Revisiting the question – Does corporate headquarters location matter for stock returns?

Abstract

We show that the ordinary least squares applied to right-hand-side variables consisting of averages of neighbouring observations such as stock returns representing a portfolio of firms in the same location or industry result in biased and inconsistent estimates. A maximum likelihood estimation procedure that will produce consistent estimates for this type of model is set forth. In addition, we show how to correctly interpret the maximum likelihood parameter estimates.

Notes

¹For the case of nonzero covariance between Wy and Vy we can rely on a partitioned matrix inverse formulation to produce a similar, but more complicated result. Least-squares estimates are still biased in this case.

²We must view W as nonstochastic sample data information and assume that as the sample size increases the number of nonzero elements in each row of the matrix W has a finite limit. These conditions would be met for the case where W is block diagonal with row i representing firms in the same MSA as observation i. The parameter ρ must obey the eigenvalue bounds to ensure bounded y.