SmB and HmL returns on Ken French's website

 Posts: 526
 Joined: Mon Jan 07, 2013 11:02 am
SmB and HmL returns on Ken French's website
I looked at the FF 3 factor returns over time and found different numbers than most S&D discussants have used in previous discussions. I used the entire monthly data series from Professor Ken French's website (http://mba.tuck.dartmouth.edu/pages/fac ... brary.html), starting in July 1926 to April 2013.
These are the monthly averages I got for the entire data series.
Monthly average
(MktRF)SMBHMLRF
0.64 0.21 0.50 0.29
When I multiply this *12 for the annual average factor return I get:
Annual average
(MktRF)SMBHMLRF
7.63 2.46 6.04 3.47
Most discussants quote a SMB return of 2% and a HML return of 4%, but I got 3.5% and 6%? Did I do something wrong with my calculation?
Or are the lower numbers assuming that more people will use the methods of Eugene Fama and Ken French and therefore the effect will get "diluted" or disappear over time?
Any help/input would be appreciated
These are the monthly averages I got for the entire data series.
Monthly average
(MktRF)SMBHMLRF
0.64 0.21 0.50 0.29
When I multiply this *12 for the annual average factor return I get:
Annual average
(MktRF)SMBHMLRF
7.63 2.46 6.04 3.47
Most discussants quote a SMB return of 2% and a HML return of 4%, but I got 3.5% and 6%? Did I do something wrong with my calculation?
Or are the lower numbers assuming that more people will use the methods of Eugene Fama and Ken French and therefore the effect will get "diluted" or disappear over time?
Any help/input would be appreciated
Re: SmB and HmL returns on Ken French's website
I did this myself to find the annual return and confirmed that they were 4%, 2%, etc. The error you are making is that it's not correct to use the arithmetic average of the numbers.
Somewhere you have to find a logarithmic average, or a geometric average. There might be a function in Excel, I can't remember.
Alternatively, you can do this. Convert each number to something multiplicative, e.g. 0.64 > 1.0064. Multiply all the numbers together, and that is the growth over the entire time period. Simulate making an investment at the beginning, multiply it by the growth rate, then withdraw it at the end of the period. Use XIRR to calculate the return.
Edit  the Excel function is GEOMEAN. I believe that you have to convert the returns to numbers such as 1.0064, 0.9965, etc, and then you can use the function.
Somewhere you have to find a logarithmic average, or a geometric average. There might be a function in Excel, I can't remember.
Alternatively, you can do this. Convert each number to something multiplicative, e.g. 0.64 > 1.0064. Multiply all the numbers together, and that is the growth over the entire time period. Simulate making an investment at the beginning, multiply it by the growth rate, then withdraw it at the end of the period. Use XIRR to calculate the return.
Edit  the Excel function is GEOMEAN. I believe that you have to convert the returns to numbers such as 1.0064, 0.9965, etc, and then you can use the function.

 Posts: 526
 Joined: Mon Jan 07, 2013 11:02 am
Re: SmB and HmL returns on Ken French's website
Thanks so much Ketawa
There were some previous discussions about using geometric mean as opposed to arithmetic mean to calculate longterm returns and I am glad that the wrong mean calculation explains the too high estimate of SmB and HmL returns.(I need to spend some time to understand why this makes such a difference).
I just ran the numbers using the Geomean function in Excel, here is what I get (Will try the XIRR function later):
Monthly
(MktRF)SMBHMLRF
0.49 0.11 0.44 0.29
Annual
(MktRF)SMBHMLRF
5.87 1.31 5.33 3.47
Using the Geomean function, the SMB premium is now 1.3% and the HML risk factor premium is 5.3%  Still seems different than the 2% and 4%....
Mathematical error again or do most S&D posters simply reduce/adjust the expected future SMB and HML riskfactor returns for their calculations as these factors are common knowledge and therefore could give lower than historic returns in the future? If so, why would one not reduce the SMB expected premium to 1%?
(Edited as I had an error in the SMB Geomean calculation (initially got 3.5% annual return))
There were some previous discussions about using geometric mean as opposed to arithmetic mean to calculate longterm returns and I am glad that the wrong mean calculation explains the too high estimate of SmB and HmL returns.(I need to spend some time to understand why this makes such a difference).
I just ran the numbers using the Geomean function in Excel, here is what I get (Will try the XIRR function later):
Monthly
(MktRF)SMBHMLRF
0.49 0.11 0.44 0.29
Annual
(MktRF)SMBHMLRF
5.87 1.31 5.33 3.47
Using the Geomean function, the SMB premium is now 1.3% and the HML risk factor premium is 5.3%  Still seems different than the 2% and 4%....
Mathematical error again or do most S&D posters simply reduce/adjust the expected future SMB and HML riskfactor returns for their calculations as these factors are common knowledge and therefore could give lower than historic returns in the future? If so, why would one not reduce the SMB expected premium to 1%?
(Edited as I had an error in the SMB Geomean calculation (initially got 3.5% annual return))
Re: SmB and HmL returns on Ken French's website
There must be some kind of error in your spreadsheet. I just ran it and came up with the following factors, 192607 through 201305.
Each cell used for GEOMEAN should be converted to a multiplicative growth number. For example, first row for 192607 should be
Run GEOMEAN for the desired time period, raise the result to the 12th power to convert to an annual value, and subtract 1 for the factor returns.
Code: Select all
MktRF SMB HML RF
6.06% 2.22% 4.01% 3.52%
Each cell used for GEOMEAN should be converted to a multiplicative growth number. For example, first row for 192607 should be
Code: Select all
1.0265 0.9784 0.9708 1.0022
Run GEOMEAN for the desired time period, raise the result to the 12th power to convert to an annual value, and subtract 1 for the factor returns.

 Posts: 526
 Joined: Mon Jan 07, 2013 11:02 am
Re: SmB and HmL returns on Ken French's website
Thanks so much Ketawa, will start from scratch later on. My July 1926 numbers are the same except for SmB, where I have 0.9784  so something is off and I will do it over. But first, I have to go to swim practice with the kids now.
Quick question though, once I have the geometric mean of the monthly data  why can't I just multiply it by 12 to get an estimated annual return?
(sorry if this question sounds trivial)
One more question  do you think that common knowledge of the FF factors could impact their future returns?
Quick question though, once I have the geometric mean of the monthly data  why can't I just multiply it by 12 to get an estimated annual return?
(sorry if this question sounds trivial)
One more question  do you think that common knowledge of the FF factors could impact their future returns?
Re: SmB and HmL returns on Ken French's website
Multiplying by 12 will be close, but not exact. It would underestimate the actual number. For example, 1% per month multiplies to 12% per year, but is really 12.7% per year.
I think knowledge of the factors could impact them. It depends on how much of the returns are due to behavioral factors vs risk. I believe they are mostly risk stories, and there is a range of views from 100% risk to 100% behavioral. Hard to say.
I think knowledge of the factors could impact them. It depends on how much of the returns are due to behavioral factors vs risk. I believe they are mostly risk stories, and there is a range of views from 100% risk to 100% behavioral. Hard to say.

 Posts: 15597
 Joined: Thu Feb 22, 2007 8:28 am
 Location: St Louis MO
Re: SmB and HmL returns on Ken French's website
Just so you all know the risk factors are ALWAYS calculated as AVERAGE ANNUAL PREMIUMS, not compound (annualized) returns
Larry
Larry
 Rick Ferri
 Posts: 8201
 Joined: Mon Feb 26, 2007 11:40 am
 Location: Vagabond, USA, Twitter: @Rick_Ferri
 Contact:
Re: SmB and HmL returns on Ken French's website
Another often confusing item about HmL and SmB factors are that they relative to growth stocks and large cap stocks, not to the market.
In other words, to have actually earn the FF 3.9% HmL Research Factor premium (June 1927 to May 2013), you would have had to short all the low booktomarket stocks and go long all the high booktomarket stocks in the FF database, and do it at no cost.
Index products (including DFA Funds) don't short growth stocks to earn a HmL risk permium. From June 1927 to May 2013, the Dimensional US Marketwide Value Index returned 11.53% while the Dimensional US Market Index returned 9.84%, a difference of 1.69%. During the same period, the Fama/French US HmL Research Factor was 3.89%. Thus, the HmL premium of a value stock portfolio is less than half the HmL research factor premium, and this is before trading costs and management fees.
The lesson is that if you're expecting a 4.0% excess return over the market because you own all value funds, you're going to be disappointed. You'll earn perhaps a 1.5% premium after fees if the 4.0% HmL premium occurs in the future is as it was in the past AND you keep your costs extremely low.
BTW, paying an adviser 1% to do HmL and SmL portfolio tilts does not add up to excess returns for you. All the excess return from tilting plus some goes to the adviser to pay their fee. If you're thinking about paying an adviser 1% just to gain access to DFA funds, you're better off selfmanaging using cheap total market funds and forgetaboutit.
Rick Ferri
In other words, to have actually earn the FF 3.9% HmL Research Factor premium (June 1927 to May 2013), you would have had to short all the low booktomarket stocks and go long all the high booktomarket stocks in the FF database, and do it at no cost.
Index products (including DFA Funds) don't short growth stocks to earn a HmL risk permium. From June 1927 to May 2013, the Dimensional US Marketwide Value Index returned 11.53% while the Dimensional US Market Index returned 9.84%, a difference of 1.69%. During the same period, the Fama/French US HmL Research Factor was 3.89%. Thus, the HmL premium of a value stock portfolio is less than half the HmL research factor premium, and this is before trading costs and management fees.
The lesson is that if you're expecting a 4.0% excess return over the market because you own all value funds, you're going to be disappointed. You'll earn perhaps a 1.5% premium after fees if the 4.0% HmL premium occurs in the future is as it was in the past AND you keep your costs extremely low.
BTW, paying an adviser 1% to do HmL and SmL portfolio tilts does not add up to excess returns for you. All the excess return from tilting plus some goes to the adviser to pay their fee. If you're thinking about paying an adviser 1% just to gain access to DFA funds, you're better off selfmanaging using cheap total market funds and forgetaboutit.
Rick Ferri
Last edited by Rick Ferri on Sat Jun 29, 2013 7:30 am, edited 4 times in total.
The views expressed by Rick Ferri are strictly his own as a private investor and author and do not reflect the views of any entity or other persons.

 Posts: 526
 Joined: Mon Jan 07, 2013 11:02 am
Re: SmB and HmL returns on Ken French's website
Thanks Rick, Larry and Ketawa,
It is kind of embarassing to have two methodical errors in trying to do my own analysis... Thinking about it it: it makes sense that using the arithmetic mean is not appropriate for relative values, if you loose 10% and then later gain 10% you don't end up at the original value, because they are relative values. So multiplying all n relative values and then calculating the nth root (=geometric mean) makes more sense.
Ketawa, thanks for your patience telling me how to get from a monthly return to an annual return  1% per month again is a relative value and therefore I need the 12th power...
Rick, thanks for the advice about the cost of this strategy  In theory, there is no difference between theory and practice, but in practice, there is...(seems to be true for other fields as well)
It is kind of embarassing to have two methodical errors in trying to do my own analysis... Thinking about it it: it makes sense that using the arithmetic mean is not appropriate for relative values, if you loose 10% and then later gain 10% you don't end up at the original value, because they are relative values. So multiplying all n relative values and then calculating the nth root (=geometric mean) makes more sense.
Ketawa, thanks for your patience telling me how to get from a monthly return to an annual return  1% per month again is a relative value and therefore I need the 12th power...
Rick, thanks for the advice about the cost of this strategy  In theory, there is no difference between theory and practice, but in practice, there is...(seems to be true for other fields as well)

 Posts: 526
 Joined: Mon Jan 07, 2013 11:02 am
Re: SmB and HmL returns on Ken French's website
Ketawa,
I finally had time to sit down and reimport the FF french data from the website and recalculate the data series. I got exactly your numbers
Thanks again for your help.
Happy 4th to everyone!
I finally had time to sit down and reimport the FF french data from the website and recalculate the data series. I got exactly your numbers
Thanks again for your help.
Happy 4th to everyone!

 Posts: 1842
 Joined: Fri Feb 23, 2007 8:21 pm
Re: Question for Rick
Rick,
I understand your point that one should not pay a1% advisor fee just for tilted DFA access. I wonder if you would modify your answer at all for the investor with predominantly taxable accounts. Seems the DFA tax managed value funds and core funds do add something extra beyond the tilt. Thanks,
Dave
I understand your point that one should not pay a1% advisor fee just for tilted DFA access. I wonder if you would modify your answer at all for the investor with predominantly taxable accounts. Seems the DFA tax managed value funds and core funds do add something extra beyond the tilt. Thanks,
Dave
 Rick Ferri
 Posts: 8201
 Joined: Mon Feb 26, 2007 11:40 am
 Location: Vagabond, USA, Twitter: @Rick_Ferri
 Contact:
Re: SmB and HmL returns on Ken French's website
Dave,
Equity ETFs are a far more taxefficient fund structure than openend funds. There are no capital gain distributions from ETFs to worry about (with seasoned ETFs that have sufficient assets). This is due to the way shares are created and redeemed "inkind". Capital gains only occur when you decide to sell shares. So, if you're looking for taxefficiency, then you should be looking at an equity ETF portfolio.
Rick Ferri
Equity ETFs are a far more taxefficient fund structure than openend funds. There are no capital gain distributions from ETFs to worry about (with seasoned ETFs that have sufficient assets). This is due to the way shares are created and redeemed "inkind". Capital gains only occur when you decide to sell shares. So, if you're looking for taxefficiency, then you should be looking at an equity ETF portfolio.
Rick Ferri
The views expressed by Rick Ferri are strictly his own as a private investor and author and do not reflect the views of any entity or other persons.
Re: SmB and HmL returns on Ken French's website
Rick Ferri wrote:Another often confusing item about HmL and SmB factors are that they relative to growth stocks and large cap stocks, not to the market.
In other words, to have actually earn the FF 3.9% HmL Research Factor premium (June 1927 to May 2013), you would have had to short all the low booktomarket stocks and go long all the high booktomarket stocks in the FF database, and do it at no cost.
Index products (including DFA Funds) don't short growth stocks to earn a HmL risk permium. From June 1927 to May 2013, the Dimensional US Marketwide Value Index returned 11.53% while the Dimensional US Market Index returned 9.84%, a difference of 1.69%. During the same period, the Fama/French US HmL Research Factor was 3.89%. Thus, the HmL premium of a value stock portfolio is less than half the HmL research factor premium, and this is before trading costs and management fees.
The lesson is that if you're expecting a 4.0% excess return over the market because you own all value funds, you're going to be disappointed. You'll earn perhaps a 1.5% premium after fees if the 4.0% HmL premium occurs in the future is as it was in the past AND you keep your costs extremely low.
BTW, paying an adviser 1% to do HmL and SmL portfolio tilts does not add up to excess returns for you. All the excess return from tilting plus some goes to the adviser to pay their fee. If you're thinking about paying an adviser 1% just to gain access to DFA funds, you're better off selfmanaging using cheap total market funds and forgetaboutit.
Rick Ferri
Very helpful, puttingthingsinperspective, post. Thanks, Rick.
Re: SmB and HmL returns on Ken French's website
Rick Ferri wrote:Dave,
Equity ETFs are a far more taxefficient fund structure than openend funds. There are no capital gain distributions from ETFs to worry about (with seasoned ETFs that have sufficient assets). This is due to the way shares are created and redeemed "inkind". Capital gains only occur when you decide to sell shares. So, if you're looking for taxefficiency, then you should be looking at an equity ETF portfolio.
Rick Ferri
On the topic, I found the following interesting.
Gus Sauter wrote: A lot of investors are familiar with the inkind redemption process that occurs in ETFs and it does lend tax efficiency to ETFs. However, I think, many investors misunderstand that that's not a tax regime that is solely available to ETFs. In fact, it's available to any mutual fund, any regulated investment company as all mutual funds are.
So, our regular mutual fund can take advantage of the same inkind process and gain tax efficiency from that. If we have a large redemption, instead of paying the redeeming investor out with cash, we might give them securities. In other words that's an inkind redemption in securities, and that turns out to be a very tax efficient way to do distribution. And so, that tax efficiency isn't just available to ETFs, it's available to all mutual funds, really, any regulated investment company.
He goes on to say that they don't do small redemptions inkind, only the larger ones. It would seem logical that a tax managed mutual fund would have a lower threshold for doing inkind redemptions, but I'm not sure if they do in practice.
Sauter: ETFs Aren't Automatically More Tax Efficient
 Rick Ferri
 Posts: 8201
 Joined: Mon Feb 26, 2007 11:40 am
 Location: Vagabond, USA, Twitter: @Rick_Ferri
 Contact:
Re: SmB and HmL returns on Ken French's website
I agree with Gus, although his comments are specific to Vanguard ETFs only, which are a share class of the openend fund. No other ETFs are structured like Vanguard ETFs.
Rick Ferri
Rick Ferri
The views expressed by Rick Ferri are strictly his own as a private investor and author and do not reflect the views of any entity or other persons.