The scale effects in the value relevance of accounting data, due to the errors of statistical inference, have attracted much attention in the empirical study of accounting. This paper first compared the implication of different deflators，and analyzed the efficiency of different deflators in mitigating the scale effect by the residual analysis. We found that the economic meaning of the variable is changed in the deflating method, which cannot really reflect the value relevance of accounting data. What’s more, the deflating method does mitigate the coefficient biases and heteroscedasticity of the large scale part, but increases the coefficient biases and the heteroscedasticity of the small scale part, which fails to really mitigate scale effects. Compared with other deflators, the number of shares is best.