VAR(1)-BEKK-GARCH(1,1) Model

[TomDoan]

Is the mean return shock ever defined? It should be (by construction) pretty close to zero if it is just what you would think.

[jack]

This is how the authors define it:

[TomDoan]

That's statistical gibberish. The residuals (for any of the dependent variables) should be roughly mean zero. (Not exactly mean zero as you would see in a linear regression, but pretty close). So basically this is taking pretty much a random number with random sign, multiplying it by a constant and saying that it has some meaning. If it were mean absolute value, that would make at least some sense.