版權說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權,請進行舉報或認領
文檔簡介
1、NEW FRONTIERS FOR ARCH MODEL S Robert Eng leProfessor of Finance, N YUProfessor of Economics, U CSD Prepared for confer ence Modeling and Forecasting Financial Vo latilityPerth, Austra liaSeptember 20 01This Draft June, 2600 2 ABSTRACT In the 20 years following the publication of the ARCH model, the
2、re has been a vast quantity of research uncovering the properties of compeatiinligty vmolodels. Wide-ranging applications to financial data have discovered important stylized facts and illustrated both the strengths and weaknesses of the models. There are now many surveys of this liter ature. This p
3、aper looks forward to identifyis inpgro mareas of new research. The paper lists five new frontiers. It briefly discusses thr eheig h frequency volatility models, large-scale multivariate ARCH models, and derivatives pricing models. Two further frontiers are examined in more detail application of ARC
4、H models to the broad class of no-negative processes, and use of Least Squares Monte Carlo to examine-l inoenar properties of any model that can be simulated. Using this methodology, the paper analyzes more general types of ARCH mtoocdheals,t ics volatility models, long memory models and breaking vo
5、latility models. The volatility of volatility is defined, estimated and compared with option implied v olatilities. Keywords: ARCH, GARCH, volatility-, linnoenar process, non-negative pcreoss, option pricing, stochastic volatility, long memory, Least Squares Monte Carlo, , ACDMultiplicative Error Mo
6、 dMeEl,M 1 Who could imagine 20 years ago, the flowering of research and applications that would develop around the ARCH model? It certainly was not an instanrst success. Yewent by before anyone except my students and I wrote a paper on ARCH. But as applications shifted to financial markets, and as
7、richer classes of models were developed, researchers saw how volatility models could be used to investigate the fundamental questions in finance. How are assets priced and what is the tradeoff between risk and return? ARCH models offered new tools for measuring risk, and its impact on return. They a
8、lso provided new tools for pricing and he-ldingeinagr ansosnets such aios nosp .t This conference and this paper are designed to reflect on these developments and look forward to the next important areas of research. In this paper I will give a rather brief idiosyncratic assessment of the important
9、accomplishmen ytse aorfs t ihne last 20volatility modeling. Then I will point to five frontiers on which I think new developments can be expected in the next few years. For two of these, I will give some new results to show the directions I see de veloping. I. WHAT WE HAVE LEARNED NI 20 YEARS The nu
10、mber of new models proposed, estimated and analyzed has been dramatic. The alphabet soup of volatility models continually amazes. The most influential models were the first: the GARCH model of Bollerslev(1986), and the EGARCH of Nelson1(991). Asymmetric models of Glosten, Jaganathan Runkle(1993) Rab
11、emananjara and Zakoian(1993), Engle and Ng(1993) and power models such as Higgins and Bera(1992), Engle and Bollerslev(1986), and Ding Granger and Engle(1993) joined models such as SWARCTHA, RSCH, QARCH and many more. The linguistic culmination might be that of Figlewski(1996), the YAARC aHn macordo
12、enl ym for Yet Another ARCH model. Coupled with these models was a sophisticated analysis of the stochastic process of data generathe dm boyd esulsc as well as estimators of the unknown parameters. Theorems for the autocorrelations, moments and stationarity and ergodicity of these processes have bee
13、n developed for many of the important cases; see for example Nelson(1990), and Ling and Mc(2A0le0e2ra,2002b). Work continues and new models are continually under development, but th-ist uisd aie dw efrllontier. The limiting distribution of the MLE for GARCH models waited for Lumsdaine(1996) and Lee
14、and Hansen(1994) for rigorous treatment as.r e T nhoewre a collection of survey articles that give a good appreciation of the scope of the research. See for example, Bollerslev, Chou and Kroner(1992), Bollerslev Engle and Nelson(1994), Bera and Higgins(1993), and recent pedagogical articles by2 E0n0
15、g1l)e a(nd Engle and Patton(2001). A very recent survey is Li, Ling and McAleer( 2002)Another topic for ARCH models is their usefulness in trading options. It was initially supposed that volatility models could give indications of mispricing in options markets leading to trading opportunities. Early
16、 studies such as Engle, Kane and Noh(1994) suggested the profitability of such strategies. More recent data fails to find evidence of significant trading opportunities, at least in the US index options market. 2 This is not surprising since GARCH models have a limited information set and are availab
17、le to all traders today. The same question is often asked in terms of forecast accuracy. Do GARCH model-sfo oruectast implied volatility models? The answer is complex depending upon the statistical approach to forecast evaluation, but generally it is found that implied volatilities are more accurate
18、 forecasts of future volatility than are GARCH models. See for example Granger and Poon(2001) for a survey.The theo royf asset pricing is based upon the reward for bearing risk. ARCH models have been developed to measure the price of risk. The first such model was the univariate ARC-MH model of Engl
19、e, Lilien and Robins(1987). Estimation of the CAPM began with Bollervs,l Engle and Wooldridge(1988) and has been extended and improved by a series of interesting papers including McCurdy and Stengos(1992), Engel Frankel, Froot and Rodriguez(1995), and de Santis, Gerard and Hillion(1997). With the in
20、troduction of Va lRuies ka,t a new role for ARCH models emerged. A variety of studies examined the usefulness of volatility models in computing VaR and comparing these methods with the exponential smoothing approach favored by Riskmetrics. See for example Christoffersiebno aldn(d2 0D00), Christoffer
21、sen, Hahn and Inoue(2001) and Alexander(1998). GARCH methods proved successful but suffered if errors were assumed to be G aussian. These chapters of research on ARCH models are full and may have reached the point of diminisgh irneturns. However, new directions are always available and these are the
22、 main focus of this p aper. II. FIVE NEW FRONTIERS Five new frontiers are identified below. These are areas where substantial research can be expected over the next fewp yreoabrles.m Ts haer e important, soluble, and already have some important new papers. For two of the areas, I will give some new
23、results suggesting a possible direction for futur e research. A. HIGH FREQUENCY VOLATILITY MODELS The study of volatility m owdiethlsin the day is in its infancy yet is a natural extension of the daily models examined so widely. Several alternative formulations have been introduced including Anderse
24、n and Bollerslev (1997) and Ghosh and Kroner(1995). Such models focus on tohf ed atiym oer “diurnal effect and have the requirement that they be useful forecasting many days in the future. These models have regularly spaced observations in calendar time but ultimately it will be desirable to find mo
25、dels based on irregularly astpaa caesd t hdis is the inherent limit of high frequency data. Engle(2000) calls such tick data “ultra high frequency data and gives some models which indicate that the arrival rate of trades, spreads and other economic variables may be important var rfioarbelceas sftoin
26、g volatility at this frequency. Such a model could give a continuous record of instantaneous volatility where events such as trades and quote revisions as well as time itself, modify the volatility estima te.Continuous time models are uubsi qinu itfionancial theory and derivative pricing. However mo
27、st estimation of these models begins with equally observed prices and focuses on the mean process possibly with jumps. Continuous time stochastic 3 volatility models, possibly with volatilityr eju ma npes wa class of models with interesting derivative implic ations.In addition to these models, there
28、 is now increasing interes-dt ainil yu sing intradata to estimate better daily models. Andersen Bollerslev Diebold and Labys(2001) for example build mso dbealsed upon “realized volatility and Andersen and Bollerslev(1998) use this measure to evaluate traditional GARCH s pecifications. B. MULTIVARIAT
29、E MODELS Although the research on multivariate GARCH models has produced a wide variety of models anedc isfipcations, these have not yet been successful in financial applications as they have not been capable of generalization to large covariance matrices. As computation becomes cheaper, and new par
30、simonious models are formulated, the potential for b euvilderin lgarger time varying conditional covariance and correlation matrices increases. Models such as the vec and BEKK model of Engle and Kroner(1995) have attractive properties as linear systems. The constant conditional correlation (CCC) mod
31、el ofl eBvo(l1le9r9s0) has the attraction of computational simplicity. A new model called Dynamic Conditional Correlation (DCC) by Engle(2002) combines some of these features to introduce a parsimonious correlation model to go with a conventional volatilitnyg lme oadnedl . ESheppard(2001) estimate a
32、nd test models of up to 100 assets. - Ledoit and SantaClara(1998) combine bivariate models to form multivariate models in a way which can be greatly expa nded.Correlation models can be estimated directly on in tHraodwaeyv edra atas. the frequency increases, the asynchronicity of trades and returns l
33、eads to a serious underestimate of comovements. This has been observed since Epps(1979) and the solutions of Scholes and Williams(1977) are widely employed in spite of both theoretical and empirical difficulties. These are not appropriate for ultra high frequency data and new solutions must be found
34、. C. OPTIONS PRICING AND HEDGING The pricing of options when the underlying asset follows a GARCH model is a topic of futures rearch. Most approaches are based on simulation but the appropriate approach to risk neutralization must be investigated. This was barely addressed in Engle and Mustafa(1992)
35、 when they simply simulated GARCH returns with the riskless rate as mean,l actainlcgu an “implied GARCH model. Duan(1995)(1997) in a series of papers has proposed a local risk neutralization which is based on the assumption that quadratic utility is a good local approximation to the representative a
36、gents risk preferences. Enndg Rleo asenberg(2000) develop hedging parameters for GARCH models and Rosenberg and Engle(2001) jointly estimate the pricing kernel and the empirical GARCH density. This paper is in the line of several which use both options and underlying data to ebsottimh athte risk neu
37、tral and objective densities with ever more complex time series properties in an attempt to understand the enormous skew in index options vola tilities. An alternative strategy is the GARCH tree proposed by Richken and Trevor (1999) whi cahdapts binomial tree methods for- dtheep epnadtehnce of GARCH
38、 4 models. These methods involve a variety of assumptions that must be examined empirically. They are computationally much faster than the simulation estimators discussed abov e. III. MULTIPLICATIVE ERROR MODELS FOR MODELING NON-NEGATIVE PROCESSES A. INTRODUCTION GARCH type models have been introduc
39、ed for other variables. Most notable is the Autoregressive Conditional Duration model ACD of Engle and Russell(1998) that surprisingly turns ou bte t isomorphic to the GARCH model. In this section, I explore a much wider range of potential applications of GARCH type mode-nlse fgoart iavney nontime s
40、eries proc ess. Frequently we seek to build models of time series t-hnaetg haative nonelement sS. uch series are generally common and are particularly common in finance. For example, one could model the volume of shares tra-dmeidn uotvee pr ear i1o0d. Or one might want to model the high price minus
41、the low price over a time period, or the ask price minus the bid price, or the time between trades, or the number of trades in a period, or many other series. There are two conventional approaches to this problem: the first is to ignore the n-noengativity, and the second is to take logs. Wed disisac
42、duvsasn tthaeg es of these approac hes. Consider a time serxies, t=1,.,T, wherex 0 for all t. Suppose in addition ttthat P(x0, for axll 0, and for all t (1) ttwhich says that the probability of observing zeros or near zeros in x is greater than zero. Let the conditional mean and variance of the proc
43、ess b e defined as: mExx?xs2(1,.,)V(x?x1,.,1)x(2) tttttA linear model is given by =m+2e,e?1D(0s,) (3) tttttIt is clear that the distribution of the disturbances must be carefully specified. Since the mean is positive and x -isn engoantive, the disturbances cannot be more negative than the mean. Thus
44、 the range of the disturbaen cdeif fweirlel nbt for every observation. The variance and other higher moments are unlikely to be constant. Efficient estimation via Maximum Likelihood is going to be very difficult, although least squares will remain consistent. The probability of a noe ias rg ziveern
45、b y Pt?1(xx=Pt)t?1(ex?m tt)hence the error distribution must be discon?tminu inou osr daet r to sati(s1f)y. t The second convenatli osnolution is to take logs. The model might then be written a slog(x)=m+u,(4) tttwher e 5 m=emtEeuts2=e2mVeu (),t(tt). (5) This solution will not work if there are any
46、exactx ze. r oSso mine times a small tconstant is added to eliminateo ethse. zHeorwever, this is more of a theoretical solution than a practical one since the finite sample estimates are typically heavily influenced by the size of this constant. Furthermore, the (a1s) stuhmatp tion in observations v
47、ery near zero are possible, rePq(uirat0 , for allA .0 This is only true of very peculiar distrib ut ions. Estimation o(f4 ) requires the specification of both m and u. Clearly the relation betweenm andm depends upon the distribution of u. Thus even one step forecasts ttrequire knowing the distribouf
48、 tuio .n B. THEORETICAL MODEL The proposed model, which solves all these problems, is the multiplicative error model that could be abbreviated as MEM. This model specifies an error that is multiplied times the mean. The specific ation isx=me?Df2, 1,ttt1(t)(6) thereby automatically sati(s2f)y. in Tgh
49、 e range of the disturbance would naturally be from zero to infinity thereby sat(i1s)f.y iInf gt he disturbancie.i .ids. , then the variance of x is proportional to qthuea rse of its mean. This is a strong restriction but is not in conflict with other parts of the model. Iif. ii.td i.s t hneont a n-
50、onnegative distribution with a unit mean and time varying variance can be specified. There are many candidates. Theid rueasl inx is naturally measured as the proportional deviation from the estimated mean as the standardizedx re/?msid. u Tahl,i s would be homoskedastic ttalthough the additive resxid
51、?uam?l =m?(e?1would no t.tttt) Vector modlse can be formulated in just the same way. Let an arrow represent a column vector ande le rte present the Hadamard product of two matrices, which is elemen-bty-element multiplica tion.rr xm, andV xmerrr=()=V(e)=diagm(V(e)diamg( )(7) tttttttttThus the positiv
52、e definiteness of the covariance matrix is automatically guaranteed by the MEM structure . Estimation of the MEM can typically pbryo cmeeadx imum likelihood once the distribution of the disturbance has been specified. A natural choice of distribution is the exponential as it ha-sn neognative support
53、. An exponential random variable with mean one is called a unit exponential. Assautm tihneg dtihsturbance is a unit exponential, then the univariate log likelihood is simplyT()?x?Lq=?log(mq?tt()?(8) t=1?m(qt)?where theta repre stehnet vsector of parameters to be estimated. The first order conditions
54、 for a maximum of this likelihood fu nction are 6 ?LT?x?m?m?=ttt ?q?2? (9) t=1?m?qt?By the law of iterated expectations, the expected value of the first order condition when evaluated at the true parameter value will be zero regardless of whether the density of x is truly unit exponential. This impl
55、ies that the log lik(8e) lichaono db ein i nterpreted as a Quasi Likelihood function and that parameters that maximize this are QMLE. Application of the theorem originally given in White(198re0g) urleaqriutiyr ecso nditions on the mean function and its determinants, and gives general expressions for
56、 the covariance mat rix. A fairly general class of mean functions can be entertained for this problem. Suppose the mean is linear in lagged x and cinto ar okfx 1p rveedetermined or weakly exogenous variabzle. s T hen a (p,q) mean specification w ould betpq m=w+ax?+bm?+gz(10) tjtjjtjtj=1j=1The parame
57、ters of this model may be restricted to ensure positive means for all possible realizations, and to ensure stationary distributions for x. If z are positive variables, then sufficient conditions for-n neognativity are clearly a tlhl aptarameters are positive. However these are not necessary. See Nel
58、son and Cao(1992) for an examination of sufficient conditions. Sufficient conditions for the covariance stationarity of x from Bollerslev, Engle and Nelson(1994) are that zn icse c sotvaatiroianary a nd pq a+b1.(11) jjj=1j=1 This result can be formalized from Engle and Russell(1998) based upon a the
59、orem in Lee and Hansen(1994) for GARCH models. In their case, x is the duration between successive events, but the theorem applies-n teog antiyv en opnrocess. In this theorem, the process is assumed to be a first order GARCH type model possibly with unit or exploisve roots . Corollary to Lee and Han
60、sen(1 994)If 1) E?1(x)m0=w+axtt,t00t?1+bm00?, t,1ex/mtt0 is i) strictly stationary andic ergod,t ii) nondeeragten 2) iii) hansd beodu conditional second moments iuvp) Eslnb(0+a0e)?1?+? where 1 for t1log(m)t tt=1?mt?m=w/(1?b) for =t1tThen : 7 a) the maximizeLr owfi ll be consistent asnydm aptotically
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
- 4. 未經(jīng)權益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負責。
- 6. 下載文件中如有侵權或不適當內(nèi)容,請與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。
最新文檔
- 聘用上課合同范例
- 錫焊加工合同范例
- 肖邦銷售合同范例
- 資金變動協(xié)議合同范例
- 古建石塔采購合同范例
- 美發(fā)店長合同范例范例
- 養(yǎng)殖馬蜂出租合同范例
- 預售房屋合同范例
- 施工打墻合同范例
- 農(nóng)村房屋轉(zhuǎn)讓協(xié)議合同范例
- MOOC 數(shù)字邏輯電路實驗-東南大學 中國大學慕課答案
- 內(nèi)鏡逆行沖洗治療闌尾炎
- 2024年福建德化閩投抽水蓄能有限公司招聘筆試參考題庫附帶答案詳解
- 2024年貴安新區(qū)產(chǎn)業(yè)發(fā)展控股集團有限公司招聘筆試參考題庫含答案解析
- 山東省青島市李滄區(qū)2023-2024學年七年級上學期期末道德與法治試題
- 設備工程行業(yè)年終述職報告總結
- 《軍事理論》教案第五章
- 團隊能力建設分析報告
- 四川省成都市雙流區(qū)2023-2024學年三年級上學期期末數(shù)學試卷
- 2024年1月浙江首考高考選考歷史試卷試題真題(含答案)
- 部編版語文四年級上冊期末復習專項1:生字詞語句子復習課件
評論
0/150
提交評論