Comparison of explanatory power of linear and nonlinear models predicts expected stock returns

One of the most challenging topics in finance and accounting is balancing returns and risk. If so, it is important for the market to identify trends in stock returns to predict the future. Although most research on stock return changes has been based on the use of linear models, there is little evidence that stock return fluctuations may follow nonlinear patterns. Therefore, this study seeks to compare the explanatory power of linear and nonlinear models of expected stock returns. In this regard, information about 102 companies listed on the Tehran Stock Exchange during the years 2009 to 2019 has been analyzed. The results showed that among the linear models, the coefficients of market variables, size and value in the Karhart model were higher than the coefficients of other models used. The results of estimating nonlinear models showed that threshold self-explanatory models have higher coefficients than smooth logistic transmission models. Also, using the homogeneity test of mean averages, the results indicate that the nonlinear self-explanatory threshold model based on trading volume (TARVOL) had the lowest standard mean error, which indicates that this model is more accurate in explaining stock returns.