Ceris-Cnr, W.P. N° 2/2002 from the conditional distribution of uft given ψft, which incorporates whatever information ψft contains concerning uft. The conditional density function of uft given Ψft = ψft is given by83 fU | Ψ = ψ (u ) = f Ψ ,U (ψ , u ) f Ψ (ψ ) , [A.17] thus, using equations [A.6B] and [A.9b], 1 exp− {(u − µ * ) 2 / σ *2 } 2 . f U | Ψ = ψ (u ) = 2π σ * Φ[ µ * / σ * ] [A.18] The overall cost efficiency of the f th firm at the tth observation, CEft, may be expressed as the ratio of stochastic frontier minimum cost (with uft = 0) to observed cost, which is equal to84 CE ft = 1 = exp{−u ft } . exp{u ft } [A.19] This measure is bounded between zero (uft → ∞) and one (uft = 0), and can be predicted in a similar way to that described for technical efficiency in the stochastic production frontier case analyzed by Battese and Coelli (1993). Using the conditional distribution of uft given ψft defined by equation [A.17], the authors derive an expression for the conditional expectation of the technical efficiency for the f th firm at the tth observation, conditional upon the observed value of ψft = (vft - uft). This expression, E(exp{-uf}|Ψft = ψft), is a generalization of the results presented in Jondrow et al. (1982) and Battese and Coelli (1988). The prediction of the individual cost efficiencies relative to a stochastic cost frontier, i.e. expression [A.19], can be obtained by minor sign alterations of the technical efficiency point estimator in Battese and Coelli (1993). It is derived using the conditional density function of uft given Ψft = ψft specified in equation [A.18] and is given by  Φ[( µ *ft / σ * ) − σ * ]   * 1 2 ˆ CE ft = (exp{−u ft } | Ψ ft = ψ ft ) =   exp− µ ft + σ *  * 2   Φ[ µ ft / σ * ]   83 84 [A.20] Again the subscripts, f and t, are omitted in the following expressions for convenience in the presentation. Expression [A.19] is appropriate for CEft only if the general specification of the stochastic frontier cost model is given by equations [1]-[2] in the text. 45