Correlation Measures.pptVIP

Correlation Measures Intuition: Correlation Measures seek to quantify statistically how closely related variables are. Correlation Limitations Covariance stationarity for a time series, y(t), is defined as: Constant, finite mean Constant, finite variance Covariance (y(t), y(t-s)) depends only on the lag s Covariance is a linear relationship as demonstrated by the associated regression relationship: Y(t) = βx(t) + ε(t) Pearson’s Correlation r(x,y) = (∑(x - xavg)(y – yavg))/(nsxsy) The range of this statistic is between -1 and 1, and if x and y are unrelated, it will equal zero. Spearman’s Correlation r = 1 – 6 (∑d2)/(N(N2-1) Where: d is the difference in the relative ranking of two associated variates x and y. N is the number of observations. Kendalls Tau Coefficient  Let inv := number of inversions, i.e. reversals of pair-wise rank orders between n pairs. Equal rankings need an adjustment. ? = 1 - 2* inv/(number of pairs of objects) = 1 - 2 * inv/ (n*(n-1)/2) = 1 – 4* inv/(n*(n-1)) Spearman’s r treats ranks as scores and computes the correlation between two sets of ranks. Kendall’s tau is based on the number of inversions in rankings. Usefulness of foregoing definitions Historical definitions may give viable long term estimates of correlation, especially if phenomena is stationary. If phenomena has varying correlation, constant correlation may not properly value financial instruments. Correlation Definitions Used in Hedging Instantaneous Correlation ?ij(u) := instantaneous correlation between variables with indices i,j a moment u. Terminal Correlation term_?ij(T) = [?0T?i(u)?j(u)?ij(u)du] / ?(vivj) where: vi = ?0T ?i(u) 2 du Implied Correlation The implied volatility on the cross rate on a foreign exchange process may be expressed as: ?2x-y = ?2x + ?2y -2?x?y? which may be solved when options are available on the f/x rates and cross, i.e.: ? = (?2x + ?2y - ?2x-y)/(2?x?y) In the above example, ? is the implied correlation.