Earnings Volatility
factor.formula
Fit a time series regression model to company j's annual earnings.
Calculate the residual standard deviation, which is the earnings volatility factor.
in:
- :
The annual profit of company j in year t can be selected from indicators such as earnings per share (EPS) and net profit attributable to shareholders, which must remain consistent.
- :
The constant term in the first-order autoregressive model of firm j’s earnings time series.
- :
The autoregression coefficient of the first-order autoregression model of company j's earnings time series measures the persistence of earnings. The larger the absolute value, the greater the impact of the current earnings on the previous period's earnings. When the coefficient is positive, it usually indicates that the earnings have a certain inertia.
- :
The residual term of the first-order autoregressive model of the earnings time series of company j in year t reflects the part of earnings volatility that the model cannot explain, and it is assumed to follow a distribution with a mean of 0.
- :
The residual variance of the first-order autoregressive model of firm j's earnings time series measures the unpredictable part of the earnings series.
- :
The residual standard deviation of the first-order autoregressive model of company j's earnings time series, i.e., earnings volatility, measures the amplitude of earnings volatility.
factor.explanation
The smaller the value of the earnings volatility factor, the smaller the time series volatility of the company's earnings, the higher the predictability of earnings, and the more stable its profitability. This factor can be used in stock selection strategies to screen companies with stable profitability and high quality. Conversely, the higher the factor value, the greater the earnings volatility and the more unstable the profitability. In practical applications, the choice of earnings indicators (such as EPS, net profit, etc.) and the regression model used should be considered (in addition to the first-order autoregressive model, other models can also be tried). This factor is a low-frequency factor and is usually updated annually.