The Journal of Finance - December 1993 - ENGLE

THE JOURNAL OF FINANCE v VOL. XLVIII, NO.
5 v DECEMBER 1993
Measuring and Testing the Impact of

News on Volatility
ROBERT F. ENGLE and VICTOR K. NG*
ABSTRACT
This paper defines the news impact curve which measures how new information is
incorporated into volatility estimates. Various new and existing ARCH models
including a partially nonparametric one are compared and estimated with daily
Japanese stock return data. New diagnostic tests are presented which emphasize
the asymmetry of the volatility response to news. Our results suggest that the
model by Glosten, Jagannathan, and Runkle is the best parametric model. The
EGARCHalso can capture most of the asymmetry; however, there is evidence that
the variability of the conditional variance implied by the EGARCHis too high.
THE ABILITY TO FORECAST financial market volatility is important for portfolio

selection and asset management as well as for the pricing of primary and
derivative assets. While most researchers agree that volatility is predictable
in many asset markets (see for example the survey by Bollerslev et al.
(1992)), they differ on how this volatility predictability should be modeled. In
recent years the evidence for predictability has led to a variety of approaches,
some of which are theoretically motivated, while others are simply empirical
suggestions. The most interesting of these approaches are the "asymmetric"
or "leverage" volatility models, in which good news and bad news have
different predictability for future volatility. These models are motivated by
the empirical work of Black (1976), Christie (1982), French, Schwert, and
Stambaugh (1987), Nelson (1990), and Schwert (1990). Pagan and Schwert
(1990) provide the first systematic comparison of volatility models. This paper
builds on their results, focusing on the asymmetric effect of news on volatil-
ity. Specifically, we provide new diagnostic tests, a partially nonparametric
model for discovering the empirical relations between news and volatility,
and a metric for interpreting the differences between volatility models.
* Engle is from the Department of Economics, University of California, San Diego, and Ng is
from the School of Business Administration, University of Michigan, Ann Arbor.We thank John
Campbell, David Hsieh, Gautam Kaul, Laura Kodres, Stanley Kon, M. P. Narayanan, Bill
Schwert, Paul Seguin, Rob Stambaugh, Rene Stulz, an anonymous referee, and workshop
participants at the University of California, San Diego, the University of Michigan, the Univer-
sity of Southern California and conference participants at the National Bureau of Economic
Research, Inc., and the Southern Finance Association Meetings for helpful comments. We also
thank Zhuanxin Ding for his excellent research assistance in carrying out the simulation
experiment in this paper.
1749
15406261, 1993, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.1993.tb05127.x by Cochrane Peru, Wiley Online Library on [17/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1750 The Journal of Finance
The importance of a correctly specified volatility model is clear from the

range of applications requiring estimates of conditional volatilities. In the
valuation of stocks, Merton (1980) shows that the expected market return
is related to predictable stock market volatility. French, Schwert, and
Stambaugh (1987) and Chou (1988) also find empirical evidence for this
relationship, although Chou, Engle, and Kane (1992) indicate that the rela-
tion may be more complex. Ferson and Harvey (1991) provide evidence that
much of the predictability of a sample of monthly portfolio returns can be
related to the predictability of risk premiums. Schwert and Seguin (1990) and
Ng, Engle, and Rothschild (1992) show that individual stock return volatility
is driven by market volatility, with individual stock return premiums affected
by the predictable market volatility.
In the valuation of stock options, Wiggins (1987) and Hull and White (1987)
suggest that stochastic stock return volatility might be the source of some
documented pricing biases of the Black-Scholes option-pricing formula. Fur-
thermore, the research of Day and Lewis (1992) shows that implied volatility
from the Black-Scholes model cannot capture the entire predictable part of
future volatility relative to some GARCH and EGARCH models. Harvey and
Whaley (1992) also find some predictability in changes in implied volatilities,
and profits can be earned by trading on this information, although only gross
of transaction costs. Kuwahara and Marsh (1992) find predictable volatility
models like the EGARCH useful in the valuation of warrants. Amin and Ng
(1993) show that option valuation under predictable volatility is different
from option valuation under unpredictable volatility.
Ross (1989) argues that volatility can be regarded as a measure of informa-
tion flow, and Conrad, Gultekin, and Kaul (1991) find evidence that informa-
tion flow is from large to small firms. Hamao, Masulis, and Ng (1990) find
evidence of volatility spillover from the U.S. stock market to both the U.K.
and the Japanese stock markets. However, Susmel and Engle (1992) and Lin,
Engle, and Ito (1992) show that such spillover could be very short lived.
Masulis and Ng (1992) find evidence of overnight information affecting
daytime volatility.
Finally, the predictability of volatility is important in designing optimal
dynamic hedging strategies for options and futures (Baillie and Myers (1991)
and Engle, Hong, Kane, and Noh (1992)). The predictability of volatility
might also affect the results of event studies (see, for example, Connolly
(1989)).
In the next section, we discuss several models of predictable volatility and
present the idea of a news impact curve which characterizes the impact of
past return shocks on the return volatility implicit in a volatility model. In
Section II, we suggest several new diagnostic tests based on the news impact
curve. We also perform a small Monte Carlo experiment to examine the finite
sample properties of the test statistics. In Section III, a partially nonparamet-
ric ARCH model is introduced. Using a Japanese stock return series, in
Section IV we compare the GARCH (1, 1) model with several other volatility
models that allow for asymmetry in the impact of news on volatility. To check
Measuring and Testing the Impact of News on Volatility 1751
the adequacy of the models, we employ the diagnostic tests developed in

Section II. In Section V, the partially nonparametric model is estimated and
compared with the others. In Section VI, the best models are reestimated on a
precrash sample period. Section VII concludes the paper.
I. Models of Predictable Volatility

Given the importance of predicting volatility in many asset-pricing and
portfolio management problems, many approaches of forecasting volatility
have been proposed in the literature. The most popular one is the class of
autoregressive conditional heteroskedasticity (ARCH) models originally intro-
duced by Engle (1982). In a recent survey by Bollerslev et al. (1992) more
than 200 papers are cited applying ARCH and related models to financial
time series. In this section, we will review some of the more popular pre-
dictable volatility models in the ARCH class. We will define a news impact
curve which can be used to evaluate and compare the properties of different
volatility models in the ARCH class.
Let Yt be the rate of return of a particular stock or the market portfolio
from time t - 1 to time t. Also, let Ft- 1 be the past information set contain-
ing the realized values of all relevant variables up to time t - 1. Since
investors know the information in Ft-1 when they make their investment
decision at time t - 1, the relevant expected return and volatility to the
investors are the conditional expected value of Yt, given Ft -, and the
conditional variance of Yt, given Ft -. We denote these by mt and ht
respectively. That is, mt = E(yt IFt- 1) and ht = Var(yt I-1). Given these
definitions, the unexpected return at time t is ?t-Yt-mt. In this paper, et
is treated as a collective measure of news at time t. A positive et (an
unexpected increase in price) suggests the arrival of good news, while a
negative et (an unexpected decrease in price) suggests the arrival of bad
news. Further, a large value of ltl implies that the news is "significant" or
"big" in the sense that it produces a large unexpected change in price.
Engle (1982) suggests that the conditional variance ht can be modeled as a
function of the lagged ?'s. That is, the predictable volatility is dependent on
past news. The most detailed model he develops is the pth order autoregres-
sive conditional heteroskedasticity model, the ARCH(p):
p
ht ='+ E at-X (2)
i=l1
where al,..., ap, and to are constant parameters. The effect of a return
shock i periods ago (i < p) on current volatility is governed by the parameter
ai. Normally, we would expect that ai < aj for i > j. That is, the older the
news, the less effect it has on current volatility. In an ARCH(p) model, old
news which arrived at the market more than p periods ago has no effect at
all on current volatility.
Bollerslev (1986) generalizes the ARCH(p) model to the GARCH(p, q)

model, such that
p q
ht =tc' + ,ai?t2-1l + E 3jht - i (3)
i=l i=l
where a1,..., ap, I1, ..., ,fp, and to are constant parameters. The GARCH
model is an infinite order ARCH model. In the GARCH(1, 1) model, the effect
of a return shock on current volatility declines geometrically over time.
Empirically, the family of GARCH models has been very successful. Of these
models, the GARCH (1, 1) is preferred in most cases (see the survey by
Bollerslev et al. (1992)).
Despite the apparent success of these simple parameterizations, the ARCH
and GARCH models cannot capture some important features of the data. The
most interesting feature not addressed by these models is the leverage or
asymmetric effect discovered by Black (1976), and confirmed by the findings
of French, Schwert, and Stambaugh (1987), Nelson (1990), and Schwert
(1990), among others.1 Statistically, this effect occurs when an unexpected
drop in price (bad news) increases predictable volatility more than an unex-
pected increase in price (good news) of similar magnitude. This effect sug-
gests that a symmetry constraint on the conditional variance function in past
?'s is inappropriate. One method proposed to capture such asymmetric effects
is Nelson's (1990) exponential GARCH or EGARCH model
log(ht) = o + 3-log(ht-1) + Y?
1 + a [ et-1
8t_ -2/T (4)
where to, /3, y, and a are constant parameters. The EGARCH model is
asymmetric because the level of et - 1/ ht1 is included with a coefficient y.
Since this coefficient is typically negative, positive return shocks generate
less volatility then negative return shocks, all else being equal.
A comparison between the GARCH(1, 1) model and the EGARCH(1, 1)
suggests an interesting metric by which to analyze the effect of news on
conditiona-l heteroskedasticity. Holding constant the information dated t - 2
and earlier, we can examine the implied relation between et -1 and ht. We
call this curve, with all lagged conditional variances evaluated at the level of
the unconditional variance of the stock return, the news impact curve be-
cause it relates past return shocks (news) to current volatility. This curve
measures how new information is incorporated into volatility estimates. It is
similar in spirit to Figure 2 in Pagan and Schwert (1990). In the GARCH
model, this curve is a quadratic function centered on et -1 = 0. For the
EGARCH, it has its minimum at et- 1 = 0, and is exponentially increasing in
both directions but with different parameters. In particular, the news impact
1 It is not yet clear in the finance literature that the asymmetric properties of variances are
due to changing leverage. The name "leverage effect" is used simply because it is popular among
researchers when referring to such a phenomenon.
curve for the EGARCH model when the lagged conditional variance is
evaluated at its unconditional level, o-2 is given by
(y + a) 1
ht= A exp[ .t1]| for et-1>, and
(e-a) 1
ht =A .exp( t- |] for ?t-1 < ?
where A u- 2P3.exp[wIto a, 2/I]. (5)
In Figure 1, the news impact curve of the EGARCH(1, 1) is compared with

the news impact curve of the GARCH(1, 1) for y < 0 but a + y > 0. If the
curves were extrapolated, the EGARCH would have higher variances in both
directions because the exponential curve eventually dominates the quadratic.
Thus, we can see from the news impact curve that the EGARCH model differs
from the standard GARCH model in two main respects:
1. The EGARCH model allows good news and bad news to have a different
impact on volatility, while the standard GARCH model does not, and
2. the EGARCH model allows big news to have a greater impact on
volatility than the standard GARCH model.
The news impact curve can be constructed for many other models, some of
which are outlined in Table I. While the functional form of these models is
rather complicated, the qualitative differences between these models can be
compared by contrasting their implied news impact curves.
The news impact curve of the nonlinear model of Engle and Bollerslev
(1986) is symmetric. However, it implies a reduced response to extreme news
if y < 2. The news impact curve of the multiplicative ARCH model of Mihoj
(1987), Geweke (1986), and Pantula (1986), is symmetric and passes through
the origin. Depending on the value of the ai's, the two sides of the news
impact curve can be either steeper or less steep than the GARCH(1, 1) news
impact curve. The news impact curve of the autoregressive standard devia-
tion model of Schwert (1990) is symmetric and centered at et-1 = 0. The
news impact curve of the asymmetric GARCH model of Engle (1990),
AGARCH, is asymmetric and centered at et -1 = --y, which is to the right of
the origin when y < 0. The news impact curve of the GJR model of Glosten,
Jagannathan, and Runkle (1989) is centered at et-1 = 0, but has different
slopes for its positive and negative sides. The news impact curves of both the
nonlinear asymmetric GARCH model (NGARCH) and the VGARCH model
are symmetric and centered at et 1 = (-y)- ht1. However, the slope of
the two upward-sloping portions of the VGARCH is steeper than that of the
NGARCH.
To summarize, the news impact curves of the above asymmetric volatility
models capture the leverage or asymmetric effect by allowing either the slope
of the two sides of the news impact curve to differ or the center of the news
A GRH(1, ews Impat ure

h \
News ImpactCure
A GARCH(1,1)
01
Figure 1. The news impact curves of the GARCH(1, 1) model and the EGARCH(1, 1)
model. The solid line is the GARCH(1, 1) news impact curve. The dashed line is the EGARCH(1,
1) news impact curve. The equation for the GARCH(1, 1) news impact curve is
ht = A
2
? a -1,
where h t is the conditional variance at time t, 1is ~

the unpredictable return at time t -1,
A w ?/3 o2 o- is the unconditional return standard deviation, w is the constant tr,ad/
is the parameter corresponding to h t-1 in the GARCH variance equation. The shape of the above
GARCH(1, 1) news impact curve is indicative of cases with wo> 0, 0 </13 < 1, o-> 0, 0 < a <1
and a ? /3 < 1.
The equations for the EGARCH(1, 1) news impact curve are
ht=AA exp[
ry a)
*8t-i,j
1
foret-i>O, and
ht=A* exp[, forseti<0,
where A =- exp[ w - a - 2/ir]1, o- is the unconditional return standard deviation, w is the

constant term, /3 is the parameter for the log(hht - 1) term, a is the parameter for the I t - 11I/ rh t _1
term, and y is the parameter for the -- - 1 / rh t_1 term in the EGARCH log-variance equation
(as given in Table I). The shape of the above EGARCH(1, 1) news impact curve is indicative of
cases with w > 0, 0 ? /3 < 1, o-> 0, 0 < a < 1, and a ? /3 < 1 and, importantly, y < 0.
impact curve to locate at a point where et -1 is positive. In Figure 2, the news

impact curve is plotted for the GJR model and the AGARCH model. We can
see that the GJR news ~impact curve captures the asymmetry in the effect of
news on volatility because it has a steeper slope in its negative side than on
Table I
Some Alternative Predictable Volatility Models
In the following model specifications, ht is the conditional variance at time t and Et- 1 is the
unpredictable return (the residual) at time t - 1. w, a, fi, and y are constant parameters in each
of the models.
Nonlinear ARCH model (Engle and Bollerslev (1986))
h, = w ? aetJ_117 + 3ht-1
Multiplicative ARCH (Mihoj (1987), Geweke (1986), Pantula (1986))

p
log(ht) = a + aai log(2-1)
i~ 1
GJR model (Glosten, Jagannathan, and Runkle (1989), Zakoian (1990))
1+a8t2 1 + yS[ 18t2 1, where St = 1 if Et < 0, St = otherwise

ht=c)+,Ph
EGARCH model (Nelson (1989))
log(ht) = + log(ht-1) + [7 + a 2/
Autoregressive Standard Deviation model (Schwert (1990))
? P
ht = c)+ Eail-t-il
Asymmetric GARCH model (Engle (1990))
ht = w + a(et- + Y)2 + 3ht-I

Nonlinear Asymmetric GARCH model
ht =w+ 3ht- + a(et-l + - 1)2
VGARCH model
ht +Pht- + a(t1/h1 + y)2
its positive side. The AGARCH, on the other hand, captures the asymmetry
by allowing its new impact curve to be centered at a positive gt- 1.
These differences between the news impact curves of the models have
important implications for portfolio selection and asset pricing. For instance,
after a major unexpected price drop, like the 1987 crash, the predictable
market volatilities given by the GARCH and the EGARCH are very different,
as implied by their news impact curves. Since predictable market volatility is
related to market risk premium, the two models imply very different market
risk premiums, and hence different risk premiums for individual stocks
under a conditional version of the capital asset-pricing model. The differences
in the news impact curves implied by the two models also have important
ht r
AnAGARCHTNwsImpatGure
AGJRNews ImpactCurve
Figure 2. The news impact curves of the AGARCH(1, 1) model and the GJR model.
The solid line is the AGARCH(l, 1) news impact curve. The dashed line is the GJR news impact
curve. The equation for the AGARCH(1,1) news impact curve is
h, = A + a(t1 +
where ht is the conditional variance at time t, E 1 is the unpredictable return at time t - 1,

A o + , o 2,o is the unconditional return standard deviation, o is the constant term, ,B is
the parameter for the ht-1 term, and a and y are the parameters in the quadratic term,
a(et 1 + Y)2, in the AGARCHvariance equation (as given in Table I). The shape of the above
AGARCH(l, 1) news impact curve is indicative of cases with to > 0, 0 </ < 1, u-> 0, 0 < a < 1
a + ,8 < 1 and y < 0.
The equations for the GJR news impact curve are
ht =A + a- 21, for Et-l > 0, and
ht = A + (a + y) 2, for 8t- < 0,

where A =o + 8 ro- 2o is the unconditional return standard deviation, w is the constant term,
,/ is the parameter for the h t_ term, a is the parameter for the 8t2 1 term, and y is the
parameter for the St-- 8t2 1 term in the GJR variance equation where S- 1 takes a value of 1
when Et-1 is negative and a value of 0 when - 1 is positive or zero. The shape of the above GJR
news impact curve is indicative of cases with to > 0, 0 < , < 1, u-> 0, 0 < a < 1, a + , < 1 and
y> O.
implications for option pricing. Stock return volatility is a major factor in

determining option prices. A significant difference in predicted volatility after
the arrival of some major news leads to a significant difference in the current
option price. Furthermore, since the EGARCH and the GARCH models imply
very different volatilities following major bad news, the dynamic hedging
strategies implied by the two sets of volatility estimates would be very
different.
All these concerns point to the need to have a correct understanding of the
impact of news on volatility. The news impact curve we have introduced is a
convenient way to summarize the effect of news on volatility implied by a
parametric model of predictable volatility. By comparing the news impact
curves of alternative predictable volatility models, we can highlight the
differences between the models. By testing whether the news impact curve of
a model offers a good fit to the data, we can understand the quality of the
model.
II. Diagnostic Tests Based on the News Impact Curve

As discussed in Section I, implicit in any choice of a volatility model is a
particular news impact curve. The standard GARCH model has a news
impact curve which is symmetric and centered at Et -I = 0. That is, positive
and negative return shocks of the same magnitude produce the same amount
of volatility. Also, larger return shocks forecast more volatility at a rate
proportional to the square of the size of the return shock. If a negative return
shock causes more volatility than a positive return shock of the same size, the
GARCH model underpredicts the amount of volatility following bad news and
overpredicts the amount of volatility following good news. Furthermore, if
large return shocks cause more volatility than a quadratic function allows,
then the standard GARCH model underpredicts volatility after a large return
shock and overpredicts volatility after a small return shock.
These observations suggest three new diagnostic tests for volatility models:
the Sign Bias Test, the Negative Size Bias Test, and the Positive Size Bias
Test. These tests examine whether we can predict the squared normalized
residual by some variables observed in the past which are not included in the
volatility model being used. If these variables can predict the squared nor-
malized residual, then the variance model is misspecified. The sign bias test
considers the variable St- a dummy variable that takes a value of one when
Et- 1 is negative- and zero otherwise. This test examines the impact of positive
and negative return shocks on volatility not predicted by the model under
consideration. The negative size bias test utilizes- the variable S- . It
focuses on the different effects that large and small negative return shocks
have on volatility which is not predicted by the volatility model. The positive
size bias test utilizes the variable S+- 1 -1 where St+ 1 is defined as 1 minus
S- 1. It focuses on the different impacts that large and small positive return
shocks may have on volatility, which are not explained by the volatility
model. Since an important piece of bad news might have a very different
impact on volatility than an important piece of good news, it is critical to
distinguish between positive and negative return shocks while examining the
effects of the magnitude of a piece of news. Using the same approach, we can
create a number of closely related tests. For example, we can examine even
more extreme values of e for particular biases by using variables, such as

St- le 2 or order statistics, such as Da(st_), where Da(,t_) is one if E
exceeds the a th percentile of the set of {E}. All these tests can be carried out
individually or jointly. They can also be applied to volatility models which are
not members of the GARCH family.
To derive the optimal form of these tests, we assume that the volatility
model under the null hypothesis is a special case of a more general model of
the following form:
log(ht) = log(hot(5o'zot)) + a Zat, (6)
where hot(50'zo0) is the volatility model hypothesized under the null; 8 is

the k x 1 vector of parameters under the null; Zot is the k x 1 vector of
explanatory variables under the null; and 8a is the m x 1 vector of addi-
tional parameters corresponding to Zat which is the m x 1 vector of missing
explanatory variables. This particular maintained hypothesis about the gen-
eral form of the model is chosen over the commonly used linear form,
ht= hot(50 Zot) + 8a'Zat in our specification tests because it encompasses
both the GARCH and EGARCH classes of variance models.2
Since the volatility model under the null hypothesis is obtained when the
parameters for the additional explanatory variables, Zat, are zeros, we can
test the model by testing whether these parameter restrictions (6a = 0) are
satisfied.
Let vt be the normalized residual corresponding to observation t under the
volatility model hypothesized. That is, vt- Et/ ho. Following Engle and
Kraft (1983), Engle (1984), and Bollerslev (1986), the Lagrange multiplier
test statistic for H *8 = O in equation (6) is.-simply a test of 8 = 0 in the
auxiliary regression
2 * -at-a + u, (7)
vt =oto +zata +t
where zt =h_18htl/8_, 4 *t holdht/88a, and ut is the residual. Both

dhtl/8o and dht/88a are evaluated at 8 = 0 and 80 (the maximum likeli-
hood estimator of 8 under Ho). Theoretically, v?2 is orthogonal to z*t by the
first-order conditions for maximum likelihood. So, if the parameter restric-
tions are met, the right-hand side variables in (7) should have no explanatory
-z
2For example, if hot(-) is the usual GARCH(1, 1) form such that: hot(80z0t) z0 ,8t [t),
/3, a]' and zot [1, ht_i, stJ-J'; and ba and Zat are the EGARCH parameters and variables:
8a = [ a}* y*] 'fj and Zat = [log(ht-1), t-t/ /t1, (It-1I/ t1 - 2-)] then the en-
compassing model is
log(ht) = log[ to + 8ht ?+ as2 1 ] + 8*log(ht-1) + y*Bt j

ht1
? a*(IBtl h
-hl - ).2/I)
So, when /3 = a = 0, the model is the EGARCH. But when /3* = -* = a* = 0, the model is the
standard GARCH(1, 1).
power at all. Thus, the test is often computed as
LM= -T*R', (8)

where R2 is the squared multiple correlation of (7), and T is the number of
observations in the sample. However, for highly nonlinear models, the numer-
ical optimization algorithm generally does not guarantee exact orthogonality.
Therefore, the procedure we use throughout this paper is to: (i) regress y' on
Zot alone, and (ii) use the residuals from this regression (which are now
guaranteed to be orthogonal to zot) in place of v' in (7). In a Monte Carlo
experiment, this procedure provides a test size closer to the one given by
asymptotic theory.
The LM test statistic is asymptotically distributed as chi-square with m
degrees of freedom when the null hypothesis is true, where m is the number
of parameter restrictions. It is asymptotically equivalent to the likelihood
ratio test and hence is also asymptotically the most powerful test.
Under our maintained encompassing model (6), dht/d8a evaluated under
the null is equal to hot:Zat* Hence, Z* =at The regression actually involves
regressing v' on a constant z*t and Zat By selecting different measures of
ZatI we construct different tests. By testing one variable at a time, we
formulate each test against a particular alternative. Finally, by allowing gat
to include several variables, we can construct joint tests. In this paper, we
consider the variables St- 1, S1 -1, and S Tests for higher order
asymmetry can also be constructed based on the same principle by consider-
ing the variables S-j St- jt_, and St 8t-j' j = 1, ... , p.
The above discussion suggests that the optimal forms of the regressions for
conducting the sign bias test, the negative size bias test, and the positive size
bias test are respectively,
vt =a+b.S7 1+I3'z*t+et, (9a)
2Vt=
a + b * S[11t_ + il3'tj;t + et, (9b)
(9c)
v2 = a + b *St+-t1 +: 13Z* + et '9c)
where a and b are constant parameters, ,B is a constant parameter vector,

and et is the residual. The sign bias test statistic is defined as the t-ratio for
the coefficient b in regression equation (9a). The negative size bias test
statistic is defined as the t-ratio of the coefficient b in regression equation
(9b). The positive size bias test statistic is defined as the t-ratio of the
coefficient b in regression equation (9c). To conduct these tests jointly, we can
consider the regression
V2 = a + b1S- l + b2Sj71rt_ +b3S1+ 1 + lZ*t + et (10)
where a-, b1, b2, and b3 are constant coefficients, ,B is a vector of constant
coefficients, and et is the residual. The t-ratios for b1, b2, and b3 are the sign
bias, the negative size bias, and the positive size bias test statistics, respec-
tively. The joint test is the LM test for adding the three variables in the
variance equation under the maintained specification (6). The test statistic is
equal to T times the R-squared from this regression. If the volatility model
being used is correct, then b1 = b2= b3 = 0, /3 = 0 and et is i.i.d. Thus, the
t-statistics and the LM test statistic have the standard limiting distributions.
In particular, the LM test statistic follows a chi-square distribution with
three degrees of freedom. If z * is not included in (10), the test will be
conservative; the size will be less than or equal to the nominal size, and the
power may be reduced.
These diagnostic test statistics can also be used as summary statistics on
the raw data to explore the nature of time-varying volatility in the data
series, without first imposing a volatility model. In this case, et and vt would
simply be defined as follows:
et - Yt - 11 (lla)
vt 8t/o, (I1lb)
where ,tu and cr are the unconditional mean and standard deviation of Yt,
respectively. Since multiplying the dependent variable in the regression by a
constant will not change the results when the null is constant variance, we
can simply use 8 2 instead of v2 as the dependent variable in the regression.
The t-statistics and the LM test statistic, which are both scale invariant, will
give us the three individual tests and the joint test.
To examine the performance of the test statistics, we conduct a small
Monte Carlo experiment. From this experiment we find that the size is rather
close to the nominal value and the power to detect departures is reasonable,
at least for the larger (1000) sample size. Since the sample size of our daily
stock return series is rather large (2532), we use the asymptotic critical
values in our empirical section.
The Monte Carlo experiment for checking the size of the tests is based on a
GARCH(1, 1) data-generating process, as in (12):
Yt ?t
ht= w+ /38h-1 + a t-1
et= vh,t , (12)
where vt i.i.d. N(0, 1), and o, /3, and a are constant parameters. Three
sets of parameter values are considered: (1) model H (for high persistence),
where Go, /3, ;) = (0.01, 0.9, 0.09) and a + / = 0.99; (2) model M (for
medium persistence), where (to, /3, a) = (0.05, 0.9, 0.05) and a + /3= 0.95;
and (3) model L (for low persistence), where (Go, /3, a) = (0.2, 0.75, 0.05) and
a + /3= 0.8. For each model, samples of size 100 and 1000 are drawn with
10,000 replications. For each replication, a GARCH(1, 1) model is estimated
and the sign bias test, the negative size bias test, the positive size bias test,
and the joint test are conducted. The actual rejection frequencies based on the
1, 5, and 10 percent critical values under the asymptotic distribution are

reported in Table II.
As is usual in Monte Carlo experiments with nonlinear models, the conver-
gence criterion is not satisfied in a fraction of cases.3 In such cases, we add
new replications to ensure that we have 10,000 converged replications. The
results reported in Table II are based on the converged replications. Approxi-
mately 7.2 percent of all replications do not converge. The test statistics from
these nonconverged replications are analyzed separately as if they had
converged. The results, not reported, are similar to those for the converged
replications.
The simulated size of the test is quite close to the proposed size under the
asymptotic distribution of the test statistics for the larger sample size of
1000. The results for the smaller sample size of 100 are also reasonable. For
the proposed 1 percent level, there are only one or two cases where the actual
rejection frequency is about 0.6 percent. Since the actual rejection frequencies
are quite close to the proposed ones for the 5 percent and the 10 percent
levels, these exceptions might be due to the small number of observations
from the tails of the distribution.
The Monte Carlo experiment for checking the power of the tests is based on
an EGARCH(1, 1) data-generating process as in (13), and a data-generating
process, as in (14), which corresponds to the GJR model. We call these
processes model E and model G, respectively.
Model E
Yt=t
In ht =-0.23 + 0.9 *-lnht- 1 + 0.25 * [ IV2_1l -0.3 *vt- 1]
et = h vt, (13)
where vt i.i.d. N(0, 1) and
Model G
Yt=t
ht = 0.005 + 0.7* ht-1 + 0.28* [t- I - 0.23 St_]2
et = Xl- vt, (14)

where vt i.i.d. N(0, 1).
Again, for each model, samples of size 100 and 1000 are drawn with 10,000
replications. For each replication, a GARCH(1, 1) model is estimated and the
tests are conducted. The actual rejection frequencies based on the 1, 5, and 10
percent critical values under the asymptotic distribution are reported in
Table III for the joint test. As before, Table III is based on only the converged
3 The convergence criterion we use in the simulation experiment is that the R-squared from a
regression of a vector of ones on the scores is less than or equal to 0.0000001.

Table II
Actual Rejection Frequencies When the Null is True
(The Simulated Size)
To check the size of the tests, we perform a Monte Carlo experiment based on the GARCH(1, 1)
data-generating process below. et is the unpredictable return (the residual) at time, t, ht is the
conditional variance at time t, vt is the normalized residual at time t which is generated from a
standard normal random number generator, Yt is the simulated return at time t (which, for
simplicity is assumed to be entirely unpredictable and hence is equal to et), and cv, /3, and a are
constant parameters.
Yt = t; et--:: h vt, where vt -i.i.d. N(0, 1)
ht= co +? ht-1 + a- 2t- 1
Three sets of parameter values are considered. They are: (1) model H (for high persistence),
where (co, ,3, a) = (0.01, 0.9, 0.09) and a + /3 = 0.99; (2) model M (for medium persistence),
where (to, /3, a) = (0.05, 0.9, 0.05) and a + /3 = 0.95; and (3) model L (for low persistence), where
(c, /3, a) = (0.2, 0.75, 0.05) and a + /3 = 0.8. For each model, samples of size 100 and 1000 are
drawn with 10,000 replications. For each replication, a GARCH(1, 1) model is estimated and the
tests are conducted. This table reports the actual rejection frequencies based on the 1, 5, and 10
percent critical values under the asymptotic distribution.
Model and Sample Size

H H M M L L
Test 1000 100 1000 100 1000 100
Actual Rejection Frequencies (%)

Sign bias 1% 1.11 1.04 0.92 1.00 1.11 1.18
5% 4.94 4.67 4.70 4.34 5.22 5.21
10% 9.99 10.19 9.78 8.93 10.07 9.19
Negative size bias 1% 0.72 1.40 1.03 1.47 0.87 0.93
5% 4.27 5.11 5.25 5.70 4.79 4.80
10% 8.79 10.58 10.35 10.52 9.45 9.26
Positive size bias 1% 0.96 0.90 0.96 0.64 0.95 0.94
5% 4.33 5.18 4.51 5.01 5.41 5.08
10% 8.95 10.18 9.26 10.61 10.11 9.74
Joint 1% 0.82 10.20 0.94 0.61 0.90 0.85
5% 4.45 4.87 4.98 4.72 4.69 4.72
10% 9.02 10.05 9.38 10.78 9.50 9.81
replications. The nonconverged replications, not reported, yield similar re-

sults.
The power of the joint test is reasonably good for the larger sample size of
1000. Based on the 5 percent asymptotic critical value, the test rejects the
null hypothesis of no asymmetry 67.05 percent of the time with data gener-
ated from model E, and 41.5 percent of the time with data generated from
model G. However, the power of the joint test is weak when the sample size is
small. This weakness is expected as both asymmetric effects and time-
varying variance are hard to detect in small samples. Furthermore, the power
of the individual tests is weak due to collinearity between the misspecifica-
Table III
Actual Rejection Frequencies When the Null is Not True
(The Simulated Power)
To check the power of the joint test, we perform a Monte Carlo experiment based on the two
data-generating processes below. et is the unpredictable return (the residual) at time t, ht is the
conditional variance at time t, vt is the normalized residual at time t which is generated from a
standard normal random number generator, and Yt is the simulated return at time t (which, for
simplicity is assumed to be entirely unpredictable and hence is equal to et).
Model E
= = where i.i.d. 1)
Yt ?t; ?t rh *
vt, Vt N(0,
log(ht) = -0.23 +0.9 log(ht -1)+ 0.25 [V 2 1 - 0.3 vt-l].
Model G
= = Xh where i.i.d.
Yt ?t; ?t *
Vt, Vt N(0, 1)
ht = 0.005 + 0.7 ht-1 + 0.28* [It11 - 0.23 t_1]
for each model, samples of size 100 and 1000 are drawn with 10,000 replications. For each
replication, a GARCH(1, 1) model is estimated and the tests are conducted. This table reports the
actual rejection frequencies based on the 1, 5, and 10 percent critical values under the asymp-
totic distribution.
Model and Sample Size

E E G G
1000 100 1000 100
Actual Rejection Frequencies (%)
Joint Test 1% 40.29 1.70 19.59 1.61
5% 67.05 7.51 41.50 6.04
10% 78.64 14.42 55.18 12.83
tion indicators, particularly that of the sign bias test and the negative size
bias test.
III. A Partially Nonparametric News Impact Model

An alternative approach to estimating the news impact curve is to imple-
ment a nonparametric procedure which allows the data to reveal the curve
directly. Several approaches are available in the literature, including notably,
Pagan and Schwert (1990) and Gourieroux and Monfort (1992). In both cases
the methods are developed only for low order autoregressive variance equa-
tions and in the fourier case, only OLS estimation is considered. Gourieroux
and Monfort essentially specify a histogram for the response of volatility to
lags of the news which they estimate by maximum likelihood. In their most
successful model however, they introduce a GARCH term to capture the long
memory aspects.
Here we introduce a new partially nonparametric model. This will allow

consistent estimation of the news impact curve under a range of conditions. It
is labeled partially nonparametric because the long memory in the variance
equation is given by a parametric component.
Let the range of {et} be divided into m intervals with break points Ti. Let
m - be the number of intervals in the range where Et- 1 is negative. Also, let
m+ be the number of intervals in the range where Et- 1 is positive, so that
m = m++ m-. Denote these boundaries by the numbers {T-rm- III T-1,
T? T1 ... XTm+1. These intervals need not be equal size, nor do we need the
same number on each side of To. For convenience and the ability to test
symmetry, we select To = 0.
If we define
Pit 1 if et > Ti
= 0 otherwise, and
Nit I if Ct < T_i
= 0 otherwise, (16)
then a piecewise linear specification of the heteroskedasticity function is
m+ m-
ht + 3h t 1 + E OiPit-( -- ) + E Nit- -1 i), (17)

i=O i=O
where co, ,3, Oi (i = 0,..., m+), and 8i (i = 0,..., m-) are constant parame-
ters. This functional form, which is really a linear spline with knots at the
Ti's, is guaranteed to be continuous. Between 0 and z1 the slope is 00 while
between -1 and T2 it is 00 + 01, and so forth. Above Tm+, the slope is the sum
of all the 0's. If the partial sums at each point are of the same sign, the shape
of the curve is monotonic.
To obtain better resolution with larger samples, we increase m. This is an
example of the method of sieves approach to nonparametric estimation. A
larger value of m can be interpreted as a smaller bandwidth, which will give
lower bias and higher variance to each point on the curve. If m is increased
slowly as a function of sample size, the procedure should asymptotically give
a consistent estimate of any news impact curve. However, the rate of conver-
gence and the standard errors may both be different from standard maximum
likelihood results. Conversely, if m is held fixed, the estimator produces a
consistent estimate of the news impact curve only if (17) is correctly specified.
In such cases, the standard errors are given in their usual form.
We should point out that although the specification in (17) is capable of
generating a wide range of news impact curves, it is very simple with respect
to the impact of older information. All information is assumed to decay in an
exponential fashion with decay rate /3. Other terms could be added to the
model, but they would substantially increase the computational complexity.
Two simple approaches to choosing the Ti's could be used. The <r's could be
unequally spaced, based on the order statistics, or equally spaced. In our
example, we use equally spaced bins with break points at or i for i = 0, +1,
+2, +3, +4, where o- is the unconditional standard deviation of the depen-
dent variable. Thus:
m+ m
ht= + 8ht-+ + Pit- - i E 8 iit-4(t - + i). (18)
i=0 i=O
With m+= m-= 4, there are ten coefficients in the news impact curve.
Figure 3 gives an example of the graph of a partially nonparametric, or PNP,
news impact curve.
IV. Estimation of Japanese Stock Return Volatility:

1980 to 1988
To compare and demonstrate the empirical properties of the GARCH(1, 1)
with some of the above-mentioned leverage or asymmetric volatility models,
we apply the models to the daily returns series of the Japanese TOPIX index.
The data were obtained from the PACAP Databases provided by the Pacific
Basin Capital Market Research Center at the University of Rhode Island. In
this section, we report or estimation and testing results for the parametric
models for the full sample period from January 1, 1980 to December 31, 1988.
In the next section, we estimate conditional volatility and the news impact
curve using a nonparametric approach, and compare the news impact curve
obtained from the nonparametric method to those obtained from the various
parametric volatility models. In Section VI, we check the robustness of our
results by reestimating some of our models using a shorter sample period
from January 1, 1980 to September 30, 1987.
Since our focus is on the conditional variance, rather than the conditional
mean, we concentrate on the unpredictable part of the stock returns, as
obtained through a procedure similar to the one in Pagan and Schwert (1990).
The procedure involves a day-of-the-week effect adjustment and an autore-
gressive regression which removes the predictable part of the return series.
Let Yt be the daily return of the TOPIX index for day t. We first regress Yt
on a constant and five day-of-the-week dummies (for Tuesday through Satur-
day) to get the residual, ut. The ut is then regressed on a constant and
ut_ 1 ... ut6 to obtain the residual, Et, which is our unpredictable stock
return data.
The results for the above adjustment regressions and some summary
statistics for our unpredictable stock return series are reported in Table IV.
From the Ljung-Box test statistic for twelfth-order serial correlation for the
levels, we find no significant serial correlation left in the stock returns series
after our adjustment procedure. The coefficients of skewness and kurtosis
both indicate that the unpredictable stock returns, the 8 's, have a distribu-
tion which is skewed to the left and significantly flat tailed. Furthermore, the
Ljung-box test statistic for twelfth-order serial correlations in the squares
strongly suggests the presence of time-varying volatility. The sign bias,
negative size bias, and positive size bias test statistics introduced in the
ht
L
? gt-1
Figure 3. A partially nonparametric news impact curve. The partially nonparametric
news impact curve is a piecewise linear function. The equation for a typical partially nonpara-
metric news impact curve is
m+ m
ht =A + E OiPat-(et-i - io) + E 81Nit_j(8t_j + io).
i=O i=O
where ht is the conditional variance at time t, et-1 is the unpredictable return at time t - 1,
A w+ f * 2-o is the unconditional return standard deviation, w is the constant term, and 3
is the parameter for the ht 1 term in the PNP variance equation (equation 18 in the text).
Oi(i = 0,..., m+) and 8i(i = 0,..., m-) are constant parameters. The above partially nonpara-
metric news impact curve is indicative of cases with I I> Ioi1,1I I> 15i- 11,and IOiI > IOi- 11,for
all i.
previous section are also computed. The sign bias and negative size bias tests
are both highly significant. The positive size bias test is not particularly
significant, although if the size term were dropped it would be significant.
These statistics strongly indicate that the value of Et -1 influences current
volatility: positive return shocks appear to increase volatility regardless of
the size, while large negative return shocks cause more volatility than small
ones.
Using the unpredictable stock returns series as the data series, we esti-
mate the standard GARCH(1, 1) model, as well as five other parametric
models from the first section which are capable of capturing the leverage and
size effects. The five additional models are: the Exponential-GARCH(1, 1), the
Asymmetric-GARCH(1, 1), the VGARCH(1, 1), the Nonlinear-Asymmetric-
u.
Yt This
takes
ut a is Therereturn
Then TOPIX
variables Research Schwert
theare
= u table
correlation.
adjustment data.
Yt for is index
= value
rate2532 The (1990).
of Center andreports
-0.0002
(0.016) 1 of
at to
+ on Tuesday, daily First, the
[ regression regressed
Joint Sign -0.0119
(0.039) theTOPIX
Mean
- the return on the
Positive
Negative Number
Variance (0.02) return a remove
test bias 0.1231 of results
Ljung-Box
Ljung-Box
Coefficient
Coefficient
of index
p-Value]
sizesize the daily
the of
test of of 0.0907 (standard
(12)(12) -ut1 (0.055)
- Wednesday, an
biasbias data
University
forfor - constant
errors TOPIX of are return
testtest thethekurtosis TUEt corresponding andy
observations
skewnessobservations
Summary + in
(0.02)
0.0802 in is predictable
* index Rhodethe
Thursday, adjustment
levels the obtainedpart Mean
squares 0.2025 first
Ut-2 (0.055) weekday from of
Statistics
+
Friday, Island. six regressed
from
for and daysample
parentheses). the
a Thethe on procedure
WEDt and lagsa
the (0.02)
0.0456 + Day-of-the-Week to
- t-1 of
Autocorrelation return Table
value period. u Adjustment
to PACAP
sample
Ut_3 Effect Ljung-Box
of to IV
0.0953
(0.055) remove
constant
- . 0 Saturday
day series.
(12) t. the
and
periodobtain
Unpredictable THUt is The
(0.02) Adjustment
0.0526 is
- + Databases
thefive
the TUEt,
Adjustmentotherwise.
Stock from
respectively.
Ut_4 ut Regressions
+ 0.1100
(0.055)- procedure
is WEDt, providedresidual
Each is day-of-the-week
Returns Ljung-Box
the by ?,
(0.02) FRIt of January
0.0772
- + THUt,1, the day-of-the-week
- effects
2532 these which
Ut_5 is analogous
1.56
139.00 -20.30
-6.26 - 0.1629
(0.060)
statisticsFRIt, 1980
residual
Pacific to from
[0.000] 1.3251 1.8927
71.3511
406.6700 0.6397
0.0000 to
forof ourdummies
and thethe
(0.02) SATt the Basin to
+ one
SATt get daily
in
0.0706ut-6 ut December
day-of-the-week
+ are Capitalthe
et twelfth-order 31, unpredictable return
Pagan
of
serial dummies1988.
day-of-the-week
dummy residual
stock
Market andthe
GARCH(1, 1), and the Glosten-Jagannathan-Runkle (GJR) model. The esti-

mation is performed using the Bollerslev-Wooldridge quasi maximum likeli-
hood approach. The adequacy of these volatility models is then checked using
the sign bias, the negative size bias, and the positive size bias tests, as well
as the commonly used Ljung-Box test for serial correlation in the squared
normalized residuals. The estimation and diagnostic results for each of these
models are presented in Table V. As a convention, the asymptotic standard
errors are reported in parentheses () and the Bollerslev-Wooldridge robust
standard errors are reported in brackets [ ].
The estimation results in Table V indicate that the parameters correspond-
ing to the et-,/ ht 1 term in the EGARCH, the constant in the quadratic
form in the AGARCH, the constant in the quadratic form in the VGARCH,
and the ht1 term in the NGARCH, are all significant and negative using
both standard and robust standard errors. The parameter corresponding to
the S- 1 2- 1 term in the GJR is significant and positive using both standard
and robust standard errors. All these results are consistent with the hypothe-
sis that negative return shocks cause higher volatility than positive return
shocks. We can also see that the standard GARCH(1, 1) has a lower log-likeli-
hood than most of the leverage or asymmetric models with the exception of
the VGARCH. The GJR, the NGARCH, and the EGARCH yield .the highest
log-likelihood.
In diagnostic checks, the Ljung-Box statistic for twelfth-order serial corre-
lations in the squared normalized residuals is significant at the 5 percent
level for only the VGARCH. On the other hand, the negative size bias test
statistics, as well as the joint test statistics, are significant for all models,
with the EGARCH and the GJR being only marginally rejected by the joint
test. All the models seem to have some problem in capturing the correct
impact of news on volatility. Furthermore, the results indicate that the
Ljung-Box test, which is commonly used as a specification check for volatility
models, does not have much power in detecting misspecifications related to
the leverage or asymmetric effects.
Overall, the Exponential GARCH model and the GJR model seem to
outperform all other models in capturing the dynamic behavior of the
Japanese stock returns, with the GJR model having a higher log-likelihood.
To further our understanding of these different volatility models, some
summary statistics, including the mean, standard deviation, minimum, maxi-
mum, skewness, and kurtosis, are produced for each of the estimated condi-
tional variance series. They are reported in Table VI.
The conditional variance series produced by the best models, the EGARCH
model and the GJR model, have the highest variation over time. The esti-
mated conditional variance ranges from a low of 0.0491 to a high of 485.27,
compared to 0.0842 and 90.83 under the standard GARCH model. The
standard deviation of the EGARCH conditional variance, 10.555, is more
than three times that of the standard GARCH model and twice that of the
squared residual itself. The EGARCH conditional variance also has a much
more skewed and flat tailed distribution than the other conditional variance
series and the squared return shocks. The fact that the unconditional vari-
ance of the EGARCH conditional variance is larger than the unconditional
variance of the squared residual can actually be interpreted as evidence
against the EGARCH model. To see this point, note that we can always write
? = ht + (e2 - ht). (19)
Since ht = - ( 82), the two terms on the right-hand side of (19) are uncorre-
lated. So
Var(e2) = Var(ht) + Var(82 - ht). (20)
Hence, Var(c 2) 2 Var(ht) if ht is correctly specified and if the unconditional
variances exist. As we can see, the EGARCH model fails this test.4
V. Partially Nonparametric ARCH Estimation

We now estimate the news impact curve by fitting a partially nonparamet-
ric model of the form given in (18). The exact specification and the estimation
results are reported in Table VII. The specification is a piecewise linear
model with kinks at et> 1 equal to 0, o-, 2o-, 3o-, and 4u. If we compare the
values of the coefficients corresponding to the terms Pit(et-1 - io-),i = 0, 1,
2, to their counterparts Nit(et-, + io-),i = 0, 1, 2, we can see that negative
- l's cause more volatility than positive e l's of equal absolute size. More-
over, the rate of increase in volatility, as we move towards et l's with bigger
absolute magnitude, is higher for the negative e's than for the positive ones.
This finding suggests a sign or asymmetric effect, as well as a size effect, that
differs for negative and positive e's. The estimated parameter values for the
terms Pit(et-, - ior) and Nit(et-, + io-) for i = 3, 4 have somewhat unex-
pected signs and magnitudes. Since these terms are for the extreme e's, they
might be driven by only a few outliers. Indeed, even though they are
significant using the traditional asymptotic standard errors, they are all
insignificant using the Bollerslev-Wooldridge robust standard errors. Thus,
the nonparametric estimation results indicate that the true slope of the news
impact curve is probably steeper on the negative side.
To cornpare the news impact curve of the nonparametric model with those
implied by the various volatility models, we compute the implied volatility
level for each model at several prespecified values for et_, assuming that
ht_ = cn = 0.63966. The results are summarized in Table VIII.
If we confine ourselves to et- 1 in the range (-2.5, 2.5), we see that, relative
to the EGARCH model, the standard GARCH model tends to understate ht
for large negative et_'s and overstate h for large positive et -'s. These
results are also true for the AGARCH, VGARCH, and NGARCH models. Of
all six parametric models, the EGARCH and the GJR have news impact
curves closest to the one suggested by the nonparametric estimation. How-
ever, if we consider the very extreme values for et-,, we see that the
4We thank Rob Stambaugh for pointing this out to us.

Table V
Estimation Results and Diagnostics
This table reports the estimation and diagnostic test results of various predictable volatility
models for the daily return of the TOPIX index. Day-of-the-week effects and a predictable
component in the daily return series have been removed. The estimation is performed by the
method of quasi maximum likelihood using the BHHH numerical optimization algorithm. The
sample period is from January 1, 1980 to December 31, 1988. In the estimation results part of
the table, the numbers in parentheses () are the asymptotic standard errors and the numbers in
brackets [] are the Bollerslev-Wooldridge robust standard errors. In the test results part,
Ljung-Box (12) is the Ljung-Box statistics for twelfth-order serial correlations in the squared
normalized residuals. Also, one and two asterisks indicate significance at the 5 and 1 percent
levels respectively.
ht is the conditional variance on day t and ?- is the unpredictable return on day t - 1. The
unpredictable return is obtained from the adjustment regressions in Table IV.
Estimation Results
GARCH(1, 1)
ht = 0.0238 + 0.6860 - ht 1 + 0.3299 -t-, 1

(0.003) (0.011) (0.008)
[0.005] [0.059] [0.097]
logL =-2356.03
EGARCH(1, 1)
log(ht) = -.0668 +0.9012 log(ht 1) + 0.4927 [ -0.1450

[ht-1 IT]Vht _
(0.010) (0.007) (0.016) (0.011)
[0.020] [0.022] [0.104] [0.048]
logL = -2344.03
AGARCH(1, 1)
ht = 0.0216 + 0.6896 ht 1 + 0.3174 - (et 1- 0.1108)2

(0.003) (0.012) (0.009) (0.017)
[0.005] [0.055] [0.088] [0.030]
logL = -2345.12
VGARCH(1, 1)
ht = 0.0192 + 0.6754 - ht- 1 + 0.1508 - (et- / - 0.1458)2

(0.005) (0.014) (0.004) (0.031)
[0.013] [0.071] [0.047] [0.052]
logL =-2424.63
NGARCH(1, 1)
ht = 0.0199 + 0.7253 ht- 1 + 0.2515 - (et 1 - 0.2683- / t_1)2

(0.002) (0.010) (0.008) (0.036)
[0.005] [0.060] [0.083] [0.061]
logL = -2335.34
Table V-Continued
Estimation Results
GJR
ht = 0.0241 + 0.7053 ht -1 + 0.1672 t21 + 0.2636 . S-le
(0.003) (0.013) (0.018) (0.020)
[0.005] [0.045] [0.036] [0.102]
logL= -2333.11
Diagnostic Test Results

Model Ljung-Box(12) Sign Bias Negative Size Bias Positive Size Bias Joint Test
GARCH(1, 1) 12.18 -0.30 -3.22* -0.59 16.05**

EGARCH(1, 1) 12.14 -0.50 -2.79* 0.01 8.38*
AGARCH(1, 1) 11.43 -0.77 -3.21* -0.58 12.57**
VGARCH(1, 1) 26.64* -2.38* -4.75* -0.32 23.85**
NGARCH(1, 1) 11.19 -1.06 -3.22* -0.61 11.92**
GJR 12.37 -0.67 -2.59* -0.54 7.91*
Table VI
Summary Statistics of the Conditional Variance Estimates
This table reports the summary statistics of the estimated conditional variances of the daily
TOPIX returns from various predictable volatility models. The TOPIX data are obtained from the
PACAP Databases provided by the Pacific Basin Capital Market Research Center at the
University of Rhode Island. The sample period is from January 1, 1980 to December 31, 1988.
The statistic "Skew." is the coefficient of skewness and the statistic "Kurto." is the coefficient of
kurtosis. For a standard normal random variable, the value of the coefficient of skewness is 0
and the value of the coefficient of kurtosis is 3.
st2 is the squared unpredictable return obtained from the adjustment regressions in Table IV.
htGARCH, htEGARCH, htAGARCH, htVGARCH, htNGARCH, and htGJR, are conditional
variances estimated from the GARCH(1, 1), EGARCH(1, 1), AGARCH(1, 1), VGARCH(1, 1),
NGARCH(1, 1), and GJR models in Table V.
Mean Std. Dev. Min. Max. Skew. Kurto.

2
0.6397 5.366 2.8e-8 236.60 37.038 1543.47
htGARCH 0.7483 3.124 0.0842 90.83 21.173 523.20
htEGARCH 0.8669 10.555 0.0491 485.27 40.843 1799.19
htAGARCH 0.7367 3.047 0.0807 87.78 21.014 515.64
htVGARCH 0.5243 0.674 0.0943 20.16 15.279 373.39
htNGARCH 0.6961 2.574 0.0847 64.47 18.215 392.61
htGJR 0.7561 3.492 0.0885 104.21 21.950 559.01
EGARCH and the GJR are indeed very different. In fact, because of its
exponential functional form, the EGARCH model produces a ridiculously high
ht of 1225.1 for an Et- 1 equal to -10 which is about three thousand times the
value of the unconditional variance. Since the Japanese stock market volatil-
ity after the 1987 crash was not that high, the EGARCH model might be too
Table VII
Partially Nonparametric Estimation
This table reports the estimation results of a partially nonparametricARCH volatility model for
the daily return of the TOPIXindex. Day-of-the-weekeffects and a predictable componentin the
daily return series have been removed. The estimation is performed by the method of quasi
maximum likelihood using the BHHH numerical optimization algorithm. The raw data are
obtained from the PACAP Database provided by the Pacific Basin Capital Market Research
Center at the University of Rhode Island. The sample period is from January 1, 1980 to
December 31, 1988. The numbers in parentheses () are the asymptotic standard errors and the
numbers in brackets [ ] are the Bollerslev-Wooldridgerobust standard errors.
ht is the conditional variance on day t and et- 1 is the unpredictable return on day t - 1
which is obtained from the adjustment regressions in Table IV. a is the unconditional standard
deviation of -t, Pit(i = 0, 1, 2, 3, 4) is a dummy variable that takes a value of 1 if et is greater
than i (o- and a value of 0 otherwise, and Nit(i = 0, 1, 2, 3, 4) is a dummy variable that takes a
value of 1 if et is less than - i *or and a value of 0 otherwise.
Partially NonparametricARCH(PNP) (logL = -2310.72)

ht= 0.0039 + 0.8015 *h
(0.002) (0.013)
[0.012] [0.040]
+ 0.0897 *Po t- let-1 + 0.2269* Pit-1(8t - f )
(0.014) (0.088)
[0.043] [0.172]
+ 0.6666 *P2 t_1(t1 - 2a) - 3.7664 P3t-1(et-1 o-3 )
-3
(0.353) (1.096)
[0.720] [1.991]
+ 3.6915 *P4t_1(et1 - 4cr)
(1.540)
[2.327]
-0.1536 *No t-l ?t-, - 0.3312 N1t-(et-1 + a)
(0.014) (0.093)
[0.053] [0.203]
-3.1194 N2t_1(-t_1 + 2cr) + 7.3481 N3t-1(et-1 + 4or)
(0.278) (0.959)
[4.143] [8.699]
-5.4904 N4t_1(st-1 + 4(r)
(1.679)
[5.769]
extreme in the tails. Consequently, the GJR model, which also has a higher
log-likelihood than the EGARCH, might be a more reasonable model to use.
VI. Subsample Robustness Check

To judge the sensitivity of our results to the extreme observations around
the 1987 crash, we repeat part of our analysis for the subsample period from
January 1, 1980 to September 30, 1987, excluding the crash. The results for
Table VIII
The News Impact Curves
This table gives the value of the current volatility, ht as a function of the past return shock,
Et- 1, holding past conditional variance, ht 1, fixed at its unconditional mean level. The values
are given for various predictable volatility models for the daily return of the TOPIX Index. The
TOPIX data are obtained from the PACAP Database provided by the Pacific Basin Capital
Market Research Center at the University of Rhode Island. The sample period is from January 1,
1980 to December 31, 1988.
GARCH EGARCH AGARCH VGARCH NGARCH GJR PNP

8t- 1 ht ht ht ht ht ht ht
- 10.0 33.450 1225.100 32.910 24.580 26.730 43.550 12.793

-5.0 8.710 22.739 8.753 6.623 7.323 11.245 4.061
-2.5 2.524 3.098 2.626 2.065 2.337 3.167 3.533
-2.0 1.782 2.079 1.877 1.507 1.717 2.198 2.470
- 1.0 0.793 0.937 0.854 0.745 0.855 0.906 0.736
-0.5 0.545 0.629 0.581 0.541 0.612 0.583 0.593
0.0 0.463 0.422 0.467 0.454 0.495 0.475 0.517
0.5 0.545 0.525 0.511 0.486 0.504 0.517 0.561
1.0 0.793 0.652 0.714 0.635 0.639 0.642 0.652
2.0 1.782 1.007 1.596 1.287 1.286 1.144 1.235
2.5 2.524 1.251 2.275 1.790 1.797 1.520 1.348
5.0 8.710 3.710 8.050 6.073 6.243 4.655 1.038
10.0 33.453 32.616 31.503 23.480 24.566 17.195 5.579
the day-of-the-week and autocorrelation adjustments, as well as some sum-

mary statistics for the residuals, are reported in Table IX. The Ljung-Box(12)
statistic for the squared residuals strongly suggests the existence of autocor-
relation in the squared residuals (and hence time-varying conditional volatil-
ity of the autoregressive type). The sign bias test statistic is significant and
the two size bias test statistics are also highly significant, with the negative
size bias test statistic having a higher value. These results indicate a size
effect of news, which is stronger for bad news than for good news. Given the
superiority of the EGARCH and the GJR model over the other asymmetric
volatility models, we repeat our estimation for the standard GARCH, the
EGARCH and the GJR models only. The results are reported in Table X.
Several results in Table X are worth special notice. First, the parameter
corresponding to the et- 1/ h term in the EGARCH and the parameter
corresponding to the S[- 1e 2 1 term in the GJR are both highly significant,
even using the Bollerslev-Wooldridge t-test. Second, the joint test is signifi-
cant for the standard GARCHI model but not for the EGARCH and GJR
models. The log-likelihoods of both the EGARCH and the GJR models are
substantially higher than the log-likelihood of the standard GARCH model.
All of these results point to the presence of a leverage effect in the data. In
terms of the size effect, the positive size bias test is insignificant for all three
models, indicating that there is not much size effect for positive return
u.
Yt This
takes
a is Therereturn
Then TOPIX
variables Research Schwert
ut=
correlation.
adjustment theare u table
for data.
is
Yt value index
= rate2192 The (1990).
of Center andreports
-0.0002
(0.014) 1 of
at to
+ on Tuesday,daily First, the
[ 0.0162 regression regressed
Sign (0.035) theTOPIX
Joint Mean - the return on the
Positive Number (0.02)
test Negative
bias Variance 0.2491 of return a remove
results
Ljung-Box
Ljung-Box
Coefficient
Coefficient
of index
p-Value]
sizesize of of 0.1088 (standard the daily
the of Mean
test (0.049)
(12)(12) ut-1 Wednesday, an
biasbias data
University
forfor - constant
of are return
TUEt errors TOPIX
testtest thethekurtosis corresponding andy
observations
skewnessobservations + in
(0.02) in Rhode is predictable
0.0614- Thursday,
index the adjustment
levels
0.1412 this obtained part Adjustme
squares (0.049)
- first
Ut-2 weekday from Island. of
-
Friday, six regressed
from
and day
parentheses). the
WEDt Thethe on procedure
a andt precrash
Summary
(0.02) + Day-of-the-Week- lagsa to
0.0275
1 of return Table
Autocorrelation value to u
ut-3 0.0682
(0.049) Effect Ljung-Box
of
sample
PACAP
to Regressio
IX
- remove
constant
Statistics
+ 0 Saturday
daysubsample series.for
for (12) t. the
and
periodobtain
e THUt is The
(0.02) Adjustment
0.0496 + is the
Databases
period. thefive
the
Adjustment TUEt,
otherwise. from
ut-4 respectively.
0.1008
(0.049) ut
+ - procedure
is WEDt, provided
residual
Each is day-of-the-week
FRIt Ljung-Box
the of
by e,
January Precrash
(0.02) + 1,
THUt, the day-of-the-week
- effects
2192 these which
0.0019Ut5 1980 is analogous
77.00
5.6413.90
-2.30 0.1411
(0.053) statistics
residual
FRIt, Period
[0.000] 540.8552
8.7135
5.9197 0.4302
0.09470.0000
- -
of to Pacific to from
for ourdummies
and thethe
(0.02) SATt the Basin to
+ one
SATt get daily
0.0490ut-6
Ut in
September
day-of-the-week
+ are Capitalthe
et
twelfth-order 30, unpredictable return
Pagan
of
serial dummies1987.
day-of-the-week
dummy stock
Market andthe
residual
Table X
Estimation Results and Diagnostics for the Precrash Period
This table reports the estimation and diagnostic test results of various predictable volatility
models for the daily return of the TOPIX index. Day-of-the-week effects in the daily return series
have been adjusted for and the predictable component in the daily return series has been
removed. The estimation is performed by the method of quasi maximum likelihood using the
BHHH numerical optimization algorithm. This precrash subsample period is from January 1,
1980 to September 30, 1987.
In the estimation results part of the table, the numbers in parentheses () are the asymptotic
standard errors and the numbers in squared brackets [ ] are the Bollerslev-Wooldridge robust
standard errors. In the test results part, Ljung-Box(12) is the Ljung-Box statistics for twelfth-
order serial correlations in the squared normalized residuals. Also, one asterisk indicates
significance at the 5 percent level.
ht is the conditional variance on day t and et 1 is the unpredictable return on day t - 1. The
unpredictable return is obtained from the adjustment regressions in Table IX.
Estimation Results
GARCH(1, 1)
= 0.0129 + 0.8007 + 0.1829 -2
(0.002) (0.013) ht-1 (0.014)
[0.003] [0.025] [0.026]
logL = -1829.50
EGARCH(1, 1)
t
log(ht) = -.0350 + 0.9579 log(ht 1) + 0.2955 - ] - 0.0615
(0.007) (0.005) (0.019) (0.010)

[0.014] [0.011] [0.037] [0.024]
logL = -1822.30
GJR
ht = 0.1093 + 0.8181 ht 1 + 0.1130 e21 + 0.1048 S-

le 2
(0.002) (0.012) (0.014) (0.019)
[0.003] [0.021] [0.023] [0 038]
logL = -1819.23
Diagnostic Test Results

Model Ljung-Box(12) Sign Bias Negative Size Bias Positive Size Bias Joint Test
GARCH(1, 1) 13.23 -0.01 -1.98 -0.33 8.15*

EGARCH(1, 1) 21.30* -0.46 -2.59* 0.47 7.38
GJR 15.47 -0.19 -1.78 0.61 3.97
shocks. However, the negative size bias test statistics are marginally signifi-
cant for the standard GARCH and significant for the EGARCH, but insignifi-
cant for the GJR. The failure of the EGARCH model to capture the size effect
is probably due to two factors: the quadratic function dominates the exponen-
tial for small r's, and the Japanese stock market was quite calm before the
1987 crash. The only model that seems to do well in both normal and
abnormal times is the GJR model, which also has the highest log-likelihood in
both periods.
VII. Summary and Conclusion

This paper recommends the news impact curve as a standard measure of
how news is incorporated into volatility estimates. In order to better estimate
and match news impact curves to the data, several new candidates for
modeling time-varying volatility are introduced and contrasted. These models
allow several types of asymmetry in the impact of news on volatility. Further-
more, some new diagnostic tests are presented which are designed to deter-
mine whether the volatility estimates adequately represent the data. Finally,
a partially nonparametric model is suggested which allows the data to
determine the news impact curve directly.
These models are fitted to daily Japanese stock returns from 1980 to 1988.
All the models find that negative shocks introduce more volatility than
positive shocks, with this effect particularly apparent for the largest shocks.
The diagnostic tests however, indicate that in many cases the modeled
asymmetry is not adequate. The best model is the one proposed by Glosten,
Jagannathan, and Runkle (GJR).
The partially nonparametric (PNP) ARCH model, when fitted to the data,
confirm this behavior. For reasonable shock values, the volatilities forecast by
EGARCH, GJR, and PNP are similar. However, for more extreme shocks,
these forecasts differ dramatically. In fact, the standard deviation of the
EGARCH estimated conditional variance is even higher than that of the
squared residual itself. This result could be interpreted as evidence against
the EGARCH, because the variability of the conditional variance, if correctly
specified, should not be higher than that of the squared residual.
The results are similar, although less dramatic, when the same analysis is
conducted excluding the October 1987 crash. Overall, these results show a
greater impact on volatility of negative, rather than positive, return shocks.
The results indicate that, of the variance parametric models, the GJR is the
best at parsimoniously capturing this asymmetric effect. Finally, the PNP
model successfully reveals the shape of the news impact curve and is a useful
approach to modeling conditional heteroskedasticity.
REFERENCES
Amin, K., and B. Ng, 1993, Option valuation with systematic stochastic volatility, Journal of
Finance 48, 881-910.
Baillie, R., and R. Myers, 1991, Modeling commodity price distributions and estimating the
optimal futures hedge, Journal of Applied Econometrics 6, 109-124.
Black, Fisher, 1976, Studies in stock price volatility changes, Proceedings of the 1976 Business
Meeting of the Business and Economics Statistics Section, American Statistical Association,
pp. 177-181.
Bollerslev, T., 1986, Generalized autoregressive conditional heteroskedasticity, Journal of

Econometrics 31, 307-327.
, R. Chou, and K. Kroner, 1992, ARCH modeling in finance: A review of the theory and
empirical evidence, Journal of Econometrics 52, 5-59.
Chou, R., 1988, Volatility persistence and stock valuations: Some empirical evidence using
GARCH, Journal of Applied Econometrics 3, 279-294.
, R. Engle, and A. Kane, 1992, Measuring risk aversion from excess returns on a stock
index, Journal of Econometrics 52, 201-224.
Christie, A., 1982, The stochastic behavior of common stock variance: Value, leverage and
interest rate effects, Journal of Financial Economics 10, 407-432.
Connolly, R., 1989, An examination of the robustness of the weekend effect, Journal of Financial
and Quantitative Analysis 24, 133-169.
Conrad, J., M. Gultekin, and G. Kaul, 1991, Asymmetric predictability of conditional variances,
Review of Financial Studies 4, 597-622.
Day, T., and C. M. Lewis, 1992, Stock market volatility and the information content of stock
index options, Journal of Econometrics 52, 267-288.
Engle, R., 1982, Autoregressive conditional heteroskedasticity with estimates of the variance of
U.K. inflation, Econometrica 50, 987-1008.
, 1984, Wald, likelihood ratio, and Lagrange multiplier tests in econometrics, in: Z.
Griliches and M. D. Intriligator, eds.: Handbook of Econometrics, Vol. II (North Holland,
Amsterdam).
and T. Bollerslev, 1986, Modeling the persistence of conditional variances, Econometric
Review 5, 1-50.
Engle, R., C. Hong, A. Kane, and J. Noh, 1992, Arbitrage valuation of variance forecasts with
simulated options, Discussion paper 92-19, Department of Economics, University of Califor-
nia, San Diego.
Engle, R., and D. Kraft, 1983, Multiperiod forecast error variances of inflation estimated from
ARCH Models, in A. Zellner, ed.: Applied Time Series Analysis of Economic Data (Bureau of
the Census, Washington, D.C.), 293-302.
Ferson, W., and C. R. Harvey, 1991, The variation of economic risk premiums, Journal of
Political Economy 99, 385-415.
French, K., G. W. Schwert, and R. Stambaugh, 1987, Expected stock returns and volatility,
Journal of Financial Economics 19, 3-29.
Geweke, J., 1986, Modeling the persistence of conditional variances: A comment, Econometric
Review 5, 57-61.
Glosten, L., R. Jagannathan, and D. Runkle, 1989, Relationship between the expected value and
the volatility of the nominal excess return on stocks, Working paper, Department of Finance,
Columbia University.
Gourieroux, C., and A. Monfort, 1992, Qualitative threshold ARCH models, Journal of Economet-
rics 52, 159-200.
Hamao, Y., R. W. Masulis, and V. Ng, 1990, Correlations in price changes and volatility across
international stock markets, Review of Financial Studies 3, 281-308.
Harvey, C., and R. Whaley, 1992, Market volatility prediction and the efficiency of the S & P 100
Index Option Market, Journal of Financial Economics 31, 43-74.
Hull, J., and A. White, 1987, The pricing of options on assets with stochastic volatilities, Journal
of Finance 42, 281-300.
Kuwahara, H., and T. Marsh, 1992, The pricing of Japanese equity warrants, Management
Science, Forthcoming.
Lin, W., R. Engle, and T. Ito, 1992, Do bulls and bears move across borders?, Discussion paper,
Department of Economics, University of Wisconsin, Madison.
Masulis, R. W., and V. Ng, 1992, Stock return dynamics over intra-day trading and nontrading
periods in the London stock market, Mitsui Life Financial Research Center, Working paper,
University of Michigan.
Merton, Robert C., 1980, On estimating the expected return on the market: An exploratory
investigation, Journal of Financial Economics 8, 323-361.
Mihoj,A., 1987, A multiplicative parameterizationof ARCHmodels, Workingpaper, Department

of Statistics, University of Copenhagen.
Nelson, D., 1990, Conditionalheteroskedasticity in asset returns: A new approach,Econometrica
59, 347-370.
Ng, V., R. Engle, and M. Rothschild, 1992, A multi-dynamic-factormodel for stock returns,
Journal of Econometrics 52, 245-266.
Pantula, S. G., 1986, Modelingthe persistence of conditionalvariances: A comment, Econometric
Review 5, 71-74.
Pagan, A., and G. W. Schwert, 1990, Alternative models for conditional stock volatility, Journal
of Econometrics 45, 267-290.
Ross, S., 1989, Informationand volatility: The non-arbitragemartingale approachto timing and
resolution irrelevancy, Journal of Finance 44, 1-17.
Schwert, G. W., 1990, Stock volatility and the Crash of 87, Review of Financial Studies 3,
77-102.
, and P. Seguin, 1990, Heteroskedasticity in stock returns, Journal of Finance 45,
1129-1155.
Susmel, R., and R. Engle, 1992, Hourly volatility spillovers between international equity mar-
kets, Discussion paper 92-08, Department of Economics,University of California, San Diego.
Wiggins, J. B., 1987, Option values under stochastic volatility: Theory and empirical tests,
Journal of Financial Economics 19, 351-372.
Zakoian, J., 1990, Threshold heteroskedasticity model, Unpublished manuscript, INSEE.

The Journal of Finance - December 1993 - ENGLE

Uploaded by

Copyright:

Available Formats

The Journal of Finance - December 1993 - ENGLE

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Journal of Finance - December 1993 - ENGLE

Uploaded by

Copyright:

Available Formats

THE JOURNAL OF FINANCE v VOL. XLVIII, NO.

Measuring and Testing the Impact of

ROBERT F. ENGLE and VICTOR K. NG*

THE ABILITY TO FORECAST financial market volatility is important for portfolio

The importance of a correctly specified volatility model is clear from the

the adequacy of the models, we employ the diagnostic tests developed in

I. Models of Predictable Volatility

Bollerslev (1986) generalizes the ARCH(p) model to the GARCH(p, q)

where A u- 2P3.exp[wIto a, 2/I]. (5)

In Figure 1, the news impact curve of the EGARCH(1, 1) is compared with

A GRH(1, ews Impat ure

where h t is the conditional variance at time t, 1is ~

ht=A* exp[, forseti<0,

where A =- exp[ w - a - 2/ir]1, o- is the unconditional return standard deviation, w is the

impact curve to locate at a point where et -1 is positive. In Figure 2, the news

Nonlinear ARCH model (Engle and Bollerslev (1986))

Multiplicative ARCH (Mihoj (1987), Geweke (1986), Pantula (1986))

1+a8t2 1 + yS[ 18t2 1, where St = 1 if Et < 0, St = otherwise

Autoregressive Standard Deviation model (Schwert (1990))

Asymmetric GARCH model (Engle (1990))

ht = w + a(et- + Y)2 + 3ht-I

ht =w+ 3ht- + a(et-l + - 1)2

ht +Pht- + a(t1/h1 + y)2

where ht is the conditional variance at time t, E 1 is the unpredictable return at time t - 1,

ht = A + (a + y) 2, for 8t- < 0,

implications for option pricing. Stock return volatility is a major factor in

II. Diagnostic Tests Based on the News Impact Curve

more extreme values of e for particular biases by using variables, such as

log(ht) = log(hot(5o'zot)) + a Zat, (6)

where hot(50'zo0) is the volatility model hypothesized under the null; 8 is

where zt =h_18htl/8_, 4 *t holdht/88a, and ut is the residual. Both

log(ht) = log[ to + 8ht ?+ as2 1 ] + 8*log(ht-1) + y*Bt j

power at all. Thus, the test is often computed as

LM= -T*R', (8)

vt =a+b.S7 1+I3'z*t+et, (9a)

where a and b are constant parameters, ,B is a constant parameter vector,

V2 = a + b1S- l + b2Sj71rt_ +b3S1+ 1 + lZ*t + et (10)

1, 5, and 10 percent critical values under the asymptotic distribution are

et = Xl- vt, (14)

regression of a vector of ones on the scores is less than or equal to 0.0000001.

Yt = t; et--:: h vt, where vt -i.i.d. N(0, 1)

ht= co +? ht-1 + a- 2t- 1

Model and Sample Size

Actual Rejection Frequencies (%)

replications. The nonconverged replications, not reported, yield similar re-

log(ht) = -0.23 +0.9 log(ht -1)+ 0.25 [V 2 1 - 0.3 vt-l].

ht = 0.005 + 0.7 ht-1 + 0.28* [It11 - 0.23 t_1]

Model and Sample Size

III. A Partially Nonparametric News Impact Model

Here we introduce a new partially nonparametric model. This will allow

ht + 3h t 1 + E OiPit-( -- ) + E Nit- -1 i), (17)

IV. Estimation of Japanese Stock Return Volatility:

GARCH(1, 1), and the Glosten-Jagannathan-Runkle (GJR) model. The esti-

V. Partially Nonparametric ARCH Estimation

4We thank Rob Stambaugh for pointing this out to us.

ht = 0.0238 + 0.6860 - ht 1 + 0.3299 -t-, 1

log(ht) = -.0668 +0.9012 log(ht 1) + 0.4927 [ -0.1450

ht = 0.0216 + 0.6896 ht 1 + 0.3174 - (et 1- 0.1108)2

ht = 0.0192 + 0.6754 - ht- 1 + 0.1508 - (et- / - 0.1458)2

ht = 0.0199 + 0.7253 ht- 1 + 0.2515 - (et 1 - 0.2683- / t_1)2

log(ht) = log[ to + 8ht ?+ as2 1 ] + 8log(ht-1) + yBt j