I've rewritten my Matlab MEX Kappa Ratio function to take vectors arguments for the threshold return and Kappa order arguments, allowing me now to easily generate a matrix with the N-th order Kappa Ratio in the N-th row over M threshold returns.  This comes at the cost of being able to calculate a fixed order Kappa for multiple assets simultaneously, but now I can do something like this:

r = linspace(mean(iwn)-3*std(iwn), mean(spy)+3*std(iwn));
n = linspace(1,10);
K = kappaCX(iwn, r, n);

    This gives me one of these matrices, which I will hereafter call a Kappa Ratio Surface.   As  successive Kappa orders increasingly punish kurtosis in the return distribution, the "columns" of this surface are increasingly kurtosis sensitive.  As Kappa is also insensitive to kurtosis in the interval around and above the threshold return, the "rows" of this surface likewise are increasingly kurtosis sensitive. 

    Now, it's just as easy to calculate the difference of two of these surfaces for two return distributions.  For example:

D = kappaCX(iwn, r, n) - kappaCX(iwo, r, n);

    This now I will call a Kappa Ratio Difference Surface.  Now, supposing there is not constant relative normalized kurtosis below each return threshold between the two return distributions, then the correlation of these columns should represent the difference in kurtosis risk below the threshold level between the i- and j-th threshold levels.  Such correlation values could be theoretically be used in generating weights for a strategy to arbitrage difference in long-term volatility pricing.

    The following is an example correlation over the Kappa Ratio Difference Surface between the Russell 2000 Value and Growth ETFs, as shown in the examples above.