Fama and French: The Five-Factor Model Revisited

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

vanbogle59 wrote: Fri Jan 21, 2022 8:23 pm
willthrill81 wrote: Fri Jan 21, 2022 5:09 pm Can we capture the SCV premium by holding a total market index, or is it just that relative to the weighting of large caps in the index it would be barely noticeable? Dumb question that I should probably know, what is the reason that SCV is a non diversifiable risk factor?
No, you have no exposure to SCV by holding TSM because in addition to SCV companies you are also holding the antithesis to SCV, which is LCG, as well as SCG, LCV, etc.

I don't know quite what you mean by your last question.
OK, this is getting way too hand-wavy for me.
It is nonsensical to say that you have no exposure to stuff you own.

That can only be asserted in the view that SCV is a thing and the actual companies are not things.
I accept that the Factor model shows no deviation if all you own is TSM. But the model can't come first, the actual stocks have to have epistemological priority. :shock:

So, of course owners of TSM are exposed to SCV stocks (albeit very little).
But the model attempting to describe the behavior of a TSM portfolio cannot invoke the SCV factor if the composition of the portfolio is pure market weight.
[/quote]

Right, owning SCV companies doesn’t mean you load on the HmL and SmB factors, because in the total market you also own SCG. You cancel out the SCV because the market is only the market.

Allocating such that you load on the HmL and SmB factors is asking “how much MORE than the market do you hold in small and value companies?” Thus to get the risk premium you need to over-allocate with respect to the market. TSM holds SCV companies, but by definition doesn’t hold more than the SCV in the total market
Crom laughs at your Four Winds
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

km91 wrote: Fri Jan 21, 2022 11:00 pmLol I listen to his podcast, this definitely wasn't the face I would picture when I'd hear his voice. At the 10 minute mark he talks about the persistence of the small and value factors, but seems to point the the historical premiums in the back test as evidence of their continued existence. This doesn't explain to me why value in particular is a systematic risk that the market compensates investors for. Couldn't an investor just hold growth stocks in combination with value and diversify away whatever risk there might be. It's not clear to me what the actual risk being taken is when overweighting to towards value.
I'm still very much a novice, but as I understand it value stocks tend to be companies or sectors that most investors have skepticism or doubts about even though they have strong fundamentals. For instance, many investors expect EVs to dominate the future so have moved away from oil companies, thus reducing share-price even though they're still highly profitable. If EVs are less successful than hoped/expected oil companies could make a robust resurgence, but if EVs are successful they probably won't.

In more probabilistic terms, the value premium seems positive about 80% of the time over rolling 10-yr periods. This means there's about an 80% chance value stocks will beat the market and a 20% chance it wont; this doesn't mean a negative return, but returns for value would be equal to or less than the overall market.

The highlighted statement is essentially what cap-weight index investing is. Holding all equities in market-cap weights diversifies away all risk except the total market. As many of us have said, bonds are less risky than stocks but stocks are likely to have higher long-term returns. Value stocks are riskier than growth stocks but likely to have higher returns, esp SCV. By allocating more towards SCV you're taking on more risk in exchange for likely higher returns, but higher returns aren't guaranteed the same way stocks aren't guaranteed to out-perform bonds.

All this said, while most academics seem to think the 5-factor model is credible, some are skeptical. There's also concern as factors become more well known and accepted their premium will diminish. Some of us think they'll persist since many investors seem persistently overly optimistic regarding the return prospects of expensive growth stocks like Tesla. Only time will tell who's right...
Last edited by Apathizer on Sat Jan 22, 2022 11:49 am, edited 1 time in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Northern Flicker
Posts: 15354
Joined: Fri Apr 10, 2015 12:29 am

Re: Fama and French: The Five-Factor Model Revisited

Post by Northern Flicker »

The larger a company is, the more difficult it is to grow the business. For Amazon to grow their business by 10%, they have to capture alot more new market share than a microcap company needs to grow by 10%.
Northern Flicker
Posts: 15354
Joined: Fri Apr 10, 2015 12:29 am

Re: Fama and French: The Five-Factor Model Revisited

Post by Northern Flicker »

Fremdon Ferndock wrote: Tue Jan 18, 2022 12:27 pm Using Portfolio Visualizer, I found that the annualized compound return from TSM from 1972 was 10.93% and for SCV is was 14.01%. If you look at the year by year returns, TSM and SCV pretty much go up and down together. I would have liked that juicy 14% from SCV but, you know what? I would have been happy enough with the 11% from TSM. I don't want to be the richest stiff in the cemetery anyway.

As they say: "don't let the search for the perfect distract you from the good."
PV asset class backtests do not take transaction costs into account-- definitely much higher for small caps over much of the period, and SCV has to do more trading.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

Northern Flicker wrote: Sat Jan 22, 2022 2:31 am
Fremdon Ferndock wrote: Tue Jan 18, 2022 12:27 pm Using Portfolio Visualizer, I found that the annualized compound return from TSM from 1972 was 10.93% and for SCV is was 14.01%. If you look at the year by year returns, TSM and SCV pretty much go up and down together. I would have liked that juicy 14% from SCV but, you know what? I would have been happy enough with the 11% from TSM. I don't want to be the richest stiff in the cemetery anyway.

As they say: "don't let the search for the perfect distract you from the good."
PV asset class backtests do not take transaction costs into account-- definitely much higher for small caps over much of the period, and SCV has to do more trading.
The ER for AVUV is only 0.25, so if CAGR is about 2-3% higher than the total market that seems worth it to me.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 11:55 am The ER for AVUV is only 0.25, so if CAGR is about 2-3% higher than the total market that seems worth it to me.
That's a big "if". Could end up 2-3% lower.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Fama and French: The Five-Factor Model Revisited

Post by willthrill81 »

rkhusky wrote: Sat Jan 22, 2022 1:21 pm
Apathizer wrote: Sat Jan 22, 2022 11:55 am The ER for AVUV is only 0.25, so if CAGR is about 2-3% higher than the total market that seems worth it to me.
That's a big "if". Could end up 2-3% lower.
Of course, the returns don't need to be that much higher to compensate. In fact, even if the long-term average returns of the fund were equal to that of a TSM fund, it could still be worthwhile for the diversification benefit. Paying roughly 20 basis points more than for a TSM fund isn't much added expense.
The Sensible Steward
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

rkhusky wrote: Sat Jan 22, 2022 1:21 pm
Apathizer wrote: Sat Jan 22, 2022 11:55 am The ER for AVUV is only 0.25, so if CAGR is about 2-3% higher than the total market that seems worth it to me.
That's a big "if". Could end up 2-3% lower.
They could be, but it's not likely based on asset pricing theory and historical data. It's about 80% likely the TSM will under-perform SV.
willthrill81 wrote: Sat Jan 22, 2022 1:29 pm
rkhusky wrote: Sat Jan 22, 2022 1:21 pm
Apathizer wrote: Sat Jan 22, 2022 11:55 am The ER for AVUV is only 0.25, so if CAGR is about 2-3% higher than the total market that seems worth it to me.
That's a big "if". Could end up 2-3% lower.
Of course, the returns don't need to be that much higher to compensate. In fact, even if the long-term average returns of the fund were equal to that of a TSM fund, it could still be worthwhile for the diversification benefit. Paying roughly 20 basis points more than for a TSM fund isn't much added expense.
Exactly. So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.

But yes, as I've said almost all investors will do just fine with index funds or a single-fund composed of index funds that automatically re-balances.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
No it doesn't. There is no growth risk premium, and small caps as a whole are riskier and don't produce better returns than the TSM. Beneficial diversification means weighting towards potential return sources that are likely to beat the TSM. SG is likely to under-perform the market.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 3:27 pm
rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
No it doesn't. There is no growth risk premium, and small caps as a whole are riskier and don't produce better returns than the TSM. Beneficial diversification means weighting towards potential return sources that are likely to beat the TSM. SG is likely to under-perform the market.
What asset pricing model are you using? The FF model is symmetric with respect to growth/value and small/big (HmL, B_hml -> LmH, -B_hml, SmB, B_smb -> BmS, -B_smb). If there is no growth premium, there is no value premium.
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

rkhusky wrote: Sat Jan 22, 2022 4:16 pm
Apathizer wrote: Sat Jan 22, 2022 3:27 pm
rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
No it doesn't. There is no growth risk premium, and small caps as a whole are riskier and don't produce better returns than the TSM. Beneficial diversification means weighting towards potential return sources that are likely to beat the TSM. SG is likely to under-perform the market.
What asset pricing model are you using? The FF model is symmetric with respect to growth/value and small/big (HmL, B_hml -> LmH, -B_hml, SmB, B_smb -> BmS, -B_smb). If there is no growth premium, there is no value premium.
I think his point is that the growth premium is negative, not that it is zero
Crom laughs at your Four Winds
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
It doesn't, because the value factor (like most other robust factors) is negatively correlated with market beta. As a consequence, SCG stocks are more highly correlated with total stock market funds than SCV stocks.

Using annual returns from 1999 to 2021, the correlation between Vanguard Total Stock and Vanguard Small Cap Growth Index is 0.91 versus just 0.77 for Vanguard Small Cap Value Index. Volatility of the two is very similar, so SCV is the better diversifier for TSM. Link
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

muffins14 wrote: Sat Jan 22, 2022 4:21 pm
rkhusky wrote: Sat Jan 22, 2022 4:16 pm
Apathizer wrote: Sat Jan 22, 2022 3:27 pm
rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
No it doesn't. There is no growth risk premium, and small caps as a whole are riskier and don't produce better returns than the TSM. Beneficial diversification means weighting towards potential return sources that are likely to beat the TSM. SG is likely to under-perform the market.
What asset pricing model are you using? The FF model is symmetric with respect to growth/value and small/big (HmL, B_hml -> LmH, -B_hml, SmB, B_smb -> BmS, -B_smb). If there is no growth premium, there is no value premium.
I think his point is that the growth premium is negative, not that it is zero.
Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

Apathizer wrote: Sat Jan 22, 2022 1:15 am
km91 wrote: Fri Jan 21, 2022 11:00 pmLol I listen to his podcast, this definitely wasn't the face I would picture when I'd hear his voice. At the 10 minute mark he talks about the persistence of the small and value factors, but seems to point the the historical premiums in the back test as evidence of their continued existence. This doesn't explain to me why value in particular is a systematic risk that the market compensates investors for. Couldn't an investor just hold growth stocks in combination with value and diversify away whatever risk there might be. It's not clear to me what the actual risk being taken is when overweighting to towards value.
I'm still very much a novice, but as I understand it value stocks tend to be companies or sectors that most investors have skepticism or doubts about even though they have strong fundamentals. For instance, many investors expect EVs to dominate the future so have moved away from oil companies, thus reducing share-price even though they're still highly profitable. If EVs are less successful than hoped/expected oil companies could make a robust resurgence, but if EVs are successful they probably won't.

In more probabilistic terms, the value premium seems positive about 80% of the time over rolling 10-yr periods. This means there's about an 80% chance value stocks will beat the market and a 20% chance it wont; this doesn't mean a negative return, but returns for value would be equal to or less than the overall market.

The highlighted statement is essentially what cap-weight index investing is. Holding all equities in market-cap weights diversifies away all risk except the total market. As many of us have said, bonds are less risky than stocks but stocks are likely to have higher long-term returns. Value stocks are riskier than growth stocks but likely to have higher returns, esp SCV. By allocating more towards SCV you're taking on more risk in exchange for likely higher returns, but higher returns aren't guaranteed the same way stocks aren't guaranteed to out-perform bonds.

All this said, while most academics seem to think the 5-factor model is credible, some are skeptical. There's also concern as factors become more well known and accepted their premium will diminish. Some of us think they'll persist since many investors seem persistently overly optimistic regarding the return prospects of expensive growth stocks like Tesla. Only time will tell who's right...
This is my understanding of the value premium too, and I find it quite believable. Investors are overly optimistic and over pay for growth, and overly pessimistic and under pay for value. Makes sense to me. But the mispricing between growth and value would seem more like an anomaly caused by investor bias rather than a true systematic risk, with the implication being that we shouldn't expect it to persist because eventually the market will arbitrage it away. To try to give this some real world context, Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete. Does this make Macy's fundamentally more "risky" than Amazon that the market provides an additional expected return for investors who are willing to hold this "risk"? I don't see why it should, after all if you hold Macy's stock you can hedge away the risk that it's business model gets disrupted by holding the company who disrupted it.

I work in credit where everything is about risk. If we believe even the slightest risk exists in a deal, that risk is clearly defined and we try to mitigate against it. If the factors are truly risks that exist in addition to the market factor the first thing I want to do is qualify them so I can determine if the expected premium is compensating me for the additional risk I'm taking. The intuition behind the market factor is pretty straightforward, we all understand at some level stocks are "risky" and investors demand more than the risk free rate for in return for holding that risk. When it comes to the other factors, it's not clear to me why owning a bunch of stock with certain financial ratios adds systematic risk.

It seems like the best proof we have of the factors is in the research but the factors haven't performed as well since the research was published. The case for factors existing would say that the factors are uncorrelated to market risk and we should be fine with periods of underperformance as that's exactly what we would expect in an uncorrelated return stream. The case against would say that the AQRs and DFAs of the world have captured whatever factor premiums existed in the data and we shouldn't expect factors to provide a consistent premium over the market factor going forward. I think I tend to lean this way. Fama and French originally published in the early 90's. From 1995 through 2021, TSM and SCV basically performed identically, in fact SCV performed a tiny bit better. But it did so with a lower Sharpe ratio and no combination of TSM and SCV improved on on the Sharpe of holding 100% TSM. Maybe this is cherry picking the holding period but its the only out of sample data we have to test FF against. 26 years and no sign of the improved risk adjusted returns we'd expect to see from holding a return factor uncorrelated to the market factor.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

km91 wrote: Sat Jan 22, 2022 5:42 pm Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete. Does this make Macy's fundamentally more "risky" than Amazon that the market provides an additional expected return for investors who are willing to hold this "risk"? I don't see why it should, after all if you hold Macy's stock you can hedge away the risk that it's business model gets disrupted by holding the company who disrupted it.
I think it's more about price relative to fundamentals, and sometimes particular circumstances seem to temporarily depress certain companies. For instance, since Covid more people are staying home and using Amazon and fewer people are going out and shopping in person at retailers like Macy's. At some point when Covid alleviates in-person shopping will likely improve, thus benefiting retailers like Macy's, but we don't know if and when.

It's not that Macy's is better overall, but it seems to have a low share-price relative to fundamentals. In terms of expected returns, paying a low price for an average company might be better than buying an exorbitant price for a good company.
km91 wrote: Sat Jan 22, 2022 5:42 pmIt seems like the best proof we have of the factors is in the research but the factors haven't performed as well since the research was published.
That's true in the US, but factors have had a positive premium in ex-US markets. I suspect the US factor under-performance is a rare, but occasionally expected occurrence. Only time will tell...
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

km91 wrote: Sat Jan 22, 2022 5:42 pm But the mispricing between growth and value would seem more like an anomaly caused by investor bias rather than a true systematic risk, with the implication being that we shouldn't expect it to persist because eventually the market will arbitrage it away. To try to give this some real world context, Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete. Does this make Macy's fundamentally more "risky" than Amazon that the market provides an additional expected return for investors who are willing to hold this "risk"? I don't see why it should, after all if you hold Macy's stock you can hedge away the risk that it's business model gets disrupted by holding the company who disrupted it.
I'm more guilty than most people of throwing around the concept of "rational" and "irrational" behaviors, but we have to remember that it is always safer to assume that people are attempting to maximize utility than to assume they are attempting to maximize wealth.

To draw an analogy to food, most people don't consume foods solely (or even mostly) based on some objective measure like "macronutrients per dollar" or something. We eat to survive, yes, but we also tend to eat more of things we enjoy than of things we hate.

The preference that some investors have for growth stocks over value stocks might not be wealth maximizing, but if we assume that most people aren't stupid then the implication is that those investors have a preference or taste for something that growth stocks do that value stocks don't. One thing they have is negative coskewness with the overall stock market, and some investors seem to favor this coskewness enough to bid up (down) the price (return) of growth stocks. See Book-to-Market Ratio and Skewness of Stock Return by Zhang for more information.

In short, we have to separate behavioral biases which are motivated by incomplete information and/or inaccurate knowledge from behavioral biases which are motivated by tastes and/or preferences. The former may not be very durable or dependable, but the later can be.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Northern Flicker
Posts: 15354
Joined: Fri Apr 10, 2015 12:29 am

Re: Fama and French: The Five-Factor Model Revisited

Post by Northern Flicker »

km91 wrote: Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete.
This might be the behavior of individual investors who pick stocks with the erroneous belief that they can outsmart the market, but how much influence do they actually have on stock prices? I think that the overwhelming majority of trading that determines stock prices is done by institutional investors wielding sophisticated models of future revenue and the risk that the future revenue will fall short of projections.

Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

Northern Flicker wrote: Sat Jan 22, 2022 7:32 pm
km91 wrote: Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete.
Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
Why is this a risk that the market compensates investors for holding in addition to the market factor? This seems like plain old operating risk to me which should be captured by beta. By implication, shouldn't we expect ARKK or small cap biotech to provide better risk adjusted returns as the probability of their expected earnings potential is more uncertain than Macy's?
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

vineviz wrote: Sat Jan 22, 2022 4:37 pm It doesn't, because the value factor (like most other robust factors) is negatively correlated with market beta. As a consequence, SCG stocks are more highly correlated with total stock market funds than SCV stocks.

Using annual returns from 1999 to 2021, the correlation between Vanguard Total Stock and Vanguard Small Cap Growth Index is 0.91 versus just 0.77 for Vanguard Small Cap Value Index. Volatility of the two is very similar, so SCV is the better diversifier for TSM. Link
And SCG slightly outperformed SCV over that time period.

Also, those correlations are very time dependent. I note that there seems to be a trend in recent years for SCG to get less correlated and SCV to be more correlated with TSM. And the SCV premium also appears to be going negative.

Picking some random dates:
(TSM correlation, return)

1/1/99 - 12/31/03
SCG: 0.96, 7.86%
SCV: 0.56, 11.02%

1/1/99 - 12/31/07
SCG: 0.94, 9.48%
SCV: 0.52, 10.46%

1/1/03 - 12/31/07
SCG: 0.95, 17.21%
SCV: 0.90, 14.79%

1/1/07 - 12/31/21
SCG: 0.92, 10.91%
SCV: 0.91, 8.47%

1/1/17 - 12/31/21
SCG: 0.72, 16.77%
SCV: 0.92, 10.19%

Edit: I just noticed the correlation was set for yearly returns. If I put in monthly returns the correlations become much more similar. For example, for 1/1/99 - 12/31/21 the correlations of SCG and SCV with TSM are both 0.88.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
No, it's about both. Expecting higher returns of SV is based on both empirical evidence and logic as I think we've explained many times. Seriously, are just being pointlessly contrarian/trolling?
Last edited by Apathizer on Sat Jan 22, 2022 10:26 pm, edited 1 time in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

rkhusky wrote: Sat Jan 22, 2022 8:03 pm
vineviz wrote: Sat Jan 22, 2022 4:37 pm It doesn't, because the value factor (like most other robust factors) is negatively correlated with market beta. As a consequence, SCG stocks are more highly correlated with total stock market funds than SCV stocks.

Using annual returns from 1999 to 2021, the correlation between Vanguard Total Stock and Vanguard Small Cap Growth Index is 0.91 versus just 0.77 for Vanguard Small Cap Value Index. Volatility of the two is very similar, so SCV is the better diversifier for TSM. Link
And SCG slightly outperformed SCV over that time period.
Okay, but that has nothing to do with correlations.
rkhusky wrote: Sat Jan 22, 2022 8:03 pm Also, those correlations are very time dependent. I note that there seems to be a trend in recent years for SCG to get less correlated and SCV to be more correlated with TSM. And the SCV premium also appears to be going negative.
"Seems to be a trend" is a subjective evaluation. The correlations are what they are: expensive stocks have historically been more highly correlated with the market than value stocks over the full period for which we have data. And the correlations were higher in the first half of the data (1927 to 1973) than in the second half (1974 to 2020)
rkhusky wrote: Sat Jan 22, 2022 8:03 pm Edit: I just noticed the correlation was set for yearly returns. If I put in monthly returns the correlations become much more similar. For example, for 1/1/99 - 12/31/21 the correlations of SCG and SCV with TSM are both 0.88.
The measurement period for the correlation calculations should match the investment horizon of the investor in the asset. The implied investment horizon for equity investors is an average of about 7 to 10 years, so rolling 10 year returns would be the ideal period over which to calculate correlations. But annual returns are far more accurate for the purpose than monthly or daily returns. Unless your a day trader.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
Crom laughs at your Four Winds
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

vineviz wrote: Sat Jan 22, 2022 7:11 pm
km91 wrote: Sat Jan 22, 2022 5:42 pm But the mispricing between growth and value would seem more like an anomaly caused by investor bias rather than a true systematic risk, with the implication being that we shouldn't expect it to persist because eventually the market will arbitrage it away. To try to give this some real world context, Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete. Does this make Macy's fundamentally more "risky" than Amazon that the market provides an additional expected return for investors who are willing to hold this "risk"? I don't see why it should, after all if you hold Macy's stock you can hedge away the risk that it's business model gets disrupted by holding the company who disrupted it.
I'm more guilty than most people of throwing around the concept of "rational" and "irrational" behaviors, but we have to remember that it is always safer to assume that people are attempting to maximize utility than to assume they are attempting to maximize wealth.

To draw an analogy to food, most people don't consume foods solely (or even mostly) based on some objective measure like "macronutrients per dollar" or something. We eat to survive, yes, but we also tend to eat more of things we enjoy than of things we hate.

The preference that some investors have for growth stocks over value stocks might not be wealth maximizing, but if we assume that most people aren't stupid then the implication is that those investors have a preference or taste for something that growth stocks do that value stocks don't. One thing they have is negative coskewness with the overall stock market, and some investors seem to favor this coskewness enough to bid up (down) the price (return) of growth stocks. See Book-to-Market Ratio and Skewness of Stock Return by Zhang for more information.

In short, we have to separate behavioral biases which are motivated by incomplete information and/or inaccurate knowledge from behavioral biases which are motivated by tastes and/or preferences. The former may not be very durable or dependable, but the later can be.
This makes sense to me that investors prefer stocks with certain characteristics. I think the obvious one would be liquidity. Liquidity comes at a cost, an investor selling into an illiquid market needs to pay up, so I can see why investors would pay a premium for liquid blue chip names. I need to think a bit further on the conclusion from the paper that investors prefer the lottery-like returns of growth stocks. I don't doubt the finding and I've heard it before, I'm just not sure I fully grasp why this is a systematic risk premium in the same way liquidity is. There is no alternative to liquidity when you need to sell. Leverage or derivatives seem like they would be an alternative to growth stocks if I am understanding the conclusion of the paper correctly that investors seek out investments with greater opportunities for right tail outcomes.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

muffins14 wrote: Sat Jan 22, 2022 8:44 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
Yep. I'd be happy to build an asset pricing model using future data instead of historical data. I'm just waiting for someone to supply me with the data from 2022 to 2121.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

km91 wrote: Sat Jan 22, 2022 8:57 pm Leverage or derivatives seem like they would be an alternative to growth stocks if I am understanding the conclusion of the paper correctly that investors seek out investments with greater opportunities for right tail outcomes.
Leverage isn't available to all investors, and it is almost never free or even cheap. Plus, the relationship of growth stock returns to broad market returns isn't linear so the full character can't be mimicked solely with leverage.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 8:23 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
No, it's about both. Expecting higher returns of SV is based on both on empirical evidence and logic as I think we've explained many times. Seriously, are just being pointlessly contrarian/trolling?
The FF model says nothing about whether the HmL or SmB premiums will be positive or negative in the future. They are an input to the model. So how are you using asset pricing models to get your expectations for higher returns from SCV?
Last edited by rkhusky on Sat Jan 22, 2022 9:28 pm, edited 1 time in total.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

muffins14 wrote: Sat Jan 22, 2022 8:44 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
The FF model says nothing about whether the HmL or SmB premiums will be positive or negative in the future. They are an input to the model. If you provide an estimate of future premiums you can obtain an estimate of investment return.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

vineviz wrote: Sat Jan 22, 2022 9:00 pm
muffins14 wrote: Sat Jan 22, 2022 8:44 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
Yep. I'd be happy to build an asset pricing model using future data instead of historical data. I'm just waiting for someone to supply me with the data from 2022 to 2121.
One can use parameters obtained from past correlation data along with estimates for future premiums to calculate estimates for future investment returns.
klaus14
Posts: 953
Joined: Sun Nov 25, 2018 6:43 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by klaus14 »

vineviz wrote: Sat Jan 22, 2022 4:37 pm
rkhusky wrote: Sat Jan 22, 2022 3:21 pm
Apathizer wrote: Sat Jan 22, 2022 1:47 pm So many who dismiss SV don't seem to consider the diversification benefits. Sometimes the market lags SV. In fact, historically value beats growth about as often as the market beats bonds (about 80%). So a well-diversified portfolio should balance potential sources of return: total market, HML, SMB, RMW, CMA, and diversified bonds.
SG provides just as much diversification benefit.
It doesn't, because the value factor (like most other robust factors) is negatively correlated with market beta. As a consequence, SCG stocks are more highly correlated with total stock market funds than SCV stocks.

Using annual returns from 1999 to 2021, the correlation between Vanguard Total Stock and Vanguard Small Cap Growth Index is 0.91 versus just 0.77 for Vanguard Small Cap Value Index. Volatility of the two is very similar, so SCV is the better diversifier for TSM. Link
if you exclude 1999-2001, the effect disappears. Was it a one off divergence or not?
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

vineviz wrote: Sat Jan 22, 2022 8:32 pm
"Seems to be a trend" is a subjective evaluation. The correlations are what they are: expensive stocks have historically been more highly correlated with the market than value stocks over the full period for which we have data. And the correlations were higher in the first half of the data (1927 to 1973) than in the second half (1974 to 2020)
I doubt that early data has much relevance to today’s investing with cloud computing, machine learning, FF modeling, etc etc

All that simple value investing stuff is going by the wayside.
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

vineviz wrote: Sat Jan 22, 2022 9:06 pm
km91 wrote: Sat Jan 22, 2022 8:57 pm Leverage or derivatives seem like they would be an alternative to growth stocks if I am understanding the conclusion of the paper correctly that investors seek out investments with greater opportunities for right tail outcomes.
Leverage isn't available to all investors, and it is almost never free or even cheap. Plus, the relationship of growth stock returns to broad market returns isn't linear so the full character can't be mimicked solely with leverage.
Is this the right way to interpret the paper: book to market represents a lower bound on stock prices since book to market >1 implies that the company is worth more in liquidation than the market is currently valuing it at. As book to market decreases (stock price increases) the stock's relationship to the market return increases by more than suggested by beta alone and investors are willing to pay a premium for that convexity
Northern Flicker
Posts: 15354
Joined: Fri Apr 10, 2015 12:29 am

Re: Fama and French: The Five-Factor Model Revisited

Post by Northern Flicker »

km91 wrote: Sat Jan 22, 2022 7:56 pm
Northern Flicker wrote: Sat Jan 22, 2022 7:32 pm
km91 wrote: Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete.
Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
Why is this a risk that the market compensates investors for holding in addition to the market factor? This seems like plain old operating risk to me which should be captured by beta. By implication, shouldn't we expect ARKK or small cap biotech to provide better risk adjusted returns as the probability of their expected earnings potential is more uncertain than Macy's?
It is how a stock becomes a value stock in the first place-- the market discounting projected future cash flows back to a present value using a larger discount rate to reflect the higher risk premium, resulting in a lower present value, and hence lower valuation relative to the projected cash flows.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

vineviz wrote: Sat Jan 22, 2022 9:00 pm
muffins14 wrote: Sat Jan 22, 2022 8:44 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
Yep. I'd be happy to build an asset pricing model using future data instead of historical data. I'm just waiting for someone to supply me with the data from 2022 to 2121.
:D LOL :D
Yes, that's largely how scientific models work. Attempting to understand commonalities of equities that have had the highest historical returns is as reasonable a predictor of future returns we have. Of course even the best models are imperfect, simplified representation of the facet of complex reality they endeavor to understand. All we can do is use models shown to be the most successful. The best investment model we have is FF asset pricing.
Last edited by Apathizer on Sat Jan 22, 2022 11:14 pm, edited 1 time in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

Northern Flicker wrote: Sat Jan 22, 2022 10:47 pm
km91 wrote: Sat Jan 22, 2022 7:56 pm
Northern Flicker wrote: Sat Jan 22, 2022 7:32 pm
km91 wrote: Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete.
Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
Why is this a risk that the market compensates investors for holding in addition to the market factor? This seems like plain old operating risk to me which should be captured by beta. By implication, shouldn't we expect ARKK or small cap biotech to provide better risk adjusted returns as the probability of their expected earnings potential is more uncertain than Macy's?
It is how a stock becomes a value stock in the first place-- the market discounting projected future cash flows back to a present value using a larger discount rate to reflect the higher risk premium, resulting in a lower present value, and hence lower valuation relative to the projected cash flows.
I understand the discounting but why would the uncertainty of future cash flows be expressed as a factor separate from beta? Uncertainty of future cash flow is a fundamental risk when holding equity. Every stock has uncertain cash flows by it's very nature. Why would stocks with particularly high uncertainty earn a return greater than what is implied by their beta?
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

rkhusky wrote: Sat Jan 22, 2022 9:16 pm
muffins14 wrote: Sat Jan 22, 2022 8:44 pm
rkhusky wrote: Sat Jan 22, 2022 7:48 pm
Apathizer wrote: Sat Jan 22, 2022 4:41 pm Exactly. I should have stated this clearly as you did. I suspect rkhusky knows this and is just being contrarian. But yes, the idea is to increase diversity as to likely increase returns relative to TSM. Since growth stocks are likely to under-perform the market, having a growth slant doesn't make sense, esp since the TSM is currently overweight LG.
So, it's really all about past performance, not asset pricing models.
The model estimates past returns. You can’t learn an asset pricing model without a dataset of price performance, which by definition must be from the past
The FF model says nothing about whether the HmL or SmB premiums will be positive or negative in the future. They are an input to the model. If you provide an estimate of future premiums you can obtain an estimate of investment return.
The variable is an input, the coefficient is learned and is an output.

Like y = m*x + b in regression, x is an input variable, or predictor, and m is a coefficient that is learned.

HmL or SmB are inputs variables, and with your dataset you learn the coefficient values. If HmL is positive, there is a value premium. That is what is learned on a given dataset. So far, we observe a positive coefficient and thus a value premium. If the coefficient were zero, there is no premium. If negative, there is a growth premium.

Likely since 2010 that coefficient is negative, since value performed worse than the market
Last edited by muffins14 on Sat Jan 22, 2022 11:39 pm, edited 1 time in total.
Crom laughs at your Four Winds
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Apathizer »

km91 wrote: Sat Jan 22, 2022 11:13 pmI understand the discounting but why would the uncertainty of future cash flows be expressed as a factor separate from beta? Uncertainty of future cash flow is a fundamental risk when holding equity. Every stock has uncertain cash flows by it's very nature. Why would stocks with particularly high uncertainty earn a return greater than what is implied by their beta?
It depends on share price. Cheap stocks aren't the same as value stocks. Some stocks have a low share price because their future prospects are dismal. Other relatively low price stocks appear to have solid fundamentals; these are value stocks. A retailer on the verge of bankruptcy isn't a value stock since their future prospects are dismal. Contrast this with Chevron: a highly profitable company with a relatively low share-price. Presumably since many investors expect EVs to eventually dominate, future earnings are considered riskier.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

km91 wrote: Sat Jan 22, 2022 11:13 pm
Northern Flicker wrote: Sat Jan 22, 2022 10:47 pm
km91 wrote: Sat Jan 22, 2022 7:56 pm
Northern Flicker wrote: Sat Jan 22, 2022 7:32 pm
km91 wrote: Macy's is a top holding of AVUV. It's business and stock price have been in a down trend for a while now. On the other hand, Amazon's stock price and business have done great over that same time, largely at the expense of businesses like Macy's. I can see why investors might irrationally favor Amazon to Macy's, who wants to own a company whose whole business model was made obsolete when you can own the fast growing tech company that made it obsolete.
Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
Why is this a risk that the market compensates investors for holding in addition to the market factor? This seems like plain old operating risk to me which should be captured by beta. By implication, shouldn't we expect ARKK or small cap biotech to provide better risk adjusted returns as the probability of their expected earnings potential is more uncertain than Macy's?
It is how a stock becomes a value stock in the first place-- the market discounting projected future cash flows back to a present value using a larger discount rate to reflect the higher risk premium, resulting in a lower present value, and hence lower valuation relative to the projected cash flows.
I understand the discounting but why would the uncertainty of future cash flows be expressed as a factor separate from beta? Uncertainty of future cash flow is a fundamental risk when holding equity. Every stock has uncertain cash flows by it's very nature. Why would stocks with particularly high uncertainty earn a return greater than what is implied by their beta?
The risk is not the same as the overall market risk. It’s a distinct variable
Crom laughs at your Four Winds
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

muffins14 wrote: Sat Jan 22, 2022 11:38 pm
km91 wrote: Sat Jan 22, 2022 11:13 pm
Northern Flicker wrote: Sat Jan 22, 2022 10:47 pm
km91 wrote: Sat Jan 22, 2022 7:56 pm
Northern Flicker wrote: Sat Jan 22, 2022 7:32 pm

Thus, my view is that the difference in valuation between Amazon and Macy's reflects differing probabilities of future projected revenue materializing. This is the systematic risk factor explanation. I'm generally very suspicious of market narratives that suscribe market action to the beliefs and actions of individual investors.
Why is this a risk that the market compensates investors for holding in addition to the market factor? This seems like plain old operating risk to me which should be captured by beta. By implication, shouldn't we expect ARKK or small cap biotech to provide better risk adjusted returns as the probability of their expected earnings potential is more uncertain than Macy's?
It is how a stock becomes a value stock in the first place-- the market discounting projected future cash flows back to a present value using a larger discount rate to reflect the higher risk premium, resulting in a lower present value, and hence lower valuation relative to the projected cash flows.
I understand the discounting but why would the uncertainty of future cash flows be expressed as a factor separate from beta? Uncertainty of future cash flow is a fundamental risk when holding equity. Every stock has uncertain cash flows by it's very nature. Why would stocks with particularly high uncertainty earn a return greater than what is implied by their beta?
The risk is not the same as the overall market risk. It’s a distinct variable
Which is exactly what I'm trying to get at. If value represents a risk separate from market risk, what is it? What is a qualitative description of the risk being taken when tilting towards SCV? Market risk is the risk from holding an asset that exposes you to the operating losses of the underlying business, credit risk is the risk of loss from counterparty non payment. Value is supposed to represent a systematic risk yet a simple description of that risk is hard to come by. Seems like the only way to understand the FF risk factors is to have a PhD from the University of Chicago
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

klaus14 wrote: Sat Jan 22, 2022 9:22 pm if you exclude 1999-2001, the effect disappears. Was it a one off divergence or not?
No it doesn’t disappear. It’s a robust to time period. It’s true from 2002 to present, and it’s true before 1999 as well.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

muffins14 wrote: Sat Jan 22, 2022 11:35 pm So far, we observe a positive coefficient and thus a value premium. If the coefficient were zero, there is no premium. If negative, there is a growth premium.

Likely since 2010 that coefficient is negative, since value performed worse than the market
As you say, the results depend on the time frame. The future premium over your remaining time frame could be positive, negative or zero.
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

Apathizer wrote: Sat Jan 22, 2022 11:36 pm
km91 wrote: Sat Jan 22, 2022 11:13 pmI understand the discounting but why would the uncertainty of future cash flows be expressed as a factor separate from beta? Uncertainty of future cash flow is a fundamental risk when holding equity. Every stock has uncertain cash flows by it's very nature. Why would stocks with particularly high uncertainty earn a return greater than what is implied by their beta?
It depends on share price. Cheap stocks aren't the same as value stocks. Some stocks have a low share price because their future prospects are dismal. Other relatively low price stocks appear to have solid fundamentals; these are value stocks. A retailer on the verge of bankruptcy isn't a value stock since their future prospects are dismal. Contrast this with Chevron: a highly profitable company with a relatively low share-price. Presumably since many investors expect EVs to eventually dominate, future earnings are considered riskier.
Sounds like you are talking about fundamental analysis and stock picking. Most that use FF models in SCV investing have simple metrics to screen out undesirable stocks. And if one screens out undesirable stocks, does that make the resulting basket less risky and one should expect a lower risk premium?
rkhusky
Posts: 17756
Joined: Thu Aug 18, 2011 8:09 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by rkhusky »

vineviz wrote: Sun Jan 23, 2022 12:12 am
klaus14 wrote: Sat Jan 22, 2022 9:22 pm if you exclude 1999-2001, the effect disappears. Was it a one off divergence or not?
No it doesn’t disappear. It’s a robust to time period. It’s true from 2002 to present, and it’s true before 1999 as well.
From the PV link you provided with yearly correlation:
(TSM correlation, annual return)

1/1/02- 12/31/21
SCG: 0.92, 11.00%
SCV: 0.91, 9.58%

Not seeing much of an extra diversification or SCV premium for that 20 year period.
muffins14
Posts: 5519
Joined: Wed Oct 26, 2016 4:14 am
Location: New York

Re: Fama and French: The Five-Factor Model Revisited

Post by muffins14 »

rkhusky wrote: Sun Jan 23, 2022 8:05 am
muffins14 wrote: Sat Jan 22, 2022 11:35 pm So far, we observe a positive coefficient and thus a value premium. If the coefficient were zero, there is no premium. If negative, there is a growth premium.

Likely since 2010 that coefficient is negative, since value performed worse than the market
As you say, the results depend on the time frame. The future premium over your remaining time frame could be positive, negative or zero.

Correct. So far, the premium has been positive on average before publication and after publication. The uncertainty of the estimate is somewhat large though, such that in th US it may currently be both consistent with zero in some post-publication window and also consistent with being the same as it was in the pre-publication time period.
Crom laughs at your Four Winds
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Fama and French: The Five-Factor Model Revisited

Post by vineviz »

rkhusky wrote: Sun Jan 23, 2022 8:13 am Not seeing much of an extra diversification or SCV premium for that 20 year period.
If you go around excluding data that isn't consistent with your hypothesis you're less likely to every have to change your mind, I guess.

Even excluding those two years (again, why?) the correlation of SCG is still higher than SCV from 2002 to present. And it's not the only data we have, as I already pointed out. Over the entire 1927 to 2021 the effect was the same.

You can see it in the factor data directly, by observing the correlation between market beta and HmL is negative over the full 1964 to 2021 period as well. It's true using both the FF5 and AQR definition of HmL.

Image
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
km91
Posts: 1388
Joined: Wed Oct 13, 2021 12:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by km91 »

muffins14 wrote: Sun Jan 23, 2022 9:04 am
rkhusky wrote: Sun Jan 23, 2022 8:05 am
muffins14 wrote: Sat Jan 22, 2022 11:35 pm So far, we observe a positive coefficient and thus a value premium. If the coefficient were zero, there is no premium. If negative, there is a growth premium.

Likely since 2010 that coefficient is negative, since value performed worse than the market
As you say, the results depend on the time frame. The future premium over your remaining time frame could be positive, negative or zero.

Correct. So far, the premium has been positive on average before publication and after publication. The uncertainty of the estimate is somewhat large though, such that in th US it may currently be both consistent with zero in some post-publication window and also consistent with being the same as it was in the pre-publication time period.
Isn't there a risk that in the time since FF published the factors firms like AQR have arbitraged whatever factor premiums existed out of the market? Basically the Moneyball effect, once everyone is using the same model it loses its advantage. If one just needs to screen TSM for stocks with certain valuation ratios to identify value, shouldn't we expect that firms like AQR and DFA will bid up the price of these stocks until the premium disappears? I'm perfectly willing to look past the last 10 years of factor underperformance if the factors truly are systematic risks uncorrelated to the market factor. Is the rise of quant funds and the simultaneous underperformance of the factors just a coincidence?
User avatar
hillclimber
Posts: 136
Joined: Wed Mar 10, 2021 2:32 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by hillclimber »

km91 wrote: Sun Jan 23, 2022 10:44 am Isn't there a risk that in the time since FF published the factors firms like AQR have arbitraged whatever factor premiums existed out of the market? Basically the Moneyball effect, once everyone is using the same model it loses its advantage. If one just needs to screen TSM for stocks with certain valuation ratios to identify value, shouldn't we expect that firms like AQR and DFA will bid up the price of these stocks until the premium disappears? I'm perfectly willing to look past the last 10 years of factor underperformance if the factors truly are systematic risks uncorrelated to the market factor. Is the rise of quant funds and the simultaneous underperformance of the factors just a coincidence?
I think that risk is an underappreciated one. I recently ran a screen for etfs, and there was about a trillion of net assets in value funds, and that is just looking at etfs. DFA has 679 billion AUM. Obviously that includes fixed income, and not all of the stocks are in value strategies, but that's still a lot of money chasing value.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Fama and French: The Five-Factor Model Revisited

Post by Random Walker »

km91 wrote: Sun Jan 23, 2022 10:44 am
Isn't there a risk that in the time since FF published the factors firms like AQR have arbitraged whatever factor premiums existed out of the market? Basically the Moneyball effect, once everyone is using the same model it loses its advantage. If one just needs to screen TSM for stocks with certain valuation ratios to identify value, shouldn't we expect that firms like AQR and DFA will bid up the price of these stocks until the premium disappears? I'm perfectly willing to look past the last 10 years of factor underperformance if the factors truly are systematic risks uncorrelated to the market factor. Is the rise of quant funds and the simultaneous underperformance of the factors just a coincidence?
I’ve fallen into this trap before myself. If the factors represent risks, then we would not expect the premiums to be arbitraged away. Markets price risk as opposed to expected returns. We expect an efficient market to result in about the same Sharpe ratios for different investments, but not the same returns.

Some people dispute whether size is a risk, so that factor can be clouded. I believe it is a risk, and it seems that screening for investment and profitability saves the size premium. Value has both risk based and behavioral based components. We would expect the risk portion to not be arbitraged away. I think we should expect the behavioral portion to be partially arbitraged away. Certainly if knowledgeable investors know about it, they will jump on the anomaly. But there are limits to arbitrage that make totally capitalizing on a behavioral anomalies less likely.

Dave
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Fama and French: The Five-Factor Model Revisited

Post by willthrill81 »

km91 wrote: Sun Jan 23, 2022 10:44 am Isn't there a risk that in the time since FF published the factors firms like AQR have arbitraged whatever factor premiums existed out of the market?
For that to happen, the market had to be unaware of the premium before it was published, which would require the market to be quite inefficient.
The Sensible Steward
Post Reply