Heteroscedasticity and interval effects in estimating beta: UK evidence 

The paper compares beta estimates obtained from OLS regression with estimates corrected  for heteroscedasticity  of the error term using ARCH models,  for  145 UK shares. The differences are  mainly  less than 0.10, for betas calculated using daily returns, but even such small differences can  matter in practice. OLS tends to overestimate the beta coefficients compared  with ARCH models, and selecting  an ARCH­type estimate makes most difference for large­cap  shares. Regarding the measurement interval, the downward bias in  betas from daily returns is associated not only  with thin trading but also with the volatility of the share’s daily returns. We infer that the idiosyncratic component in daily  returns,  as well as lack  of trading, is responsible for low daily betas.