbmoney
Posts: 2292
Joined: Sun Mar 16, 2008 8:58 pm

Post by dumbmoney » Wed Jan 06, 2010 1:11 pm

Tracking error is not a measure of trading costs. If someone is trading in sync with the index, the index ceases to be an ideal measure of the market, since the index returns are influenced by index fund trading.
I am pleased to report that the invisible forces of destruction have been unmasked, marking a turning point chapter when the fraudulent and speculative winds are cast into the inferno of extinction.

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 1:16 pm

To ddb, there could have been some extra revenue for a fund from securities lending. Also, could Wisdom Tree have been capping costs (i.e. actual costs were higher than what investors paid)?

I agree that we should not select our AA based on backtesting, but the fact is that most of us do so, explicitly or implicitly, and a good number of forum members enthusiastically so :) . There is not much mention of how extremely unreliable deep backtesting (20 or more years in the past) is for more exotic asset classes like SCV or EM.

And even considering SmB premium itself -- how can we know its existence or magnitude without having a good hande on the difference in historic trading costs for small and large cap issues?

User avatar
BlueEars
Posts: 3634
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Wed Jan 06, 2010 1:20 pm

Interesting objection by fluffy to SV premium in far off past. I'm guessing that the objection to the value premium in mid and large caps is not as pronounced because trading costs were more reasonable. Correct me if I've interpreted the objection incorrectly.

I think the SV premium might have been captured by active funds in the 1970's. I recall that small cap funds did well in the 1970's after 1974. One example that comes to mind is Pennsylvania Mutual run by Royce, PENNX. I seem to recall it had a very large number of stocks in it so was probably an index like fund. You can see the chart going back to 1971 at M* here: http://quote.morningstar.com/fund/chart ... ture=en-US

Image

P.S. I do not think there is anyone investing who has not used some backtested results.

dkturner
Posts: 1358
Joined: Sun Feb 25, 2007 7:58 pm

Post by dkturner » Wed Jan 06, 2010 1:31 pm

fluffyistaken wrote:ddb, these results are for a large cap fund from the past decade. Do we have any evidence that the same was true for, say, small caps in 1980s? I don't see how any fund manager can overcome market spread and commission costs. Even now Vanguard's SCV fund has 30% annual turnover, I doubt it would've been less in 1970s or 1980s.


Here are the 1, 3, 5 and 10 year performance results for Vanguard Small Cap Value Fund, versus its benchmark:

Small-Cap Value Index Fund.....30.34% –6.29% 0.80% 7.69%
Spliced Small Cap Value Index* 30.29% –6.28% 0.88% 7.44%

Do you see any meaningful drag fronm market spreads and trading commissions? It looks like they cover a substantial portion of their expense ratios too.
Last edited by dkturner on Wed Jan 06, 2010 1:33 pm, edited 2 times in total.

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

Post by Rodc » Wed Jan 06, 2010 1:32 pm

richard wrote:
fluffyistaken wrote:Yes. But what struck me is the magnitude of trading costs for smaller issues. We're not talking a fraction of a percent here -- it's a cost of 5-10% or even more for a round-trip trade for a small cap trade in 1983. What practical use is FF data from 1940s for small caps given that even in 1983 the relatively higher trading costs come close to or exceed the expected SV premium?
Agreed. This is very significant data.

On first glance one may think the problem is solved due to today's lower trading costs and therefore the historic index spread between SV and large indexes should continue. It's important to realize that one reason SV had to offer such large returns in the past was to overcome costs. Therefore, today's lower costs does not mean we should expect the historic premiums to continue.
Agree that this is an issue. My take is that it is an issue for the small premium, not the value premium, since as far as I know, there is not particular reason to think in the past a value stock was more expensive to trade than a similar sized growth stock.

However, people used to trade a lot less (sorry no link to a source handy). Kind of like in cars, back when breaks were less reliable people left more room. As breaks got better they drive more close together. I read that the same is true after airbags: the safer people feel the more chances they take: conservation of danger or something (not entirely true, but a fair amount of the advance in safer cars is lost due to this behavior). So, if trades are expensive you trade less and keep your costs down. If trades get cheap, you trade more and still keep your costs in a similar range (perhaps). The explosion in trading only came with the advent of cheaper trading.

So, if the Total cost has not changed particularly it may not matter too much in the grand scheme of things. That of course is pure speculation as I have never seen any data one way or the other.

There are other factors too, like tax changes on capital gains, tax changes on dividends, etc. All of which work to render backtesting a shaky proposition.

Also, if this is a big concern we should see the premium (as seen in the FF data) change over time and I don't think we see that with value, might with small to some degree:

http://www.bogleheads.org/forum/viewtopic.php?t=12741
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.

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 1:33 pm

Les wrote:Interesting objection by fluffy to SV premium in far off past. I'm guessing that the objection to the value premium in mid and large caps is not as pronounced because trading costs were more reasonable. Correct me if I've interpreted the objection incorrectly.
Correct. I don't have any evidence that LCV has had higher historic trading costs than LCG (or SCV than SCG). So in a way, this data presents an objection to SmB premium, and not to HmL. Though I prefer to not take on FF just yet :D -- I just want to clear up some things about the ever-popular backtesting charts.

User avatar
nick22
Posts: 859
Joined: Sun Mar 04, 2007 11:00 am
Location: Ohio

Backtesting

Post by nick22 » Wed Jan 06, 2010 1:35 pm

why are we worried about the cost of trading 20-30 years ago? Anyone have a time machine?

Ok, just kidding. I think it has been well documented that the historical returns listed for SV, EM, etc. have obvious limitations. Certainly trading costs narrow the listed return gap probably by 1-2% (when compared to LCB) and risk premiums may explain the rest on any retrun gap. Several books discuss this at length, including a couple by Larry.

Backtesting is fun on the forum, but obviously should be used with disclaimers and caution. All we really know about investing is that we should save a lot, minimize expenses, diversify our holdings and rebalance. That is my summary of the evidence-base after 50 years of portfolio research.
Nick22

Trev H
Posts: 1874
Joined: Fri Mar 02, 2007 10:47 pm

Post by Trev H » Wed Jan 06, 2010 2:04 pm

Simba's original data set was one I provided (back around 2005 or 2006 best I remember).

I am not sure about the investment cost reflected in that older set.

In the set that I keep I use actual Vanguard Fund Returns (investor share class) where possible history wise, so the Investor Share Class ER is included in those annual returns.

Where I have to change to data provided by some index, I back out the ER of the same share class in the current vanguard fund.

I don't care what it cost to invest in SV or ISV back in 1970 or 1980, I simply reflect current day expenses (appropriate for us all going forward).

For example below for TSM prior to 1993 I use CRSP D1-10 but back out the ER of Vanguards TSM fund (investor share ER Cost).

TSM = US Cap Weighted Market:
=============================
1970-1992 CRSP Market Decile 1-10
1993-Forward Vanguards Total Stock Market Index Fund (VTSMX)

In the case of SV below, the same is done. Where Russell and Ibbotson data have to be used, the ER of VISVX (investor share class) is backed out.

SV = US Small Value:
====================
1970-1978 Ibbotson
1979-1998 Russell 2000 Value Index
1999-Forward Vanguards Small Cap Value Index Fund (VISVX)

Below shows ISV and the bp deducted for cost - the older stuff is from IFA.com, the more recent actual fund data includes the DFA funds ER.

ISV – International Small Value (including EM):
===============================================
1970-1994 85% ISV, 15% ESV (ifa.com)
ISV 1970-1981 Dimensional Intl Small Cap Index -4.58 bp/month
--- 1982-1994 Dimensional Intl Small Value Index -5.75 bp/month
--- 1995-Forward – DFA Intl Small Cap Value Fund (DISVX)
ESV = 50% EV, 50% ES *calculated* (ifa.com)
--- EV 1970-1988 IFA EM Index
------ 1989-1998 Fama/French EM Value Index -5.0 bp/month
------ 1999-Forward DFA EM Value Fund (DFEVX)
--- ES 1970-1988 IFA EM Index
------ 1989-1998 Fama/French EM Small Index -6.5 bp/month
------ 1999-Forward DFA EM Small Cap Fund (DEMSX)


===

User avatar
fluffyistaken
Posts: 1435
Joined: Fri Apr 04, 2008 1:32 pm

Post by fluffyistaken » Wed Jan 06, 2010 2:20 pm

Hi Trev, based on the data I see, I think the ER of Vanguards TSM fund is much less than what the real-life trading costs of a small-cap fund would have been (at least prior to 1983). As you can see from my OP trading costs were much higher for small caps compared to large caps. For a small cap index fund in 1983, we'd be talking 5-10% round-trip trade cost * annual portfolio turnover (15% for SCB now, 30% for SCV) or 0.75-3.00% range for ER without even including manager compensation, etc.

User avatar
BlueEars
Posts: 3634
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Wed Jan 06, 2010 2:24 pm

Just want to clarify what my PENNX example shows. There was a way to participate in the SV premium (I think) way back in the 1970's. But you didn't have the FF theory back then, nor did you have SV index funds. Plus you didn't have Bogleheads telling you about this stuff, etc. :D I think maybe AAII was the source way back in the 1980's for discussion of some sort of small cap premium and how individual investors could capture it with a list of individual stocks. Lots of things have changed and will continue to do so.

Lucio
Posts: 286
Joined: Wed Apr 08, 2009 10:31 pm

Post by Lucio » Wed Jan 06, 2010 2:26 pm

ddb wrote:
I have no idea what effect trading costs would have had on a small cap index fund in the 80s, but I also don't really care since I'm not making portfolio decisions based on how a non-existent fund performed in the 80s.
The DFA CRSP 9-10 fund (DFSCX) began operation in 1981. Perhaps that's a reasonable "index" fund to compare.
ddb wrote: The other issue is that it's hard to even evaluate with current funds the impact of trading costs. Take the WisdomTree International SmallCap Dividend Fund (DLS) - inception-to-date returns of -0.31%/year, but the index has had a return over the same time frame of 0.33%/year, so the fund lagged the index by 64bps per year. But, the expense ratio is 58bps, so there's only 6bps of unexplained tracking error there.
What about the revenue from securities lending, depreciation, interest and foreign currency gains? I suspect that you understand how to account for those factors when discussing tracking error, but I have no idea. Would you mind explaining that?

Lucio

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Post by ddb » Wed Jan 06, 2010 3:12 pm

Lucio wrote:
ddb wrote:
I have no idea what effect trading costs would have had on a small cap index fund in the 80s, but I also don't really care since I'm not making portfolio decisions based on how a non-existent fund performed in the 80s.
The DFA CRSP 9-10 fund (DFSCX) began operation in 1981. Perhaps that's a reasonable "index" fund to compare.
Yeah, but DFSCX isn't an index fund, so any difference between fund performance and CRSP 9-10 performance could easily be a result of the security-selection methodology that DFA uses, plus you have securities lending revenue. Would be impossible (I think) to pinpoint any performance differential to transaction costs alone.
Lucio wrote:
ddb wrote:The other issue is that it's hard to even evaluate with current funds the impact of trading costs. Take the WisdomTree International SmallCap Dividend Fund (DLS) - inception-to-date returns of -0.31%/year, but the index has had a return over the same time frame of 0.33%/year, so the fund lagged the index by 64bps per year. But, the expense ratio is 58bps, so there's only 6bps of unexplained tracking error there.
What about the revenue from securities lending, depreciation, interest and foreign currency gains? I suspect that you understand how to account for those factors when discussing tracking error, but I have no idea. Would you mind explaining that?
Not sure what impact depreciation/interest/currency changes have. The underlying index performance is reported in local currency on a total return basis, as is the fund tracking the index. The securities lending can certainly help offset trading costs, but I don't know to what extent.

It's easier to evaluate a fund that doesn't use sampling but actually replicates the full index, but we typically only see this with the types of funds that wouldn't be expected to pose a problem in terms of high internal transaction costs.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

Trev H
Posts: 1874
Joined: Fri Mar 02, 2007 10:47 pm

Post by Trev H » Wed Jan 06, 2010 5:38 pm

fluffy said...

==
Hi Trev, based on the data I see, I think the ER of Vanguards TSM fund is much less than what the real-life trading costs of a small-cap fund would have been (at least prior to 1983). As you can see from my OP trading costs were much higher for small caps compared to large caps. For a small cap index fund in 1983, we'd be talking 5-10% round-trip trade cost * annual portfolio turnover (15% for SCB now, 30% for SCV) or 0.75-3.00% range for ER without even including manager compensation, etc.
==

I agree that it may have cost more for SV in 1983, than it did for TSM but I don't really care about that.

Today - going forward - investor share classs for TSM cost .18 and SV .28

The CRSP D1-10 returns in my annual return data spreadsheet have .18 backed out, the Russell SV and Ibbotson SV have .28 backed out.

The difference in cost of SV and TSM back in 1983 may have been 5.00 (I don't really know, don't really care).

Today the cost difference in TSM and SV (investor share class) is 0.10.

My backtesting data reflects that.

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Post by ddb » Wed Jan 06, 2010 5:41 pm

Trev H wrote:fluffy said...

==
Hi Trev, based on the data I see, I think the ER of Vanguards TSM fund is much less than what the real-life trading costs of a small-cap fund would have been (at least prior to 1983). As you can see from my OP trading costs were much higher for small caps compared to large caps. For a small cap index fund in 1983, we'd be talking 5-10% round-trip trade cost * annual portfolio turnover (15% for SCB now, 30% for SCV) or 0.75-3.00% range for ER without even including manager compensation, etc.
==

I agree that it may have cost more for SV in 1983, than it did for TSM but I don't really care about that.

Today - going forward - investor share classs for TSM cost .18 and SV .28

The CRSP D1-10 returns in my annual return data spreadsheet have .18 backed out, the Russell SV and Ibbotson SV have .28 backed out.

The difference in cost of SV and TSM back in 1983 may have been 5.00 (I don't really know, don't really care).

Today the cost difference in TSM and SV (investor share class) is 0.10.

My backtesting data reflects that.
Trev, I think you're missing fluffy's point about the non-reported (or more specifically, not included in the fund's expense ratio) internal transaction costs, which are comprised of direct brokerage costs, bid-ask spreads, and market impact costs.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

matt
Posts: 2305
Joined: Sun Mar 04, 2007 3:47 pm

Post by matt » Wed Jan 06, 2010 6:01 pm

The high costs of small cap trading and the severe limitations of backtesting are well known by, well, investors who know investing well.

I don't believe there is any meaningful premium that will be earned through small cap indexing. Patient investors who try to buy at the bid and sell at the ask have a chance of earning a liquidity premium, but they also risk missing out on reversals in market trends. Since indexing is not patient, there is no reason to believe it will capture any liquidity premium; more likely to lose the premium, in fact.

Value stocks, on average, are smaller than growth stocks, so there is a bias to the historical data as well, but I don't think it is quite as problematic. I still don't think value indexes are particularly good at value investing, but at least there is a reasonable chance to outperform the broad market.

Personally, I think indexing really only works well with the total market. For extra return potential, I prefer active value management over small/value index tilts.

Indexer88
Posts: 419
Joined: Mon Jan 05, 2009 5:32 pm

Post by Indexer88 » Wed Jan 06, 2010 6:15 pm

Thank you, fluffyistaken (whatever that alias means!!) for this real-world investment observation vis-a-vis the overly optimistic charts (based upon academic studies) that promise SCV will triple our investments. This ought to shake up the SCV kool-aid. Just joking.... Really, though I have to say TrevH's charts are great for graphics but highly misleading for investments.

There was a great thread about how SCV incurrs greater tax liabilities. Now we see that there probably was no real-world way of indexing SCV in the past b/c of trading costs. In other words, the sellers of SCV stocks demanded up front a higher price, i.e. Mr. Market wasn't going to just let you buy these small companies for nothing. There's no free lunch folks -- take SCV volatility for real.

Beware, Mr. Market appears to give SCV a couple percentage points over 40 years, but for five or ten or twenty, be prepared for major losses. Real returns are different than "expected" returns.

However, buying SCV on a major downturn, when the volatility is baked in, would be a different story, but that involves market timing, and that is generally not a good idea.

SCV is 4.5% of the market. When you see "SCV," think potential bubble. Mr. Market is always willing to sell you some sector or factor that is overvalued.

Get back to no debt, emergency fund, TSM, US and World, age in bonds and sleep better.

User avatar
chuck-lyn
Posts: 294
Joined: Wed Apr 15, 2009 12:54 pm

Post by chuck-lyn » Wed Jan 06, 2010 8:45 pm

Speaking of bubbles, what about the bubble that Mr. Market blew during the dot.com craze? That sort of problem is the real risk, IMHO, and it is the major reason I am listening to the sweet call of Mel's beloved mid-caps .... not too big, not too small, not too valuey, not too growthy, but just right. :roll:

Cheers,

charlie

User avatar
snodog
Posts: 267
Joined: Fri Dec 21, 2007 5:33 pm
Location: Pennsylvania

Post by snodog » Wed Jan 06, 2010 10:11 pm

I must be missing something but how can SCV be a bubble? When a stock gets too overpriced the index will drop it.

Indexer88
Posts: 419
Joined: Mon Jan 05, 2009 5:32 pm

Post by Indexer88 » Thu Jan 07, 2010 7:58 am

Mid-caps "...not too big, not too small, not too valuey, not too growthy, but just right." That sounds like a good philosophy for life, but I don't think it matches a more general approach to the markets. Your trying to beat the market and you may fall behind it.

SCV can't be in a bubble? Just because it's "value" doesn't mean there are times when it sells too high. And it can always go very, very low and still remain "value."

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Post by ddb » Thu Jan 07, 2010 8:38 am

matt wrote:Value stocks, on average, are smaller than growth stocks, so there is a bias to the historical data as well, but I don't think it is quite as problematic.
According to Morningstar, the geometric average weighted market cap of underlying holdings of the MSCI US Large Cap Growth Index is $53.8 billion, compared to $57.0 billion for the MSCI US Large Cap Value Index. The figures for the MidCap indexes are $4.6 billion for growth and $4.4 billion for value.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

matt
Posts: 2305
Joined: Sun Mar 04, 2007 3:47 pm

Post by matt » Thu Jan 07, 2010 8:54 am

ddb wrote:
According to Morningstar, the geometric average weighted market cap of underlying holdings of the MSCI US Large Cap Growth Index is $53.8 billion, compared to $57.0 billion for the MSCI US Large Cap Value Index. The figures for the MidCap indexes are $4.6 billion for growth and $4.4 billion for value.
This points out one of the weaknesses in broad indexing of growth vs. value. Most of the stocks in a large or mid value index aren't really value stocks, they're just less "growthy" than the growth stocks. I approach investing from an absolute value, absolute return viewpoint. Simply having a lower valuation than the broad market does not make a value stock; it is the absolute valuation that matters. DFA and indexes such as S&P's Style Pure Value are improvements over the inclusive index groups such as MSCI since they reduce the number of stocks that qualify as value, but they are still based on relative valuations. I want to buy value stocks that are cheap, period; not just cheap compared to expensive stocks.

User avatar
ddb
Posts: 5509
Joined: Mon Feb 26, 2007 12:37 pm
Location: American Gardens Building, West 81st St.

Post by ddb » Thu Jan 07, 2010 9:26 am

matt wrote:ddb wrote:
According to Morningstar, the geometric average weighted market cap of underlying holdings of the MSCI US Large Cap Growth Index is $53.8 billion, compared to $57.0 billion for the MSCI US Large Cap Value Index. The figures for the MidCap indexes are $4.6 billion for growth and $4.4 billion for value.
This points out one of the weaknesses in broad indexing of growth vs. value. Most of the stocks in a large or mid value index aren't really value stocks, they're just less "growthy" than the growth stocks. I approach investing from an absolute value, absolute return viewpoint. Simply having a lower valuation than the broad market does not make a value stock; it is the absolute valuation that matters. DFA and indexes such as S&P's Style Pure Value are improvements over the inclusive index groups such as MSCI since they reduce the number of stocks that qualify as value, but they are still based on relative valuations. I want to buy value stocks that are cheap, period; not just cheap compared to expensive stocks.
Well, I wouldn't call it a weakness, but rather that broad cap-weighted value indices just don't reach as "deeply" into the value space as the alternatives that you mention. One isn't really better than the other, it just depends how much risk one is looking to take. FWIW, my entire US equity position tracks the S&P 600 Pure Value index.

- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB

Indexer88
Posts: 419
Joined: Mon Jan 05, 2009 5:32 pm

Post by Indexer88 » Thu Jan 21, 2010 7:29 am

Bid/ask spreads are probably not fully accounted for in historical reconstructions of small cap growth. I can't even say small cap funds, for in actual history, there were no major ones.

Post Reply