The aim of this paper is to determine the extent of undercounting of Mäori and Pacific deaths in New Zealand during the 1980s and 1990s, and to calculate corrected ethnic mortality and life expectancy trends. We calculated adjustment ratios for undercounting of Mäori and Pacific deaths (and over-counting of non-Mäori non-Pacific (nMnP) deaths) using the linked census-mortality data. These ratios were then used to calculate corrected mortality rates and life expectancies. Mäori deaths were underestimated by a quarter, and Pacific deaths by a third, during the 1980s and early 1990s. Undercounting was minor in the late 1990s following alignment of ethnicity collection on mortality data to approximate the census. Corrected mortality rates demonstrated 30% (males) and 26% (females) decreases among nMnP from 1980-84 to 1996-99, smaller decreases among Mäori (8% and 7%) and no clear change among Pacific people (9% decrease for males, 4% increase for females). The gap in life expectancy increased from an average of 7.7 years in 1980-84 to 10.8 years in 1996-99 for Mäori, and from 3.3 to 7.7 years for Pacific people, in comparison to nMnP people. Deaths among 45-64 and 65 plus year olds, and cardiovascular disease and cancer deaths, were the main contributors to these disparities. The economic reforms in New Zealand during the 1980s and early 1990s impacted harder upon Mäori and Pacific people in terms of unemployment and income, and are a likely explanation for the diverging mortality trends in this period. Both behavioural factors and health services probably also play a role, but in the absence of trend data by ethnicity, their contribution to diverging mortality trends is unknown. Internationally, our study demonstrates marked undercounting of Mäori and Pacific deaths. We strongly encourage researchers and custodians of vital statistics in other countries to investigate the possibility of undercounting of deaths by ethnicity.