The Application of Errors-in-Variables Methodology to Capital Market Research: Evidence on the Small-Firm Effect

Errors in variables due to nonsynchronous trading and benchmark error are significant problems for capital market research. This paper develops the use of direct and reverse regression to bound true coefficient estimates when the data exhibit error structures arising from these two sources both separately and jointly. The approach appears to have broad applicability for capital markets research. As an example, the paper reexamines the small-firm effect to show that it cannot be attributed to nonsynchronous trading or benchmark error in the estimated variance of the market portfolio. This result is shown to hold even when the tax-selling effect is controlled for by excluding January returns.