Estimating the knowledge-capital model for multiple parents and hosts: taking the cross-classified structure of the data into account

Abstract

Data on bilateral FDI and trade flows are clustered within parent and host countries and parent-host combinations. Failure to take into account these forms of clustering will lead to biased estimates of the coefficients due to an omitted variables problem and to an underestimation of the coefficients’ standard errors. Hence coefficients are incorrectly classified as significant. A cross-classified multilevel estimation procedure takes care of both problems. This paper estimates the knowledge-capital model of bilateral FDI for a sample of multiple parent and host countries by means of the cross-classified multilevel approach and compares the estimates with other techniques.