a comparison between performance of linear and nonlinear capital asset pricing model in tehran stock exchange

capital asset pricing model (capm) has been among the common models to estimate expected returns rate. since the linearity assumption is considered in the standard version of the capital asset pricing model, estimating beta in nonlinear setting will be inconsistent and bias-oriented. therefore, this study tries to evaluate predictive power of nonlinear capital asset pricing model as well as standard capital asset pricing model. semiparametric method and local kernel regression are utilized in order to estimate nonlinear model. for this purpose, expected returns has estimated with regard to two abovementioned models during period of the study and the results have compared with realized returns. mean absolute percentage error and especially diebold-mariano test are used to measure predictive power of the models. the results indicate that considering nonlinearity relation between stock returns and market returns increases predictive power of realized returns.key words: kernel regression, linear capital asset pricing model, nonlinear capital asset pricing model, nonparametric model.