Professional Documents
Culture Documents
Textbook Empirical Asset Pricing Models Jau Lian Jeng Ebook All Chapter PDF
Textbook Empirical Asset Pricing Models Jau Lian Jeng Ebook All Chapter PDF
Jau-Lian Jeng
Visit to download the full and correct content document:
https://textbookfull.com/product/empirical-asset-pricing-models-jau-lian-jeng/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
https://textbookfull.com/product/contemporaneous-event-studies-
in-corporate-finance-methods-critiques-and-robust-alternative-
approaches-jau-lian-jeng/
https://textbookfull.com/product/continuous-time-asset-pricing-
theory-robert-a-jarrow/
https://textbookfull.com/product/financial-asset-pricing-
theory-1st-edition-munk-claus/
https://textbookfull.com/product/pricing-analytics-models-and-
advanced-quantitative-techniques-for-product-pricing-first-
edition-paczkowski/
Asset pricing and portfolio choice theory Second
Edition Kerry E. Back
https://textbookfull.com/product/asset-pricing-and-portfolio-
choice-theory-second-edition-kerry-e-back/
https://textbookfull.com/product/financial-decisions-and-markets-
a-course-in-asset-pricing-john-y-campbell/
https://textbookfull.com/product/analytical-corporate-valuation-
fundamental-analysis-asset-pricing-and-company-valuation-
pasquale-de-luca/
https://textbookfull.com/product/econophysics-and-capital-asset-
pricing-splitting-the-atom-of-systematic-risk-1st-edition-james-
ming-chen-auth/
https://textbookfull.com/product/financial-modelling-with-
forward-looking-information-an-intuitive-approach-to-asset-
pricing-1st-edition-nadi-serhan-aydin-auth/
Jau-Lian Jeng
Empirical Asset
Pricing Models
Data, Empirical Verification, and Model Search
Jau-Lian Jeng
School of Business and Management
Azusa Pacific University
Stevenson Ranch, CA, USA
v
vi PREFACE
vii
INTRODUCTION
Ever since the pioneering work of the capital asset pricing model, theo-
retical and empirical discussion on the pricing kernel of asset returns has
been huge in the financial economics literature. Although many alternative
methodologies and theories have been devised, the difficulty in empirical
application of asset pricing models still remains unresolved in many areas
such as model instability over different time horizons, variable selection on
proxies for factors, and (possibly) applicable robust statistical inferences. It
is likely that we will discover that an empirical asset pricing model, once
selected, can only apply to a certain time period before the model validity
quickly disappears when an extended time horizon or data set is considered.
Unfortunately, this phenomenon seems to prevail in many data sets
(domestic or foreign) that are applied. The disappointing results in turn
lead to the pervasive discontent with the theoretical foundation of asset
pricing models. Emphasis on (time series) predictability becomes the norm
for model validity for empirical asset pricing models. With the keen demand
for validating empirical asset pricing models, statistical verification (with
predictability and specification tests) when certain proxies or variables of
interest are used becomes the mainstream for financial time series modeling
on asset returns.
Essentially, emphases in finding the common features or characters of
asset returns (in an attempt to reduce the dimensionality, for instance)
through statistical significance should be dealt with using additional cau-
tion since these features, once identified, may only prevail tentatively (or
contempornaeously) over the selected time horizon.
ix
x INTRODUCTION
left-over data in the data set which the modeler has already obtained. The
major dilemma lies in the trade-offs as to whether the model specification
on empirical asset pricing models is to find something that may help
to describe the (short-run) dynamics of asset returns or to identify the
quintessence of pricing kernels when short-run predictability could be
sacrificed.
Although these trade-offs are not immediately clear-cut, given the
notorious time-changing nature of financial markets, it is unlikely that there
exists an omnipotent model that encompasses all others across all time
horizons. To the best that can be shown, the winning model (through
statistical verification or otherwise) only represents a tentative explanation
or approximation for the underlying pricing kernel of asset returns. Time
changes everything.
Hence, even with the contemporaneous model that encompasses all
other competitive alternatives, the empirical result only shows the current
notion for the underlying determinant of asset returns. What is more criti-
cal, however, is whether the tentative model obtained helps us understand
more about the pricing kernel of the asset returns or not. And perhaps
more essentially, it helps us to modify diligently the model(s) for different
time horizons or data sets.
In Chap. 1 of Part I, the discussions focus on the conventional linear
models for asset returns. Given the enormous volume of literature on asset
pricing models, this book only surveys and develops the discussions on
parametric model building and variable selection. The recent developments
on semi-parametric (factor) modeling for asset pricing are also briefly
discussed.
Starting from the capital asset pricing model (CAPM), the methods
for reduction of dimensionality are covered where factor-pricing models
are typical examples. It is not too difficult to find that the empiricist in
applied finance may criticize these models as somewhat useless in the usage
of profit-taking transactions. Nonetheless, from the perspectives of the
financial economist, this is precisely the result of a properly working market
mechanism where the advantage in any attempt at speculative opportunity
should quickly resolve to zero. Does this mean that these theories are
all useless in empirical application? We can only be sure if we have some
better theories to explain the mechanism of capital markets and the ultimate
determinants for pricing kernels of stock returns.
Although many alternative approaches such as the nonlinearity and
behavioral assumptions are developed, the question to ask is “Are these
xii INTRODUCTION
xv
xvi CONTENTS
Bibliography 257
Index 259
PART I
The author surveys and discusses linear asset pricing models with the intent
to identify some sets of variables or factors with reduced dimensionality
to approximate the core or pricing kernel of asset returns. A theoretical
foundation may start with discussion on factor pricing models where
asset returns are projected onto some lower-dimensional sets of factors
that possibly explain the major variations of asset returns. The aim is to
identify major determinants for the fluctuations of asset returns where
these determinants satisfy some systematic properties that ensures their
indispensable roles.
Controversies begin with questions of measurability of factors and their
justification. Classical issues such as the measurability of market portfolio
in the capital asset pricing model (CAPM) and selection of market indices,
for instance, all incur the problems of measurability and representation for
the verification of a theoretical framework. Developments and extensions
of arbitrage pricing theory (APT) and multi-factor asset pricing models do
not make the hope of attaining robust asset pricing models any brighter.
Statistical inferences do not always mediate the severity of problems
mentioned if caution regarding their limitations is not taken into account.
Given that all measurable factors presumed for asset pricing models may
contain some measurement errors, it is unlikely that empirical asset pricing
models will resolve the difficulty of completeness in model justification. At
their best extent, empirical asset pricing models can only mimic the sys-
tematic patterns or properties of asset returns that provide the tractability
EŒRi D Rf C ˇi ŒE(Rm ) Rf ;
i ;Rm )
where Rf stands for the risk-free rate and ˇi D Cov(R
m2
as the systematic
risk, Rm stands for the rate of return of the market portfolio. In brief,
ASSET PRICING MODELS: SPECIFICATION, DATA AND THEORETICAL… 7
the asset’s risk premium depends on the systematic risk and the market
premium E(Rm ) Rf : In applying the CAPM onto the stock return data,
conventional studies consider the time series regression model as
where rit D Rit Rft is the excess return of asset i at time t; and rmt D Rmt Rft :
The time series regression will give the estimates for the “betas” of the
excess returns of included assets. Accordingly, if the theory holds true, the
intercept in the time series regression should be close to zero. Ideally, if
the market portfolio is correctly identified then the theory should result in
the second-pass regression, such that for i D 1; : : : ; n;
ri D o C 1 ˇOi C i ;
where ri is the (time-series) average excess return for asset i; and ˇOi is the
estimate of beta for asset i from the first-pass regression. Under the model
CAPM, the coefficient o should be equal to zero, and 1 is the coefficient
for the market premium.
However, since the market indices may not precisely represent the mar-
ket portfolio and the cross-sectional dependence, and since heteroskedas-
ticity may make the conclusion of second-pass regression misleading, Fama
and MacBeth (1973) has developed the “grouping” portfolios as applying
the estimates of “betas” from the time series regressions so as to consider
the second-pass regression such that
rp D C ˇp C p ;
ri D EŒri C bi f C ei ;
where ri is the excess return for asset i; ei is the idiosyncratic risk, and f is
the systematic factor, EŒei D EŒ f D CovŒei ; f D 0: Under no arbitrage
condition, the expected premium for asset i should be expressed as
EŒri bi :
Reisman (1992) extends the analysis to consider the “beta” when defined
on the reference variable(s), which establishes a similar result. That is, given
the beta bQ i (with respect to the reference variable g), it is feasible (if the
factor structure is correct) to write
bQ i Cov( f ; g)bi :
ASSET PRICING MODELS: SPECIFICATION, DATA AND THEORETICAL… 9
which is equivalent to stating that EŒri bQ i : That is, there exists a pricing
functional for expected premiums when using the reference variables.
Following Chamberlain and Rothschild (1983), Reisman (1992), the
excess returns for all assets are projected onto the set of factors (or reference
variables) f f1 ; : : : ; fk gsuch that
X
k
ri D ˛i C ˇij fj C i ;
jD1
where the fi giD1;:::;n represent the idiosyncratic risk of the return processes.
The factor structure (with k factors) will hold if the eigenvalues of the
covariance matrix of fi giD1;:::;n are all bounded from above. Given the
factor structure (and under the continuity assumption), it is feasible for
the expected returns (or risk premiums) to be expressed as
X
k
EŒri ˇij ( fj );
jD1
For the first issue, according to Clements and Hendry (1999), the unpre-
dictability (for a stochastic process t ) is defined as
D t ( t jIt1 ) D D t ( t );
insights into understanding the pricing kernel of asset returns. One of the
possible reasons is that these verifications are usually based on different
information sets over time horizons, data sources, and constructions of
relevant variables. In addition, the notion of (un)predictability can also be
considered as time-varying such that for time index H 1;
D t ( t jIt1 ) D D t ( t ); t D 1; : : : ; T;
D t ( t jIt1 ) ¤ D t ( t ); t D T C 1; : : : ; T C H;
or,
D t ( t jIt1 ) ¤ D t ( t ); t D 1; : : : ; T;
D t ( t jIt1 ) D D( t ); t D T C 1; : : : ; T C H:
are available in a short period of time, this so-called predictability may vary
and/or vanish over different time periods.
Various methods have been designated to the search, including the high-
end technology of financial econometrics. However, since any investor will
almost always endeavor to pursue speculative profits once the technology
is discovered, it is inconceivable that the findings will be published and
become enunciable. Hence, it is perhaps not too surprising to say that if
one successfully develops a predictive mechanism that belittles others in
the market, one should keep it completely confidential.
In particular, under the pavilion of contemporaneous financial econo-
metrics, predictability is essential in either the model verifications or the
empirical applications. Unfortunately, even though predictability is only
the verification of the model’s validity, the provision of some empirical
studies may lead to misunderstanding and the attempt to search for the
possibility of speculative profits with better forecasts. More strikingly,
overemphasis on the predictions and forecasts may lead the financial
modeling toward tracking asset returns with devices or mechanisms of
short-term validity where no plausible explanation (in financial economics
or otherwise) may be feasible. In fact, this kind of emphasis and motivation
(for new tracking methods) may simply destroy the validity of verifying
model specification (with predictability) since the devised mechanism
is only for “tracing and chasing the prey”. More specifically, it is not
merely due to the possibility of a time-varying pricing mechanism that the
theoretical models may not perform better in predictions or forecasts. As
a matter of fact, this lack of soundness in predictability simply shows that
all models are only approximations for the data generating mechanism.
Namely, theoretical models are not developed solely for predictability.
Developing models or hypotheses in asset pricing models is to improve
the search for understanding and approaches toward some better direction
for decision making, if not more.
Thirdly, notice that the ultimate objective of asset pricing models is
to identify effectively a tractable explanation for the pricing mechanism
(and perhaps on its changing nature) based on the accessible infor-
mation. Emphasis should be on the continuous effort and work (of
finance professionals or academics) of searching for the determinants of
the pricing mechanism of stock returns with rigor and explanation—
provided that the underlying system of asset returns is almost always
evolving through different time periods, regimes, systems, or economies.
The goal, however, is to search for any tractable mechanism from which
14 J.-L. JENG
y D x1 ˇ1 C x2 ˇ2 C "; (1.2.1)
and x D (x01 ; x02 )0 is the vector of x1 ; x2 , and X is the design matrix of x: The
prediction error can be expressed as
E(y yO )0 (y yO )
(1.2.3)
D 2 (1 C x0 (X 0 X)1 x):
y D x1 1 C ; (1.2.4)
16 J.-L. JENG
E(y yO )0 (y yO )
The Tool will Pick Up a Drop of Oil and Deposit It Where Wanted
The oiling tool is dipped in light oil and a drop applied to each
bearing. Replace the works in the case and the job is finished. A
reliable jeweler will charge very little for this work, but the more crafty
ones may ask a good price for this “mysterious” process. If the works
are not dirty, apply the oil with the tool. Anyone who has tried to oil a
clock with an ordinary spout oilcan knows the futility of the attempt.
The object of the tool is to pick up and carry a drop of oil and deposit
it where wanted. A can, a feather, or a match will do, but any one of
them is apt to carry dirt, flood the dry part, or smear up nonmoving
parts.
Easily Constructed Wall Shelves
Shelves for Books Supported with Picture-Frame Wire to the Wall
All that is necessary to make and support the simple set of wall
shelves, shown in the illustration, is lumber for the shelves, four
screw eyes, four screw hooks, sufficient picture-frame wire to form
the braces and supports, and wood screws for attaching the wire. On
the top side of the upper shelf are fastened the four screw eyes, two
near the wall edge and the others near the outer edge. To support
the upper shelf four screw hooks are used; two placed in the wall
and spaced to match the set of screw eyes nearest the wall, the
others being placed above the first and connected to the outer set of
screw eyes with the wire, thereby forming strong inclined supports.
The remaining shelves can be hung to suit by the supporting wires,
which are fastened with screws to the end of each shelf.
Showing the Strength of a Giant
This trick is not so well known as it might be, although for a while it
was quite a popular drawing attraction for circus side shows and
other amusement places. It is one of the favorite Hindu tricks. The
performer passes for examination two pieces of rope 10 ft. long. In
one end of each rope a large ring is fastened. Taking a ring in each
hand the performer commands three or four men at each end of the
rope to take hold of it and at a signal they pull as hard as possible.
They pull until they are exhausted as in a tug of war, but the
performer only appears a trifle exerted and finds no difficulty in
holding the men.
The Performer Seems to Hold the Ones Pulling on the Ropes without Any
Effort, Producing an Effect That cannot be Readily Understood, and Making
an Excellent Trick for the Lawn Party
Toouter
enjoy a vacation in the woods thoroughly, it is essential that the
be provided with the right kind of an outfit. The
inexperienced are likely to carry too much rather than too little to the
woods; to include many unnecessary luxuries and overlook the more
practical necessities. However, camp life does not mean that one
must be uncomfortable, but rather implies plain and simple living
close to nature. An adequate shelter from the sun and rain, a
comfortable bed, a good cooking kit, and plenty of wholesome food,
are the important things to consider. No man or woman requires
more, and if unwilling to share the plain fare of the woodsman, the
pampered ones should be left at home, for the grouchy, complaining
individual makes, of all persons, the very worst of camping
companions.
The Old Hand at the Camping Game Prefers
to Cut Poles on the Camping Site and Set
Them Up on the Outside for the Camp-Fire
Tent
There are tents and tents, but for average outings in what may be
considered a permanent camp, the regulation wall, or army, tent is
generally used to make a comfortable shelter. It is a splendid utility
tent, with generous floor space and plenty of headroom. For the
permanent camp, the wall tent is often provided with a fly, which may
be set up as an extra covering for the roof, or extended over the front
to make a kind of porch. An extension may also be purchased to
serve the same purpose. The 7 by 9-ft. wall tent will shelter two
persons comfortably, but when the camp is seldom moved, the 9 by
12-ft. size, with a 3¹⁄₂-ft. wall, will afford more room. The regulation 8-
oz. duck is heavy enough, or the same tent may be obtained in tan
or dark green khaki, if preferred. In any case the tent should have a
sod cloth, from 6 to 12 in. wide, extending around the bottom and
sewed to the tent. An extra piece of canvas or floor cloth is desirable,
but this as well as the fly are extras, and while convenient, are by no
means necessary. The wall tent may be erected with the regular
poles, or it may be ordered with tapes along the ridge and erected by
suspending between two trees. The old hand at the camping game
rarely uses the shop poles supplied with most tents, but prefers to
cut them at the camping site and rig them up on the outside, one
slender pole fastened with tapes along the ridge and supported at
either end in the crotch formed by setting up two poles, tripod or
shear-fashion.
The “Baker” style is a popular tent, giving a large sleeping
capacity, yet folding compactly. The 7 by 7-ft. size, with a 2-ft. wall,
makes a good comfortable home for two, and will shelter three, or
even four, if required. The entire front may be opened to the fire by
extending it to form an awning, or it may be thrown back over the
ridge to form an open-front lean-to shelter.
The “Dan Beard,” or camp-fire, tent is a modification of the Baker
style, having a slightly steeper pitch, with a smaller front opening.
The dimensions are practically the same as the Baker, and it may be
pitched by suspending between two trees, by outside poles, or the
regular poles may be used.
For traveling light by canoe or pack, a somewhat lighter and less
bulky form of tent than the above styles may be chosen, and the
woodsman is likely to select the forester’s or ranger types. The
ranger is a half tent with a 2-ft. wall and the entire front is open; in
fact, this is the same as the Baker tent without the flap. If desired,
two half ranger tents with tapes may be purchased and fastened
together to form an A, or wedge, tent. This makes a good tent for two
on a hike, as each man carries his own half, and is assured a good
shelter in case one becomes separated from his companion, and a
tight shelter when the two make camp together.
The forester’s tent is another good one, giving good floor space
and folding up very compactly, a 9 by 9-ft. tent weighing about 5¹⁄₂ lb.
when made of standard-weight fabric. It may be had either with or
without hood, and is quickly erected by using three small saplings,
one along the ridge, running from peak to ground, and one on each
side of the opening, to form a crotch to support the ridge pole, shear-
fashion. These tents are not provided with sod or floor cloths,
although these may be ordered as extras if wanted.
The canoe or “protean” tents are good styles for the camper who
travels light and is often on the move. The canoe tent has a circular
front, while the protean style is made with a square front, and the
wall is attached to the back and along the two sides. Both tents are
quickly set up, either with a single inside pole or with two poles set
shear-fashion on the outside. A 9 by 9-ft. canoe or protean tent with
a 3-ft. wall makes a comfortable home in the open.
Whatever style of tent is chosen, it is well to pay a fair price and
obtain a good quality of material and workmanship. The cheaper
tents are made of heavier material to render them waterproof, while
the better grades are fashioned from light-weight fabric of close
weave and treated with a waterproofing process. Many of the
cheaper tents will give fair service, but the workmanship is often
poor, the grommets are apt to pull out, and the seams rip after a little
hard use. All tents should be waterproofed, and each provided with a
bag in which to pack it. An ordinary tent may be waterproofed in the
following manner: Dissolve ¹⁄₂ lb. of ordinary powdered alum in 4 gal.
of hot rain water, and in a separate bucket dissolve ¹⁄₂ lb. of acetate
of lead—sugar of lead—in 4 gal. of hot rain water. The acetate of
lead is poisonous if taken internally. When thoroughly dissolved, let
the solution stand until clear, then pour the alum solution into a tub
and add the lead solution. Let the solution stand for an hour or two,
then pour off the clear water and thoroughly soak the fabric in the
waterproofing mixture by rubbing and working the material with the
hands. Hang the cloth up without wringing it out.
The Forester’s Tent is Quickly Erected by
Using Three Small Saplings, One along the
Ridge, and One on Each Side of the Opening
to Form a Crotch for the Ridge Pole
The Ranger’s or Hiker’s Tent Comes in The Canoe or Protean Tents
Halves. Each Half may be Used Are Good Styles for the
Independently as a Lean-To Shelter for One Camper Who Travels Light
Man, or Both Joined Together to Make Room and Is Often on the Move,
for Two Persons and They can be Quickly Set
Up with a Single Inside Pole
The camping kit, including the few handy articles needed in the
woods, as well as the bedding and cooking outfit, may be either
elaborate or simple, according to the personal experience and ideas
of the camper. In making up a list, it is a good plan to remember that
only comparatively few articles are really essential for a comfortable
vacation in the wilderness. A comfortable bed must be reckoned one
of the chief essentials, and one may choose the de-luxe couch—the
air mattress or sleeping pocket—use the ordinary sleeping bag, or
court slumber on one of the several other styles of camp beds. The
fold-over combination bed, the stretcher bed, or a common bag
made of ticking, 6¹⁄₂ ft. long by 2 ft. wide, which is stuffed with
browse or leaves, will suffice for the average person. Folding camp
cots, chairs, tables, and other so-called camp furniture, have their
places in the large, fixed camps, but the woodsman can manage to
live comfortably without them. A good pair of warm blankets should
be included for each person, providing the sleeping bag is not taken
along. The regulation army blankets are a good choice and
reasonable in price, or the blankets used at home may be pressed
into service.
A good ax is the woodsman’s everyday companion, and a good-
weight tool, weighing 3 or 4 lb., and a smaller one of 1¹⁄₂ lb. should
be carried. When going light, the belt ax will suffice.
The oil lantern is only suited for the fixed camp, since the fuel is
difficult to transport unless it is placed in screw-top cans. The
“Stonbridge” and other folding candle lanterns are the most
convenient for the woods and give sufficient light for camp life.
The aluminum cooking outfits are light in weight, nest compactly,
and will stand many years of hard usage, but like other good things,
they are somewhat expensive. A good substitute, at half the price,
may be obtained in tin and steel, having the good feature of nesting
within each other, but, of course, not being quite so light nor so
attractive in appearance as the higher-priced outfits. Both the
aluminum and steel outfits are put up in canvas carrying bags, and
an outfit for two includes a large and a small cooking pot coffee pot;
frying pan with folding or detachable handle; two plates; cups knives;
forks, and spoons. Outfits may be bought for any number of persons
and almost all sporting-goods stores carry them. The two-man outfit
in heavy aluminum will cost $9 or $10, while the same outfit
duplicated in steel is priced at $3.35.