A performance moment integration method is proposed to accurately estimate the first two statistical moments, mean and variance, of electromagnetic performance functions in the present of uncertainties. To maximize computational efficiency, its numerical integration is executed not on the input design domain but on the output performance domain, where quadrature points are explored by the first order reliability analysis method. For better understating between statistical moment analysis methods, two different numerical integration schemes of dimension reduction method and performance moment integration method are compared with each other. Finally, a mathematical model and a loudspeaker design model are tested to demonstrate the features of two methods and to examine their numerical accuracy and efficiency.