A disaggregate approach to accounting-based measures of systematic risk

Studies have found an association between Accounting Risk Measures (ARM's) and the {dollar}beta{dollar} of the market model in cross-section and over long time periods. The purpose of this study was to build a useful individual firm predictive model for {dollar}beta{dollar} for quarterly intervals using only accounting data from 10Q and Annual Reports. {dollar}beta{dollar} was estimated using the Scholes & Williams technique and became the dependent variable in a stepwise regression model fitting procedure. The independent variable set was generated from 40 ARM's by principal components analysis. This model was used to predict {dollar}beta{dollar} for the following quarter on an individual firm basis.; The experimental model failed to consistently outperform naive models on a mean absolute error basis. This indicates that for these firms for this time period and for these ARM's, accounting data alone was not sufficient to predict {dollar}beta{dollar}.; In cross-sectional studies, the larger the firm, the smaller the {dollar}beta{dollar} in general. This study showed a significant negative correlation over time between firm size and {dollar}beta{dollar} for half the firms in the sample. This indicates that as a firm grows, {dollar}beta{dollar} decreases.