The purpose of this paper is to develop a new transformation (the revised Box and Cox transformation) to solve the zero value problem when using log specification for examining the technology of insurance companies producing multiproducts. It is impossible to compute the asymptotic covariance matrix of the parameter estimates with the traditional Box and Cox transformation when an insurance company produces only a subset of the feasible outputs. The limitation is easily solved with the new transformation developed in this paper. The new transformation enables sample size to be sufficient in the estimation of quadratic translog functional form. Therefore the limitation in terms of a degree of freedom can be solved, especially for the companies producing multiproducts such as insurance companies.