Research article Special Issues

The memory of volatility

  • The focus of the volatility literature on forecasting and the predominance of the conceptually simpler HAR model over long memory stochastic volatility models has led to the fact that the actual degree of memory estimates has rarely been considered. Estimates in the literature range roughly between 0.4 and 0.6 -that is from the higher stationary to the lower non-stationary region. This difference, however, has important practical implications -such as the existence or nonexistence of the fourth moment of the return distribution. Inference on the memory order is complicated by the presence of measurement error in realized volatility and the potential of spurious long memory. In this paper we provide a comprehensive analysis of the memory in variances of international stock indices and exchange rates. On the one hand, we find that the variance of exchange rates is subject to spurious long memory and the true memory parameter is in the higher stationary range. Stock index variances, on the other hand, are free of low frequency contaminations and the memory is in the lower non-stationary range. These results are obtained using state of the art local Whittle methods that allow consistent estimation in presence of perturbations or low frequency contaminations.

    Citation: Kai R. Wenger, Christian H. Leschinski, Philipp Sibbertsen. The memory of volatility[J]. Quantitative Finance and Economics, 2018, 2(1): 622-644. doi: 10.3934/QFE.2018.1.137

    Related Papers:

    [1] Lukáš Pichl, Taisei Kaizoji . Volatility Analysis of Bitcoin Price Time Series. Quantitative Finance and Economics, 2017, 1(4): 474-485. doi: 10.3934/QFE.2017.4.474
    [2] Isaac O. Ajao, Hammed A. Olayinka, Moruf A. Olugbode, OlaOluwa S. Yaya, Olanrewaju I. Shittu . Long memory cointegration and dynamic connectedness of volatility in US dollar exchange rates, with FOREX portfolio investment strategy. Quantitative Finance and Economics, 2023, 7(4): 646-664. doi: 10.3934/QFE.2023031
    [3] Xiaohang Ren, Weixi Xu, Kun Duan . Fourier transform based LSTM stock prediction model under oil shocks. Quantitative Finance and Economics, 2022, 6(2): 342-358. doi: 10.3934/QFE.2022015
    [4] Changhong Guo, Shaomei Fang, Yong He, Yong Zhang . Some empirical studies for the applications of fractional G-Brownian motion in finance. Quantitative Finance and Economics, 2025, 9(1): 1-39. doi: 10.3934/QFE.2025001
    [5] Elvira Caloiero, Massimo Guidolin . Volatility as an Alternative Asset Class: Does It Improve Portfolio Performance?. Quantitative Finance and Economics, 2017, 1(4): 334-362. doi: 10.3934/QFE.2017.4.334
    [6] Wenhui Li, Normaziah Mohd Nor, Hisham M, Feng Min . Volatility conditions and the weekend effect of long-short anomalies: Evidence from the US stock market. Quantitative Finance and Economics, 2023, 7(2): 337-355. doi: 10.3934/QFE.2023016
    [7] Per B. Solibakke . Forecasting hourly WTI oil front monthly price volatility densities. Quantitative Finance and Economics, 2024, 8(3): 466-501. doi: 10.3934/QFE.2024018
    [8] Francesco Audrino, Lorenzo Camponovo, Constantin Roth . Testing the Lag Structure of Assets' Realized Volatility Dynamics. Quantitative Finance and Economics, 2017, 1(4): 363-387. doi: 10.3934/QFE.2017.4.363
    [9] Tetsuya Takaishi . Volatility estimation using a rational GARCH model. Quantitative Finance and Economics, 2018, 2(1): 127-136. doi: 10.3934/QFE.2018.1.127
    [10] Lorna Katusiime . Time-Frequency connectedness between developing countries in the COVID-19 pandemic: The case of East Africa. Quantitative Finance and Economics, 2022, 6(4): 722-748. doi: 10.3934/QFE.2022032
  • The focus of the volatility literature on forecasting and the predominance of the conceptually simpler HAR model over long memory stochastic volatility models has led to the fact that the actual degree of memory estimates has rarely been considered. Estimates in the literature range roughly between 0.4 and 0.6 -that is from the higher stationary to the lower non-stationary region. This difference, however, has important practical implications -such as the existence or nonexistence of the fourth moment of the return distribution. Inference on the memory order is complicated by the presence of measurement error in realized volatility and the potential of spurious long memory. In this paper we provide a comprehensive analysis of the memory in variances of international stock indices and exchange rates. On the one hand, we find that the variance of exchange rates is subject to spurious long memory and the true memory parameter is in the higher stationary range. Stock index variances, on the other hand, are free of low frequency contaminations and the memory is in the lower non-stationary range. These results are obtained using state of the art local Whittle methods that allow consistent estimation in presence of perturbations or low frequency contaminations.


    Modeling and forecasting asset volatility is one of the central topics of financial econometrics. While the early literature has focused on short memory GARCH models, today it is well established that financial market volatility typically exhibits long memory. Standard models that capture the long-memory feature are, for example, ARCH() and LARCH models (Giraitis et al., 2007, 2009), as well as stochastic volatility models that make use of ARFIMA processes. The conceptually simpler HAR model of Corsi (2009) can also approximate long memory by using a regression with overlapping averages of past volatilities. While in the HAR model the actual degree of memory remains unknown, the other models provide estimates of the memory parameter d.

    However, a problem arises when the volatility series are contaminated by level shifts or deterministic trends, known as low frequency contaminations. In this case standard estimation methods for d are positively biased.

    The issue is that both long memory and mean shifts generate similar time series features such as significant autocorrelations at large lags or a pole in the periodogram at Fourier frequencies local to zero (Diebold and Inoue, 2001; Granger and Hyung, 2004; Mikosch and Stărică, 2004). If long memory is falsely detected in a short-memory time series subject to low frequency contaminations, it is referred to as 'spurious long memory'. However, recently several methods have been proposed that allow for robust estimation of d under these circumstances (Iacone, 2010; McCloskey and Perron, 2013; Hou and Perron, 2014).

    Another issue frequently discussed for volatility series is the effect of perturbations. Deo and Hurvich (2001) and Arteche (2004) show that standard estimation methods are negatively biased if a noisy volatility proxy is used. Local Whittle based methods that reduce this bias are proposed by Hurvich et al. (2005) and Frederiksen et al. (2012), among others.

    Several studies have estimated the memory parameter in index-and exchange rate variance. However, most of them use standard estimation methods that do not account for the issues discussed above. The estimates achieved are roughly in the range of 0.4<d<0.6, so in the higher stationary or in the lower non-stationary region (Andersen et al., 2003; Hurvich and Ray, 2003; Martens et al., 2009; among others). It is an important question whether d>0.5 since the features of the underlying processes are substantially different. In particular, if d>0.5 the variance of the variance series is infinite, so that the kurtosis of the returns does not exist. To see this, denote the continuously compounded asset returns by rt and assume that they are mean zero with conditional heteroscedasticity of the form

    rt=σtηt, (1)

    where σt denotes the volatility at day t and it is assumed that ηtiid(0,1) with finite kurtosis Kη. Then the return kurtosis (Kr) can be decomposed into the kurtosis of the volatility process (Kσ) and that of the innovation sequence as follows

    Kr=E(r4t)E(r2t)2=E(σ4t)E(σ2t)2E(η4t)E(η2t)2=KσKη. (2)

    If d>0.5, we have

    Var(σ2t)=E[σ4t]E[σt]4= (3)

    with E[σt]4=E[σ2t]2<. This implies

    Kr=(Var(σ2t)2E[σ2t]2+1)Kη=. (4)

    Here, we provide a comprehensive analysis of the memory of a wide range of international stock indices and exchange rates using recently published robust estimation methods. We find that the variance of exchange rates is in the higher stationary range while the variance of stock indices is in the lower non-stationary range. Additionally, we find that exchange rates are likely to be subject to low frequency contaminations which bias standard estimation methods upwards, whereas the stock index variances are free of spurious long memory.

    The rest of the paper is structured as follows. In Section 2 we review the methodological issues associated with the estimation of long memory in realized volatility time series. This motivates the use of robust estimation methods reviewed in Section 3. Here, we also provide a Monte Carlo simulation that analyzes the performance of the robust estimation methods if there is potential for both -low frequency contaminations and perturbations. Section 4 contains our empirical contribution that analyzes the memory parameters of a large set of international stock indices and exchange rates. Finally, Section 5 concludes.

    ARCH-type models usually assume that the daily variance σ2t is some function of the past squared returns, so that σ2t=h(rt1,rt2,...). On the contrary, the stochastic volatility literature usually assumes that the log-variance logσ2t is a function of the lagged returns as well as an additional innovation sequence εt that is specific to the volatility process

    logσ2t=g(rt1,rt2,...,εt,εt1,...). (5)

    To fit these models, one either has to rely on complicated unobserved component models or it is necessary to employ a proxy for the unobserved volatility process σt.

    Since high frequency data has become widely available, it has become standard practice to use realized variance as a proxy. Realized variance was popularized (among others) by Andersen et al.(2001, 2003). Recent examples of long memory models for realized variance include Deo et al. (2006), Martens et al. (2009) and Chiriac and Voev (2011).

    Let the log-price pt of an asset be observed at regular intervals -N times per trading day-and denote the i-th intraday log-return by ri,t=pi,tpi1,t, then the realized variance is given by

    zt=Ni=1r2i,t, (6)

    so that zt=σ2t(1+wt), for some error sequence wt and therefore

    logzt=logσ2t+log(1+wt)logσ2t+wt, (7)

    for small wt. It is clear from (7) that zt can be regarded as a perturbed version of the underlying volatility process. The influence of this estimation error in the volatility proxy on the accuracy of the estimated memory parameter is an important topic in the long memory stochastic volatility literature.

    Note however, that Barndorff-Nielsen and Shephard (2002) show that plim zt=σ2t, as N, so that wt0. Therefore, zt is a relatively precise estimate of σ2t and it is sometimes treated as if it was a direct observation of the variance process. Nevertheless, a careful analysis of the memory of volatilities as intended here should take these effects into account.

    Another issue that has to be taken into account is the possible presence of low frequency contaminations such as level shifts or deterministic trends. Especially log-squared returns have been a prominent example in the literature on spurious long memory, from early contributions such as Granger and Ding (1996), Mikosch and Stărică (2004) or Granger and Hyung (2004) to more recent contributions such as Lu and Perron (2010) or Xu and Perron (2014).

    We therefore consider the following model for the log realized variance

    zt=c+yt+uT,t+wt, (8)

    where the variance process zt consists of a short-or long-memory process yt, a constant c, a so called low frequency contamination uT,t (e.g. a level shift process or trend), and the additive, mean zero, short memory noise term wt with variance σ2w<.

    A possible data generating process for the low frequency contamination uT,t is a random level shift process as given by

    uT,t=Tt=1δT,t,whereδT,t=πT,tξt, (9)

    with ξtN(0,σ2ξ) and πT,tiidB(p/T,1), for p0. Here, πT,t and ξt are mutually independent and they are also independent of yt and wt.

    To estimate the unknown memory parameter d in applications, it is common to use the local Whittle estimator of Künsch (1987) and Robinson (1995a). Compared to ARFIMA models this semiparametric approach has the advantage that it is consistent irrespective of the form of the short run dynamics. Furthermore, the asymptotic variance of the local Whittle estimator is smaller than that of the log-periodogram estimator of Geweke and Porter-Hudak (1983) and Robinson (1995b). The discussion in this paper is therefore focused on the local Whittle estimator that is discussed in detail in the next section.

    In absence of low frequency contaminations in model (8) -that is if ut,T=0 for all t=1,...,T -Arteche (2004) shows that the local Whittle estimator is biased downwards. Bias corrected versions of the estimator have been proposed, among others, by Hurvich et al. (2015) and Frederiksen et al. (2012).

    Similarly, in absence of perturbations in (8) -that is when wt=0 for all t=1,...,T -Perron and Qu (2010) and McCloskey and Perron (2013) show that the periodogram of zt can be decomposed into

    Iz,T(λj)=Iy,T(λj)+Iu,T(λj)+Iyu,T(λj) (10)
    =12πT|Tt=1yteiλjt|2+12πT|Tt=1uT,teiλjt|2+22πTTt=1Ts=1ytuT,tcos(λj(ts)). (11)

    For λj=o(1) they show that

    Iz,T=Op(1λ2dj)+Op(1Tλ2j)+Op(1Tλ1+dj). (12)

    It follows that the part corresponding to the random level shift process uT,t dominates for j=o(T(12d)/(22d)) while the part corresponding to the short-or long-memory process yt dominates for jT(2d1)/(22d).

    Hence, local Whittle estimates are biased upwards especially when small bandwidths are used. This results in the aforementioned effect of spurious long memory.

    For the model in (8), it is therefore well established that there are potential effects that cause both upwards bias as well as downwards bias in the estimated memory parameters if standard methods such as the local Whittle estimator are used.

    Remark: The random level shift specification in (9) is just one example out of a larger class of processes that generates the behavior of the periodogram specified in (12). Other examples from this class comprise smooth trends, as shown in Qu (2011) or deterministic structural breaks (McCloskey and Perron, 2013) and the processes considered by Gourieroux and Jasiak (2001). Recently, Leschinski and Sibbertsen (2017) provide results that show that this behavior extends to a large class of level shift processes that are non-degenerate in the sense that the level change is asymptotically rare and occurs after a non-zero fraction of the sample. The methods discussed here are therefore not restricted to the random level shift model, but can be applied for a wide range of structural change models.

    As argued in the previous section, it is likely that volatility measures are perturbed (even though the perturbation is less pronounced when the realized variance is used) and subject to low frequency contaminations. Therefore, robust methods against these issues have to be used to estimate d.

    The spectral density function fz(λ) of the perturbed volatility measure process under low frequency contaminations in (8) at frequency λ is given by

    fz(λ)=ϕy(λ)λ2d+ϕw(λ)+ϕu(λ)λ2/T (13)

    where ϕa with a{y,w,u} corresponds to the spectral density of the short run components in yt, wt, and ut,T. All local Whittle estimation methods are based on the concentrated local log-likelihood as given by

    Ra(d,θ)=logˆGa2dml+1mj=1logλj+1mmj=1log(ga), (14)

    where ˆGa approximates the spectral density local to zero, m=Tb is the bandwidth, l is a trimming parameter which is equal to one except for the trimmed local Whittle estimator, λj=(2πj/T) are the Fourier frequencies, ga is a function that controls for perturbations and/or low frequency contaminations, and a{LW,LPWN,mLW,tLW}.

    Depending on whether perturbations, low frequency contaminations, or both are present in the volatility series, the spectral density in (13) has to be approximated differently local to zero. Therefore, several estimators can be derived.

    For the standard local Whittle estimator it is assumed that there are no perturbations and low frequency contaminations. Hence, the short memory dynamics of the spectral density of the series zt in (13) are approximated by a constant G. It follows that

    ˆGLW=1mmj=1λ2djIz(λj) and gLW(d,θ,λ)=0, (15)

    where the periodogram Iz(λ) is given by Iz(λ)=(2πT)1|Tt=1zteitλ|2. The standard local Whittle estimator suggested by Küunsch (1987) is then given by

    ˆdLW=argmindRLW(d). (16)

    For 1/m+m/T, as T and d(0.5,0.5) consistency of ˆdLW is shown by Robinson (1995a). Under strengthened assumptions (especially on the bandwidth choice) it is also shown that m(ˆdLWd)aN(0,1/4). Velasco (1999) extends these results and shows that the local Whittle estimator is consistent for d(0.5,1] and asymptotically normal for d(0.5,0.75].

    In the case where the volatility measure exhibits short memory dynamics and perturbations, the local polynomial Whittle with noise (LPWN) estimator of Frederiksen et al. (2012) can be applied. They extend the idea of Andrews and Sun (2004) who approximate the spectral density local to zero by a polynomial instead of a constant to reduce the finite sample bias of the local Whittle estimator. Frederiksen et al. (2012) add an additional polynomial to approximate the spectrum ϕw(λ) of the perturbation in (13). Other approaches that try to approximate the perturbation by a constant rather than a polynomial are, for example, suggested by Hurvich and Ray (2003) who proposed the local polynomial Whittle estimator with noise (LWN). This estimator is nested in the LPWN estimator if the polynomials are chosen of order zero.

    Precisely, Frederiksen et al. (2012) fit the following two polynomials

    logϕy(λ)logG+hy(θy,λ) (17)
    logϕw(λ)logG+logθp+hw(θw,λ). (18)

    to approximate the logarithms of ϕy(λ) and ϕw(λ) in (13). Here θ=(θy,θρ,θw), θρ=ϕw(0)/ϕy(0) is the long-run signal-to-noise ratio, ha(θa,λ)=Rar=1θa,rλ2r, and a{y,w}. Therefore,

    ˆGLPWN=1mmj=1λ2djIz(λj)gLPWN(d,θ,λj) (19)
    and gLPWN(d,θ,λ)=exp(hy(θy,λ))+θρλ2dexp(hw(θw,λ)). (20)

    The estimator is given by

    (ˆdLPWN,ˆθ)=argmind[d1,d2],θΘRLPWN(d,θ), (21)

    where 0<d1<d2<1 is assumed to be stationary, and Θ is a compact and convex set in RRy×(0,)×RRw. For a properly chosen m and d(0,1) Frederiksen et al. (2012) show in their Theorem 2 under some regularity conditions that the memory estimator is consistent. They further show for d(0,0.75) that it converges in distribution to the normal distribution when perturbations are present. Compared with the local Whittle estimator the asymptotic variance increases by a multiplicative constant, but the bias through perturbation and short memory dynamics is reduced.

    In case that the volatility series is subject to level shifts or other low frequency contaminations, the trimmed local Whittle (tLW) estimator of Iacone (2010) and the modified local Whittle (mLW) estimator of Hou and Perron (2014) can be applied.

    The idea of the trimmed local Whittle estimator of Iacone (2010) is to use a trimming of the (l1) lowest frequencies where the contaminations have their biggest effect on the spectral density of the series according to (12). Therefore, we obtain

    ˆGtLW=1ml+1mj=lλ2djIz(λj)andgtLW(d,θ,λ)=0, (22)

    where 1l<mT/2, so that

    ˆdtLW=argmindRtLW(d). (23)

    In case l=1 the estimator is reduced to the standard local Whittle. Under suitable assumptions on the bandwidth and trimming parameter Iacone (2010) shows consistency and asymptotic normality of ˆdtLW for d(0,0.5). Its asymptotic variance is the same as that of the local Whittle estimator.

    The modified local Whittle estimator uses another approach to achieve consistent estimates of d under low frequency contaminations. It adds an additional term to account for ϕu(λ)λ2/T-the influence of the low frequency contamination in the spectral density function of the variance zt. Hou and Perron (2014) also provide an additional extension of the modified local Whittle estimator to account for both the perturbation and the low frequency contamination. In this case they approximate the spectral density ϕw(λ) of the perturbation by a constant term following the approach of Hurvich et al. (2005). Denoting θ=(θw,θu) as the signal-to-noise ratios of the perturbation and the low frequency contamination the estimator uses

    ˆGmLW=1mmj=1Iz(λj)gmLW(d,θ,λj)andgmLW(d,θ,λ)=(λ2d+θw+θuλ2/T) (24)

    and is given by

    (ˆdmLW,ˆθ)=argmind,θRmLW(d,θ). (25)

    If θw=0, we have the modified local Whittle estimator (mLW) and if θw0 we have the modified local Whittle plus noise estimator (mLWN). For θw=0, a properly chosen m which needs to be larger than T5/9 and d(0,0.5), Hou and Perron (2014) show consistency and under strengthened assumptions asymptotic normality of the estimator. The estimator possesses the same asymptotic variance as the local Whittle estimator. The consistency of these methods for d>0.5 is addressed in our simulations below.

    A prominent test to distinguish true from spurious long memory is the Lagrange multiplier-type test of Qu (2011). Its null hypothesis incorporates all second-order stationary short-or long-memory processes. Under the alternative, the process is contaminated by some low frequency contamination, for example a random level shift as given in (9). The test statistic uses the difference between the rates given in (12) and it is based on the local Whittle likelihood function. It is given by

    W=supr[ϵ,1](mj=1ν2j)1/2|[mr]j=1νj(Iz(λj)G(ˆdLW)λ2ˆdLWj1)|, (26)

    with νj=logλj(1/m)mj=1logλj, and a small trimming parameter ϵ. Qu (2011) derives the consistency and the limiting distribution of (26) for d(0,0.5). Sibbertsen et al. (2017) show via simulations that the test also works in the low non-stationary range for d. Qu (2011) reports critical values for ϵ{0.02,0.05}, where the first value is recommended for sample sizes T>500. It is further recommended to use m=T0.7 frequency ordinates. The test of Qu (2011) has several desirable properties such as not requiring Gaussianity, allowing for conditional heteroskedasticity, not requiring a precise specification of the low frequency contamination due to its score-type nature and displaying high finite sample power results compared to competing tests (Leccadito et al., 2015).

    To evaluate the finite sample performance of the estimators discussed above in the situation we are facing in our empirical application, we conduct a small Monte Carlo simulation. The data generating process (DGP) is based on (8), where yt is a fractionally integrated process of order d, wt is white noise with variance σ2w, p=5, and T=4000, which approximately mirrors the sample sizes in our empirical application. Since the variance of the perturbations can be expected to be small, we set σw{0,1/10,1/4,1/2}, whereas yt is scaled so that the variance of the process is one. Furthermore, we set σ2η{0,1,2}, since we do not have any a priori knowledge about the magnitude of potential mean shifts. The bandwidth parameters are chosen as in the empirical application following the recommendations of the authors who proposed the respective methods. For the local Whittle estimator we set m=T0.7.

    The results reported in Table 1 are based on 5000 Monte Carlo replications. Starting with the local Whittle estimator, we can observe the expected result that there is a positive bias if level shift components are present and the perturbations cause only a slight negative bias due to their moderate scale. It is worth noting that the bias of the local Whittle estimator is smaller for d=0.6 than for d=0.4, so that all the estimated ds are in the range between 0.5 and 0.7.

    Table 1.  Bias and standard deviation of the long-memory estimators.
    Bias Standard Deviation
    d 0.4 0.6 0.4 0.6
    σw/ση 0 1 2 0 1 2 0 1 2 0 1 2
    LW 0 0.00 0.08 0.17 0.00 0.04 0.10 0.03 0.06 0.09 0.03 0.05 0.07
    0.1 0.00 0.08 0.17 0.00 0.04 0.10 0.03 0.06 0.09 0.03 0.05 0.07
    0.25 -0.01 0.07 0.16 -0.02 0.03 0.09 0.03 0.06 0.09 0.03 0.05 0.07
    0.5 -0.03 0.05 0.14 -0.06 -0.02 0.05 0.03 0.06 0.09 0.02 0.05 0.07
    mLW 0 -0.01 -0.01 -0.01 -0.02 -0.01 0.02 0.02 0.03 0.04 0.02 0.04 0.07
    0.1 -0.01 -0.01 -0.01 -0.03 -0.02 0.01 0.02 0.03 0.04 0.02 0.04 0.07
    0.25 -0.03 -0.03 -0.03 -0.08 -0.08 -0.08 0.02 0.03 0.04 0.04 0.05 0.08
    0.5 -0.07 -0.07 -0.08 -0.20 -0.22 -0.24 0.02 0.03 0.04 0.06 0.06 0.07
    tLW 0 -0.01 0.01 0.04 -0.01 0.01 0.04 0.03 0.04 0.05 0.03 0.04 0.05
    0.1 -0.01 0.01 0.04 -0.01 0.00 0.03 0.03 0.04 0.05 0.03 0.04 0.05
    0.25 -0.02 -0.01 0.03 -0.05 -0.04 0.00 0.03 0.03 0.05 0.03 0.04 0.05
    0.5 -0.06 -0.05 -0.02 -0.15 -0.14 -0.10 0.03 0.03 0.05 0.04 0.04 0.06
    LWN 0 0.02 0.12 0.23 0.01 0.06 0.12 0.03 0.08 0.10 0.03 0.05 0.07
    0.1 0.02 0.12 0.23 0.01 0.06 0.12 0.03 0.08 0.10 0.03 0.05 0.07
    0.25 0.01 0.12 0.23 0.01 0.06 0.13 0.04 0.08 0.11 0.04 0.06 0.08
    0.5 0.01 0.13 0.24 0.01 0.06 0.13 0.04 0.09 0.11 0.04 0.06 0.08
    mLWN 0 0.01 0.03 0.11 0.01 0.05 0.11 0.03 0.07 0.15 0.03 0.06 0.08
    0.1 0.01 0.03 0.11 0.01 0.04 0.11 0.03 0.07 0.15 0.03 0.06 0.09
    0.25 0.00 0.03 0.11 0.00 0.03 0.10 0.04 0.07 0.16 0.05 0.07 0.10
    0.5 -0.01 0.02 0.11 -0.01 0.03 0.11 0.05 0.09 0.17 0.06 0.09 0.11

     | Show Table
    DownLoad: CSV

    Turning to the mLW estimator of Hou and Perron (2014), we observe that the estimator successfully mitigates the bias caused by the level shift components. However, with increasing magnitude of the perturbation, the estimator suffers a strong negative bias -much stronger than the original local Whittle estimator.

    Similar results hold true for the tLW estimator of Iacone (2010), but the magnitude of the perturbation bias is considerably smaller than that of the mLW estimator.

    The LWN estimator behaves similarly, but in the contrary direction. It successfully mitigates the downward bias caused by perturbations, but it suffers from a stronger upward bias in case of level shift components than the local Whittle estimator.

    Finally, the mLWN estimator seems to be mitigating the perturbation bias, but it does not control the spurious long memory bias to its full extend.

    The results of the mLW and tLW estimators for d=0.6 show that the consistency extends to the lower non-stationary region -exactly like that of the LW and LPWN.

    With regard to the variation of the estimators, we can observe that all methods become increasingly variable as the influence of the level shifts increases. The mLW estimator turns out to have slightly less variance than the tLW estimator of Iacone (2010) for d=0.4, but higher variance for d=0.6. The LWN estimator is more variable than the LW estimator and the mLWN estimator is extremely volatile in presence of level shifts in a stationary long-memory sequence with d=0.4.

    We consider daily realized variances of 41 major stock indices and 10 nominal exchange rates relative to the US Dollar. The data for the indices is obtained from the 'Oxford-Man Institute's realised library' and was compiled by Heber et al. (2009). The series start between 1996 and 2000 and they end on 9 June 2017. An overview of the symbols is given in Table 5 and a summary of the start and end dates as well as the length of the series is given in Table 7 in the appendix.

    Table 2.  Estimated long-memory coefficients of the log-realized variances of the indices.
    LW0.6 LW0.7 LW0.8 LWN LPWN(1,0) LPWN(0,1) LPWN(1,1) mLW mLWN tLW Qu0.75
    AEX 0.625 0.635 0.577 0.659 0.661 0.670 0.669 0.568 0.658 0.567 0.575
    (0.542 0.709) (0.581 0.688) (0.543 0.612) (0.624 0.694) (0.626 0.696) (0.636 0.705) (0.634 0.704) (0.533 0.602) (0.623, 0.693) (0.506, 0.627)
    AORD 0.553 0.586 0.460 0.650 0.611 0.617 0.611 0.344 0.648 0.444 1.061
    (0.469 0.637) (0.532 0.640) (0.426 0.495) (0.615 0.685) (0.576 0.646) (0.582 0.652) (0.576 0.646) (0.309 0.379) (0.613, 0.683) (0.383, 0.505)
    BVSP 0.549 0.521 0.475 0.561 0.530 0.522 0.530 0.462 0.561 0.484 0.555
    (0.464 0.633) (0.467 0.575) (0.440 0.510) (0.526 0.597) (0.495 0.565) (0.487 0.558) (0.495 0.565) (0.426 0.497) (0.526, 0.597) (0.423, 0.545)
    DJI 0.622 0.586 0.534 0.624 0.594 0.572 0.594 0.488 0.624 0.539 0.610
    (0.538 0.706) (0.533 0.640) (0.499 0.569) (0.589 0.659) (0.559 0.629) (0.537 0.607) (0.559 0.629) (0.453 0.523) (0.589, 0.659) (0.479, 0.600)
    FCHI 0.605 0.624 0.564 0.643 0.646 0.656 0.655 0.518 0.642 0.550 0.616
    (0.521 0.688) (0.571 0.678) (0.529 0.599) (0.609 0.678) (0.611 0.681) (0.621 0.691) (0.620 0.689) (0.483 0.552) (0.607, 0.677) (0.490, 0.610)
    FTSE 0.650 0.640 0.569 0.672 0.677 0.691 0.691 0.487 0.671 0.544 0.677
    (0.566 0.733) (0.586 0.694) (0.534 0.604) (0.637 0.707) (0.643 0.712) (0.656 0.726) (0.656 0.726) (0.452 0.522) (0.636, 0.706) (0.483, 0.604)
    FTSEMIB 0.600 0.607 0.547 0.632 0.626 0.631 0.627 0.522 0.631 0.548 0.557
    (0.517 0.684) (0.553 0.660) (0.512 0.582) (0.597 0.667) (0.591 0.661) (0.596 0.666) (0.592 0.662) (0.488 0.557) (0.596, 0.665) (0.487, 0.609)
    GDAXI 0.653 0.637 0.564 0.663 0.667 0.728 0.728 0.471 0.662 0.499 0.722
    (0.569 0.736) (0.584 0.691) (0.529 0.598) (0.629 0.698) (0.632 0.702) (0.693 0.762) (0.694 0.763) (0.436 0.506) (0.627, 0.697) (0.438, 0.559)
    GSPTSE 0.603 0.565 0.490 0.642 0.647 0.639 0.651 0.375 0.593 0.451 1.264*
    (0.514 0.691) (0.509 0.622) (0.453 0.527) (0.605 0.679) (0.610 0.684) (0.602 0.676) (0.614 0.689) (0.338 0.412) (0.556, 0.630) (0.387, 0.515)
    HSI 0.640 0.557 0.503 0.640 0.646 0.649 0.649 0.384 0.638 0.446 1.461*
    (0.554 0.726) (0.502 0.613) (0.467 0.539) (0.604 0.676) (0.610 0.682) (0.613 0.685) (0.613 0.686) (0.348 0.420) (0.602, 0.674) (0.383, 0.509)
    IBEX 0.593 0.596 0.545 0.611 0.617 0.632 0.631 0.510 0.576 0.525 0.858
    (0.509 0.677) (0.542 0.649) (0.510 0.580) (0.576 0.645) (0.583 0.652) (0.597 0.666) (0.597 0.666) (0.475 0.544) (0.541, 0.611) (0.465, 0.586)
    IXIC 0.644 0.598 0.552 0.628 0.623 0.579 0.625 0.493 0.625 0.542 0.535
    (0.560 0.728) (0.545 0.652) (0.517 0.587) (0.593 0.662) (0.588 0.657) (0.544 0.614) (0.590 0.660) (0.458 0.527) (0.590, 0.660) (0.481, 0.603)
    KS11 0.692 0.622 0.548 0.686 0.691 0.718 0.743 0.407 0.684 0.510 1.181
    (0.607 0.776) (0.567 0.676) (0.512 0.583) (0.651 0.721) (0.656 0.726) (0.683 0.753) (0.708 0.778) (0.372 0.442) (0.649, 0.720) (0.448, 0.571)
    MIB30 0.608 0.613 0.550 0.662 0.656 0.661 0.657 0.510 0.661 0.562 0.640
    (0.516 0.700) (0.553 0.673) (0.511 0.590) (0.623 0.702) (0.617 0.695) (0.622 0.700) (0.618 0.696) (0.471 0.549) (0.622, 0.700) (0.458, 0.596)
    MIBTEL 0.667 0.622 0.532 0.735 0.686 0.693 0.687 0.343 0.735 0.536 1.002
    (0.562 0.773) (0.552 0.692) (0.486 0.579) (0.689 0.782) (0.640 0.733) (0.646 0.739) (0.640 0.733) (0.296 0.389) (0.689, 0.782) (0.493, 0.630)
    MID 0.712 0.637 0.564 0.727 0.730 0.754 0.754 0.373 0.726 0.530 1.160
    (0.620 0.804) (0.577 0.697) (0.525 0.603) (0.687 0.766) (0.691 0.770) (0.714 0.793) (0.714 0.793) (0.333 0.412) (0.687, 0.765) (0.453, 0.618)
    MSCIAU 0.659 0.606 0.499 0.731 0.698 0.701 0.698 0.297 0.728 0.470 0.721
    (0.555 0.762) (0.538 0.674) (0.454 0.545) (0.685 0.776) (0.653 0.743) (0.655 0.746) (0.653 0.743) (0.252 0.343) (0.682, 0.773) (0.460, 0.599)
    MSCIBE 0.730 0.656 0.536 0.805 0.738 0.733 0.790 0.160 0.804 0.501 1.349*
    (0.628 0.832) (0.589 0.723) (0.492 0.581) (0.760 0.849) (0.693 0.782) (0.688 0.777) (0.745 0.834) (0.116 0.204) (0.759, 0.848) (0.389, 0.550)
    MSCIBR 0.629 0.567 0.501 0.677 0.670 0.671 0.675 0.358 0.675 0.503 0.628
    (0.511 0.747) (0.488 0.645) (0.448 0.555) (0.624 0.730) (0.617 0.723) (0.618 0.724) (0.622 0.729) (0.304 0.411) (0.622, 0.728) (0.422, 0.580)
    MSCICA 0.681 0.614 0.503 0.765 0.680 0.683 0.676 0.215 0.763 0.512 0.636
    (0.572 0.789) (0.543 0.686) (0.454 0.551) (0.716 0.813) (0.632 0.728) (0.635 0.731) (0.628 0.724) (0.167 0.263) (0.715, 0.811) (0.407, 0.599)
    MSCICH 0.730 0.676 0.565 0.782 0.724 0.720 0.719 0.263 0.782 0.546 1.103
    (0.629 0.832) (0.609 0.743) (0.521 0.610) (0.738 0.827) (0.680 0.768) (0.675 0.764) (0.675 0.764) (0.218 0.307) (0.737, 0.826) (0.426, 0.598)

     | Show Table
    DownLoad: CSV
    Table 3.  Estimated long-memory coefficients of the log-realized variances of the indices.
    LW0.6 LW0.7 LW0.8 LWN LPWN(1,0) LPWN(0,1) LPWN(1,1) mLW mLWN tLW Qu0.75
    MSCIDE 0.680 0.639 0.521 0.781 0.695 0.670 0.680 0.191 0.780 0.493 1.468*
    (0.578 0.782) (0.573 0.706) (0.476 0.565) (0.737 0.826) (0.650 0.739) (0.626 0.715) (0.635 0.724) (0.147 0.236) (0.736, 0.825) (0.467, 0.625)
    MSCIES 0.640 0.623 0.519 0.716 0.653 0.656 0.651 0.359 0.714 0.524 0.616
    (0.539 0.742) (0.556 0.690) (0.474 0.563) (0.672 0.761) (0.608 0.697) (0.611 0.700) (0.606 0.695) (0.314 0.403) (0.669, 0.758) (0.414, 0.572)
    MSCIFR 0.656 0.619 0.522 0.761 0.680 0.645 0.680 0.237 0.761 0.537 1.054
    (0.554 0.758) (0.553 0.686) (0.477 0.566) (0.717 0.806) (0.635 0.724) (0.601 0.690) (0.635 0.724) (0.193 0.281) (0.717, 0.806) (0.445, 0.603)
    MSCIGB 0.693 0.643 0.544 0.773 0.697 0.687 0.676 0.246 0.772 0.542 1.013
    (0.591 0.794) (0.576 0.710) (0.499 0.588) (0.729 0.817) (0.652 0.741) (0.643 0.731) (0.631 0.720) (0.202 0.290) (0.728, 0.816) (0.458, 0.615)
    MSCIIT 0.633 0.628 0.532 0.728 0.674 0.680 0.673 0.352 0.727 0.532 1.113
    (0.532 0.735) (0.562 0.695) (0.488 0.577) (0.684 0.773) (0.630 0.718) (0.635 0.724) (0.629 0.718) (0.307 0.396) (0.683, 0.772) (0.463, 0.621)
    MSCIJP 0.685 0.568 0.514 0.619 0.626 0.691 0.696 0.437 0.594 0.455 1.372*
    (0.581 0.790) (0.499 0.637) (0.468 0.560) (0.573 0.665) (0.580 0.672) (0.645 0.737) (0.650 0.742) (0.391 0.483) (0.548, 0.640) (0.454, 0.611)
    MSCIKR 0.695 0.625 0.534 0.691 0.638 0.639 0.625 0.407 0.689 0.512 0.785
    (0.591 0.800) (0.556 0.694) (0.488 0.580) (0.645 0.737) (0.592 0.684) (0.593 0.685) (0.579 0.671) (0.361 0.453) (0.643, 0.735) (0.373, 0.536)
    MSCIMX 0.669 0.603 0.494 0.768 0.705 0.709 0.714 0.184 0.767 0.491 0.967
    (0.552 0.786) (0.525 0.681) (0.441 0.547) (0.715 0.821) (0.652 0.758) (0.656 0.762) (0.661 0.767) (0.132 0.237) (0.715, 0.820) (0.431, 0.593)
    MSCINL 0.672 -0.635 0.534 0.772 0.704 0.631 0.704 0.208 0.772 0.574 0.729
    (0.571 0.774) (0.569 0.702) (0.490 0.579) (0.728 0.817) (0.660 0.749) (0.586 0.675) (0.660 0.749) (0.163 0.252) (0.727, 0.816) (0.396, 0.585)
    MSCIWO 0.600 0.528 0.445 0.661 0.582 0.582 0.591 0.291 0.658 0.468 0.482
    (0.494 0.707) (0.457 0.599) (0.398 0.493) (0.614 0.708) (0.534 0.629) (0.535 0.630) (0.543 0.638) (0.243 0.338) (0.610, 0.705) (0.495, 0.652)
    MXX 0.575 0.503 0.441 0.601 0.576 0.576 0.624 0.343 0.599 0.418 0.848
    (0.491 0.659) (0.449 0.557) (0.406 0.476) (0.566 0.636) (0.541 0.611) (0.541 0.611) (0.589 0.659) (0.308 0.378) (0.564, 0.634) (0.357, 0.479)
    N2252 0.618 0.557 0.514 0.589 0.598 0.636 0.636 0.477 0.588 0.504 0.949
    (0.533 0.703) (0.502 0.611) (0.478 0.549) (0.554 0.625) (0.563 0.633) (0.601 0.672) (0.601 0.672) (0.442 0.513) (0.553, 0.624) (0.442, 0.565)
    NSEI 0.572 0.515 0.497 0.537 0.549 0.601 0.617 0.470 0.465 0.455 0.622
    (0.484 0.660) (0.458 0.572) (0.460 0.534) (0.500 0.574) (0.512 0.586) (0.564 0.638) (0.580 0.654) (0.433 0.507) (0.428, 0.502) (0.390, 0.520)
    RUA 0.646 0.583 0.539 0.663 0.664 0.667 0.665 0.440 0.662 0.541 0.534
    (0.554 0.738) (0.524 0.643) (0.499 0.578) (0.623 0.702) (0.624 0.703) (0.627 0.706) (0.626 0.704) (0.400 0.479) (0.622, 0.701) (0.385, 0.552)
    RUI 0.644 0.580 0.537 0.657 0.660 0.665 0.664 0.444 0.657 0.539 0.520
    (0.551 0.736) (0.521 0.640) (0.498 0.577) (0.618 0.697) (0.621 0.700) (0.626 0.705) (0.625 0.703) (0.405 0.484) (0.618, 0.697) (0.472, 0.611)
    RUT2 0.549 0.551 0.501 0.574 0.549 0.527 0.549 0.471 0.573 0.494 0.658
    (0.465 0.633) (0.497 0.605) (0.466 0.536) (0.539 0.609) (0.515 0.584) (0.492 0.562) (0.515 0.584) (0.436 0.506) (0.538, 0.608) (0.434, 0.555)
    SPTSE 0.681 0.600 0.508 0.742 0.660 0.643 0.652 0.275 0.742 0.485 1.169
    (0.581 0.781) (0.534 0.665) (0.464 0.551) (0.698 0.785) (0.617 0.704) (0.599 0.687) (0.608 0.696) (0.231 0.318) (0.698, 0.785) (0.470, 0.608)
    SPX2 0.623 0.586 0.548 0.610 0.591 0.573 0.568 0.534 0.608 0.546 0.562
    (0.539 0.707) (0.532 0.640) (0.514 0.583) (0.575 0.645) (0.556 0.626) (0.538 0.608) (0.533 0.603) (0.499 0.569) (0.573, 0.643) (0.485, 0.607)
    SSMI 0.656 0.671 0.590 0.694 0.695 0.707 0.705 0.581 0.693 0.573 1.119
    (0.572 0.740) (0.617 0.725) (0.555 0.625) (0.659 0.729) (0.660 0.730) (0.672 0.742) (0.670 0.740) (0.546 0.616) (0.659, 0.728) (0.512, 0.634)
    STOXX50E 0.591 0.594 0.527 0.613 0.618 0.637 0.637 0.480 0.611 0.494 0.828
    (0.507 0.675) (0.540 0.647) (0.492 0.562) (0.578 0.647) (0.584 0.653) (0.603 0.672) (0.602 0.672) (0.445 0.514) (0.576, 0.646) (0.433, 0.554)

     | Show Table
    DownLoad: CSV
    Table 4.  Estimated long-memory coefficients of the log-realized variances of the exchange rates.
    LW0.6 LW0.7 LW0.8 LWN LPWN(1,0) LPWN(0,1) LPWN(1,1) mLW mLWN tLW Qu0.75
    AUD 0.648 0.558 0.404 0.782 0.625 0.507 0.889 0.051 0.782 0.367 2.765**
    (0.575 0.722) (0.512 0.605) (0.375 0.433) (0.753 0.811) (0.596 0.655) (0.477 0.536) (0.859 0.918) (0.022 0.080) (0.753, 0.811) (0.317, 0.417)
    BRL 0.615 0.554 0.465 0.635 0.641 0.630 0.640 0.329 0.634 0.400 1.924**
    (0.530 0.699) (0.500 0.608) (0.430 0.500) (0.600 0.670) (0.606 0.677) (0.595 0.665) (0.605 0.675) (0.293 0.364) (0.599, 0.670) (0.339, 0.461)
    CAD 0.676 0.560 0.438 0.766 0.693 0.680 0.838 0.140 0.765 0.335 3.447**
    (0.603 0.749) (0.514 0.606) (0.408 0.467) (0.737 0.795) (0.663 0.722) (0.651 0.709) (0.809 0.867) (0.111 0.170) (0.735, 0.794) (0.285, 0.385)
    CHF 0.586 0.522 0.406 0.674 0.601 0.597 0.678 0.222 0.673 0.384 2.542**
    (0.513 0.659) (0.476 0.568) (0.377 0.435) (0.645 0.703) (0.572 0.630) (0.567 0.626) (0.649 0.707) (0.192 0.251) (0.644, 0.703) (0.334, 0.434)
    EUR 0.619 0.508 0.372 0.765 0.646 0.602 0.794 0.090 0.764 0.316 3.132**
    (0.544 0.695) (0.460 0.555) (0.342 0.403) (0.734 0.795) (0.616 0.677) (0.571 0.632) (0.764 0.825) (0.059 0.120) (0.733, 0.794) (0.264, 0.369)
    GBP 0.662 0.566 0.406 0.824 0.713 0.676 0.835 0.023 0.823 0.355 3.388**
    (0.590 0.735) (0.520 0.611) (0.377 0.435) (0.794 0.853) (0.684 0.742) (0.647 0.705) (0.806 0.864) (-0.006 0.052) (0.794, 0.852) (0.305, 0.405)
    INR 0.517 0.498 0.445 0.539 0.549 0.596 0.626 0.375 0.380 0.384 2.259**
    (0.436 0.597) (0.447 0.550) (0.412 0.478) (0.506 0.572) (0.516 0.582) (0.563 0.629) (0.593 0.659) (0.342 0.408) (0.347, 0.413) (0.327, 0.441)
    JPY 0.561 0.462 0.416 0.558 0.499 0.462 0.499 0.342 0.555 0.405 1.466*
    (0.488 0.633) (0.417 0.508) (0.387 0.445) (0.529 0.587) (0.470 0.528) (0.433 0.491) (0.470 0.528) (0.313 0.371) (0.526, 0.584) (0.356, 0.455)
    RUB 0.708 0.567 0.503 0.708 0.707 0.852 0.919 0.286 0.703 0.334 2.908*
    (0.617 0.798) (0.509 0.626) (0.464 0.541) (0.669 0.746) (0.669 0.746) (0.813 0.890) (0.881 0.957) (0.248 0.324) (0.664, 0.741) (0.267, 0.401)
    ZAR 0.758 0.679 0.538 0.842 0.751 0.726 0.886 0.113 0.842 0.481 3.178**
    (0.682 0.833) (0.631 0.727) (0.508 0.569) (0.811 0.872) (0.721 0.782) (0.696 0.757) (0.856 0.916) (0.083 0.143) (0.811, 0.872) (0.429, 0.534)

     | Show Table
    DownLoad: CSV
    Table 5.  Identification codes of the indices.
    Codes Indices Codes Indices
    AEX AEX Index MSCIDE MSCI Germany
    AORD All Ordinaries MSCIES MSCI Spain
    BVSP Bovespa Index MSCIFR MSCI France
    DJI Dow Jones Industrials MSCIGB MSCI UK
    FCHI CAC 40 MSCIIT MSCI Italy
    FTSE FTSE 100 MSCIJP MSCI Japan
    FTSEMIB FTSE MIB MSCIKR MSCI South Korea
    GDAXI German DAX MSCIMX MSCI Mexico
    GSPTSE S & P/TSX Composite Index MSCINL MSCI Netherlands
    HSI Hang Seng MSCIWO MSCI World
    IBEX Spanish IBEX MXX IPC Mexico
    IXIC Nasdaq 100 N2252 Nikkei 250
    KS11 KOSPI Composite Index NSEI S & P CNX Nifty
    MIB30 Milan MIB 30 RUA Russell 3000
    MIBTEL Italian MIBTEL RUI Russell 1000
    MID S & P 400 Midcap RUT2 Russell 2000
    MSCIAU MSCI Australia SPTSE S & P TSE
    MSCIBE MSCI Belgium SPX S & P 500
    MSCIBR MSCI Brazil SSMI Swiss Market Index
    MSCICA MSCI Canada STOXX50E Euro STOXX 50
    MSCICH MSCI Switzerland

     | Show Table
    DownLoad: CSV
    Table 6.  Identification codes of the exchange rates.
    AUD BRL CAD CHF EUR GBP INR JPY RUB ZAR
    USD/Australian Dollar USD/Brazilian Real USD/Canadian Dollar USD/Swiss Franc USD/Euro USD/British Pound USD/Indian Rupee USD/Japanese Yen USD/Russian Rouble USD/South African Rand

     | Show Table
    DownLoad: CSV
    Table 7.  Starting dates and available observations of the indices.
    Symbol Start Date Obs Symbol Start Date Obs
    AEX 03-01-2000 4440 MSCIDE 02-07-1999 2427
    AORD 04-01-2000 4358 MSCIES 02-07-1999 2441
    BVSP 03-01-2000 4266 MSCIFR 02-07-1999 2416
    DJI 03-01-2000 4360 MSCIGB 09-06-1999 2448
    FCHI 03-01-2000 4441 MSCIIT 02-07-1999 2443
    FTSE 04-01-2000 4379 MSCIJP 05-12-1999 2430
    FTSEMIB 03-01-2000 4398 MSCIKR 06-12-1999 2231
    GDAXI 03-01-2000 4413 MSCIMX 07-10-2002 2253
    GSPTSE 02-05-2002 3772 MSCINL 02-07-1999 1602
    HSI 03-01-2000 4026 MSCIWO 12-02-2001 2447
    IBEX 03-01-2000 4406 MXX 03-01-2000 4361
    IXIC 03-01-2000 4362 N2252 04-01-2000 4225
    KS11 04-01-2000 4290 NSEI 06-01-2000 3780
    MIB30 03-01-1996 3261 RUA 03-01-1996 2091
    MIBTEL 04-07-2000 3289 RUI 03-01-1996 3262
    MID 03-01-1996 2176 RUT 03-01-2000 4359
    MSCIAU 05-12-1999 3258 SPTSE 04-01-1999 3262
    MSCIBE 02-07-1999 2314 SPX 03-01-2000 4357
    MSCIBR 07-10-2002 2435 SSMI 04-01-2000 4364
    MSCICA 13-02-2001 1577 STOXX50E 03-01-2000 4417
    MSCICH 10-06-1999 2003

     | Show Table
    DownLoad: CSV

    To construct similar series for the exchange rates, we use 5-minute returns obtained from the Thomson Reuters Tick History database. The data is cleaned following the recommendations of Barndorff-Nielsen et al. (2009) to account for the typical high frequency data quality issues. Similar to the procedure of Heber et al. (2009), some additional manual edits are made so that the data is suitable for statistical inference. Since there is no market closure for exchange rates the log-realized variance is calculated based on all 5-minute log-returns within each 24-hour period. As for the indices, an overview of the meaning of the symbols is given in Table 6 and starting dates of the resulting series and the number of observations are given in Table 8. The last observation of all exchange rate series is from 31 January 2017.

    Table 8.  Starting dates and available observations of the exchange rates.
    Symbol Start Date Obs
    AUD 01-01-1996 6700
    BRL 27-10-2000 4311
    CAD 02-01-1996 6754
    CHF 02-01-1996 6781
    EUR 05-05-1998 6090
    GBP 01-01-1996 6856
    INR 01-01-1998 5001
    JPY 02-01-1996 6866
    RUB 06-01-2005 3483
    ZAR 15-02-1996 6087

     | Show Table
    DownLoad: CSV

    The results of the different long-memory estimators applied to the log-realized variances log zt of the indices are given in Tables 2 and 3. Theoretical confidence intervals are given in brackets below. For the Qu-test bold-faced values indicate significance at the nominal 10% level; an additional * (**) indicates significance at the nominal 5% (1%) level. Starting with the local Whittle estimates, we observe that they tend to decrease as the bandwidth increases from m=T0.6 to m=T0.8. Even though this decrease has a magnitude of up to 0.2 for some of the series, it is moderate for the majority of them. Nevertheless, this could be seen as an indication for low frequency contaminations in the respective series. The intuition behind this is given in Section 2: the long-memory component of a contaminated series dominates the low frequency contamination at higher frequencies such that the positive bias of the estimated d decreases if a larger bandwidth is chosen.

    Nearly all point estimates are within the lower non-stationary region between 0.5 and 0.6 and the vast majority of asymptotic confidence intervals are completely in the non-stationary region as well. If the impact of perturbations and low frequency contaminations is low, this is strong evidence that the memory of the indices is larger than 0.5.

    Turning to the LPWN estimates, we first observe that the estimates are very stable across the different specifications of the estimator. The level of the estimates tends to be higher than that of the local Whittle estimates. Most of them are in the range between 0.6 and 0.7. This further supports the previous finding that the index variances possess non-stationary long memory and points to the fact that the measurement error in the log-realized variance still has a magnitude so that it causes downward bias in the local Whittle estimator.

    The right hand side of Tables 2 and 3 shows the results of the mLW and tLW estimators of Hou and Perron (2014) and Iacone (2010) as well as the mLWN estimator. In all cases the mLW estimates are smaller than the local Whittle estimates and in some series the memory drops by a considerable amount. The same holds true for the tLW estimator of Iacone (2010), but the reduction in memory compared to the local Whittle estimates is of a smaller magnitude. This could be seen as evidence for low frequency contaminations in the variances. However, the results of the mLWN estimator are more in line with those of the LPWN estimators and the test of Qu (2011) fails to reject the null hypothesis of true long memory for the vast majority of index series. At the 5% significance level the test of Qu (2011) only rejects for GSPTSE, HSI, MSCIBE, MSCIDE and MSCIJP.

    Having in mind that the mLW and tLW estimators are severely downward biased in the presence of moderate perturbations, we therefore conclude that there is no evidence for spurious long memory. Hence, the LPWN estimator is most suitable to give the best estimate of the true memory of the variances of the stock indices under consideration.x

    We therefore find that the memory of stock index variance is non-stationary. Most stock index variances display memory parameters in the range between 0.6 and 0.7, which is far in the non-stationary region.

    In Table 4 we report the results for the realized variance of exchange rates. Here, we observe some major differences compared to the results for the variances of indices, discussed above. First of all, the local Whittle estimates decrease heavily when the bandwidth increases. Again, this can be seen as an indication of low frequency contaminations by the same arguments as discussed above.The assertion that the exchange rate variances exhibit spurious long memory is further supported by the fact that both the mLW estimator of Hou and Perron (2014) and the tLW estimator of Iacone (2010) are reduced compared to the local Whittle estimates. But most importantly, the Qu (2011) test rejects strongly for all currencies considered. This is clear evidence for the presence of spurious long memory. Only the mLWN estimator does not show evidence for a lower degree of memory. However, the results show a very high variability and we know from the simulations in the previous section that the estimator fails to control the spurious long memory bias, if the level shift component is large. Finally, the LPWN estimates are much higher compared to the local Whittle estimates, which is also consistent with the observation that the LPWN estimators have a larger spurious long memory bias than the standard local Whittle estimator, as shown in our simulations.

    If we now consider the mLW and tLW estimators that are most likely to give consistent estimates in this setup, we observe that all estimates are in the stationary region. In particular the estimator of Iacone (2010) gives estimates that lie consistently between 0.3 and 0.4. The results for the Hou and Perron (2014) estimator are a bit more variable and are found to be in the range between 0.05 and 0.4.

    To ensure the robustness of our findings, we consider a variety of alternative specifications. The results of these exercises are given in the appendix. First, we consider realized kernels instead of realized variances for the indices. Realized kernels are a measure of variance that is more robust to market microstructure effects. The results are given in Tables 9 and 10 in the appendix. Furthermore, in Tables 11 and 12, we construct confidence intervals for the local Whittle estimator using the frequency domain bootstrap procedure of Arteche and Orbe (2016). Finally, we apply the trimmed log-periodogram regression of McCloskey and Perron (2013) as an alternative to estimate the memory robust to spurious long memory. These results can be found in Tables 13 and 14. The results of all these analyses are remarkably similar to those presented here, which highlights the robustness of our findings.

    Table 9.  Estimated long-memory coefficients of the log-realized kernels.
    LW0.6 LW0.7 LW0.8 LWN LPWN(1,0) LPWN(0,1) LPWN(1,1) mLW mLWN tLW Qu0.75
    AEX 0.618 0.630 0.576 0.654 0.652 0.656 0.654 0.566 0.654 0.573 0.561
    (0.535 0.702) (0.576 0.683) (0.541 0.610) (0.619 0.689) (0.618 0.687) (0.621 0.691) (0.619 0.688) (0.531 0.601) (0.619, 0.689) (0.513, 0.634)
    AORD 0.555 0.593 0.471 0.647 0.613 0.619 0.613 0.364 0.645 0.458 1.003
    (0.471 0.639) (0.539 0.647) (0.436 0.506) (0.612 0.682) (0.578 0.648) (0.584 0.654) (0.578 0.648) (0.329 0.399) (0.610, 0.680) (0.398, 0.519)
    BVSP 0.548 0.522 0.475 0.566 0.532 0.530 0.532 0.458 0.566 0.482 0.521
    (0.464 0.633) (0.468 0.577) (0.440 0.510) (0.531 0.601) (0.496 0.567) (0.495 0.565) (0.496 0.567) (0.423 0.493) (0.531, 0.601) (0.421, 0.544)
    DJI 0.631 0.605 0.561 0.631 0.610 0.590 0.610 0.548 0.630 0.568 0.730
    (0.547 0.715) (0.551 0.659) (0.526 0.596) (0.596 0.666) (0.576 0.645) (0.555 0.624) (0.576 0.645) (0.513 0.583) (0.595, 0.665) (0.507, 0.629)
    FCHI 0.608 0.623 0.553 0.645 0.648 0.659 0.658 0.494 0.644 0.536 0.668
    (0.525 0.692) (0.569 0.676) (0.518 0.587) (0.610 0.680) (0.613 0.683) (0.624 0.693) (0.623 0.692) (0.459 0.528) (0.609, 0.678) (0.476, 0.597)
    FTSE 0.650 0.650 0.591 0.673 0.674 0.671 0.671 0.578 0.671 0.588 0.663
    (0.566 0.733) (0.596 0.704) (0.557 0.626) (0.638 0.707) (0.639 0.709) (0.636 0.706) (0.636 0.706) (0.543 0.613) (0.636, 0.706) (0.527, 0.649)
    FTSEMIB 0.597 0.624 0.555 0.643 0.622 0.630 0.622 0.532 0.642 0.565 0.581
    (0.513 0.680) (0.570 0.677) (0.520 0.589) (0.608 0.678) (0.588 0.657) (0.595 0.665) (0.588 0.657) (0.497 0.566) (0.607, 0.676) (0.504, 0.625)
    GDAXI 0.649 0.643 0.573 0.668 0.673 0.715 0.715 0.487 0.667 0.524 0.714
    (0.566 0.733) (0.590 0.697) (0.538 0.607) (0.634 0.703) (0.638 0.707) (0.680 0.749) (0.681 0.750) (0.452 0.522) (0.633, 0.702) (0.464, 0.585)
    GSPTSE 0.612 0.578 0.509 0.640 0.648 0.653 0.653 0.404 0.563 0.462 1.281*
    (0.524 0.701) (0.521 0.635) (0.472 0.546) (0.603 0.677) (0.611 0.685) (0.616 0.690) (0.616 0.690) (0.367 0.441) (0.526, 0.600) (0.398, 0.527)
    HSI 0.641 0.561 0.522 0.616 0.624 0.669 0.701 0.433 0.503 0.460 1.261*
    (0.555 0.727) (0.506 0.617) (0.486 0.558) (0.580 0.652) (0.588 0.660) (0.633 0.705) (0.665 0.737) (0.397 0.470) (0.467, 0.539) (0.397, 0.523)
    IBEX 0.593 0.596 0.552 0.608 0.615 0.624 0.624 0.518 0.604 0.533 0.763
    (0.509 0.676) (0.542 0.650) (0.517 0.586) (0.574 0.643) (0.581 0.650) (0.589 0.659) (0.589 0.659) (0.483 0.553) (0.570, 0.639) (0.473, 0.594)
    IXIC 0.657 0.608 0.570 0.634 0.625 0.594 0.603 0.545 0.584 0.562 0.578
    (0.573 0.741) (0.554 0.661) (0.535 0.605) (0.599 0.669) (0.590 0.660) (0.559 0.629) (0.568 0.638) (0.510 0.580) (0.549, 0.619) (0.501, 0.623)
    KS11 0.699 0.632 0.558 0.680 0.684 0.735 0.739 0.430 0.678 0.518 1.420*
    (0.614 0.783) (0.578 0.686) (0.523 0.594) (0.644 0.715) (0.649 0.719) (0.699 0.770) (0.704 0.774) (0.395 0.465) (0.643, 0.713) (0.457, 0.579)
    MIB30 0.601 0.605 0.530 0.671 0.651 0.657 0.650 0.444 0.670 0.541 0.653
    (0.509 0.694) (0.545 0.665) (0.491 0.569) (0.632 0.711) (0.612 0.690) (0.618 0.696) (0.611 0.690) (0.405 0.483) (0.631, 0.709) (0.466, 0.604)
    MIBTEL 0.659 0.611 0.513 0.737 0.686 0.693 0.686 0.297 0.736 0.515 1.038
    (0.554 0.765) (0.541 0.681) (0.466 0.559) (0.691 0.784) (0.640 0.733) (0.647 0.740) (0.639 0.732) (0.251 0.344) (0.689, 0.782) (0.472, 0.610)
    MID 0.713 0.634 0.560 0.729 0.733 0.756 0.756 0.362 0.728 0.524 1.204
    (0.621 0.806) (0.575 0.694) (0.521 0.600) (0.690 0.768) (0.693 0.772) (0.717 0.796) (0.716 0.795) (0.323 0.401) (0.689, 0.768) (0.433, 0.598)
    MSCIAU 0.638 0.588 0.476 0.727 0.684 0.687 0.683 0.265 0.724 0.445 0.696
    (0.535 0.742) (0.520 0.656) (0.431 0.521) (0.682 0.773) (0.639 0.729) (0.642 0.732) (0.638 0.728) (0.220 0.311) (0.679, 0.770) (0.455, 0.593)
    MSCIBE 0.715 0.631 0.511 0.793 0.731 0.731 0.757 0.154 0.792 0.467 1.240
    (0.613 0.817) (0.564 0.698) (0.466 0.555) (0.749 0.837) (0.687 0.776) (0.686 0.775) (0.712 0.801) (0.109 0.198) (0.747, 0.836) (0.364, 0.525)
    MSCIBR 0.603 0.548 0.490 0.644 0.649 0.654 0.653 0.382 0.641 0.481 0.447
    (0.485 0.720) (0.469 0.626) (0.437 0.543) (0.590 0.697) (0.596 0.703) (0.601 0.707) (0.600 0.706) (0.329 0.435) (0.588, 0.694) (0.388, 0.546)
    MSCICA 0.660 0.598 0.483 0.749 0.673 0.677 0.670 0.223 0.747 0.484 0.641
    (0.551 0.769) (0.526 0.670) (0.435 0.531) (0.701 0.797) (0.624 0.721) (0.629 0.726) (0.622 0.718) (0.175 0.271) (0.699, 0.795) (0.385, 0.577)
    MSCICH 0.706 0.646 0.537 0.772 0.720 0.722 0.743 0.237 0.771 0.505 1.125
    (0.604 0.807) (0.579 0.713) (0.492 0.581) (0.727 0.816) (0.675 0.764) (0.677 0.766) (0.698 0.787) (0.192 0.281) (0.727, 0.815) (0.398, 0.570)

     | Show Table
    DownLoad: CSV
    Table 10.  Estimated long-memory coefficients of the log-realized kernels.
    LW0.6 LW0.7 LW0.8 LWN LPWN(1,0) LPWN(0,1) LPWN(1,1) mLW mLWN tLW Qu0.75
    MSCIDE 0.666 0.630 0.502 0.779 0.711 0.707 0.761 0.179 0.778 0.443 0.678
    (0.564 0.768) (0.563 0.696) (0.457 0.546) (0.734 0.823) (0.666 0.755) (0.663 0.752) (0.717 0.806) (0.135 0.224) (0.733, 0.822) (0.426, 0.584)
    MSCIES 0.628 0.604 0.499 0.706 0.654 0.657 0.653 0.339 0.703 0.488 0.673
    (0.526 0.730) (0.537 0.671) (0.454 0.543) (0.662 0.751) (0.609 0.698) (0.613 0.702) (0.609 0.698) (0.295 0.384) (0.658, 0.747) (0.364, 0.522)
    MSCIFR 0.649 0.614 0.501 0.767 0.679 0.675 0.726 0.194 0.766 0.490 1.302*
    (0.548 0.751) (0.548 0.681) (0.457 0.546) (0.723 0.811) (0.635 0.724) (0.631 0.720) (0.682 0.770) (0.150 0.239) (0.722, 0.810) (0.409, 0.568)
    MSCIGB 0.674 0.629 0.524 0.765 0.696 0.696 0.720 0.237 0.764 0.514 1.113
    (0.572 0.776) (0.562 0.696) (0.480 0.568) (0.720 0.809) (0.651 0.740) (0.652 0.741) (0.676 0.765) (0.193 0.281) (0.719, 0.808) (0.411, 0.568)
    MSCIIT 0.625 0.609 0.510 0.725 0.672 0.677 0.675 0.315 0.723 0.496 1.260*
    (0.524 0.727) (0.542 0.676) (0.466 0.555) (0.680 0.769) (0.628 0.717) (0.633 0.722) (0.631 0.720) (0.271 0.360) (0.679, 0.767) (0.436, 0.593)
    MSCIJP 0.673 0.568 0.508 0.628 0.634 0.682 0.686 0.457 0.624 0.453 1.362*
    (0.568 0.777) (0.499 0.637) (0.462 0.554) (0.582 0.674) (0.588 0.680) (0.636 0.728) (0.640 0.732) (0.411 0.503) (0.578, 0.670) (0.417, 0.575)
    MSCIKR 0.678 0.602 0.519 0.673 0.621 0.605 0.601 0.400 0.671 0.494 0.925
    (0.574 0.782) (0.533 0.671) (0.473 0.564) (0.628 0.719) (0.575 0.667) (0.559 0.651) (0.555 0.647) (0.354 0.446) (0.625, 0.716) (0.371, 0.534)
    MSCIMX 0.640 0.576 0.470 0.748 0.669 0.673 0.678 0.189 0.747 0.464 0.930
    (0.523 0.757) (0.499 0.654) (0.417 0.523) (0.696 0.801) (0.616 0.722) (0.620 0.725) (0.626 0.731) (0.136 0.242) (0.695, 0.800) (0.414, 0.575)
    MSCINL 0.662 -0.621 0.522 0.755 0.678 0.641 0.678 0.252 0.755 0.532 0.825
    (0.561 0.764) (0.555 0.688) (0.478 0.566) (0.711 0.800) (0.634 0.722) (0.597 0.685) (0.634 0.722) (0.208 0.296) (0.710, 0.799) (0.370, 0.559)
    MSCIWO 0.596 0.526 0.428 0.675 0.599 0.600 0.597 0.247 0.672 0.432 0.540
    (0.490 0.703) (0.455 0.597) (0.381 0.475) (0.628 0.722) (0.552 0.646) (0.553 0.647) (0.550 0.644) (0.200 0.294) (0.625, 0.719) (0.454, 0.611)
    MXX 0.585 0.522 0.447 0.620 0.629 0.625 0.634 0.329 0.618 0.404 0.692
    (0.501 0.669) (0.468 0.576) (0.412 0.482) (0.585 0.655) (0.594 0.664) (0.590 0.660) (0.599 0.669) (0.294 0.364) (0.583, 0.653) (0.343, 0.465)
    N2252 0.627 0.552 0.512 0.579 0.588 0.643 0.643 0.498 0.577 0.496 1.005
    (0.542 0.712) (0.498 0.607) (0.477 0.548) (0.543 0.614) (0.552 0.623) (0.607 0.678) (0.607 0.678) (0.462 0.533) (0.542, 0.613) (0.434, 0.558)
    NSEI 0.579 0.518 0.523 0.539 0.552 0.596 0.611 0.500 0.500 0.492 0.628
    (0.492 0.667) (0.461 0.575) (0.486 0.560) (0.502 0.576) (0.515 0.589) (0.559 0.633) (0.574 0.648) (0.463 0.537) (0.462, 0.537) (0.427, 0.557)
    RUA 0.648 0.582 0.540 0.659 0.664 0.669 0.668 0.444 0.659 0.539 0.558
    (0.556 0.741) (0.523 0.642) (0.501 0.579) (0.620 0.699) (0.625 0.703) (0.630 0.709) (0.629 0.707) (0.404 0.483) (0.620, 0.699) (0.348, 0.516)
    RUI 0.646 0.580 0.538 0.655 0.660 0.668 0.667 0.446 0.655 0.536 0.555
    (0.554 0.738) (0.520 0.640) (0.498 0.577) (0.616 0.694) (0.621 0.700) (0.628 0.707) (0.627 0.706) (0.406 0.485) (0.616, 0.694) (0.470, 0.608)
    RUT 0.567 0.564 0.514 0.589 0.563 0.540 0.563 0.479 0.587 0.506 0.654
    (0.483 0.651) (0.510 0.618) (0.479 0.549) (0.554 0.624) (0.529 0.598) (0.505 0.575) (0.529 0.598) (0.444 0.514) (0.552, 0.622) (0.445, 0.567)
    SPTSE 0.672 0.589 0.503 0.729 0.653 0.633 0.641 0.290 0.728 0.479 0.909
    (0.572 0.772) (0.523 0.655) (0.459 0.547) (0.685 0.773) (0.609 0.697) (0.589 0.676) (0.597 0.685) (0.246 0.333) (0.684, 0.772) (0.467, 0.606)
    SPX 0.628 0.596 0.564 0.618 0.603 0.588 0.581 0.550 0.617 0.567 0.480
    (0.544 0.712) (0.542 0.650) (0.529 0.599) (0.583 0.653) (0.568 0.638) (0.553 0.623) (0.546 0.616) (0.516 0.585) (0.582, 0.652) (0.506, 0.628)
    SSMI 0.658 0.688 0.615 0.706 0.705 0.714 0.712 0.560 0.706 0.611 0.861
    (0.574 0.742) (0.634 0.741) (0.580 0.650) (0.671 0.741) (0.670 0.740) (0.679 0.749) (0.677 0.747) (0.525 0.595) (0.671, 0.741) (0.550, 0.672)
    STOXX50E 0.598 0.594 0.533 0.614 0.619 0.638 0.639 0.484 0.612 0.503 0.678
    (0.514 0.681) (0.541 0.648) (0.499 0.568) (0.579 0.648) (0.585 0.654) (0.604 0.673) (0.604 0.673) (0.450 0.519) (0.577, 0.647) (0.443, 0.564)

     | Show Table
    DownLoad: CSV
    Table 11.  Estimated long-memory coefficients of the log-realized variances. Confidence intervals using the bootstrap procedure of Arteche and Orbe (2016) are given in brackets below.
    LW0.6 LW0.7 LW0.8 LW0.6 LW0.7 LW0.8
    AEX 0.625 0.635 0.577 MSCIDE 0.680 0.639 0.521
    (0.525 0.721) (0.580 0.693) (0.540 0.613) (0.616 0.818) (0.607 0.742) (0.525 0.612)
    AORD 0.553 0.586 0.460 MSCIES 0.640 0.623 0.519
    (0.466 0.629) (0.526 0.635) (0.424 0.498) (0.583 0.788) (0.574 0.711) (0.463 0.573)
    BVSP 0.549 0.521 0.475 MSCIFR 0.656 0.619 0.522
    (0.450 0.626) (0.466 0.587) (0.440 0.509) (0.531 0.732) (0.555 0.683) (0.460 0.571)
    DJI 0.622 0.586 0.534 MSCIGB 0.693 0.643 0.544
    (0.535 0.696) (0.527 0.637) (0.502 0.569) (0.533 0.763) (0.546 0.688) (0.466 0.573)
    FCHI 0.605 0.624 0.564 MSCIIT 0.633 0.628 0.532
    (0.526 0.681) (0.569 0.676) (0.530 0.599) (0.596 0.792) (0.570 0.713) (0.493 0.593)
    FTSE 0.650 0.640 0.569 MSCIJP 0.685 0.568 0.514
    (0.559 0.730) (0.579 0.693) (0.533 0.606) (0.541 0.728) (0.560 0.691) (0.479 0.581)
    FTSEMIB 0.600 0.607 0.547 MSCIKR 0.695 0.625 0.534
    (0.512 0.686) (0.549 0.656) (0.510 0.584) (0.571 0.794) (0.488 0.641) (0.466 0.560)
    GDAXI 0.653 0.637 0.564 MSCIMX 0.669 0.603 0.494
    (0.544 0.757) (0.573 0.697) (0.523 0.601) (0.596 0.794) (0.555 0.687) (0.492 0.581)
    GSPTSE 0.603 0.565 0.490 MSCINL 0.672 -0.635 0.534
    (0.508 0.681) (0.506 0.621) (0.452 0.528) (0.553 0.777) (0.523 0.675) (0.436 0.550)
    HSI 0.640 0.557 0.503 MSCIWO 0.600 0.528 0.445
    (0.546 0.724) (0.500 0.604) (0.467 0.541) (0.562 0.765) (0.573 0.695) (0.480 0.590)
    IBEX 0.593 0.596 0.545 MXX 0.575 0.503 0.441
    (0.498 0.668) (0.542 0.651) (0.504 0.576) (0.506 0.661) (0.448 0.558) (0.406 0.470)
    IXIC 0.644 0.598 0.552 N2252 0.618 0.557 0.514
    (0.564 0.708) (0.545 0.648) (0.516 0.586) (0.509 0.701) (0.497 0.609) (0.474 0.549)
    KS11 0.692 0.622 0.548 NSEI 0.572 0.515 0.497
    (0.604 0.767) (0.570 0.672) (0.510 0.582) (0.475 0.656) (0.456 0.567) (0.464 0.533)
    MIB30 0.608 0.613 0.550 RUA 0.646 0.583 0.539
    (0.537 0.703) (0.511 0.633) (0.491 0.567) (0.489 0.729) (0.429 0.615) (0.381 0.511)
    MIBTEL 0.667 0.622 0.532 RUI 0.644 0.580 0.537
    (0.509 0.681) (0.551 0.670) (0.512 0.589) (0.558 0.722) (0.517 0.641) (0.502 0.575)
    MID 0.712 0.637 0.564 RUT 0.549 0.551 0.501
    (0.571 0.746) (0.558 0.683) (0.475 0.582) (0.459 0.625) (0.497 0.603) (0.468 0.534)
    MSCIAU 0.659 0.606 0.499 SPTSE 0.681 0.600 0.508
    (0.626 0.802) (0.574 0.701) (0.526 0.603) (0.554 0.719) (0.517 0.634) (0.500 0.571)
    MSCIBE 0.730 0.656 0.536 SPX 0.623 0.586 0.548
    (0.547 0.741) (0.540 0.672) (0.452 0.544) (0.540 0.693) (0.528 0.638) (0.516 0.581)
    MSCIBR 0.629 0.567 0.501 SSMI 0.656 0.671 0.590
    (0.636 0.823) (0.594 0.714) (0.490 0.577) (0.561 0.742) (0.618 0.718) (0.558 0.624)
    MSCICA 0.681 0.614 0.503 STOXX50E 0.591 0.594 0.527
    (0.552 0.753) (0.489 0.642) (0.448 0.557) (0.498 0.685) (0.541 0.644) (0.491 0.560)
    MSCICH 0.730 0.676 0.565
    (0.542 0.782) (0.524 0.689) (0.444 0.574)

     | Show Table
    DownLoad: CSV
    Table 12.  Estimated long-memory coefficients of the log-realized kernels. Confidence intervals using the bootstrap procedure of Arteche and Orbe (2016) are given in brackets below.
    LW0.6 LW0.7 LW0.8 LW0.6 LW0.7 LW0.8
    AEX 0.618 0.630 0.576 MSCIDE 0.680 0.639 0.521
    (0.512 0.711) (0.573 0.687) (0.536 0.611) (0.616 0.818) (0.607 0.742) (0.525 0.612)
    AORD 0.555 0.593 0.471 MSCIES 0.640 0.623 0.519
    (0.464 0.635) (0.534 0.647) (0.436 0.507) (0.583 0.788) (0.574 0.711) (0.463 0.573)
    BVSP 0.548 0.522 0.475 MSCIFR 0.656 0.619 0.522
    (0.459 0.620) (0.468 0.577) (0.436 0.510) (0.531 0.732) (0.555 0.683) (0.460 0.571)
    DJI 0.631 0.605 0.561 MSCIGB 0.693 0.643 0.544
    (0.539 0.702) (0.548 0.655) (0.525 0.599) (0.533 0.763) (0.546 0.688) (0.466 0.573)
    FCHI 0.608 0.623 0.553 MSCIIT 0.633 0.628 0.532
    (0.518 0.692) (0.569 0.676) (0.515 0.587) (0.596 0.792) (0.570 0.713) (0.493 0.593)
    FTSE 0.650 0.650 0.591 MSCIJP 0.685 0.568 0.514
    (0.552 0.728) (0.590 0.701) (0.556 0.627) (0.541 0.728) (0.560 0.691) (0.479 0.581)
    FTSEMIB 0.597 0.624 0.555 MSCIKR 0.695 0.625 0.534
    (0.509 0.683) (0.565 0.672) (0.519 0.590) (0.571 0.794) (0.488 0.641) (0.466 0.560)
    GDAXI 0.649 0.643 0.573 MSCIMX 0.669 0.603 0.494
    (0.53 0.763) (0.577 0.702) (0.533 0.612) (0.596 0.794) (0.555 0.687) (0.492 0.581)
    GSPTSE 0.612 0.578 0.509 MSCINL 0.672 -0.635 0.534
    (0.508 0.688) (0.522 0.630) (0.470 0.545) (0.553 0.777) (0.523 0.675) (0.436 0.550)
    HSI 0.641 0.561 0.522 MSCIWO 0.600 0.528 0.445
    (0.543 0.715) (0.509 0.609) (0.486 0.556) (0.562 0.765) (0.573 0.695) (0.480 0.590)
    IBEX 0.593 0.596 0.552 MXX 0.575 0.503 0.441
    (0.493 0.675) (0.544 0.646) (0.514 0.587) (0.506 0.661) (0.448 0.558) (0.406 0.470)
    IXIC 0.657 0.608 0.570 N2252 0.618 0.557 0.514
    (0.577 0.722) (0.552 0.659) (0.536 0.602) (0.509 0.701) (0.497 0.609) (0.474 0.549)
    KS11 0.699 0.632 0.558 NSEI 0.572 0.515 0.497
    (0.608 0.777) (0.576 0.679) (0.526 0.591) (0.475 0.656) (0.456 0.567) (0.464 0.533)
    MIB30 0.601 0.605 0.530 RUA 0.646 0.583 0.539
    (0.495 0.676) (0.542 0.661) (0.489 0.564) (0.489 0.729) (0.429 0.615) (0.381 0.511)
    MIBTEL 0.659 0.611 0.513 RUI 0.644 0.580 0.537
    (0.560 0.738) (0.545 0.672) (0.458 0.557) (0.558 0.722) (0.517 0.641) (0.502 0.575)
    MID 0.713 0.634 0.560 RUT 0.549 0.551 0.501
    (0.624 0.796) (0.566 0.693) (0.518 0.597) (0.459 0.625) (0.497 0.603) (0.468 0.534)
    MSCIAU 0.638 0.588 0.476 SPTSE 0.681 0.600 0.508
    (0.535 0.727) (0.521 0.646) (0.430 0.518) (0.554 0.719) (0.517 0.634) (0.500 0.571)
    MSCIBE 0.715 0.631 0.511 SPX 0.623 0.586 0.548
    (0.607 0.814) (0.564 0.695) (0.462 0.553) (0.540 0.693) (0.528 0.638) (0.516 0.581)
    MSCIBR 0.603 0.548 0.490 SSMI 0.656 0.671 0.590
    (0.526 0.713) (0.471 0.627) (0.436 0.548) (0.561 0.742) (0.618 0.718) (0.558 0.624)
    MSCICA 0.660 0.598 0.483 STOXX50E 0.591 0.594 0.527
    (0.521 0.771) (0.517 0.670) (0.423 0.544) (0.498 0.685) (0.541 0.644) (0.491 0.560)
    MSCICH 0.706 0.646 0.537
    (0.607 0.793) (0.582 0.708) (0.494 0.583)

     | Show Table
    DownLoad: CSV
    Table 13.  Estimated long-memory coefficients of the log-realized variances applying the estimator of McCloskey and Perron (2013) with theoretical confidence intervals in brackets below.
    Variances Confidence Variances Confidence
    AEX 0.570 MSCIDE 0.510
    (0.525, 0.614) (0.537, 0.651)
    AORD 0.433 MSCIES 0.507
    (0.385, 0.481) (0.452, 0.569)
    BVSP 0.467 MSCIFR 0.541
    (0.420, 0.514) (0.448, 0.566)
    DJI 0.559 MSCIGB 0.601
    (0.514, 0.604) (0.484, 0.597)
    FCHI 0.569 MSCIIT 0.557
    (0.525, 0.614) (0.544, 0.658)
    FTSE 0.554 MSCIJP 0.538
    (0.510, 0.599) (0.500, 0.614)
    FTSEMIB 0.557 MSCIKR 0.566
    (0.512, 0.601) (0.478, 0.597)
    GDAXI 0.573 MSCIMX 0.509
    (0.529, 0.618) (0.507, 0.625)
    GSPTSE 0.458 MSCINL 0.563
    (0.409, 0.508) (0.438, 0.579)
    HSI 0.503 MSCIWO 0.401
    (0.455, 0.550) (0.506, 0.620)
    IBEX 0.557 MXX 0.444
    (0.513, 0.602) (0.397, 0.491)
    IXIC 0.575 N2252 0.487
    (0.531, 0.620) (0.440, 0.534)
    KS11 0.541 NSEI 0.478
    (0.496, 0.586) (0.429, 0.527)
    MIB30 0.578 RUA 0.569
    (0.522, 0.623) (0.332, 0.469)
    MIBTEL 0.548 RUI 0.567
    (0.528, 0.629) (0.518, 0.619)
    MID 0.579 RUT2 0.506
    (0.488, 0.608) (0.460, 0.552)
    MSCIAU 0.487 SPTSE 0.513
    (0.528, 0.629) (0.517, 0.618)
    MSCIBE 0.545 SPX2 0.564
    (0.427, 0.547) (0.520, 0.609)
    MSCIBR 0.481 SSMI 0.596
    (0.488, 0.602) (0.551, 0.641)
    MSCICA 0.503 STOXX50E 0.526
    (0.408, 0.553) (0.481, 0.570)
    MSCICH 0.594
    (0.439, 0.567)

     | Show Table
    DownLoad: CSV
    Table 14.  Estimated long-memory coefficients of the log-realized variances applying the estimator of McCloskey and Perron (2013) with theoretical confidence intervals in brackets below.
    AUD BRL CAD CHF EUR GBP INR JPY RUB ZAR
    0.406 0.441 0.390 0.352 0.260 0.313 0.433 0.401 0.423 0.506
    (0.366, 0.447) (0.394, 0.489) (0.349, 0.431) (0.310, 0.394) (0.211, 0.309) (0.269, 0.357) (0.388, 0.478) (0.361, 0.441) (0.370, 0.476) (0.467, 0.546)

     | Show Table
    DownLoad: CSV

    Considering these results, we are able to establish a number of key findings. First of all, there is a considerable difference between the behavior of stock index variances and exchange rate variances. The stock index variances exhibit true long memory in the non-stationary range between 0.6 and 0.7. In contrast to that, the exchange rate variances show clear signs of spurious long memory and the true long memory of the series is only around 0.3.

    In Section 2 we discuss the effect of measurement error and level shifts on estimates of the memory parameter in log-realized variances using the local Whittle estimator of Küunsch (1987) and Robinson (1995a). In the recent literature a large number of new local Whittle estimators has been proposed that are robust to these effects, most importantly those of Hurvich et al. (2005), Frederiksen et al. (2012), Iacone (2010) and Hou and Perron (2014). These are discussed in Section 3, where we also conduct a simulation study to evaluate the performance of these methods if both of these complications are incurred at the same time. We find that, while the estimators are successful in mitigating the bias they are build to address, they become more vulnerable to the bias they do not account for. That means the LPWN estimator has a larger bias due to spurious long memory than the standard local Whittle estimator and the modified and trimmed local Whittle estimators have a larger bias in presence of perturbations. In our empirical application we are able to establish some new stylized facts about the memory in realized variances. Considering a wide range of stock indices, we find that the index variances are true long-memory processes with a memory parameter between 0.6 and 0.7, which is in the non-stationary range. As discussed in the introduction, this means that long memory stochastic volatility models are able to reproduce the finding that the kurtosis of stock market returns is infinite. Exchange rate variances, however, exhibit spurious long memory and the true memory parameters are between 0.3 and 0.4, which is far in the stationary region.

    Financial support of the Deutsche Forschungsgesellschaft (DFG) is gratefully acknowledged. We would like to thank the anonymous referee for his review. We highly appreciate his comments and suggestions.

    All authors declare no conflicts of interest in this paper.

    [1] Andersen TG, Bollerslev T, Diebold FX, et al. (2001) The distribution of realized exchange rate volatility. J Am Statist Associ 96: 42-55.
    [2] Andersen TG, Bollerslev T, Diebold FX, et al. (2003) Modeling and forecasting realized volatility. Econom 71: 579-625.
    [3] Andrews DWK, Sun Y (2004) Adaptive Local Polynomial Whittle Estimation of Long-range Dependence. Econom 72: 569-614.
    [4] Arteche J (2004) Gaussian semiparametric estimation in long memory in stochastic volatility and signal plus noise models. J Econom 119: 131-154.
    [5] Arteche J, Orbe J (2016) A bootstrap approximation for the distribution of the Local Whittle estimator. Comput Stat Data Anal 100: 645-660.
    [6] Barndorff-Nielsen OE, Hansen PR, Lunde A, et al. (2009) Realized kernels in practice: Trades and quotes. Econom J 12: C1-C32.
    [7] Barndorff-Nielsen OE, Shephard N (2002) Econometric analysis of realized volatility and its use in estimating stochastic volatility models. J Royal Stat Soc: Ser B (Statist Methodol) 64: 253-280.
    [8] Chiriac R, Voev V (2011) Modelling and forecasting multivariate realized volatility. J Appl Econom 26: 922-947.
    [9] Corsi F (2009) A simple approximate long-memory model of realized volatility. J Financ Econom 7: 174-196.
    [10] Deo RS, Hurvich CM, Lu Y (2006) Forecasting realized volatility using a long-memory stochastic volatility model: estimation, prediction and seasonal adjustment. J Econom 131: 29-58.
    [11] Deo RS, Hurvich CM (2001) On the log periodogram regression estimator of the memory parameter in long memory stochastic volatility models. Econom Theory 17: 686-710.
    [12] Diebold FX, Inoue A (2001) Long memory and regime switching. J Econom 105: 131-159.
    [13] Frederiksen P, Nielsen FS, Nielsen M (2012) Local polynomial Whittle estimation of perturbed fractional processes. J Econom 167: 426-447.
    [14] Geweke J, Porter-Hudak S (1983) The estimation and application of long memory time series models. J Time Ser Anal 4: 221-238.
    [15] Giraitis L, Leipus R, Surgailis D (2007) Recent advances in ARCH modelling. Long Memory Econom 3-38.
    [16] Giraitis L, Leipus R, Surgailis D (2009) ARCH infinity models and long memory properties. Handb Financ Time Ser 71-84.
    [17] Gourieroux C, Jasiak J (2001) Memory and infrequent breaks. Econom Lett 70: 29-41.
    [18] Granger CWJ, Ding Z (1996) Varieties of long memory models. J Econom 73: 61-77.
    [19] Granger CWJ, Hyung N (2004) Occasional structural breaks and long memory with an application to the S & P 500 absolute stock returns. J Empir Financ 11: 399-421.
    [20] Heber G, Lunde A, Shephard N, et al. (2009) Oxford-Man Institutes realized library, version 0. 2, Oxford-Man Institute, University of Oxford.
    [21] Hou J, Perron P (2014) Modified local Whittle estimator for long memory processes in the presence of low frequency (and other) contaminations. J Econom 182: 309-328.
    [22] Hurvich CM, Moulines E, Soulier P (2005) Estimating long memory in volatility. Econom 73: 1283-1328.
    [23] Hurvich CM, Ray BK (2003) The local Whittle estimator of long-memory stochastic volatility. J Financ Econom 1: 445-470.
    [24] Iacone F (2010) Local Whittle estimation of the memory parameter in presence of deterministic components. J Time Ser Anal 31: 37-49.
    [25] Künsch HR (1987) Statistical aspects of self-similar processes. Proc First World Congr Bernoulli Soc 1: 67-74.
    [26] Leccadito A, Rachedi O, Urga G (2015) True versus spurious long memory: Some theoretical results and a monte carlo comparison. Econom Rev 34: 452-479.
    [27] Leschinski C, Sibbertsen P (2017) Origins of Spurious Long Memory. Hann Econ Pap.
    [28] Lu YK, Perron P (2010) Modeling and forecasting stock return volatility using a random level shift model. J Empir Financ 17: 138-156.
    [29] Martens M, Dijk DV, de Pooter M (2009) Forecasting S & P 500 volatility: Long memory, level shifts, leverage effects, day-of-the-week seasonality, and macroeconomic announcements. Int J Forecast 25: 282-303.
    [30] McCloskey A, Perron P (2013) Memory parameter estimation in the presence of level shifts and deterministic trends. Econom Theory 29: 1196-1237.
    [31] Mikosch T, Stărică C (2004) Nonstationarities in financial time series, the long-range dependence, and the IGARCH effects. Rev Econ Stat 86: 378-390.
    [32] Perron P, Qu Z (2010) Long-memory and level shifts in the volatility of stock market return indices. J Bus Econ Stat 28: 275-290.
    [33] Qu Z (2011) A test against spurious long memory. J Bus Econ Stat 29: 423-438.
    [34] Robinson PM (1995a) Gaussian semiparametric estimation of long range dependence. Annals Stat 23: 1630-1661.
    [35] Robinson PM (1995b) Log-periodogram regression of time series with long range dependence. Annals Stat 23: 1048-1072.
    [36] Sibbertsen P, Leschinski C, Busch M (2017) A multivariate test against spurious long memory. J Econom (in press).
    [37] Velasco C (1999) Gaussian Semiparametric Estimation of Non-stationary Time Series. J Time Ser Anal 20: 87-127.
    [38] Xu J, Perron P (2014) Forecasting return volatility: Level shifts with varying jump probability and mean reversion. Int J Forecast 30: 449-463.
  • This article has been cited by:

    1. Janis Becker, Christian Leschinski, Estimating the volatility of asset pricing factors, 2021, 40, 0277-6693, 269, 10.1002/for.2713
    2. Theoplasti Kolaiti, Mwasi Mboya, Philipp Sibbertsen, Volatility Transmission across Financial Markets: A Semiparametric Analysis, 2020, 13, 1911-8074, 160, 10.3390/jrfm13080160
    3. Atanu Das, Performance evaluation of modified adaptive Kalman filters, least means square and recursive least square methods for market risk beta and VaR estimation, 2019, 3, 2573-0134, 124, 10.3934/QFE.2019.1.124
  • Reader Comments
  • © 2018 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(4931) PDF downloads(1655) Cited by(3)

Figures and Tables

Tables(14)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog