We analyse the spatial distribution of halo merger sites using four numerical simulations with high time resolution of structure growth in a Lambda-CDM cosmology. We find no evidence for any large-scale relative bias between mergers and randomly selected haloes of the same mass at high redshift. Given a sample of galaxies that form a Poisson sampling of halo mergers, then the amplitude of the measured clustering ought to lead to a robust estimate of the median host halo mass. In the hierarchical picture of structure growth, mergers between galaxies at high redshifts are thought to create dust enshrouded starbursts leading to increased emission in the sub-mm wavebands. Hence, a measurement of the clustering strength of SCUBA galaxies, for example, can determine whether or not they are giant elliptical galaxies in the process of formation.