A 1993 paper in the journal of financial

economics with nearly 15,000 citations

today changed financial economics and

portfolio management forever. Eugene FMA

and Kenneth French found that a group of

three factors explained the vast

majority of differences in returns

across diversified portfolios. These

findings had and continue to have

sweeping implications for both the study

and practice of investment management.

You guys like to make fun of me in the

comments for talking about FMA and

French all the time, which I definitely

do. And this paper is one of the reasons

they always come up. It's research that

every investor should understand whether

they choose to apply its findings or

not. Personally, this paper is the

foundation of how I think about

investing and building portfolios. I'm

Ben Felix, chief investment officer at

PWL Capital, and I'm going to tell you

why common risk factors in the returns

on stocks and bonds changed everything

we thought we knew about finance and

investing.

I probably don't have to tell you that

this is going to be a nerdy video. Let's

be honest, that's probably why you're

here. I think it'll be worth it. I'm

going to cover this paper's methodology,

results, and its enduring impacts. I'll

also talk later in the video about how

this paper's findings apply to investors

today, so be sure to stick around for

that. The fundamental premise of this

paper is that multiple factors, a term

I'll get more into in a minute, affect

asset prices and expected returns. And

that these factors can help to explain

why different types of stocks and bonds

and therefore different investment

portfolios have different expected

returns. An expected return, as it

sounds, is the return you expect from

investing in a stock or bond. The study

of expected returns is often referred to

as asset pricing because those two

things are directly related. An asset's

price is based on its expected return

and its expected return can be inferred

from its price. Expected returns are not

guaranteed outcomes, but as we will see,

they do contain information. This makes

sense. You would pay more for a safer

asset than a riskier one, all else

equal. It's commonly known that stocks

are riskier than bonds and therefore

have higher expected returns. But what

if some types of stocks have

systematically higher expected returns

than others? In Fman French's original

framing, common undiversifiable risks

that a lot of investors are sensitive to

help to explain why some stocks have

higher returns than others. I do want to

note that while this paper looks at both

stock and bond factors, I will focus

mostly on stocks in this video. If a

stock is exposed to more of a certain

type of risk that a lot of investors are

sensitive to, that stock needs to have a

higher expected return to entice

investors to buy it. All else equal. The

idea that investors might care about

multiple types of risk was not new when

this paper came out. Robert Mertton and

others had written about this idea

before. But figuring out what those

risks might be and how to measure them

was new. To understand why this

perspective was so impactful, it's

important to understand what came before

it. In 1964 and 1965, researchers

developed the capital asset pricing

model or CAPM. This model connected a

stock's expected return to how its price

moves relative to the overall market

expressed as its market beta or commonly

just beta. A beta of one means that a

stock tends to move up and down by a

similar amount to the market. A higher

beta more than one means that the stock

will tend to be up more when the market

is up and down more when the market is

down and vice versa for a beta less than

one. According to this single factor

model where exposure to market risk is

the only risk that investors care about,

stocks with higher betas should on

average deliver higher returns. The CAPM

was a big deal in finance. I mean big

enough to win Bill Sharp the Nobel

Memorial Prize in economic sciences in

1990 for his work on it because it

formalized a relationship between risk

and expected return and it did a pretty

good job of explaining the observed

returns of stocks. Even though the CAPAM

defined and then dominated the study of

asset pricing from its inception through

the 70s and 80s, research had

consistently come out showing that

certain types of stocks had higher

returns than what could be explained by

their CAPM betas. Since these

observations could not be explained by

an asset pricing model, by the asset

pricing model at the time, they were

referred to as anomalies. An asset

pricing anomaly could mean one of two

things. If we believe the CAPM is a

perfect model for explaining expected

returns, anomalies must mean that

markets are not efficient. Or if we know

that markets are efficient, the CAPM

must be wrong. The asset pricing model

must be wrong. This point turns out to

be impossible to resolve. It's actually

called the joint hypothesis problem. We

can't say whether markets are efficient

without having an asset pricing model to

test market efficiency. And we can't

prove whether an asset pricing model is

right without knowing whether markets

are efficient. Anyway, this initial

asset pricing research was important and

it's where FUM and French start in their

1993 paper. They focus on three key

problems with the CAPM. One, small

stocks earned higher average returns

than large stocks, unexplained by

differences in their CAPM betas. Two,

high booktomarket or value stocks earned

higher average returns than low

booktomarket or growth stocks. again

unexplained by differences in their cap

and betas. And three, the relationship

between beta and average returns was

weaker than CAPM predicted with low beta

stocks earning higher returns than the

model suggested. Basically, CAPM claimed

that two companies with the same betas

should have similar expected returns

regardless of their size or value

characteristics or I mean any other

characteristics for that matter.

Anything not following the model was

considered a mispricing or an asset

pricing anomaly. And Fman and French

wanted to explain the known anomalies at

the time. In this paper, Fman and French

noted these patterns that small stocks

and value stocks tended to perform

better on average than CAPM would

predict, stating right at the top of the

paper that the cross-section of average

returns on US common stocks shows little

relation to the market betas of the

sharp litner asset pricing model. That's

blunt language for an academic paper and

was followed up with the suggestion that

maybe the standard model the CAPM was

ignoring risk factors that are in fact

priced by the market. It is on this

basis that F and French build their now

famous three factor asset pricing model.

Rather than just looking at the market

factor, F and French developed a three

factor model that relates a stock's

expected return to the market factor, a

size factor, and a relative price or

value factor. The market factor is the

same concept from the CAPM. how much the

stock or portfolio moves when the

overall market moves. This factor shows

whether stocks went up or down with the

market. This pricing factor remains

important because broad market movements

still affect most stocks regardless of

their size or value characteristics. The

size factor is pretty straightforward.

It's based on the size of companies.

More specifically, this factor is

denoted as SMB or small minus big. It

refers to the difference in returns

between small company stocks and large

company stocks. If small stocks

outperform large stocks in a given

period, the SMB premium was positive in

that period. If large stocks do better,

the premium was negative. By including

this factor in the model, the model

captures whether being a small company

is related to systematic return

variation that investors are compensated

for. Then there's the value factor

denoted as HML or high minus low. This

is referring to booktomarket valuation

ratios. Booktomarket is basically the

company's accounting value compared to

its stock market value. High

booktomarket stocks are value stocks.

They're cheaper in market price relative

to the book value of their assets. Low

booktomarket stocks are growth stocks.

Investors are paying more for future

potential, pushing prices further above

their book values or accounting values.

HML measures the difference in returns

between value stocks and growth stocks.

Again, high minus low. When value stocks

outperform, the HML premium or value

premium is positive and vice versa. This

factor captures the effects of any

return premium that investors receive

for holding value stocks rather than

growth stocks, which again that was one

of the anomalies at the time. Each of

these factors is what's called a long

short portfolio. SMB, for example, is

long small cap stocks and short big

stocks. The specific constructions of

the factor portfolios have some details

that are even too nerdy, I think, for

this video. But the main idea is that

they're capturing return variations

related to company size and relative

price which had been well documented as

return anomalies back when this paper

was written. Okay, before we get into

the next part of their methodology and

the main results, which are where things

get really interesting and practically

useful, make sure to subscribe to this

channel if you haven't already, I break

down nerdy investing topics like this to

help you make more informed decisions

with your finances. Okay, let's keep

going on the finance paper that changed

everything. Take their new asset pricing

model for a test drive. FM and French

formed diversified portfolios by sorting

stocks based on size and value

characteristics. They split stocks into

five size groups and five booktomarket

groups, creating 25 total portfolios to

test. This method allowed them to

capture every combination of these

factors, small value stocks, big growth

stocks, and everything in between. They

show that these various groups of stocks

have significant variation in average

returns. And then they test whether

their three-factor asset pricing model

is able to explain that variation. This

is where Fman French did something

pretty groundbreaking at the time that

still shows up in tons of academic

finance papers and practical investment

analysis today. They used time series

regressions to test how well their asset

pricing factors explained the returns of

their test portfolios. A time series

regression looks at the returns of a

portfolio over time and asks how much of

the variation in returns is explained by

the asset pricing factors being used.

Running a time series regression is

pretty straightforward with the tools

available to investors today. You can do

it in Excel or using free online tools

like portfolio visualizer. The output of

a time series regression includes factor

loadings or coefficients which tell you

how your portfolio moves relative to

each factor and an alpha which is the

portion of returns that is not explained

by the asset pricing model. That last

term alpha is really important. It was

first used to describe excess risk

adjusted returns in a 1968 academic

paper. Another banger if you're into

this kind of thing. That paper used the

single factor capital asset pricing

model to ask whether actively managed

mutual fund managers were generating

returns in excess of what should be

expected based on the amount of risk

they were taking. It's worth mentioning

that active managers in the study were

not able to beat the market, but that's

another topic. And remember that model

only accounted for market risk. In the

context of F and French's 1993 paper,

they took each of the 25 test portfolios

and tracked their performance from July

1963 to December 1991 and then used

their model to try to explain the

differences in their returns. For each

portfolio, the regression estimated how

its returns co-moved with the three

factors in the model and whether there

were any large alphas, excess returns

unexplained by the model. They also

measured the R squared values of the

regressions. R 2 measures the

explanatory power of a regression. What

percentage of a portfolio's ups and

downs over time can be attributed to the

factors in the model? An R squared of 1

or 100% would mean perfect explanatory

power and zero would mean the factors

explained nothing. What FMA and French

found in terms of alphas and R 2 values

across their 25 test portfolios was

pretty incredible. They found that three

factor regressions on these portfolio

constructions resulted in R squared

values ranging from 0.83 to 0.97,

about 0.93 on average. In fact, 21 of

the 25 portfolios tested had R squared

values over 0.9. That means that this

model explains around 90% of the

differences in returns of diversified

stock portfolios. This test also

provided two other key data points.

First, every test portfolio combination

has a beta value close to one. This

consistency reinforces the fact that

variation in returns is explained by

more than just exposure to market beta

since the market beta exposures were

nearly the same in every case and yet

returns varied widely. That would be

surprising if market beta fully priced

all assets, but it followed the pattern

that FAM and French had predicted. In

their tests, FM and French found that

market beta could only explain around

60% of the differences in returns across

their test portfolios. Let that sink in

for a minute. We went from being able to

explain about 60% of the differences in

returns between diversified portfolios

with the CAPM to being able to explain

around 90% with the FMAN French

three-factor model. Another important

observation was that these three-factor

regressions resulted in near zero

intercepts or alphas on almost all test

portfolios. The one exception was small

cap growth stocks which had much lower

returns than the three-factor model

could explain, leaving lots of room for

future research, which I'll talk about

in a minute. The finding of minimal

alphas is important and might be one of

the most compelling validations of the

model. They demonstrate that once you

account for market size and value

exposures, there are basically no

persistent unexplained returns remaining

in most of their portfolio sorts. Other

than the 25 test portfolios formed on

size and price to book, FA and French

also tested portfolios formed on

dividends to price and earnings to

price, two other characteristics that

had been associated with higher returns.

They again found that their three-factor

model explained the returns of these

portfolios very well, shattering the

beliefs of dividend focused investors

everywhere. With their analysis, FMA and

French showed that size and value don't

need to be viewed as mysterious asset

pricing anomalies or market

inefficiencies. Rather, they can be

viewed as systematic risk factors that

together with market beta almost fully

account for return differences across

diversified portfolios. It's worth

mentioning that it's debatable whether

the factors in the model are risk

factors or artifacts of persistent

mispricing. That's one of those things

that's really hard to test in a way that

offers definitive proof one way or

another. All right, I don't mean to sit

here and nerd out about time series

regressions. Oh, okay, fine. That's

exactly what I mean to do. But this

asset pricing model and series of tests

really did change the way we viewed

markets. Our ability to explain

differences in returns between different

types of diversified portfolios went

from pretty good to in some cases near

perfect. Remember, CAPM using just

market beta typically explain only

somewhere between 60 and maybe 80% of

differences in returns between

diversified portfolios. By adding size

and value, F and French jumped to over

90% explanatory power with their

three-factor model. What this showed is

that a three-factor model could capture

fundamental systematic drivers of

returns that CAPM had completely missed.

The other area that this research

touches is traditional active

management. Remember the 1968 paper

showing that active managers don't

typically outperform after adjusting for

market risk. Well, with more factors in

the model, the data on traditional

active management only got worse. F and

French looked at that in later research

using their three-factor model.

Basically, if an active fund has beaten

the market, there's a really good chance

it did so by maintaining exposure to

some of the well-known factors. This

matters because if that is all an active

manager is doing, you could probably

replicate their factor exposure at a

lower cost than the fee that most active

managers charge. Altogether, the

findings in this paper challenged and

built on the single factor CAPM asset

pricing to create a much more

comprehensive framework for how

investors should think about building

and evaluating investment portfolios

systematically. After this paper, the

study of empirical asset pricing, that

is looking at asset returns and figuring

out which factors may be driving them,

took off. Researchers wanted to find the

best factors and create better asset

pricing models. It became a bit of a

problem or a joke. In 2011, John Cochran

in his presidential address to the

American Finance Association described

the proliferation of factors as a zoo, a

factor zoo. You could probably even call

it a factor slop to use the language

from my recent video on ETF slop. A 2016

census from Zoo, Lou, and Harvey

confirmed this concept, finding that 316

distinct factors had been published in

academic journals at that time and more

since then. This created a whole new set

of techniques for choosing factors and

comparing asset pricing models. In the

face of the slop proliferation, FMAN

French went on to update their

groundbreaking 1993 research in 2015,

which introduced two new asset pricing

factors, profitability and investment.

Profitability is expressed as RMW or

robust minus weak. That's the excess

return expected from companies with high

profitability over those with weak

profitability. Investment is expressed

as conservative minus aggressive.

Companies that grow their assets slowly,

conservative, tend to outperform those

that pour cash into rapid asset growth.

Aggressive. The addition of these two

new factors created the FMA and French

five factor model. The five-factor model

did help to solve some problems that the

three-factor model couldn't, and it

further improved the explanatory power

of the model closer to explaining 95% of

the differences in returns between

diversified portfolios. It's now what I

would call the workhorse asset pricing

model in academic finance. Don't get me

wrong, there is still plenty of ongoing

debate both in academia and in practice

about which factors make sense and which

asset pricing model we should be using.

I'm pretty comfortable saying that when

transaction costs are accounted for.

Fman French's five factor model is a

very strong foundation for thinking

about portfolio construction and

evaluating portfolio performance. Okay,

I told you I would bring this back to

practical relevance. There are fund

companies using this research, this

asset pricing research to build real

investment portfolios, real investment

products uh that you can invest in.

They're kind of like lowcost index funds

that deliver exposure to more than just

market risk, but they're not actually

index funds in most cases. Dimensional

Fund Advisors has a long history of

implementing factor investing research.

Eugene FMA, one of this paper's authors,

was a founding director and remains on

the board today. Kenneth French, the

other co-author, has long-standing

connections to Dimensional. As I

detailed in a previous video, their

track record using these strategies has

been very good over decades. One catch

is that Dimensional used to be only

available through financial advisors,

though at least in the US, they're now

available as ETFs that anybody can buy.

Not in Canada, though, unfortunately.

More recently, Avantis Investors, a

dimensional competitor, launched similar

products, and they have just launched

ETFs in Canada through CIBC. If you want

details on that, we did just release a

podcast episode with the Avantis CIO.

I'll do a deeper dive on these products

in another video. Full disclosure real

quick, my firm uses Dimensional Fund

Advisors funds pretty extensively. We

are not paid by Dimensional Orantis, and

I was not being paid to mention them in

this video. What does this all mean for

you? Evidence suggests that long-term

expected returns are driven by exposure

to specific factors. Factors that we

know about due to decades of academic

research, implying that investors may

achieve higher expected returns by

tilting toward certain types of stocks.

But which types of stocks and knowing

how to do it effectively and at a low

cost is a whole other question.

Fortunately, there are fund companies

like Dimensional Funds Advisors and

Avantis Investors that do this. creating

lowcost and broadly diversified

investment portfolios specifically built

to take academic theory and apply it to

your portfolio. Like I said at the

beginning, this seinal paper forms the

foundation for a big part of how I think

about investing. It's also part of the

reason that my firm PWL Capital has been

using Dimensional Fund Advisors products

since we helped bring them to Canada

back in 2003. If you want to learn more

about PWL and how we applied this

thinking to our clients portfolios, you

can click right here. Or if you want to

nerd out on one of the most

controversial papers in finance, you can

click here.