摘要：A very important research topic is describing the way in which different assetprice movements are correlated. Modern portfolio theory methods are basedon observed correlations between returns at daily or larger time scales. Oneexpects that coarse scale correlations originate from intraday movements thatare strongly correlated. This implies the important question of how to obtainbetter estimates of such correlations by using high-frequency data. Here, one canobserve an analogous paradox as for volatility estimation under microstructurenoise. It is surprising that the correlation coefficient is an increasing function ofthe time resolution, and that correlation very quickly decays and almost vanishesat a very high frequency. The dependence of correlations between stock priceson the sampling frequency of time series involves a phenomenon called the Eppseffect. The actual correlations between returns of stocks decrease as the samplingfrequency of data increases. The Epps effect has received considerable attention,not only from economists but also from mathematicians and theoretical physicists.But there are only a few contributions to the subject
