On the validity of the quadratic utility approach in mean-variance portfolio analysis: An empirical test

Summary This paper empirically compares the quadratic utility approach and the negative exponential utility approach within the framework of portfolio analysis. Estimates for the subjective variance-covariance matrix are obtained using the two models under consideration. The two approaches have been compared through these estimates and also by their forecasting ability. The quadratic utility approach appears to be inferior. This paper relies heavily on parts of my University of Essex Ph.D. thesis (Bhattacharyya, 1975). I am grateful to Professors Bergstrom and Parkin for their help, comments and encouragement. Mr. R. E. Bailey's help and advice is also gratefully acknowledged. In addition, the comments of the referee of this journal helped me to improve on my earlier draft. Any error remaining is mine.