The concept of protection on value-added or the effective rate of protection has received much support as a measure of the impact of protection on primary factors of production. It is predicted by H. G. Grubel and H. G. Johnson [5, p. 76] that effective rates of protection will receive at least as much attention as nominal rates in future international tariff negotiations. The analysis and applications of the concept are generally carried out with the crucial assumption that the physical input-output relationships remnain the saine before and after the imposition of tariffs. This is convenient and often necessary since information on input coefficients is available only for the protected state. It has been argued that this assumiiption introduces a bias in the measurement of effective rates; W. M. Corden [4, pp. 233-35] has suggested that the bias is in the direction of overstating the effective rates. Oine complication is, however, that once substitution is allowed, two definitions of the concept are possible. Under one definition, the overstatement occurs but not necessarily so under the other definition. The purpose of this paper is to present the different behavior of bias in the two concepts under substitution. Following a brief introduction to these two concepts, the analysis is developed using the constant return constant elasticity, (CES) production function with two inputs (one primary and one intermediate) for the sake of simplicity. The analysis is then applied in evaluating briefly the computational formulas and results of studies by B. Balassa [1] and G. Basevi [2].