Bayes Risk for Regression

What is the significance of using p(x,y) in the expression of Bayes Risk for Regression?