Factor Investing Makes Someone Money, But Maybe Not You

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
stlutz
Posts: 4665
Joined: Fri Jan 02, 2009 1:08 am

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by stlutz » Fri Jun 22, 2018 6:09 pm

To quantify this you can use the Graham formula:

P/E = 8.5 + 2g

So for every 1% reduction in growth you should expect your multiple to decline by 2. So for small changes in growth you can see large changes in the normalized difference between the growth and value stocks.

As to stock pickers using factors, it is not true. The processes of selecting stocks is fundamentally different. Factors do not take into account fundamentals they just use what has driven stock prices in the past for selection versus understanding businesses (which is what most stock pickers do) and what drives the value in the future. Just my 2 cents.
All factors are are simple formulas like this, however. Graham & Dodd advocated a pretty quantitative approach to security selection. In subsequent backtesting, some of their formulas ended up correlating to higher returns and some didn't. The ones that did are primarily the "factors" we're left with now.

A security selection process can include additional qualitative analysis or it can't. Up until now, computers have had an easier time handling numbers and that's why factor investing has been so oriented around numerical analysis. Computer-driven investment is increasingly moving in the direction of analyzing non-numeric data, so that's going to change going forward. As that gains more attention, "factors" may become yesterdays buzzword and will be replaced by some new sophisticated-sounding term to replace the plain English explanation of having a computer do what humans used to do in security selection.

User avatar
packer16
Posts: 1019
Joined: Sat Jan 04, 2014 2:28 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by packer16 » Fri Jun 22, 2018 6:29 pm

jeffyscott wrote:
Fri Jun 22, 2018 5:27 pm
packer16 wrote:
Fri Jun 22, 2018 3:14 pm
As to stock pickers using factors, it is not true. The processes of selecting stocks is fundamentally different. Factors do not take into account fundamentals they just use what has driven stock prices in the past for selection versus understanding businesses (which is what most stock pickers do) and what drives the value in the future.
There are managed value funds, small cap funds, small cap value funds. They may or may not be using "factor screens" in the way a passive fund would, but they are buying the same stocks. The end result is the same: an active fund tilted toward one factor or another.

If lots of people are selling managed small cap and value funds and buying total market index funds, that sure seems like it could lower the price of small cap value stocks due to reduced demand.
The problem is they are not buying the same stocks with the same weights. If a value manager has stocks that screen well there may be some overlap but in a good number of cases value managers are buying stocks that do not screen well and thus may not show up in the value screen. You are confusing result & process. Just because the result at one time period may be similar (& the value factor is calibrated for that period) does not mean that in subsequent periods the resulting securities will the same because the process is different.

Packer
Buy cheap and something good might happen

Random Walker
Posts: 2989
Joined: Fri Feb 23, 2007 8:21 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by Random Walker » Fri Jun 22, 2018 6:32 pm

rkhusky wrote:
Fri Jun 22, 2018 4:21 pm
The long-short formulation in FF is relatively independent. Long small or long value by itself is not independent.
A long only fund does have exposure to the independent factor, usually about half the exposure of the pure definition,because as you say, you’re only taking the long side. So a long only fund has the factor somewhat diluted and takes on additional market factor exposure compared to the pure long short definition.

Dave

2pedals
Posts: 543
Joined: Wed Dec 31, 2014 12:31 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by 2pedals » Fri Jun 22, 2018 7:25 pm

Rick Ferri wrote:
Thu Jun 21, 2018 12:46 pm
All the chatter about tilting to factors today reminds me of the excitement of tilting to industry sector funds when they flourished in the 1990s.

Rick Ferri
+1, the sector funds of the 2010s. No one knows which factor to use or how much tilt belongs in the allocation. How does one know which one will (if any) outperform the total stock market index fund in the long run after costs? Brings back memories of the sector rotation strategy.

MrPotatoHead
Posts: 429
Joined: Sat Oct 14, 2017 10:41 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by MrPotatoHead » Fri Jun 22, 2018 7:38 pm

Rick Ferri wrote:
Thu Jun 21, 2018 8:14 am
Factor investing is discussed ad nauseam on this forum. Lately, it has reached the point were investors who shouldn’t do factor investing become confused to the point of believing they should or they’ll miss out on something. Let’s correct that.

Factor investing is complex and the “optimal” way to do it is always changing because there is no optimal way that can be known in advance. The idea that factor premiums can be predicted is a fallacy. The small cap premium was discovered in the 1970s... and has been dead ever since.

That being said, it’s all to common for advisers to relentlessly push complex investment strategies such as “optimal factor allocation” because it makes the adviser sound like they’re smart and that brings in business to their firm.

In truth, doing these strategies through an adviser cost more money. Even if it works the net-of-fee result to an investor isn’t any better than a total market index fund.

Even if done without an adviser the strategy is riskier and more costly than using traditional index funds. Also, it’s not easy sticking with something that isn’t working, which leads to behavioral mistakes and added cost.

What’s the right answer? If you really understand factors and are committed to a strategy for at least 25 years, then do factors. For everyone eles, buy a few low cost market-tracking index funds and fahgetaboutit.
Thank you for this post. I truly thank you.

User avatar
vineviz
Posts: 1274
Joined: Tue May 15, 2018 1:55 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by vineviz » Fri Jun 22, 2018 8:38 pm

packer16 wrote:
Fri Jun 22, 2018 6:29 pm
The problem is they are not buying the same stocks with the same weights. If a value manager has stocks that screen well there may be some overlap but in a good number of cases value managers are buying stocks that do not screen well and thus may not show up in the value screen. You are confusing result & process. Just because the result at one time period may be similar (& the value factor is calibrated for that period) does not mean that in subsequent periods the resulting securities will the same because the process is different.
Whether or not a factor-managed fund owns exactly the same stocks as an actively managed fund isn't really critical (although I think you'd be surprised at just how much overlap there genuinely is), nor does it much matter exactly what process an active manager uses to arrive at their portfolio.

Rather, the critical question is whether active managers are doing something that can't be done more successfully and less expensive some other way. This question has been studied multiple times, over several decades now, and the answer is a pretty definitive "no". Active managers construct portfolios that can be very well replicated using much less expensive methods by focusing on factors.

I used Morningstar to screen ten of the best-performing active large cap value funds, which on average have just barely outperformed the S&P 500 due to this being a period that favored growth stocks over value stocks. It's an impressive group of funds: AIG Focused Dividend Strategy A, Boston Partners All Cap Value Instl, Columbia Dividend Income Inst, Copley, Delaware Value Inst, Dreyfus Equity Income I, JPMorgan Equity Income I, Macquarie Pooled Large Cap Value, Northern Income Equity, Putnam Equity Income A.

These funds are undoubtedly managed by smart people, although there is always an element of luck involved. Some of the factor data I have only goes back to 2010, but since under a CAPM model these funds generated about 129 bps of alpha after expenses which a naive investor would attribute to skill. I wouldn't.

The fund managers were picking stocks that exhibit the same sort of value characteristics a factor would would pick, including a focus on price/book (which shows up on the HmL risk factor) and dividend yield (which shows up on the term risk factor). They are screening for quality (AQR's QmJ factor). Once you account for these three additional risk factors, their performance swings from 129 bps of outperformance to 117 bps of UNDERperformance. Whatever work these active managers were doing was costing shareholders almost 250 bps per year relative to a factor-managed alternative.

Whether they were consciously picking stocks based on risk factors or not, they ended up with a portfolio that LOOKS like it was picked based on risk factors EXCEPT that it was way more expensive for investors. When you can pick a list of ten best actively managed large cap value funds and NONE of them beat a value index fund like Invesco S&P 500 Pure Value ETF (RPV), you have to wonder why investors keep paying active managers the money they do.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

User avatar
Robert T
Posts: 2521
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by Robert T » Sat Jun 23, 2018 6:16 am

.
Just had a look at the differences between my actual returns (P1 = actual in table below) and two portfolios: (i) a three fund portfolio with a similar stock:bond and US:non-US split in equities (37.5:37.5:25 Vanguard total stock market (VTSMX):Vanguard total international stock fund (VGTSX):Vanguard total bond (VBMFX) – (P2 = Three Fund in table below); and (ii) 75:25 MSCI All Cap World Index:US Aggregate Bond Index (P3 = ACWI/TBM in table below).

My portfolio has held up fairly well with higher returns over the last 15, 10, 5, and 3 year periods. This is despite views that the small cap premium is dead and the value premium is following suit. If the small cap premium is dead, it was “deader” 20 years ago as over the last 2 decades an investment in Bridgeway ultra-small company market (BRSIX) – microcaps - would have grown to more than double the same investment in Vanguard 500 Index (VFINX) https://www.portfoliovisualizer.com/bac ... ion2_2=100
  • Annualized Return (%)

    ……………………..… 15yr / 10yr / 5yr / 3yr / 2008 downside

    P1 = Acutal ……..….9.9 / 6.0 / 9.6 / 8.6 / -28.7
    P2 = Three fund ..…8.6 / 5.5 / 9.1 / 7.9 / -29.2
    P3 = ACWI/TBM....…8.2 / 5.1 / 8.7 / 7.6 / -30.3

    P1-P2 ………………+1.3/+0.5/+0.5/+0.7
    P1-P3 ………………+1.7/+0.9/+0.8/+1.0
P1 returns were relatively higher in the first 5 years than the subsequent 10 years, but were still higher in the latter period. And $1 growth in P1 = 26% more than P3 and 20% more than P2. A 75:25 Vanguard Lifestrategy Growth:Moderate Growth had the same 15 yr returns as P3, and has lower returns in all periods except last 5 yrs.

FWIW, my preferred “market” benchmark is P3. It uses the MSCI ACWI Index tracked by the iShares fund ACWI. The Vanguard International Stock fund has changed its benchmark three times over this period.

An alternative is to construct a benchmark from Ken French’s data that matches my factor load targets (within equities = 1.0, 0.2, 0.4 for market, size, and value, and within bonds = 0.0 and 0.5 for default and term). Annualized return over last 15 years = 9.8%.

The above example illustrates that costs matter - a point Rick brought up in the OP. A 1% advisor fee would more than have eliminated any premium returns over the last 10 years.

My expected long-term return at portfolio inception was 7.5% (when ‘expected return’ = ‘needed’ return). This is the most important benchmark, more so than any of the above. In this respect, doing okay - performance has been better than expected (so far).

If you are not convinced about the merits of tilting away from the market portfolio, simply hold the market portfolio. If you do tilt away from the market, then stick with it over the long-term (don’t chop and change every 3-5 years).

If I recall correctly, Wes Gray said something along the lines of the following – there is not enough time left in his life to produce enough out of sample data to disprove the existence of the value (and momentum) premiums – hence he is in it for the rest of his life. Shorter-term data 3, 5, 10 yrs etc. does not tell you a whole lot about the long-term existence of factor premiums – yet most of the discussions seem dominated by these shorter-term periods.

Obviously no guarantees.

Robert
.

User avatar
jeffyscott
Posts: 7176
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by jeffyscott » Sat Jun 23, 2018 7:46 am

Robert T wrote:
Sat Jun 23, 2018 6:16 am
My portfolio has held up fairly well with higher returns over the last 15, 10, 5, and 3 year periods.
What time period did you look at...is it through today, the end of 2017, or... ?

If you do tilt away from the market, then stick with it over the long-term (don’t chop and change every 3-5 years).

I tend to do the chop and change :twisted: , based on relative valuations. But it seems to have worked out okay.
press on, regardless - John C. Bogle

phantom cosmonaut
Posts: 15
Joined: Wed Jun 13, 2018 4:57 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by phantom cosmonaut » Sat Jun 23, 2018 9:34 am

Robert T wrote:
Sat Jun 23, 2018 6:16 am
My portfolio has held up fairly well with higher returns over the last 15, 10, 5, and 3 year periods. This is despite views that the small cap premium is dead and the value premium is following suit. If the small cap premium is dead, it was “deader” 20 years ago.
As a way of trying to replicate Robert's returns, I plugged the example DFA and RAFI portfolios from his collected thoughts thread into portfolio visualizer, along with a comparable Vanguard portfolio. You can see the results for yourself here.

Portfolio 1: Vanguard
Portfolio 2: RAFI
Portfolio 3: DFA

The five year returns are below. One takeaway is that if there are still factor premiums, they may be small enough and finicky enough that fund selection can make a significant difference. If you chose to hold a Robert T style portfolio using RAFI funds, you would have slightly underperformed the Vanguard portfolio over the last five years. If you chose to use DFA funds, you would have slightly outperformed before fees and taxes. That extra 60 basis points a year in returns will be zero (or less than zero) if you pay 50 basis points a year to your advisor and have any tax drag.

Image

Image
Last edited by phantom cosmonaut on Sat Jun 23, 2018 9:52 am, edited 1 time in total.

User avatar
matjen
Posts: 1925
Joined: Sun Nov 20, 2011 11:30 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by matjen » Sat Jun 23, 2018 9:52 am

phantom cosmonaut wrote:
Sat Jun 23, 2018 9:34 am

The five year returns are below. One takeaway is that if there are still factor premiums, they may be small enough and finicky enough that fund selection can make a significant difference.
Thanks for doing this but really the main takeaway that should never be forgotten is that five years is a rather meaningless short time period. Wes Gray's quote which Robert T. refers to is great.

I suppose I would make an exception if the markets performed in a manner where you would expect that a small cap value tilt should work but the past five years have been the opposite of that with the markets going consistently up and being led by large cap growth/tech type of story. I'm kind of surprised it is even close actually which is a credit to Robert T.'s portfolio it seems to me.
A man is rich in proportion to the number of things he can afford to let alone.

phantom cosmonaut
Posts: 15
Joined: Wed Jun 13, 2018 4:57 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by phantom cosmonaut » Sat Jun 23, 2018 10:02 am

matjen wrote:
Sat Jun 23, 2018 9:52 am
phantom cosmonaut wrote:
Sat Jun 23, 2018 9:34 am

The five year returns are below. One takeaway is that if there are still factor premiums, they may be small enough and finicky enough that fund selection can make a significant difference.
Thanks for doing this but really the main takeaway that should never be forgotten is that five years is a rather meaningless short time period. Wes Gray's quote which Robert T. refers to is great.

I suppose I would make an exception if the markets performed in a manner where you would expect that a small cap value tilt should work but the past five years have been the opposite of that with the markets going consistently up and being led by large cap growth/tech type of story. I'm kind of surprised it is even close actually which is a credit to Robert T.'s portfolio it seems to me.
I picked the last five years because it illustrates a period in which the DFA portfolio beats the Vanguard portfolio, but the equivalent RAFI portfolio trails the Vanguard portfolio.

If you look at the full sample for which all the relevant funds existed (about ten years), the tiled portfolios picked up 30-40 basis point in returns, but at the cost of extra volatility. That period is also somewhat biased towards tilting strategies, since it misses the beginning of great recession, when small and value stocks dropped more than the market as a whole.

User avatar
nisiprius
Advisory Board
Posts: 36475
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: Factor Investing Makes Someone Money, But Maybe Not You

Post by nisiprius » Sat Jun 23, 2018 10:54 am

columbia wrote:
Fri Jun 22, 2018 4:59 pm
I “get” the idea using the momentum factor, but have any real world funds been around long enough to demonstrate that it works (or doesn’t)?
Momentum is tricky, for at least two reasons.

First, it has generally shown negative correlation with value. Jared Kizer has written:
Practically, it is impossible, therefore, to own a stock portfolio that is both deeply tilted toward value and deeply tilted toward momentum.
This provokes many questions, the obvious one being "value or momentum, which is the good one?" Factor mavens often suggesting you should invest in as many different factors as possible to capture benefit of the low correlations between them, but negative correlation between factors is a problem. (Negative correlation with stocks, if it were robust and persistent and large, would be sensational because it would mean you could partially erase the risks of stocks, but negative correlation among two factors means that they are actually working against each other). In fact, Kizer says:
More generally, it is not feasible to build portfolios that capture significant amounts of multiple factor premia.
You can't have it all. You have to choose which factors to commit to or they will step on each others' toes and partly cancel each other out. Even in theory, just adding one more factor isn't necessarily or automatically going to be an improvement.

Second, momentum is sort of like intentionally investing in a mini-bubble, a "buy high, sell higher" strategy. The premise behind momentum is that because other investors believe in momentum, when a stock has been rising everyone believes that everyone else will expect it to show momentum and keep rising, therefore it will keep rising. But there can be sudden reversal and a momentum-accelerated decline. This phenomenon is called a "momentum crash." The "answer" to this problem is sometimes said to be an ability to predict and dodge momentum crashes. Momentum crashes are severe and infrequent, so the present the problem of being low-probability, high-consequences risk; it might not show up for long periods of time, leaving you to believe that a fund or ETF or strategy has "proven" itself.
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.

Ben Mathew
Posts: 96
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by Ben Mathew » Sat Jun 23, 2018 12:04 pm

Robert T wrote:
Fri Jun 22, 2018 5:07 am
From this Fama-French 1993 article : Risk factors in equities = exposure to market, value, and size; in bonds = exposure to term and default risk.

“Factor investing” is often discussed as if exposure to the market is not a risk factor, while it is the largest of those in the Fama-French article. Everyone who holds an equity market portfolio does “factor investing”, having exposure to a single risk factor.
The "market exposure" factor seems to be in a different category than value, size, momentum etc. Everyone can hold the market portfolio. But for one person to tilt towards value, size, etc., someone else somewhere has to tilt away from it. I am not against tilts. I tilt towards small and value myself. But it gives me a bit of pause to think that someone else out there has tilted away from small and value to allow me to do so. Is that person a fool? Does he or she know something I don't?

I think a lot of how we approach factor investing depends on whether we think of them as justly compensated risk factors or as underpriced assets than provide excess risk-adjusted returns (alpha). If they are risk factors, then we can be a lot more confident in them. They won't go away just because everybody knows about them--any more than the excess returns from stocks over bonds will go away because everybody knows about it. People know that stocks offer a premium, but still refuse to hold them because they don't like the risk. I have a hard time viewing small, value, momentum, etc. as genuine sources of risk. I view them as potentially underpriced assets that might provide excess returns. So I view my tilt towards small and value as a bet against others who are tilted away from it. So I have less confidence in the excess returns offered by my small and value factor tilts than in the excess returns offered by my "overall market exposure" factor.

User avatar
Taylor Larimore
Advisory Board
Posts: 27320
Joined: Tue Feb 27, 2007 8:09 pm
Location: Miami FL

Correlations

Post by Taylor Larimore » Sat Jun 23, 2018 12:09 pm

nisiprius wrote:Factor mavens often suggesting you should invest in as many different factors as possible to capture benefit of the low correlations between them, but negative correlation between factors is a problem.
nisiprius:

Nisiprius, as you know, expecting factor funds to provide low-correlation with Total Stock Market for safety is nearly always a mistake. This is because nearly all stock funds have similar correlations. Bond funds are, by far, best for reducing stock fund risks as correlations with Vanguard Total Stock Market demonstrate:

Listed below are the correlations of several Vanguard funds with Vanguard Total Stock Market Index Fund:

+97 Value Index
+94 Growth Index
+94 Total World
+93 U.S. Value
+91 Small-Cap Index
+78 Total International
-22 Total Bond Market

* Source: 2018 Independent Guide to the Vanguard Funds

Knowledgeable investors use stocks for higher returns and use good quality, low-cost, short- and intermediate term bond funds for safety.

Best wishes.
Taylor
"Simplicity is the master key to financial success." -- Jack Bogle

User avatar
vineviz
Posts: 1274
Joined: Tue May 15, 2018 1:55 pm

Re: Factor Investing Makes Someone Money, But Maybe Not You

Post by vineviz » Sat Jun 23, 2018 12:14 pm

nisiprius wrote:
Sat Jun 23, 2018 10:54 am
Momentum is tricky, for at least two reasons.

First, it has generally shown negative correlation with value. Jared Kizer has written:
Practically, it is impossible, therefore, to own a stock portfolio that is both deeply tilted toward value and deeply tilted toward momentum.
It's really not even hard, much less impossible.
    • 25.00% SPDR SSGA US Small Cap Low Volatil ETF (SMLV)
    • 25.00% SPDR S&P 600 Small Cap Growth ETF (SLYG)
    • 50.00% SPDR Dow Jones Industrial Average ETF (DIA)
    This simple portfolio gives you statistically significant exposure to size, value, and momentum without holding anything esoteric.

    https://www.portfoliovisualizer.com/fac ... tion3_1=50
    "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

    User avatar
    vineviz
    Posts: 1274
    Joined: Tue May 15, 2018 1:55 pm

    Re: Correlations

    Post by vineviz » Sat Jun 23, 2018 12:21 pm

    Taylor Larimore wrote:
    Sat Jun 23, 2018 12:09 pm
    Knowledgeable investors use stocks for higher returns and use good quality, low-cost, short- and intermediate term bond funds for safety.
    That certainly is one way to look at it.

    Some of us also like the fact that we can manage the tradeoff between safety and returns within our stock allocations and bond allocations independently. Happily, making wise choices about which stock funds to hold doesn't preclude an investor from also making wise choices about which bond funds to hold.
    "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

    jalbert
    Posts: 3585
    Joined: Fri Apr 10, 2015 12:29 am

    Re: Factor Investing Makes Someone Money, But Maybe Not You

    Post by jalbert » Sat Jun 23, 2018 12:25 pm

    vineviz wrote:
    Sat Jun 23, 2018 12:14 pm
    nisiprius wrote:
    Sat Jun 23, 2018 10:54 am
    Momentum is tricky, for at least two reasons.

    First, it has generally shown negative correlation with value. Jared Kizer has written:
    Practically, it is impossible, therefore, to own a stock portfolio that is both deeply tilted toward value and deeply tilted toward momentum.
    It's really not even hard, much less impossible.
      • 25.00% SPDR SSGA US Small Cap Low Volatil ETF (SMLV)
      • 25.00% SPDR S&P 600 Small Cap Growth ETF (SLYG)
      • 50.00% SPDR Dow Jones Industrial Average ETF (DIA)
      This simple portfolio gives you statistically significant exposure to size, value, and momentum without holding anything esoteric.

      https://www.portfoliovisualizer.com/fac ... tion3_1=50
      Just because PV computes stats that look good, does not make a regression from a 3-year sample of contiguous (neither random nor independent) data points statistically significant. In fact, the opposite is generally true.
      Index fund investor since 1987.

      garlandwhizzer
      Posts: 1939
      Joined: Fri Aug 06, 2010 3:42 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by garlandwhizzer » Sat Jun 23, 2018 12:44 pm

      Robert T wrote:
      My portfolio has held up fairly well with higher returns over the last 15, 10, 5, and 3 year periods. This is despite views that the small cap premium is dead and the value premium is following suit.
      I personally believe that Robert T is a highly accomplished factor investor who after extensive research carefully selects there right factor funds in the right proportions and at a very low cost. This IMO is likely to be the secret sauce that gives him excellent risk adjusted returns. He has done quite well, no argument there. The more important question for the rest of us is: Will the average person reading on this Forum who chooses a factor approach do similarly well? I suspect not. To determine whether a strategy is reliable for the average investor one doesn't pick an outlier at the thin and narrow end of the bell shaped curve but rather the median, right in the fat middle of the curve. I don't doubt that there were plenty of factor investor losers during this time frame who produced lower returns at higher risk that a simple Vanguard portfolio. I suspect that the average factor investor's results would be much less impressive than Robert T's. If a strategy is sound and reliable, its tide should raise all boats, not just the very adept out-performers. Any strategy including active management produces some out-performers. I seriously doubt that over the last 3,5,10,15 years that these wonderful results would have been the case for the average factor investor or that the average factor investor should expect returns like Robert T's.

      Garland Whizzer

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sat Jun 23, 2018 12:47 pm

      If I recall correctly, Wes Gray said something along the lines of the following – there is not enough time left in his life to produce enough out of sample data to disprove the existence of the value (and momentum) premiums
      In statistical analysis, the absence of a clear invalidation of a hypothesis does not constitute establishing the hypothesis. Moreover, if it takes, say 30 years to establish if the effect is observed, then a statistically significant out of sample test would involve a random sample of N such independent 30-year periods for a sufficiently large N.
      Index fund investor since 1987.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sat Jun 23, 2018 1:25 pm

      jalbert wrote:
      Sat Jun 23, 2018 12:25 pm
      Just because PV computes stats that look good, does not make a regression from a 3-year sample of contiguous (neither random nor independent) data points statistically significant. In fact, the opposite is generally true.
      You've got access to the same data I've got: if you can find an actual problem with these regressions then feel free to point it out.

      Thankfully, it only takes a single counter-factual example to falsify a claim like the one that Jared Kizer allegedly made (i.e. that constructing a portfolio which tilts to both value and momentum is impossible). If his statement either had or needs to have qualifiers then I'm open to hearing what those qualifiers might be.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      Fallible
      Posts: 6459
      Joined: Fri Nov 27, 2009 4:44 pm
      Contact:

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Fallible » Sat Jun 23, 2018 2:19 pm

      garlandwhizzer wrote:
      Sat Jun 23, 2018 12:44 pm
      Robert T wrote:
      My portfolio has held up fairly well with higher returns over the last 15, 10, 5, and 3 year periods. This is despite views that the small cap premium is dead and the value premium is following suit.
      I personally believe that Robert T is a highly accomplished factor investor who after extensive research carefully selects there right factor funds in the right proportions and at a very low cost. This IMO is likely to be the secret sauce that gives him excellent risk adjusted returns. He has done quite well, no argument there. The more important question for the rest of us is: Will the average person reading on this Forum who chooses a factor approach do similarly well? I suspect not. To determine whether a strategy is reliable for the average investor one doesn't pick an outlier at the thin and narrow end of the bell shaped curve but rather the median, right in the fat middle of the curve. I don't doubt that there were plenty of factor investor losers during this time frame who produced lower returns at higher risk that a simple Vanguard portfolio. I suspect that the average factor investor's results would be much less impressive than Robert T's. If a strategy is sound and reliable, its tide should raise all boats, not just the very adept out-performers. Any strategy including active management produces some out-performers. I seriously doubt that over the last 3,5,10,15 years that these wonderful results would have been the case for the average factor investor or that the average factor investor should expect returns like Robert T's.
      Garland Whizzer
      And I think these well-expressed concerns about the average investor's chances of succeeding with factor (and other more complex investing) bring us back to Rick's main point, as expressed in the opening graph of his original post. For those average investors not suited for, but still tempted by, factor investing, I think it bears repeating:
      Factor investing is discussed ad nauseam on this forum. Lately, it has reached the point where investors who shouldn’t do factor investing become confused to the point of believing they should or they’ll miss out on something. Let’s correct that.
      Bogleheads® wiki | Investing Advice Inspired by Jack Bogle

      Random Walker
      Posts: 2989
      Joined: Fri Feb 23, 2007 8:21 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Random Walker » Sat Jun 23, 2018 2:33 pm

      matjen wrote:
      Sat Jun 23, 2018 9:52 am

      I suppose I would make an exception if the markets performed in a manner where you would expect that a small cap value tilt should work but the past five years have been the opposite of that with the markets going consistently up and being led by large cap growth/tech type of story. I'm kind of surprised it is even close actually which is a credit to Robert T.'s portfolio it seems to me.
      I believe this is one of the advantages of factor investing. A portfolio diversified across factors, when it underperforms the market, underperforms by a modest amount. When it outperforms, it is much more significant.

      Dave

      Random Walker
      Posts: 2989
      Joined: Fri Feb 23, 2007 8:21 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Random Walker » Sat Jun 23, 2018 2:58 pm

      nisiprius wrote:
      Sat Jun 23, 2018 10:54 am
      Momentum is tricky, for at least two reasons.

      First, it has generally shown negative correlation with value. Jared Kizer has written:
      Practically, it is impossible, therefore, to own a stock portfolio that is both deeply tilted toward value and deeply tilted toward momentum.
      This provokes many questions, the obvious one being "value or momentum, which is the good one?" Factor mavens often suggesting you should invest in as many different factors as possible to capture benefit of the low correlations between them, but negative correlation between factors is a problem. (Negative correlation with stocks, if it were robust and persistent and large, would be sensational because it would mean you could partially erase the risks of stocks, but negative correlation among two factors means that they are actually working against each other). In fact, Kizer says:
      More generally, it is not feasible to build portfolios that capture significant amounts of multiple factor premia.
      You can't have it all. You have to choose which factors to commit to or they will step on each others' toes and partly cancel each other out. Even in theory, just adding one more factor isn't necessarily or automatically going to be an improvement.
      [/quote]

      Yikes! :-) I’m a client of Kizer’s firm, and I think I’ve got pretty good exposure to value and momentum. I certainly appreciate that value and momentum are naturally negatively correlated. I think most value funds on their own definitionally would have a negative momentum loading. By adding a momentum screen to a value fund, I think the momentum loading at least goes from negative to about zero.
      I believe I’ve read that it is most efficient to combine factors in a single fund via screens. It would be costly and inefficient to have one fund in a portfolio buy a given stock for one factor and another fund sell the same stick for the sake of another factor. My understanding is that if one uses multiple screens in the same multifactor fund, the fund sort of ends up with the stocks that sort of score a “B grade” in a couple of factor categories rather than an “A grade” in any single factor category. Overall this is superior fund and portfolio construction.
      Also important to remember that value and momentum are not limited to equities. These factors are present in other asset classes. Not only are the different factors within an asset class generally uncorrelated, also the same factor in different asset classes tends to be uncorrelated. For example value in equities will tend to be uncorrelated with value in bonds, currencies, or commodities. One can access multiple factors in multiple asset classes in a single fund with virtually no net equity market exposure.

      Dave

      Random Walker
      Posts: 2989
      Joined: Fri Feb 23, 2007 8:21 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Random Walker » Sat Jun 23, 2018 3:09 pm

      nisiprius wrote:
      Sat Jun 23, 2018 10:54 am

      Second, momentum is sort of like intentionally investing in a mini-bubble, a "buy high, sell higher" strategy. The premise behind momentum is that because other investors believe in momentum, when a stock has been rising everyone believes that everyone else will expect it to show momentum and keep rising, therefore it will keep rising. But there can be sudden reversal and a momentum-accelerated decline. This phenomenon is called a "momentum crash." The "answer" to this problem is sometimes said to be an ability to predict and dodge momentum crashes. Momentum crashes are severe and infrequent, so the present the problem of being low-probability, high-consequences risk; it might not show up for long periods of time, leaving you to believe that a fund or ETF or strategy has "proven" itself.
      Not sure about cross sectional momentum, but one of the strong arguments for time series momentum (absolute momentum, trend following) is for that strategy’s behavior in an extended equity bear market.

      Dave

      Random Walker
      Posts: 2989
      Joined: Fri Feb 23, 2007 8:21 pm

      Re: Correlations

      Post by Random Walker » Sat Jun 23, 2018 3:17 pm

      Taylor Larimore wrote:
      Sat Jun 23, 2018 12:09 pm
      nisiprius wrote:Factor mavens often suggesting you should invest in as many different factors as possible to capture benefit of the low correlations between them, but negative correlation between factors is a problem.
      nisiprius:

      Nisiprius, as you know, expecting factor funds to provide low-correlation with Total Stock Market for safety is nearly always a mistake. This is because nearly all stock funds have similar correlations. Bond funds are, by far, best for reducing stock fund risks as correlations with Vanguard Total Stock Market demonstrate:

      Listed below are the correlations of several Vanguard funds with Vanguard Total Stock Market Index Fund:

      +97 Value Index
      +94 Growth Index
      +94 Total World
      +93 U.S. Value
      +91 Small-Cap Index
      +78 Total International
      -22 Total Bond Market

      * Source: 2018 Independent Guide to the Vanguard Funds

      Knowledgeable investors use stocks for higher returns and use good quality, low-cost, short- and intermediate term bond funds for safety.

      Best wishes.
      Taylor
      And it’s not only the correlations of Bonds to stocks that are important, it’s also when the correlations tend to change. Overall stocks and bonds tend to be uncorrelated, but the correlation tends to go strongly negative just when you want it to the most, in a big equity bear.

      Dave

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sat Jun 23, 2018 5:01 pm

      You've got access to the same data I've got: if you can find an actual problem with these regressions then feel free to point it out.
      Sure. One example is the F-statistic which is reported in your regression but is not valid and conveys no meaning whatsoever for the sample you used (3 years of contiguous non-random data points that are not independent). In this case, sample variance is too biased in such a sample to know that it has sufficiently converged to actual variance for the F-statistic to be a reliable indicator of statistical significance of measured sample variance. It does not mean the F-statistic is inaccurate, but it also doesn’t mean it is accurate. It means it conveys no meaning at all for this sample and should not be interpreted.

      In my view, portfoliovisualizer does a disservice by reporting statistics when the sample clearly does not support the use or interpretation of the statistic. There is a human bias wherein once complex data is aggregated and presented in an easy to analyze and interpret format, we focus on the analysis of what we see, and lose sight of whether the aggregated information is accurate or supportive of our intended usage.

      If you are asking me to find a counterexample to the data fit, I’m sure you are aware that in hypothesis testing, the absence of an identified or observed counterexample does not make the hypothesis correct or even statistically significant.

      Moreover, you can only talk about a deterministic counterexample if you have a deterministic model, but factor returns are random variables not deterministic predictors of future returns.
      Index fund investor since 1987.

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sat Jun 23, 2018 5:10 pm

      And it’s not only the correlations of Bonds to stocks that are important, it’s also when the correlations tend to change. Overall stocks and bonds tend to be uncorrelated, but the correlation tends to go strongly negative just when you want it to the most, in a big equity bear.
      That behavior depends on the type of bonds, and also considerations like interest rates. If a bear market coincides with falling treasury rates, sample correlation between equities and bonds is likely to be negative. When I first started looking at modern portfolio theory, historical samples in use were more biased by the inflation of the 1970’s and investment authors of the time often indicated that the correlation of equities and treasuries is generally around 0.6.
      Index fund investor since 1987.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sat Jun 23, 2018 5:15 pm

      jalbert wrote:
      Sat Jun 23, 2018 5:01 pm
      You've got access to the same data I've got: if you can find an actual problem with these regressions then feel free to point it out.
      Sure. One example is the F-statistic which is reported in your regression but is not valid and conveys no meaning whatsoever for the sample you used (3 years of contiguous non-random data points that are not independent).
      Maybe I wasn’t clear: by “actual problem” I mean “provable “ as opposed to hypothetical.

      If the F-stat is biased because ??? then show us the corrected stat. Or show us that the returns are insufficiently i.i.d or whatever.

      Right now you’re just waving your hands, and that’s not how we do science.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sat Jun 23, 2018 7:41 pm

      If the F-stat is biased because ??? then show us the corrected stat. Or show us that the returns are insufficiently i.i.d or whatever.

      Right now you’re just waving your hands, and that’s not how we do science.
      There is no corrected stat. The F-stat is uninterpretable because your data sample does not meet the prerequisite conditions for the mathematical theorem on which the F-stat is based.
      Index fund investor since 1987.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sat Jun 23, 2018 8:26 pm

      jalbert wrote:
      Sat Jun 23, 2018 7:41 pm
      If the F-stat is biased because ??? then show us the corrected stat. Or show us that the returns are insufficiently i.i.d or whatever.

      Right now you’re just waving your hands, and that’s not how we do science.
      There is no corrected stat. The F-stat is uninterpretable because your data sample does not meet the prerequisite conditions for the mathematical theorem on which the F-stat is based.
      And you’re preparing to show us the proof of that statement, I presume.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sat Jun 23, 2018 10:09 pm

      Well, an introductory text on mathematical statistics might have one or more chapters on the estimation of parameters of probability distributions, such as mean or variance, from sample statistics. If you repeatedly sample a distribution, and compute sample statistics such as sample means or sample variances for each sample, the computed values will vary from sample to sample. The sample statistic such as the sample mean or sample variance then becomes a random variable with its own probability distribution referred to as the sampling distribution.

      An event outcome of the sample mean (or variance) random variable is generated by generating a sample and computing the mean (or variance) in the sample.

      There are central limit theorems that say that if you repeatedly sample a random variable and generate sufficiently large random samples of independent trials, then as the number of samples increases without bounds, the sampling distribution of the sample mean tends toward some limiting distribution, a normal distribution for sample means, and Wishart distribution for sample covariances as examples. There also are some requirements about distribution of the original random variable being sampled to avoid anomalous distributions where the integrals defining some moments of the distribution may not always converge.

      The theory of the t-test, F-test, and chi-squared test depend on the sampling distribution to be well approximated by the limit distribution in the limit theorem, which means that to interpret the statistic of the test, you have to meet the conditions of the relevant limit theorem.

      I can’t see any reason to believe that a sample of 3 consecutive years of return data would be a sufficiently large random sample of independent trials for appropriate limit theorems to apply. They certainly are not random or independent data points.

      In fact independence would even be contradictory with the existence of time series momentum, so that would create new problems to overcome if it were true. To test time series momentum you have to decide on the length of the time series and then collect a random sample of time periods of that length.
      Index fund investor since 1987.

      confusedinvestor
      Posts: 417
      Joined: Thu Mar 15, 2012 11:04 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by confusedinvestor » Sun Jun 24, 2018 12:10 am

      I hope those Value tilters / slicers and dicers will keep the couse for next decade of underperormance in hope to over performance over market cap beyond their lifetime. I'm one of those victims of small cap premium size factor dreamer ....
      David Jay wrote:
      Thu Jun 21, 2018 8:40 am
      What is important about any "tilt" is that you believe in it enough to convince yourself to stay the course though a decade or more of underperformance.[/color]

      User avatar
      nedsaid
      Posts: 10220
      Joined: Fri Nov 23, 2012 12:33 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by nedsaid » Sun Jun 24, 2018 12:18 am

      Rick Ferri wrote:
      Thu Jun 21, 2018 12:46 pm
      All the chatter about tilting to factors today reminds me of the excitement of tilting to industry sector funds when they flourished in the 1990s.

      Rick Ferri
      Well, actually comparing factors to sector investing isn't so far fetched. Financial stocks are a large component of what shows up on Value screens and such sectors as Utilities and Consumer Staples are associated with Low Volatility. Maybe factors are a refinement of the sector approach.
      A fool and his money are good for business.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sun Jun 24, 2018 3:20 am

      jalbert wrote:
      Sat Jun 23, 2018 10:09 pm
      Well, an introductory text on mathematical statistics might have one or more chapters on the estimation of parameters of probability distributions, such as mean or variance, from sample statistics.
      They might indeed. All the ones I’ve read have.

      Maybe you’re basing your skepticism on intuition or on something you think someone said in some other context, I don’t know. But if it was as easy to demonstrate the invalidity of these regressions as you keep implying, you’d show us the tests proving it. Right?
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      Call_Me_Op
      Posts: 6997
      Joined: Mon Sep 07, 2009 2:57 pm
      Location: Milky Way

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Call_Me_Op » Sun Jun 24, 2018 6:03 am

      Rick Ferri wrote:
      Thu Jun 21, 2018 8:14 am
      Factor investing is complex and the “optimal” way to do it is always changing because there is no optimal way that can be known in advance. The idea that factor premiums can be predicted is a fallacy. The small cap premium was discovered in the 1970s... and has been dead ever since.
      Stock market returns (CAGR) - January 1979 - May 2018:

      Total US Stock Market ........ 11.64%

      US Small Cap .................. 12.79%

      US Small Cap Value ........... 14.75%

      Source: Portfolio Visualizer
      Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein

      rkhusky
      Posts: 5409
      Joined: Thu Aug 18, 2011 8:09 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by rkhusky » Sun Jun 24, 2018 6:38 am

      Random Walker wrote:
      Sat Jun 23, 2018 2:33 pm
      I believe this is one of the advantages of factor investing. A portfolio diversified across factors, when it underperforms the market, underperforms by a modest amount. When it outperforms, it is much more significant.
      There is no basis for that in the FF formulation. Perhaps it has occasionally occurred in the past, but there is no guarantee that it will in the future.

      Increased number of factors means smaller number of stocks to invest in.

      If one factor underperforms the market and another factor equally outperforms the market, and they have equal loadings, their contributions to the FF return will cancel.

      Increasing the number of factors is not the same as increasing the types of investments (i.e. stocks and bonds or US and Int'l stocks). The former decreases the number of investments, the latter increases the number of investments.

      Choosing stocks via factors leads to a group of stocks that all have the same characteristics, i.e. the factors with which one is filtering. Increasing investment types leads to a group of investments with varied characteristics.

      User avatar
      jeffyscott
      Posts: 7176
      Joined: Tue Feb 27, 2007 9:12 am
      Location: Wisconsin

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jeffyscott » Sun Jun 24, 2018 6:48 am

      confusedinvestor wrote:
      Sun Jun 24, 2018 12:10 am
      I hope those Value tilters / slicers and dicers will keep the couse for next decade of underperormance in hope to over performance over market cap beyond their lifetime. I'm one of those victims of small cap premium size factor dreamer ....
      David Jay wrote:
      Thu Jun 21, 2018 8:40 am
      What is important about any "tilt" is that you believe in it enough to convince yourself to stay the course though a decade or more of underperformance.[/color]
      At least in the case of Vanguard funds small cap and small cap value are ahead of Total Stock Market and the S&P 500 over the past decade. And even the average small cap value fund is not far behind.

      I think sticking with an international allocation (and rebalancing into it) is where the real test has been, with 10 year returns under 3% and only about doubling from the market bottom in 2008, while US has about quadrupled.

      One of these things is not like the others....
      Image
      Last edited by jeffyscott on Sun Jun 24, 2018 6:55 am, edited 2 times in total.
      press on, regardless - John C. Bogle

      livesoft
      Posts: 62305
      Joined: Thu Mar 01, 2007 8:00 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by livesoft » Sun Jun 24, 2018 6:51 am

      phantom cosmonaut wrote:
      Sat Jun 23, 2018 9:34 am
      As a way of trying to replicate Robert's returns, I plugged the example DFA and RAFI portfolios from his collected thoughts thread into portfolio visualizer, along with a comparable Vanguard portfolio. You can see the results for yourself
      [...]
      One needs to be aware that Portfolio Visualizer only uses month-end data to do backtesting, so at least right now, it would miss in a big way every intra-month major change in values when one would (should?) be doing rebalancing and buying low.

      When I look at the days that I ended up rebalancing over the past few years which has been on RBDs, those have not occurred on the last day of the month.

      Thus, I would say that using Portfolio Visualizer for backtesting would be useless to do backtesting comparisons of my portfolio and probably everyone else's.

      But to Rick Ferri's point: I can agree that folks who tilt to small and value who just buy-and-hold probably do not make the expected outperformance touted. I think one has to actively manage the buys, and rebalancing of such a portfolio of passively-managed index funds.
      Last edited by livesoft on Sun Jun 24, 2018 6:55 am, edited 1 time in total.
      Wiki This signature message sponsored by sscritic: Learn to fish.

      User avatar
      randomizer
      Posts: 1447
      Joined: Sun Jul 06, 2014 3:46 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by randomizer » Sun Jun 24, 2018 6:53 am

      Totally agreed. Simplicity its best, probably for 99% of people, although I expect about 40% of people to disagree with me.
      87.5:12.5, EM tilt — HODL the course!

      Deighve
      Posts: 21
      Joined: Sat Nov 18, 2017 10:30 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by Deighve » Sun Jun 24, 2018 7:17 am

      Rick,

      I recently read your excellent book All about Asset Allocation

      In the multi asset class portfolio you suggest 10% small value ad 5% microcap. I liked this portfolio and was thinking of moving in that direction. Has your thinking on this changed at all?

      Thanks very much,

      Deighve

      phantom cosmonaut
      Posts: 15
      Joined: Wed Jun 13, 2018 4:57 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by phantom cosmonaut » Sun Jun 24, 2018 8:04 am

      vineviz wrote:
      Sun Jun 24, 2018 3:20 am
      jalbert wrote:
      Sat Jun 23, 2018 10:09 pm
      Well, an introductory text on mathematical statistics might have one or more chapters on the estimation of parameters of probability distributions, such as mean or variance, from sample statistics.
      They might indeed. All the ones I’ve read have.

      Maybe you’re basing your skepticism on intuition or on something you think someone said in some other context, I don’t know. But if it was as easy to demonstrate the invalidity of these regressions as you keep implying, you’d show us the tests proving it. Right?
      (1) The original point vineviz was making is that you can tilt towards multiple factors without doing anything too complicated. Given the proliferation of cheap factor factor ETFs, this strikes me as obvious.

      (2) On the other hand, it is pretty hard to strongly tilt towards multiple factors without leveraging or shorting. Just try building a portfolio with a .5 tilt towards momentum, value, and size with long-only portfolios.

      (3) Everyone should agree that stock returns are not normally distributed. Whether various statistical tests are meaningful then depends on the degree of non-normality as compared to the size of the sample. Jalbert's position is that we have to show that these conditions are probably met before we get to use the statistical tests. Vineviz thinks that we get to use the tests unless we can show that the conditions are probably not met. Myself, I would think that be caution would be warranted when applying fancy statistical tests.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sun Jun 24, 2018 8:21 am

      phantom cosmonaut wrote:
      Sun Jun 24, 2018 8:04 am

      (3) Everyone should agree that stock returns are not normally distributed. Whether various statistical tests are meaningful then depends on the degree of non-normality as compared to the size of the sample. Jalbert's position is that we have to show that these conditions are probably met before we get to use the statistical tests. Vineviz thinks that we get to use the tests unless we can show that the conditions are probably not met.
      I agree with all of this except that I DO happen to think that there are a few basic tests for validity that should be checked when doing a regression before accepting the results. In this case, I checked them before reporting the results and they were fine.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      User avatar
      FIREchief
      Posts: 2630
      Joined: Fri Aug 19, 2016 6:40 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by FIREchief » Sun Jun 24, 2018 8:32 am

      jalbert wrote:
      Sat Jun 23, 2018 10:09 pm
      Well, an introductory text on mathematical statistics might have one or more chapters on the estimation of parameters of probability distributions, such as mean or variance, from sample statistics. If you repeatedly sample a distribution, and compute sample statistics such as sample means or sample variances for each sample, the computed values will vary from sample to sample. The sample statistic such as the sample mean or sample variance then becomes a random variable with its own probability distribution referred to as the sampling distribution.

      An event outcome of the sample mean (or variance) random variable is generated by generating a sample and computing the mean (or variance) in the sample.

      There are central limit theorems that say that if you repeatedly sample a random variable and generate sufficiently large random samples of independent trials, then as the number of samples increases without bounds, the sampling distribution of the sample mean tends toward some limiting distribution, a normal distribution for sample means, and Wishart distribution for sample covariances as examples. There also are some requirements about distribution of the original random variable being sampled to avoid anomalous distributions where the integrals defining some moments of the distribution may not always converge.

      The theory of the t-test, F-test, and chi-squared test depend on the sampling distribution to be well approximated by the limit distribution in the limit theorem, which means that to interpret the statistic of the test, you have to meet the conditions of the relevant limit theorem.

      I can’t see any reason to believe that a sample of 3 consecutive years of return data would be a sufficiently large random sample of independent trials for appropriate limit theorems to apply. They certainly are not random or independent data points.

      In fact independence would even be contradictory with the existence of time series momentum, so that would create new problems to overcome if it were true. To test time series momentum you have to decide on the length of the time series and then collect a random sample of time periods of that length.
      Thank you!!! I think you've just written six paragraphs that invalidate a whole bunch of guidance provided by "academic research by Nobel prize winning authors." Best post of the thread since Rick's OP. :sharebeer
      I am not a lawyer, accountant or financial advisor. Any advice or suggestions that I may provide shall be considered for entertainment purposes only.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sun Jun 24, 2018 8:40 am

      FIREchief wrote:
      Sun Jun 24, 2018 8:32 am
      Thank you!!! I think you've just written six paragraphs that invalidate a whole bunch of guidance provided by "academic research by Nobel prize winning authors." Best post of the thread since Rick's OP. :sharebeer
      Because obviously all those academics, with their Nobel Prizes in economics and PhDs in mathematics, never learned the properties of a normal distribution.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      User avatar
      nisiprius
      Advisory Board
      Posts: 36475
      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: Factor Investing Makes Someone Money, But Maybe Not You

      Post by nisiprius » Sun Jun 24, 2018 9:04 am

      FIREchief wrote:
      Sun Jun 24, 2018 8:32 am
      jalbert wrote:
      Sat Jun 23, 2018 10:09 pm
      Well, an introductory text on mathematical statistics might have one or more chapters on the estimation of parameters of probability distributions, such as mean or variance, from sample statistics. If you repeatedly sample a distribution, and compute sample statistics such as sample means or sample variances for each sample, the computed values will vary from sample to sample. The sample statistic such as the sample mean or sample variance then becomes a random variable with its own probability distribution referred to as the sampling distribution....

      The theory of the t-test, F-test, and chi-squared test depend on the sampling distribution to be well approximated by the limit distribution in the limit theorem, which means that to interpret the statistic of the test, you have to meet the conditions of the relevant limit theorem.

      I can’t see any reason to believe that a sample of 3 consecutive years of return data would be a sufficiently large random sample of independent trials for appropriate limit theorems to apply. They certainly are not random or independent data points.
      Thank you!!! I think you've just written six paragraphs that invalidate a whole bunch of guidance provided by "academic research by Nobel prize winning authors." Best post of the thread since Rick's OP. :sharebeer
      Actually, I think this is worth pursuing a bit, because there is indeed something weird in the Fama-French paper, 1992, The Cross-Section of Expected Stock Returns.

      In Table III, page 439, they do indeed present a table of values together with "t statistics."

      Image

      As jalbert notes, the use of t statistics to show statistical significance is only valid if you have some good reason to believe that you are looking at a) independent values of b) normally distributed data. There are other statistical techniques, well-known in 1992 but for some reason (yes, I'm cynical) rarely used by financial economists, that do not depend on the assumption of normality. That might not be a big deal compared to a far more serious problem: the assumption of independence. And there is potentially a third problem, because that table does not show a single t-test, it shows twenty-six of them, all being computed simultaneously.

      Now, let's be clear. As nearly as I can tell, Fama and French are not trying to prove that the t-statistics for "Size, Book-to-Market Equity, Leverage, and E/P" are significant. They seem to be mostly concerned with proving that β has no explanatory power. They do not mention a significance level anywhere. In fact, it is not clear to me why they are even presenting them. They do mention a few of them a few times within the paper, but never with any significance level assigned to them, just as qualitative comparisons. For example, p. 441
      The average slope from the monthly regressions of returns on ln(BE/ME) alone is 0.50%, with a t-statistic of 5.71. This book-to-market relation is stronger than the size effect, which produces a t-statistic of - 2.58 in the regressions of returns on ln(ME) alone.
      You can't argue with that, they are just presenting values of a number and saying that one indicates a "stronger" relationship to another.

      But I have to ask: why are they presenting all of those t-statistics most of which are never used in the paper?

      Yeah, I think this is a little weird. And I don't think it's unreasonable to raise questions about it.

      And in general I think the endless citation of "t-statistics," sometimes by the nickname "t-stats," in financial economics, without any explanation or defense as to why this would be a valid number to test and why they think the assumptions behind the test are close to being met, is also weird.

      Vineviz and others: under what specific situations in financial economics do you think that the use of a T-test is appropriate, and under what situations is it inappropriate?
      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.

      User avatar
      vineviz
      Posts: 1274
      Joined: Tue May 15, 2018 1:55 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by vineviz » Sun Jun 24, 2018 9:37 am

      nisiprius wrote:
      Sun Jun 24, 2018 9:04 am
      And there is potentially a third problem, because that table does not show a single t-test, it shows twenty-six of them, all being computed simultaneously.
      I think you are misreading the table, which is presenting different regressions on each line. The largest number of simultaneously computed t-stats is in the last line, which has five.

      And there is no problem, inherently, in reporting the t-stat for each of your coefficients. In fact it’d be weird not to. If your model has five variables, you have five computed t-stats and report them all.

      There are obviously other tests you also run, to verify your t-stats are reliable and that your overall model has a good fit based on degrees of freedom, etc.
      "Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

      phantom cosmonaut
      Posts: 15
      Joined: Wed Jun 13, 2018 4:57 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by phantom cosmonaut » Sun Jun 24, 2018 9:42 am

      nisiprius wrote:
      Sun Jun 24, 2018 9:04 am
      As jalbert notes, the use of t statistics to show statistical significance is only valid if you have some good reason to believe that you are looking at a) independent values of b) normally distributed data. There are other statistical techniques, well-known in 1992 but for some reason (yes, I'm cynical) rarely used by financial economists, that do not depend on the assumption of normality.
      This is something I know basically nothing about, but it is curious that finance is so committed to parametric statistics when (a) non-parametric statistics is a thing and (b) finance is a paradigmatic example of an area in which we know very little about the underlying distributions.

      The following nice explanation of non-parametric statistic is from here. I'm posting it here because I found it helpful.
      Aaron Brown wrote: Non-parametric statistics do not make assumptions about the underlying distribution of the data.

      For example, suppose you have a sample of 99 independent realizations of a random variable. The mean is 50, the standard deviation is 10, the smallest value is 23 and the largest value is 68. You want to estimate the probability that the 100th observations will be outside the range of values you’ve seen to date.

      If you assume you have a Normal distribution, the answer is to look at at the -2.7 and +1.8 tails of a t-distribution with 97 degrees of freedom. You’d say it was 4.16%.

      But you can also answer this question making no assumptions about the distribution at all. Since the draws are independent, the ex ante probability that the 100th draw would be either the highest or the lowest value is exactly 2%. This is a non-parametric answer.

      Which answer is correct? Well, the non-parametric is correct ignoring any information from the actual values you have observed so far. But in order to make use of that information, you need some kind of assumption about the data. If you guess right about data, you will have a more accurate estimate. If you guess wrong, you will have a less accurate estimate.

      I generally prefer non-parametric methods. In my experience, if non-parametric methods give you a reliable enough answer, you can trust it. If they don’t, making parametric assumptions may appear to give you the reliability you need, but you can’t trust it.

      The exception is when you really do know a lot about how the data are generated. But even then, if you are relying heavily on the parametric assumption, you’re really using theory, not data, to support your conclusions.

      User avatar
      nisiprius
      Advisory Board
      Posts: 36475
      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: Factor Investing Makes Someone Money, But Maybe Not You

      Post by nisiprius » Sun Jun 24, 2018 10:02 am

      In addition to nonparametric tests which don't assume a normal distribution, people working in other fields have developed:

      -"sequential tests," where you want to be able to test continuously to see if significance has been reached while data is being acquired, and stop when it has been--in order to get an earlier result, or avoid the expense of further testing;

      --"multiple hypothesis tests," where you know that you are going to perform more than one test, but have at least stated in advance how many tests you will be performing before you begin testing the data. A trivially easy way to do this is called "the Bonferroni correction" or adjustment. There are also specialized, more powerful multiple hypothesis tests.
      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.

      phantom cosmonaut
      Posts: 15
      Joined: Wed Jun 13, 2018 4:57 pm

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by phantom cosmonaut » Sun Jun 24, 2018 10:20 am

      nisiprius wrote:
      Sun Jun 24, 2018 10:02 am
      In addition to nonparametric tests which don't assume a normal distribution, people working in other fields have developed:

      -"sequential tests," where you want to be able to test continuously to see if significance has been reached while data is being acquired, and stop when it has been--in order to get an earlier result, or avoid the expense of further testing;

      --"multiple hypothesis tests," where you know that you are going to perform more than one test, but have at least stated in advance how many tests you will be performing before you begin testing the data. A trivially easy way to do this is called "the Bonferroni correction" or adjustment. There are also specialized, more powerful multiple hypothesis tests.
      Well, now I have two more interesting ideas to spend the summer learning about. :D

      I do wonder why such tests are not more common in finance? Perhaps the cynical explanation is that standards tests are (a) easier to manipulate and (b) easier to sell to investors who work in areas where those tests are applicable (e.g. medicine).

      jalbert
      Posts: 3585
      Joined: Fri Apr 10, 2015 12:29 am

      Re: Factor Investing Makes Someone Money, But Maybe Not You

      Post by jalbert » Sun Jun 24, 2018 1:19 pm

      Maybe you’re basing your skepticism on intuition or on something you think someone said in some other context, I don’t know. But if it was as easy to demonstrate the invalidity of these regressions as you keep implying, you’d show us the tests proving it. Right?
      You seem to be confusiing a probabilistic result and deterministic result. If a statistical analysis leads to a finding that is not statistically significant, that means the test conveys very little to no information on whether or not the result is correct. Asking me to find a counterexample for the result would be a reasonable response if I stated that the result was deterministically false. I said that the F-stat is not interpretable as a measure of statistical significance, so that there is no basis to claim statistical significance. That does not even mean your result is incorrect so I could be chasing my tail looking for a counterexample. But the result equally could be wrong.

      Another point is that if a researcher uses statistical analysis to test a hypothesis, it is incumbent on him or her to use methods that can support a high confidence in the result.

      If I use inadequate methods to claim the moon is made of green cheese, the result is not entitled to be accepted as correct unless and until someone can prove it is wrong.

      But I think most people consider results derived from a 3-year sample of equity returns not to be generalizable to future returns in any case.
      Last edited by jalbert on Sun Jun 24, 2018 1:32 pm, edited 2 times in total.
      Index fund investor since 1987.

      Post Reply