Stock Price Inferencing and Prediction Based on Fama-French and Two-way Clustering Structure

This paper tests the Fama-French model using a new approach to estimate the standard error and verify the significance of different factors. Traditional standard error estimation neglects the correlation between stock return observations. As a result, the standard error will usually be underestimated, and some factors will show ostensible significance due to smaller standard error estimation and larger t-stat. This paper assumes a two-way clustering structure, assumes that stock return is correlated in industry and stock itself in two dimensions, and concludes with more decisive factors. Then this paper utilizes influential factors in stock return prediction and selection with the help of bootstrap simulation, and the result is slightly better than the standard OLS regression.