This study presents results from a quantitative analysis of a new inventory of n=232 landslide dam occurrences in New Zealand. Previously published data were expanded by documentation of recent events and evidence from a regional air-photo reconnaissance focused on the upland regions of southwestern South Island. Additional geomorphometric data on landslide dams, associated lakes, and contributing catchment characteristics, were extracted from a 25-m Digital Elevation Model (DEM), augmented by limited ground truthing, and compiled in a GIS-based inventory. The New Zealand case examples fall into the global trend, although they contain both two extremely large features in terms of landslide dam volume VD and lake volume VL. Analysis suggests that landslide dam height HD, landslide dam volume VD, lake volume VL, contributing catchment area AC, and local relief HR, are key variables for assessing landslide dam stability independently from other catchment parameters such as lithology, climate, or dam sedimentology. They may be provisionally used as representative characteristics of landslide dams, irrespective of environmental boundary conditions, such as climate, geology, or site-specific valley geomorphometry. Three newly proposed dimensionless indices, i.e., the Backstow Index Is, Basin Index Ia, and Relief Index Ir, based on landslide dam height HD allow limited, yet promising, preliminary assessments of landslide dam stability. Compared with worldwide examples, they also demonstrate a much narrower conditional range for the formation of stable landslide dams in New Zealand. Catchment parameters such as maximum elevation Emax, upstream relief HR, contributing catchment area AC, and relief ratio RR are significantly different at sites of former and existing landslide-dammed lakes, and may be used as independent variables in future terrain-based classification schemes. Generally, data are incomplete with underreporting of small and ephemeral landslide dams, and over-representation of earthquake case studies. Nonparametric correlation highlights the statistical interdependence between geomorphometric variables as an artefact of initial data calculation, while varying accuracy poses a significant drawback for complex multivariate statistical techniques such as principal component, cluster, or discriminant analyses.