The unprecedented urbanisation observed in leading developing countries has placed immense pressure on effective and efficient land management. The significance of land-use efficiency in the Chinese context has been addressed in the literature, particularly on the measurements of land-use efficiency and key influencing factors. However, quantifying the interdependence between land-use efficiency, local government revenue, employment and infrastructure development whilst controlling for significant cross-city differences remains a gap in the literature. Based on data for 272 prefecture-level Chinese cities between 2012 and 2017, this study employs a novel modelling approach, combining latent class analysis (LCA) in a generalised structural equation model. The incorporation of LCA helps to control for the significant, non-linear heterogeneity across city samples. The empirical model identifies both the direct (one-off land conveyance fee and transaction-related tax revenue from land transactions) and indirect (corporate and personal taxes generated from employment and business growth) channels, through which land development contributes to local government revenue. It also provides one of the first quantified evidence, confirming that employment growth provides higher long-term return than a one-off, land conveyance fee to government revenue in China, controlling for significant cross-city heterogeneity in land-use efficiency and wage.