Yesterday I attended New York Accounting and Finance Forum (NYAFF) Conference. You can access the program and the papers that were presented here.
The paper presented by Hank Bassembinder is very interesting. In short, the authors argue that although the stock prices are measured with random error, the returns computed from these noisy prices contain an error that is no more random. If, in asset pricing tests, the explanatory variables are correlated with the causes for the noise, the resulting coefficients will be biased upwards.
The prices are noisy for many reasons. The authors concentrate on microstructure noise caused by bid-ask bounce, nonsynchronous trading etc. It is straightforward to show that microstructure noise and the liquidity factors will be correlated in asset pricing models, such as liquidity augmented CAPM. The authors also suggest the remedy to correct the bias. Basically the remedy is to use a weighted least square (WLS) estimation rather than simple OLS.