There’s no a-priori reason why stocks should deliver a premium above long-term bonds.vineviz wrote: ↑Sun Jan 23, 2022 4:33 pmI guarantee you that 99% of investors can't articulate why stocks are riskier than bonds any more clearly than academics have articulated why the other factors exist. Heck, 99% of investors can't even properly identify the risk-free asset, much less explain what an equity risk premium is or how to calculate it.
"I don't understand it therefore it can't exist" is a pretty weak argument, if you ask me.
Fama and French: The Five-Factor Model Revisited
Re: Fama and French: The Five-Factor Model Revisited
Re: Fama and French: The Five-Factor Model Revisited
Yes, that was a very different fund.sycamore wrote: ↑Sun Jan 23, 2022 8:11 pmVanguard's Small Cap Value NAESX goes back a bit further to 1985, but I'm not sure if NAESX was one of those funds whose strategy changed over time.Northern Flicker wrote: ↑Sun Jan 23, 2022 4:45 pm Here is a small-cap value fund that predates the Fama & French (1993) publication.
https://www.portfoliovisualizer.com/bac ... ion1_1=100
- hillclimber
- Posts: 136
- Joined: Wed Mar 10, 2021 2:32 pm
Re: Fama and French: The Five-Factor Model Revisited
Here is a layman level explanation on why stocks are riskier than bonds:
The risk with stocks is that the CEO does stupid stuff and wastes your money. However, if the business does well, you get dividends, buybacks, and earnings growth. Also, every few years, the stock market goes down 10% or more.
Most of the people I know personally at least implicitly understood these concepts and could express them if given time to think it through.
Why are you being so belligerent right now? No one is arguing, just expressing skepticism.
- nisiprius
- Advisory Board
- Posts: 52216
- Joined: Thu Jul 26, 2007 9:33 am
- Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry
Re: Fama and French: The Five-Factor Model Revisited
NAESX is not small cap value, it's just small caps.sycamore wrote: ↑Sun Jan 23, 2022 8:11 pmVanguard's Small Cap Value NAESX goes back a bit further to 1985, but I'm not sure if NAESX was one of those funds whose strategy changed over time.Northern Flicker wrote: ↑Sun Jan 23, 2022 4:45 pm Here is a small-cap value fund that predates the Fama & French (1993) publication.
https://www.portfoliovisualizer.com/bac ... ion1_1=100
The name of PRSVX today is "T. Rowe Price Small-Cap Value Fund;" it would be interesting to know what its name was in 1988, and how T. Rowe Price described its strategy back then. Here's their summary today:
Here's the comparison with the Fama-French research series for small-cap value:Invests in small-company stocks that firm believes do not reflect their underlying value. Intensive in-house research considers, among many factors: low price/earnings, price/book value, and price/cash flow ratios and solid financial characteristics.
Small companies tend to be riskier than large companies.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
Re: Fama and French: The Five-Factor Model Revisited
If you want to see how things are changing over time, you don't average over the whole time period. Or do you believe that there is a "true" constant premium and correlation coefficient around which all the historical data dances? And by averaging over longer and longer periods, we get better and better estimates of these "true" values?
If you consider a correlation of 0.92 significantly higher than a correlation of 0.91. And we know that correlation isn't all that matters. Otherwise, cash or bonds would be better diversifiers. One also needs to consider return. And I would think that the average annual return advantage of SCG of 11.00% vs the 9.58% return of SCV outweighs the 0.01 correlation advantage of the latter in determining which was the better diversifier in the 20 year period between 2002 - 2021.
And if we go back to the full dataset, which is more important for determining a better diversifier, the 0.91 vs 0.77 correlation of SCG vs SCV or the 10.40% vs 9.98% return of SCG vs SCV? Well, turning to PV, if we compare a portfolio of 50/50 SCG/TSM with a portfolio of 50/50 SCV/TSM, the results (with rebalancing bands) are:
SCG/TSM: CAGR = 9.49%, STD = 17.23%, Sharpe = 0.52
SCV/TSM: CAGR = 9.33%, STD = 16.76%, Sharpe = 0.52
So, they are equally good diversifiers over 1999 - 2021. The higher return of SCG defeats the correlation in term of the portfolio return, at the expense of higher volatility.
How about over the last 20 years (2002 - 2021)?:
SCG/TSM: CAGR = 10.47%, STD = 16.70%, Sharpe = 0.61
SCV/TSM: CAGR = 9.75%, STD = 16.72%, Sharpe = 0.57
SCG was the better diversifier.
Re: Fama and French: The Five-Factor Model Revisited
Here is my explanation:hillclimber wrote: ↑Mon Jan 24, 2022 1:37 amHere is a layman level explanation on why stocks are riskier than bonds:
The risk with stocks is that the CEO does stupid stuff and wastes your money. However, if the business does well, you get dividends, buybacks, and earnings growth. Also, every few years, the stock market goes down 10% or more.
With a bond, you know exactly what you are going to get in the future (apart from defaults). But you don't whether there might have been a better bond for you to have chosen over the coming years. So, you know what you are getting, but you don't know what the competition will get.
With a stock, you don't know what you are going to get in the future. And you don't know if some other stock might have have been better. So, you don't know what you are going to get and you don't know what the competition will get.
Bonds have 1 form of uncertainty. Stocks have 2 forms of uncertainty.
And with a bond, you can dial the risk up or down by looking at the term and the credit quality. So, you have some control.
With a stock, dialing the risk up or down is very uncertain. You may think that a large cap stock is safer than a small cap stock, but we've seen lots of large caps crash over the years. So, you have much less control over safety regarding stocks.
Re: Fama and French: The Five-Factor Model Revisited
I've looked at data from 1927 to present, 1964 to present, 1999 to present: no the correlations aren't constant; yet, the correlation of SCV with the market is lower than SCG most of the time and on average in all time periods.rkhusky wrote: ↑Mon Jan 24, 2022 7:41 am If you want to see how things are changing over time, you don't average over the whole time period. Or do you believe that there is a "true" constant premium and correlation coefficient around which all the historical data dances? And by averaging over longer and longer periods, we get better and better estimates of these "true" values?
Sharpe ratio is not a measure of diversification.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
-
- Posts: 143
- Joined: Mon Jun 11, 2012 9:01 pm
Re: Fama and French: The Five-Factor Model Revisited
Bond holders get paid before stock holders. That's a pretty straightforward explanation of why they are less risky.
Re: Fama and French: The Five-Factor Model Revisited
Never said it was. But it is a factor in making something a good diversifier. I don't care about correlation or diversification except for how they affect my return. If they increase return and lower volatility, it is good. If they lower return and increase volatility, it is bad. Otherwise, you have to trade-off return and volatility, and the Sharpe Ratio is one way to do that. In this case, I agree with the Sharpe ratio - they are equivalently good diversifiers of TSM over this period.
If lack of correlation is all you are after, cash and bonds had negative correlation with TSM over 1999 - 2021.
Re: Fama and French: The Five-Factor Model Revisited
No, it is not a factor. If you want to argue that SCG>SCV in terms of diversification, which is what I think you're trying to do, you can't use Sharpe as the measure.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
I am not arguing that. I am saying that if I have a choice in choosing between two diversifiers, I will consider the return difference and volatility difference that comes with using each of them in my portfolio.vineviz wrote: ↑Mon Jan 24, 2022 9:17 amNo, it is not a factor. If you want to argue that SCG>SCV in terms of diversification, which is what I think you're trying to do, you can't use Sharpe as the measure.
Re: Fama and French: The Five-Factor Model Revisited
The problem is that in real life you don't get to choose assets based on their future returns.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
- vanbogle59
- Posts: 1314
- Joined: Wed Mar 10, 2021 7:30 pm
- willthrill81
- Posts: 32250
- Joined: Thu Jan 26, 2017 2:17 pm
- Location: USA
- Contact:
Re: Fama and French: The Five-Factor Model Revisited
Two variables having had the same Sharpe ratio does not mean that their risk/volatility has been the same for a long-term investor. For instance, the start date sensitivity, which is based on rolling 10 year returns, of SCV has been about 40% less than TSM since 1970.rkhusky wrote: ↑Mon Jan 24, 2022 9:24 amI am not arguing that. I am saying that if I have a choice in choosing between two diversifiers, I will consider the return difference and volatility difference that comes with using each of them in my portfolio.vineviz wrote: ↑Mon Jan 24, 2022 9:17 amNo, it is not a factor. If you want to argue that SCG>SCV in terms of diversification, which is what I think you're trying to do, you can't use Sharpe as the measure.
The Sensible Steward
Re: Fama and French: The Five-Factor Model Revisited
Who said that? We all know that we invest based on past performance.
2002 - 2021
50/50 portfolios
SCG/TSM: CAGR = 10.47%, STD = 16.70%, Sharpe = 0.61
SCV/TSM: CAGR = 9.75%, STD = 16.72%, Sharpe = 0.57
Last edited by rkhusky on Mon Jan 24, 2022 11:04 am, edited 2 times in total.
Re: Fama and French: The Five-Factor Model Revisited
So, you are saying that SCV is less risky than TSM, which means that SCV should return less than TSM over the long term?willthrill81 wrote: ↑Mon Jan 24, 2022 10:05 am For instance, the start date sensitivity, which is based on rolling 10 year returns, of SCV has been about 40% less than TSM since 1970.
-
- Posts: 6234
- Joined: Mon Apr 11, 2011 12:28 am
Re: Fama and French: The Five-Factor Model Revisited
Start date sensitivity is not a measure of riskrkhusky wrote: ↑Mon Jan 24, 2022 11:01 amSo, you are saying that SCV is less risky than TSM, which means that SCV should return less than TSM over the long term?willthrill81 wrote: ↑Mon Jan 24, 2022 10:05 am For instance, the start date sensitivity, which is based on rolling 10 year returns, of SCV has been about 40% less than TSM since 1970.
Having higher expected returns has the tendency of reducing recovery periods because volatility is higher. Upside volatility is a good thing
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
- willthrill81
- Posts: 32250
- Joined: Thu Jan 26, 2017 2:17 pm
- Location: USA
- Contact:
Re: Fama and French: The Five-Factor Model Revisited
That would be contrary to the last 50+ years of data we have. SCV has had higher long-term returns than TSM and lower start date sensitivity (i.e., more consistent 10 year returns). Interpret that how you will, but it was what it was.rkhusky wrote: ↑Mon Jan 24, 2022 11:01 amSo, you are saying that SCV is less risky than TSM, which means that SCV should return less than TSM over the long term?willthrill81 wrote: ↑Mon Jan 24, 2022 10:05 am For instance, the start date sensitivity, which is based on rolling 10 year returns, of SCV has been about 40% less than TSM since 1970.
The Sensible Steward
Re: Fama and French: The Five-Factor Model Revisited
Possible interpretations:willthrill81 wrote: ↑Mon Jan 24, 2022 11:05 amThat would be contrary to the last 50+ years of data we have. SCV has had higher long-term returns than TSM and lower start date sensitivity (i.e., more consistent 10 year returns). Interpret that how you will, but it was what it was.rkhusky wrote: ↑Mon Jan 24, 2022 11:01 amSo, you are saying that SCV is less risky than TSM, which means that SCV should return less than TSM over the long term?willthrill81 wrote: ↑Mon Jan 24, 2022 10:05 am For instance, the start date sensitivity, which is based on rolling 10 year returns, of SCV has been about 40% less than TSM since 1970.
SCV will return a lot less than TSM in the future until the long term average drops below TSM.
Higher SCV returns are illusory.
Start date sensitivity is not a measure of risk.
Investors are irrational.
Investors don't care about start date sensitivity.
Re: Fama and French: The Five-Factor Model Revisited
You're saying it, whether you realize it or not. It's implicit in your use of Sharpe to make the diversification claim you made earlier.
SCV has a lower correlation with TSM and higher variance than SCG. The only way to make SCG come out as the "better" diversifier in your rubric is for SCG to have higher expected returns.
Using Sharpe ratio as an optimization metric has a well-documented problem in that it functions very much like a "past performance" optimizer .
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
I am all about higher returns with little change in volatility. If all you care about is correlations, might I suggest cash under the mattress. It has a very low correlation with TSM.vineviz wrote: ↑Mon Jan 24, 2022 11:15 amYou're saying it, whether you realize it or not. It's implicit in your use of Sharpe to make the diversification claim you made earlier.
SCV has a lower correlation with TSM and higher variance than SCG. The only way to make SCG come out as the "better" diversifier in your rubric is for SCG to have higher expected returns.
Using Sharpe ratio as an optimization metric has a well-documented problem in that it functions very much like a "past performance" optimizer .
Anyway, here is what was said:
Note therkhusky wrote: ↑Mon Jan 24, 2022 8:59 amNever said it was. But it is a factor in making something a good diversifier. I don't care about correlation or diversification except for how they affect my return. If they increase return and lower volatility, it is good. If they lower return and increase volatility, it is bad. Otherwise, you have to trade-off return and volatility, and the Sharpe Ratio is one way to do that. In this case, I agree with the Sharpe ratio - they are equivalently good diversifiers of TSM over this period.
If lack of correlation is all you are after, cash and bonds had negative correlation with TSM over 1999 - 2021.
I'm not a slave to Sharpe. I look at the return and the volatility and make a judgement. In this case, the return difference was small and the volatility difference was small. And the Sharpe ratio was the same. So, looks about the same to me.Otherwise, you have to trade-off return and volatility, and the Sharpe Ratio is one way to do that.
Re: Fama and French: The Five-Factor Model Revisited
You're unlikely to get that. Higher volatility is generally a trade-off for higher returns. Higher volatility doesn't necessarily mean higher returns as is the case with small-growth. SG is more volatile with lower returns than the TSM while SV is more volatile but has higher returns than the TSM. You don't seem to understand that.
Last edited by Apathizer on Mon Jan 24, 2022 2:21 pm, edited 1 time in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Re: Fama and French: The Five-Factor Model Revisited
Reality sometimes doesn't follow what we expect.Apathizer wrote: ↑Mon Jan 24, 2022 11:56 amYou're unlikely to get that. Higher volatility is generally a trade-off for higher returns. Higher volatility, doesn't necessarily higher returns as is the case with small-growth. SG is more volatile with lower returns than the TSM while SV is more volatile but has higher returns than the TSM. You don't seem to understand that.
2002 - 2021
50/50 portfolios
SCG/TSM: CAGR = 10.47%, STD = 16.70%, Sharpe = 0.61
SCV/TSM: CAGR = 9.75%, STD = 16.72%, Sharpe = 0.57
Adding SCG to TSM gave higher return and lower volatility than adding SCV, over this 20 year period. I would go with SCG over SCV as a diversifier if I expected this to continue.
- willthrill81
- Posts: 32250
- Joined: Thu Jan 26, 2017 2:17 pm
- Location: USA
- Contact:
Re: Fama and French: The Five-Factor Model Revisited
Of course. We expect stocks to have higher returns than bonds, but that's not always the case. Bonds trounced stocks from 2000-2009. But that doesn't mean that we use only single, isolated periods to form our investment strategies.rkhusky wrote: ↑Mon Jan 24, 2022 12:15 pmReality sometimes doesn't follow what we expect.Apathizer wrote: ↑Mon Jan 24, 2022 11:56 amYou're unlikely to get that. Higher volatility is generally a trade-off for higher returns. Higher volatility, doesn't necessarily higher returns as is the case with small-growth. SG is more volatile with lower returns than the TSM while SV is more volatile but has higher returns than the TSM. You don't seem to understand that.
The Sensible Steward
Re: Fama and French: The Five-Factor Model Revisited
Diversification is about more than correlations, obviously.
My main point here is that when you're talking about Sharpe ratio you're talking about something OTHER than diversification. And mostly what you're talking about is high past performance.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
Understood.
-
- Posts: 15363
- Joined: Fri Apr 10, 2015 12:29 am
Re: Fama and French: The Five-Factor Model Revisited
Actually the uncertainty about the source of the value premium carties over to the market factor. If the market is efficient, and the market prices in risk premia for the risk of individual stocks (when they are held in a diversified portfolio) then the market premium is the aggregation if that in the whole market while the value premium would be determined from aggregating across stocks defined as value stocks.km91 wrote: ↑Sun Jan 23, 2022 8:14 pmHuh? We don't why equity risk commands a premium to the risk free asset?Northern Flicker wrote: ↑Sun Jan 23, 2022 7:53 pmThe same critique applies to the market factor (equity risk premium).km91 wrote: This makes absolutely no sense. How can the factor command a risk premium in the market, yet none of the market participants are able to explain it. Sellers of the factor are paying a premium to buyers, but don't know why they're paying it?
If the behavioral story of a value premium is correct, and investors overpay for growth, then that also affects how the market establishes a risk premium for the whole market since growth is part of the whole market.
Re: Fama and French: The Five-Factor Model Revisited
Yep. Didn't come back to even until 2016. 2000 would have been a bad year to lump sum into stocks. Of course, the fundamental differences between stocks and bonds are quite a bit larger than between value stocks and growth stocks or between small stocks and big stocks.willthrill81 wrote: ↑Mon Jan 24, 2022 12:17 pmOf course. We expect stocks to have higher returns than bonds, but that's not always the case. Bonds trounced stocks from 2000-2009. But that doesn't mean that we use only single, isolated periods to form our investment strategies.rkhusky wrote: ↑Mon Jan 24, 2022 12:15 pmReality sometimes doesn't follow what we expect.Apathizer wrote: ↑Mon Jan 24, 2022 11:56 amYou're unlikely to get that. Higher volatility is generally a trade-off for higher returns. Higher volatility, doesn't necessarily higher returns as is the case with small-growth. SG is more volatile with lower returns than the TSM while SV is more volatile but has higher returns than the TSM. You don't seem to understand that.
Re: Fama and French: The Five-Factor Model Revisited
True, but there have been extended periods when bonds out performed the TSM. Other periods TSM and bonds had similar returns while SV out performed both.rkhusky wrote: ↑Mon Jan 24, 2022 2:35 pmYep. Didn't come back to even until 2016. 2000 would have been a bad year to lump sum into stocks. Of course, the fundamental differences between stocks and bonds are quite a bit larger than between value stocks and growth stocks or between small stocks and big stocks.willthrill81 wrote: ↑Mon Jan 24, 2022 12:17 pmOf course. We expect stocks to have higher returns than bonds, but that's not always the case. Bonds trounced stocks from 2000-2009. But that doesn't mean that we use only single, isolated periods to form our investment strategies.rkhusky wrote: ↑Mon Jan 24, 2022 12:15 pmReality sometimes doesn't follow what we expect.Apathizer wrote: ↑Mon Jan 24, 2022 11:56 amYou're unlikely to get that. Higher volatility is generally a trade-off for higher returns. Higher volatility, doesn't necessarily higher returns as is the case with small-growth. SG is more volatile with lower returns than the TSM while SV is more volatile but has higher returns than the TSM. You don't seem to understand that.
That's the point of diversifying across all markets and likely positive risk factors. Since we can't predict the future all we can do is apply our best understanding of markets to determine an approach most likely to produce the best returns.
Last edited by Apathizer on Tue Jan 25, 2022 10:16 pm, edited 1 time in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Re: Fama and French: The Five-Factor Model Revisited
Exactly. Sometimes Small Growth will outperform. Sometimes Small Value will outperform. Sometimes Large Growth will outperform. Sometimes Large Value will outperform. And the correlations between them range from 0.61 - 0.89. So, diversify by investing in all of them with TSM.
If you really want to tilt to one part of the market, you don't want to invest in all the factors, because they will cancel out and you will be left with TSM. And most people don't invest in factors anyway, they invest in half-factors, i.e. the long portion of the factor.
The correlation with the market and the return of various market segments over the last 20 years is below:
(note that the annualized return for TSM was 9.78%):
1/1/2002 - 12/31/2021
(correlation with TSM, annualized return)
LCG - 0.95, 10.71%
LCV - 0.95, 8.37%
SCG - 0.92, 11.00%
SCV - 0.91, 9.58%
If you weren't invested in growth during this period, you missed out on a lot of return, which will take years to make up. Best not to make bets on one part of the market, and just invest in it all.
Re: Fama and French: The Five-Factor Model Revisited
You are making exactly the opposite point of the poster you are quoting, and I think you know that.rkhusky wrote: ↑Tue Jan 25, 2022 7:38 amExactly. Sometimes Small Growth will outperform. Sometimes Small Value will outperform. Sometimes Large Growth will outperform. Sometimes Large Value will outperform. And the correlations between them range from 0.61 - 0.89. So, diversify by investing in all of them with TSM.
If you really want to tilt to one part of the market, you don't want to invest in all the factors, because they will cancel out and you will be left with TSM. And most people don't invest in factors anyway, they invest in half-factors, i.e. the long portion of the factor.
The correlation with the market and the return of various market segments over the last 20 years is below:
(note that the annualized return for TSM was 9.78%):
1/1/2002 - 12/31/2021
(correlation with TSM, annualized return)
LCG - 0.95, 10.71%
LCV - 0.95, 8.37%
SCG - 0.92, 11.00%
SCV - 0.91, 9.58%
If you weren't invested in growth during this period, you missed out on a lot of return, which will take years to make up. Best not to make bets on one part of the market, and just invest in it all.
Firstly, you know well that the TSM is not "all the factors", it is one of them, the market factor. This is by definition, like "how many ounces is a gallon".
Secondly, Growth and Value are not two factors that cancel each other out in the TSM. There is one factor, HmL, or value minus growth. The TSM has zero because it has equal parts growth and value, by definition.
Thirdly, this poster is recommending to diversify across all factors, which explicitly means NOT investing in only the TSM. Correspondingly it means not investing in equal parts growth and value, nor following market cap for small and growth. The poster is implying to have non-zero loadings on the factors in the portfolio, which means more value than growth, and more small compared to the TSM. Factor diversification implies non-zero (and positive) allocations to the factors, not zero allocations to the factors.
I think you know all of this, and are just using their quote and stating the exact opposite of their original intention as some tongue-in-cheek humor or something
Crom laughs at your Four Winds
Re: Fama and French: The Five-Factor Model Revisited
Depends on what you mean by "all the factors". There are purportedly hundreds of factors in the "factor zoo".
And factor investing doesn't just mean small/value. In that sense, large and growth are factors too. (Although, these are really half-factors, but most people only invest in the long side anyway.) So, tilting to large and/or growth is also factor investing.
Re: Fama and French: The Five-Factor Model Revisited
To my knowledge it doesn't depend on what I mean by all the factors. The TSM is by definition the market portfolio, and everything else is relative to that portfolio. I would also argue the majority of people who would say they are factor investors only meaningfully care about size and value, and then a large but non-majority also care about momentum, CmA, and RmW. I don't think many if any retail factor investors are seeking out more than that in any meaningful way. When people say "there are hundreds of factors lol look how dumb they are" they are typically trying to belittle the approach by over-generalizing.rkhusky wrote: ↑Tue Jan 25, 2022 8:37 amDepends on what you mean by "all the factors". There are purportedly hundreds of factors in the "factor zoo".
And factor investing doesn't just mean small/value. In that sense, large and growth are factors too. (Although, these are really half-factors, but most people only invest in the long side anyway.) So, tilting to large and/or growth is also factor investing.
Yes tilting to large and growth is factor investing in that it uses factors to achieve some outcome. But in some sense it's like anti-factor investing. That strategy would mean one prefers a lower expected value of long-term returns because they get some other utility from choosing those investments that is not related to the value of the investment. One example might be an extremely tax-sensitive situation, where one gets more utility from growth stocks where they avoid paying taxes on the higher dividends of value stocks.
Crom laughs at your Four Winds
Re: Fama and French: The Five-Factor Model Revisited
"lower expected value of long-term returns" is really what it is all about. People saw an over-performance of SCV in the past and expect it to occur in the future. And have convinced themselves of that with some hand-waving arguments about the market mispricing risk. That's all it is. It doesn't have anything to do with factor models or asset pricing models, which provide no advantage to big, small, value, growth.muffins14 wrote: ↑Tue Jan 25, 2022 9:05 am Yes tilting to large and growth is factor investing in that it uses factors to achieve some outcome. But in some sense it's like anti-factor investing. That strategy would mean one prefers a lower expected value of long-term returns because they get some other utility from choosing those investments that is not related to the value of the investment.
Re: Fama and French: The Five-Factor Model Revisited
Correct on all counts.muffins14 wrote: ↑Tue Jan 25, 2022 9:05 am
To my knowledge it doesn't depend on what I mean by all the factors. The TSM is by definition the market portfolio, and everything else is relative to that portfolio. I would also argue the majority of people who would say they are factor investors only meaningfully care about size and value, and then a large but non-majority also care about momentum, CmA, and RmW. I don't think many if any retail factor investors are seeking out more than that in any meaningful way. When people say "there are hundreds of factors lol look how dumb they are" they are typically trying to belittle the approach by over-generalizing.
The "hundreds of factors" objection is a strawman, overlooking the fact that many of the "hundreds" actually describe the same risk factors. It is morereasonable to count Book to price, Earnings to price, Dividend to price, Cash flow-to-price, Sales to price, and Enterprise book-to-price as six different factors, for instance, or to consider them six possible specifications of the same "value" factor?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
You have it complete backwards, actually. The factors work when markets do NOT misprice risk, and informed factor investors know that the factors likely mean they are taking on MORE risk in exchange for the higher expected return. We're assuming the market isn't dumb, in other words.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
I don't agree. The factor model only requires that there be some anomaly in the data. It doesn't matter why the anomaly is there. For example, if it was noticed that returns of stocks whose name starts with Q behaved differently than stocks whose name starts with Z, one could add the ZmQ (or QmZ) factor to the model. If the anomaly was persistent and favored one of the two, then a non-zero premium would result. If the anomaly was not persistent or fluctuated sufficiently between the two, then the premium would end up close to zero and probably removed from the model.vineviz wrote: ↑Tue Jan 25, 2022 10:50 amYou have it complete backwards, actually. The factors work when markets do NOT misprice risk, and informed factor investors know that the factors likely mean they are taking on MORE risk in exchange for the higher expected return. We're assuming the market isn't dumb, in other words.
Re: Fama and French: The Five-Factor Model Revisited
The model is what it is. What I objected to was your characterization of how "factor investors" interpret the model. Few are convinced of what you suggest, that the market is systematically mispricing risk. Instead, they are proceeding - for the most part -under a belief that the market is ACCURATELY pricing risk.rkhusky wrote: ↑Tue Jan 25, 2022 11:52 amI don't agree. The factor model only requires that there be some anomaly in the data. It doesn't matter why the anomaly is there. For example, if it was noticed that returns of stocks whose name starts with Q behaved differently than stocks whose name starts with Z, one could add the ZmQ (or QmZ) factor to the model. If the anomaly was persistent and favored one of the two, then a non-zero premium would result. If the anomaly was not persistent or fluctuated sufficiently between the two, then the premium would end up close to zero and probably removed from the model.vineviz wrote: ↑Tue Jan 25, 2022 10:50 amYou have it complete backwards, actually. The factors work when markets do NOT misprice risk, and informed factor investors know that the factors likely mean they are taking on MORE risk in exchange for the higher expected return. We're assuming the market isn't dumb, in other words.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
Yet a qualitative explanation of what the additional risk being taken when tilting towards a factor seems impossible to provide outside of the context of the model. I'm beating a dead horse here but how could the market accurately price SCV risk since the 70's yet not come up with a clear explanation of it in that time? The premium exists in the data, I don't doubt the rigor that lead FF to their conclusion, but I don't see how the market could collectively price this risk for decades but have little understanding of the economic fundamentals that produce it. I'd love to hear an explanation why ranking companies based on financial ratios then shorting the bottom and buying the top introduces an additional risk into the portfoliovineviz wrote: ↑Tue Jan 25, 2022 12:22 pmThe model is what it is. What I objected to was your characterization of how "factor investors" interpret the model. Few are convinced of what you suggest, that the market is systematically mispricing risk. Instead, they are proceeding - for the most part -under a belief that the market is ACCURATELY pricing risk.rkhusky wrote: ↑Tue Jan 25, 2022 11:52 amI don't agree. The factor model only requires that there be some anomaly in the data. It doesn't matter why the anomaly is there. For example, if it was noticed that returns of stocks whose name starts with Q behaved differently than stocks whose name starts with Z, one could add the ZmQ (or QmZ) factor to the model. If the anomaly was persistent and favored one of the two, then a non-zero premium would result. If the anomaly was not persistent or fluctuated sufficiently between the two, then the premium would end up close to zero and probably removed from the model.vineviz wrote: ↑Tue Jan 25, 2022 10:50 amYou have it complete backwards, actually. The factors work when markets do NOT misprice risk, and informed factor investors know that the factors likely mean they are taking on MORE risk in exchange for the higher expected return. We're assuming the market isn't dumb, in other words.
-
- Posts: 6234
- Joined: Mon Apr 11, 2011 12:28 am
Re: Fama and French: The Five-Factor Model Revisited
Individual security analysis is nothing new. It takes into account things like industry/sector risks, management risks, earnings risks, etc.km91 wrote: ↑Tue Jan 25, 2022 12:40 pmYet a qualitative explanation of what the additional risk being taken when tilting towards a factor seems impossible to provide outside of the context of the model. I'm beating a dead horse here but how could the market accurately price SCV risk since the 70's yet not come up with a clear explanation of it in that time? The premium exists in the data, I don't doubt the rigor that lead FF to their conclusion, but I don't see how the market could collectively price this risk for decades but have little understanding of the economic fundamentals that produce it. I'd love to hear an explanation why ranking companies based on financial ratios then shorting the bottom and buying the top introduces an additional risk into the portfoliovineviz wrote: ↑Tue Jan 25, 2022 12:22 pmThe model is what it is. What I objected to was your characterization of how "factor investors" interpret the model. Few are convinced of what you suggest, that the market is systematically mispricing risk. Instead, they are proceeding - for the most part -under a belief that the market is ACCURATELY pricing risk.rkhusky wrote: ↑Tue Jan 25, 2022 11:52 amI don't agree. The factor model only requires that there be some anomaly in the data. It doesn't matter why the anomaly is there. For example, if it was noticed that returns of stocks whose name starts with Q behaved differently than stocks whose name starts with Z, one could add the ZmQ (or QmZ) factor to the model. If the anomaly was persistent and favored one of the two, then a non-zero premium would result. If the anomaly was not persistent or fluctuated sufficiently between the two, then the premium would end up close to zero and probably removed from the model.vineviz wrote: ↑Tue Jan 25, 2022 10:50 amYou have it complete backwards, actually. The factors work when markets do NOT misprice risk, and informed factor investors know that the factors likely mean they are taking on MORE risk in exchange for the higher expected return. We're assuming the market isn't dumb, in other words.
By filtering on stocks with similar “Value” attributes you are selecting companies that the market is heavily discounting due to a wide variety of heightened perceived risks.
Why are you claiming that the market doesn’t have an understanding of these economic fundamentals? Just because it wasn’t employed at the macro level?
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
Re: Fama and French: The Five-Factor Model Revisited
IMHO, it only seems "impossible" to participants who wish there to be no explanation.
Both this thread and the finance literature are full of such explanations.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
I'm skeptical but there's no motivated reasoning going on here. In fact I would love for my skepticism to be proven wrong, now would be the perfect time to buy value since it's dirt cheap on a number of valuation metrics. Hand waving at the "finance literature" that requires a Phd in stats to parse is not a compelling explanation in my opinion. Over and over again the market has proven very smart people with very sophisticated models to be wrong. Add to the fact that the factors haven't performed well on the out of sample data leads me to believe there is a bit more model risk here than some would like to admit, so I try to approach the assumptions of the model from first principles. Maybe an explanation of the factors comes easier to those with a financial stake in them
-
- Posts: 6234
- Joined: Mon Apr 11, 2011 12:28 am
Re: Fama and French: The Five-Factor Model Revisited
Factors have performed well, out of samplekm91 wrote: ↑Tue Jan 25, 2022 1:33 pmI'm skeptical but there's no motivated reasoning going on here. In fact I would love for my skepticism to be proven wrong, now would be the perfect time to buy value since it's dirt cheap on a number of valuation metrics. Hand waving at the "finance literature" that requires a Phd in stats to parse is not a compelling explanation in my opinion. Over and over again the market has proven very smart people with very sophisticated models to be wrong. Add to the fact that the factors haven't performed well on the out of sample data leads me to believe there is a bit more model risk here than some would like to admit, so I try to approach the assumptions of the model from first principles. Maybe an explanation of the factors comes easier to those with a financial stake in them
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
-
- Posts: 15363
- Joined: Fri Apr 10, 2015 12:29 am
Re: Fama and French: The Five-Factor Model Revisited
The market does not try to price SCV. You have it backwards. The market prices stocks. You don't know which stocks are the value stocks until the market has priced them. You also don't know the market cap of a stock until you know its price.km91 wrote: Yet a qualitative explanation of what the additional risk being taken when tilting towards a factor seems impossible to provide outside of the context of the model. I'm beating a dead horse here but how could the market accurately price SCV risk since the 70's yet not come up with a clear explanation of it in that time...
The market prices risk by incorporating a risk premium into the discount rate used to discount future cash flows back to a present value. Factor models have increased our understanding of equity risk by providing a systematic framework for understanding risks beyond market risk instead of just lumping them into alpha under CAPM as a catch-all for risks other than market risk.
It really just boils down to whether you wish to take additional equity risks beyond market risk to try to boost the expected return of one's equity portfolio. It really is taking more risk. Periods of value underperformance are a testament to that. There is no imperative to do so.
Holding the market portfolio, and diversifying away all factors but the market factor and diversifying away all alpha is a very rational option. You diversify away all alpha. Factor-tilted portfolios have residual alpha. And there is no model risk-- the market portfolio gives you the market return. I do think that the robustness of factor models has been oversold. If you invest in a large market index fund implemented by full replication of the index, the sources of potential tracking error are miniscule to nil.
Here are a couple of quality-tilted portfolios to contemplate.
https://www.portfoliovisualizer.com/fac ... sion=false
Last edited by Northern Flicker on Tue Jan 25, 2022 2:27 pm, edited 3 times in total.
Re: Fama and French: The Five-Factor Model Revisited
I was a physicist, so I appreciate the yearning for understanding the root cause of things, or the underlying theory of why things are the way they are. Sometimes, we just aren't there yet -- Theory hasn't caught up with experimental evidence. The two go hand-in-hand over time. New anomalies are discovered, then we try to explain them with theory. Maybe the theory has new implications. People will test the new theory by collecting new data and some theories will still valid, while some must be ruled out with some confidence. Theories evolve. the cycle continues.
We could be at a point in finance/factor mmodels where it is difficult to describe from first principles what the unique risk of small-cap value stocks is that is unique to market risk, but we can measure the effect that risk has on returns. We know some things
it seems that SCV portfolios have higher negative skew than larger or growthier portfolios, and some people don't like that skew and want to avoid it
It seems that SCV is has higher returns than LCG in times when inflation is higher rather than lower
it seems that SCV companies pay more for their financing than growth companies
it seems that SCV companies are less "popular" than LCG companies, especially among retail investors
so there's some data -- come up with some hypotheses for why that is the case and what it means for you
We could be at a point in finance/factor mmodels where it is difficult to describe from first principles what the unique risk of small-cap value stocks is that is unique to market risk, but we can measure the effect that risk has on returns. We know some things
it seems that SCV portfolios have higher negative skew than larger or growthier portfolios, and some people don't like that skew and want to avoid it
It seems that SCV is has higher returns than LCG in times when inflation is higher rather than lower
it seems that SCV companies pay more for their financing than growth companies
it seems that SCV companies are less "popular" than LCG companies, especially among retail investors
so there's some data -- come up with some hypotheses for why that is the case and what it means for you
Crom laughs at your Four Winds
Re: Fama and French: The Five-Factor Model Revisited
I question whether all (or even most) SCV investors are doing so under the belief that they are investing in a more risky asset. Many posters seem to think they are guaranteed to get a higher return, which, if true, means there is no additional risk. And that goes for any time period. If you are guaranteed to outpeform if you only wait long enough, then there is no additional risk. Many seem to think that higher risk automatically means higher return, when instead investors should demand higher return for taking higher risk. How are SCV investors demanding higher return for taking more risk?vineviz wrote: ↑Tue Jan 25, 2022 12:22 pm The model is what it is. What I objected to was your characterization of how "factor investors" interpret the model. Few are convinced of what you suggest, that the market is systematically mispricing risk. Instead, they are proceeding - for the most part -under a belief that the market is ACCURATELY pricing risk.
Re: Fama and French: The Five-Factor Model Revisited
I think everyone you'd consider a "factor proponent" in this thread has made it clear that this is NOT the way we see it. At what point will you consider that perhaps what you think we believe isn't what we actually believe?
Anyone who says they are using SCV stocks for diversification, for instance, MUST de facto believe that there is any additional systematic risk factor involved.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: Fama and French: The Five-Factor Model Revisited
For those looking to take a deep dive from first principles on CAPM, anomalies and rational/behavioral explanations of Fama-French factors, the Nobel prize committee's 2013 compilation article detailing the contributions of Fama, Hansen & Shiller is an excellent reference.
"Understanding Asset Prices"
https://faculty.fuqua.duke.edu/~rb7/bio ... _nobel.pdf
"Understanding Asset Prices"
https://faculty.fuqua.duke.edu/~rb7/bio ... _nobel.pdf
"Take a simple idea and take it seriously" ~ Charlie Munger
-
- Posts: 15363
- Joined: Fri Apr 10, 2015 12:29 am
Re: Fama and French: The Five-Factor Model Revisited
You can demand a higher expected return for taking more risk with equities. You can only demand an actual higher return with fixed income instruments.rkhusky wrote: ↑Tue Jan 25, 2022 3:08 pmI question whether all (or even most) SCV investors are doing so under the belief that they are investing in a more risky asset. Many posters seem to think they are guaranteed to get a higher return, which, if true, means there is no additional risk. And that goes for any time period. If you are guaranteed to outpeform if you only wait long enough, then there is no additional risk. Many seem to think that higher risk automatically means higher return, when instead investors should demand higher return for taking higher risk. How are SCV investors demanding higher return for taking more risk?vineviz wrote: ↑Tue Jan 25, 2022 12:22 pm The model is what it is. What I objected to was your characterization of how "factor investors" interpret the model. Few are convinced of what you suggest, that the market is systematically mispricing risk. Instead, they are proceeding - for the most part -under a belief that the market is ACCURATELY pricing risk.
Re: Fama and French: The Five-Factor Model Revisited
I work in credit, which explains why I find the factors compelling yet baffling. We need to explain every risk, how else could we set the price?Northern Flicker wrote: ↑Tue Jan 25, 2022 4:40 pm
You can only demand an actual higher return with fixed income instruments.