【正文】 and quote responses, however, is plicated by deviations from the theoretical assumptions. The vector time series model was used to circumvent this difficulty, by way of explicitly reflecting the dynamic structure of the system and permitting calculation of both the contemporaneous and lagged impacts.Third, the structure of pricevolume interaction may be important for event studies which use a bination of price and volume data to draw inferences. Examples of such studies can be found in Richardson, Sefcik and Thompson (1986) and Lakonishok and Vermaelen (1986). If price and volume are indeed jointly determined, incorporating their structure may improve the power of the tests.Empirical research in pricevolume relation typically uses correlation analysis or regression modelling. Correlation analysis assumes that price and volume are serially uncorrelated, which may not be a valid assumption. Serial correlation has been observed in many return series. Systematic differences in the serial correlation pattern of return occur between stock index and individual stock, stock index futures and the corresponding index, and shortterm horizon and longterm horizon (see, eg, Amihud and Mendelson, 1989b。 Lo and MacKinlay, 1988。 and Fama and French, 1988). Although evidence conceming the autocorrelation structure of volume is relatively scarce, some empirical studies suggest that the autoregressive order of volume is higher than that of return (see, eg, Tse, 1991). Thus, inference using correlation analysis may be misleading if the time series structure of price and volume is not taken into account. While regression modelling can incorporate the time series structure to a certain extent through the use of distributed lags, it suffers from simultaneity bias if price and volume are jointly determined. Also, the inclusion of contemporaneous volume as an explanatory variable in a return equation does not provide a viable forecasting model unless a plete simultaneous equation system is constructed.In this paper we adopt a multiple time series approach to model the pricevolume relation. In particular, we use the modelling procedure suggested by Tiao and Box (1981). This method has the advantage of being more direct and transparent, as pared with alternatives due to Granger and Newbold (1977) and Wallis (1977). The sequential and iterative steps of tentative specification, estimation and diagnostic checking parallel those of the wellknown BoxJenkins method in the univariate domain. Empirical applications of this approach can be found in Tiao and Tsay (1983) and Heyse and Wei (1985).An alternative approach is to examine the effect of volume on return through the conditional variance. This approach has been shown to pare favourably against the Generalised Autoregressive Conditional Heteroskedasticity (GARCH) model in its ability to capture the return volatility, as demonstrated by the recent work of Lamoureux and Lastrapes (1990). Nonetheless, the volume effect is assumed to be of only the second order and no statistical feedbacks of return on volume are accounted for. Thus, this approach is not free from the simultaneity bias. In veiw of the prevalence of conditional heteroskedasticity in many financial studies, however, it may be interesting to generalise the Tiao and Box multiple time series approach to incorporate autoregressive conditional variance as a topic of future research.Results in this paper show that forecasts based on multiple time series models outdo naive forecasts and univariate time series forecasts. Thus, it may be useful to create filter trading rules based on multivariate time series forecasts of stock returns. Whether a modelbased trading rule is superior to a traditional trading rule based on following market momentum is an interesting topic awaiting further investigation.The plan of the rest of the paper is as follows. In Section 2 we outline the TiaoBox multiple time series modelling approach. Section 3 describes the data used in this study. Empirical results are summarised in Section 4. We describe in detail the results of modeling one series as an illustration of the application of the TiaoBox methodology. Results for other series are then summarised, with parisons of postsample forecasts. Some concluding remarks are given in Section 5.Source: WS Chan and YK relation in stocks: A multiple time series analysis on the Singapore market [J]．ASIA PACIFIC JOURNAL OF MANAGEMENT．2000.經(jīng)濟與管理學院畢 業(yè) 論 文文獻綜述學生姓名： 羅劍華 專 業(yè)： 經(jīng)濟學 指導教師： 雷紅霞 2010年 10月 15日一、前言股票是市場經(jīng)濟的產(chǎn)物，自1773年英國率先發(fā)行股票以來，至今己200多年，其發(fā)行與交易促進了市場經(jīng)濟的發(fā)展。股票市場作為現(xiàn)代經(jīng)濟活動中極為重要和最為復雜的金融領域之一，是現(xiàn)代經(jīng)濟系統(tǒng)的核心組成部分，對一國國民經(jīng)濟的發(fā)展具有重要意義。股票市場上的兩個最為根本的變量是股票的價和量。所以研究股票價格與成交量的關系不僅對研究股票市場有重要意義，由于價量關系是股票技術分析理論的重要基石，是投資者判斷市場或個股運行趨勢的主要手段之一，所以對投資者也有重要意義。因此，價量關系理論一直是金融領域研究的的熱點課題之一，大量的研究也應運而生。國際上關于價量關系的研究可以分為兩個階段，即依據(jù)傳統(tǒng)經(jīng)濟學理論而建立的研究和利用現(xiàn)代金融學理論進行的實證分析。由于中國股票市場起步較晚，發(fā)展過程短，所以國內(nèi)對股票價量關系的研究，特別是以研究市場行為特征為目的的價量關系的實證研究相對較少，近幾年才有部分學者對這個課題進行研究。二、主體（一）國外早期的研究價量理論最早見于美國股市分析家所寫的股票指導書《股票市場指標》中。他通過對成交量與股價趨勢關系研究后總結了九大法則，也就是著名的格蘭碧九大法則。早期的金融研究并沒有注重交易量在資產(chǎn)價格方面的作用。Baghot（1971），Copeland和Galai（1983），Easley和O’Hara（1987）等早期的金融理論和微觀結構理論研究者們也沒有或明顯揭示出交易量在資產(chǎn)價格形成過程中所起的作用。從學術角度對股票價量關系的實證研究可以追溯到Osbonre（1959），他通過建立模型，說明了股價變動是一個擴散的過程，其方差取決于交易的次數(shù)。雖然他在文章中沒有直接研究股票的價量關系，但其已經(jīng)暗示了在交易量與價格變化的絕對值之間存在正的相關關系。后來他的觀點又被Clark（1973），Tauchen和Pitts（1983），還有Harris（1986）所證實并加以發(fā)展。而關于價量關系的實證檢驗最早是由Granger和Morgenstern，在1963年做的。他們用1939年到1961年的周數(shù)據(jù)進行譜分析，發(fā)現(xiàn)SEC成分指數(shù)走勢與紐約股票交易所的總成交量之間沒有關系，來自兩個相互獨立的股票市場的數(shù)據(jù)同樣沒有價量關系。隨后許多學者對此問題進行了一系列的研究。1964年，Gdfrey，Granger和Morgenstern利用新的數(shù)據(jù)序列再次進行了論證，但他們還是沒有發(fā)現(xiàn)股價與成交量或價格差分的絕對值與成交量之間的相關關系?？伤麄儏s發(fā)現(xiàn)了另一個問現(xiàn)象：日成交量與每日最高、最低價之差存在正相關關系。后來的研究也證實了日成交量數(shù)據(jù)與日開盤價和收盤價差值的平方存在顯著關系。而在1987年Karpoff通過對月度、周和一些日股票收益進行研究，得出了交易量和股票價格波動之間具有正向變動關系的結論。支持這一結論的有Epps（1976），Jain與Joh（1988），Schwert（1989），Gallant等（1992），Lamoureux與Lastrapes（1990）以及Andersen（1996）等學者的研究。對股票價量關系早期的研究主要集中在收益率和成交量之間的同期關系上，最具代表性的觀點為Copeland（1976）的信息順序到達模型和Clark（1973）的混合分布