The public health case for concern and action on tobacco use in Aotearoa-New Zealand is overwhelming; 4500 to 5000 deaths per year in New Zealand can be attributed to tobacco use.1The long-run trends in life expectancy show continual improvement in non-M ori life expectancy and a substantial increase in M ori life expectancy since the turn of the 19th Century.2-4 Over the last century, average annual reductions in mortality have been approximately 3.5% per annum for M ori and 2.0% per annum for non-M ori (Woodward and Blakely, work in progress). Such reductions do not occur by chance, but reflect concerted policy and public health efforts, in addition to general improvements in health services and standard of living. Furthermore, closing of ethnic inequalities in life expectancy are far from guaranteed, as evidenced by a widening of M ori:non-M ori life expectancy gaps in the 1980s and 1990s associated with structural changes in New Zealand society.3,5-7The gap in life expectancy between M ori and non-M ori c.2006 remains large at 7 to 8 years, but has narrowed from the 9 to 10 year gap c.1996.3,4,7 Such a pattern suggests the possibility of a return to long-run trends of closing ethnic gaps in mortality in New Zealand.One of the greatest obstacles to extending the long-run improvements in life expectancy into the future is tobacco use. We have previously quantified the impact of smoking on ethnic and socioeconomic inequalities in New Zealand during the 1990s.8-11 A Tupeka Kore Vision has been developed which seeks the end of tobacco use in New Zealand by 2020.12This paper asks the question: if New Zealand did end tobacco smoking by 2020, what would the effect be on life expectancy by 2040, and in particular the gap between M ori and non-M ori?To attempt to answer this question, we present estimates of life expectancy for M ori and non-M ori, and current- and never-smokers, in 1996-99 and then project these life expectancy estimates out to 2040. Assuming a substantive upgrade in tobacco control such that smoking prevalence is negligible by 2020, and allowing a 20 year wash-out period for the majority of tobaccos effect on excess mortality, we assume that life expectancy in 2040 will be that projected for never-smokers.Methods In an accompanying paper in this issue of the Journal,7 we present life-tables and life expectancies for multiple combinations of time, ethnic group, income tertiles and smoking status, using mortality rates from linked census-mortality data.7 Here we use life-tables and life expectancy by sex/smoking/ethnicity, estimates of future mortality decline among never-smokers, and estimates of future rate differences for current- versus never-smokers (and ex- compared to never-smokers in one analysis), to estimate life expectancy in 2040. Life-tables and life expectancy Briefly, the method used to create the life-tables brings together three pieces of information: The official Statistics New Zealand life-tables by year and sex; The distribution of the total New Zealand population by the variables of interest (e.g. the proportion who smoke and ethnicity); and Estimates of the differences in subpopulation mortality rates (from the New Zealand Census-Mortality Study [NZCMS]). These three inputs were combined to produce mortality rates by single year of age for each subpopulation, and complete life-tables including central death rates (mx) and probabilities of death (qx), calculated for each age (x:0-100), which were then used to derive other life-table functions such as life expectancy for each subpopulation. The life-tables used in this paper were generated for males and females by ethnicity (M ori and non-M ori), and smoking status (never-, current- and ex-smoker), for the 1996-99 census-mortality cohort. For smoking life-tables, mortality rates up to age 14 were assumed to be those of the sex by ethnic group (i.e. not stratified by smoking status). (The 2006 Census, which also includes a smoking variable, is not yet linked to mortality data.) Projections We focus first on projections for never- and current- smokers. The method of estimation to 2040 involved three steps: Estimating mortality rates by single year of age for never-smokers (by sex and ethnic group) in 2040; Estimating mortality rate differences (by single year of age) between current- and never-smokers in 2040, and adding this to the never-smoker mortality rates to get the smoker mortality rates; and Deriving life-tables and life expectancy in 2040 using these estimated mortality rates (mx) to calculate mortality risk (qx) and remaining life-table parameters. These steps are outlined in more detail below. Step 1We set an initial estimate of the annual percentage reduction in never-smoker mortality rates mx of 2.5% for M ori and 1.5% for non-M ori. These figures are consistent with Statistics New Zealand projections for low, medium and high mortality scenarios of 2.1%, 1.6% and 1.0% percent per year reductions in mortality rates for males (ethnic groups combined), and 2.4%, 1.8% and 1.3% percent per year reductions for females (http://www.stats.govt.nz/methods_and_services/TableBuilder/population-projections-tables.aspx). We modified these estimates to allow for ethnic variation in mortality rate reductions (the long-run trends suggest mortality rates are falling faster for M ori), and assumed that mortality reductions in the future will be the same for males and females. Extrapolating to 2040, we multiplied the non-M ori mx in 1996, for every single year of age, by 0.98544= 0.514, where 0.985 is one minus the annual percentage reduction of 1.5% and 44 is 2040 minus 1996. The M ori mx was multiplied by 0.97544 = 0.328. We also projected more optimistic annual reductions in mortality rates of 3.5% per annum for M ori and 2.0% per annum for non-M ori using the long-run trends of improving life expectancy over the last 100 years. Step 2We have previously found that the mortality rate difference (not the rate ratio) comparing smokers to never-smokers, is consistent across time and ethnic group.11 Therefore, we set one option for the smoking:non-smoking mortality rate differences in 2040 as being the same as that observed in 1996-99. However, constant rate differences over time mean increasing rate ratios if the mortality rate among never-smokers is reducing. For example, the rate ratio comparing current- to never-smokers within M ori is roughly 1.5 in 1996-99. Under the assumption of 2.5% per annum reduction in M ori never-smoker mortality rates to 2040, a constant rate difference would see the rate ratio increase to 1+ (1.5-1)/0.328 = 2.5. We also explored the effects of alternative assumptions of 1% and 2% per annum reductions in the rate difference, applied similarly within sex by ethnic groups. Under the 2% per annum reduction in the rate difference (and the 2.5% reduction in M ori never-smoker mortality rates), the rate ratio of 1.5 in 1996 would be about 1.6 in 2040. Step 3With mortality rates by single year of age for all sex/ethnicity/smoking status groups, life-tables were easily calculated for 2040. Because of the relatively simple nature of our projections and the differential mortality rate declines by ethnic group, it was possible for estimated M ori never-smoker mortality rates to be less than non-M ori never-smoker mortality rates in 2040 for some single years of age, and likewise for estimates of M ori current-smoker mortality rates to be less than non-M ori current- smoker rates in 2040. It is possible that M ori mortality will fall below that of non-M ori at some time in the future, if the long-run trend since 1900 continues. However, for the purposes of this analysis, we assumed no more than convergence and therefore, where necessary, forced the M ori rate to equal the projected non-M ori rate in 2040. Ex-smokersThe ex-smoker compared to never-smoker mortality rate differences and rate ratios obtained from the NZCMS should be treated with caution. This is because we do not have data on the time since quitting, which makes ex-smoker mortality experience in 1996-99 difficult to interpret and potentially unreliable as the basis for future projections. Nevertheless, to provide a point of comparison, we also calculated life expectancy in 2040 for sex by ethnic groups assuming the 2006 census distribution of smoking behaviour (never-, current-, and ex-smoker). A parallel method to that described above for current-smoker mortality rate projections was used for ex-smokers. Sensitivity analyses In addition to varying the percentage annual decline in never-smoker mortality rates, and percentage decline in the smoking:non-smokiing mortality rate difference, we tested two other assumptions. First, life expectancy in the future will be influenced by mortality over the age of 100, as the proportion of centenarians in the population increases. Thus, we extended the life-tables out to age 120 by simply assuming that the mortality rate increased by 6% per year of age for every year of age over 100, where 6% is approximately the change in mortality by year of age from 90 to 100. (The methods and life-tables in the accompanying paper in this Journal7 apply to age 100, the top end of current \u2018official Statistics New Zealand life-tables.) Altering this 6% percent increase down to 4% and up to 10% had only a negligible influence of the results presented in this paper, and is therefore not discussed further here. Second, we had to estimate mortality rate ratios for smokers compared to never-smokers beyond the age of 80 in the accompanying paper,7 because the 1996-99 census-mortality cohort only included deaths up to age 77. Our assumption was that the predicted mortality rate ratio for current-smokers compared to never-smokers at age 80 reduced linearly to 1.0 by age 100. For the estimates to 2040 in this paper, we investigated setting a minimum rate ratio (and hence rate difference) at all ages - essentially ages above 80 years. Setting such a minimum at 1.2, or even 1.5, had negligible impact on the estimations in this paper, so is not presented further. A copy of the Microsoft Excel spreadsheet used to generate all estimates in this current paper is provided at the NZCMS website (www.uow.otago.ac.nz/nzcms-info.html), and allows interested users to alter any of the input assumptions specified above. Results Life epectancy 1996-9 Figure 1 shows estimates of life expectancy in 1996-99, the most recent cohort for which we have smoking data. The estimates are reproduced in Table 1, with the addition of gaps in years of life expectancy between M ori and non-M ori within smoking status groups, and conversely gaps in life expectancy between smoking status groups within M ori and non-M ori populations. Table 1. Life expectancy estimates for 1996-99 (observed) Life expectancy by smoking status Smoking gap 1996-99 (i.e. Figure 1) Never Current Total Never: Current Never: Total Males M ori 68.1 63.8 66.4 4.3 1.7 Non-M ori 78.3 70.9 75.4 7.4 2.9 Ethnic gap 10.2 7.2 9.0 Females M ori 73.4 69.5 71.4 3.9 2.0 Non-M ori 82.2 76.0 80.5 6.2 1.7 Ethnic gap 8.8 6.5 9.1 Figure 1. Life expectancy in years for 1996-99, by sex, ethnicity and smoking status Two patterns are evident. First, current-smokers have lower life expectancy among males and females for M ori and non-M ori strata. However, the gap between current- and never-smokers is less among M ori than among non-M ori (4.3 and 3.9 years among M ori males and females respectively, compared to 7.4 and 6.2 years among non-M ori). Second, M ori have lower life expectancy in all smoking strata. However, the gap between M ori and non-M ori is greater among never-smokers than among current-smokers (10.2 and 8.8 years for male and female never-smokers respectively, compared to 7.2 and 6.5 years for current-smokers). Projections to 2040 Table 2 shows our projected 2040 life expectancy estimates for never- and current-smokers for the six scenarios. Assuming New Zealand is smokefree by 2020, and allowing for a wash-out period of 20 years for past smoking-related mortality risk, then we might assume that 2040 life expectancies are approximated by the never-smokers. For all scenarios, and all strata of ethnicity by smoking, there are substantial improvements in life expectancy compared to 1996-99. Regardless of the scenario, ethnic gaps within strata of smoking are also reduced. Gaps in life expectancy between current- and never-smokers (column iv, Table 2) range from 5 to 6 years in Scenario C (2.5/1.5% annual reduction in M ori/non-M ori never-smoker mortality; 2% per annum reduction in mortality rate difference between current- and never-smokers) up to 11 to 13 years in Scenario D (3.5/2.5% annual reduction in M ori/non-M ori never-smoker mortality; 0% per annum reduction in mortality rate difference between current- and never-smokers). Regardless of the scenario, the impact of smoking on life expectancy is more similar for M ori and non-M ori than that observed in 1996-99. \u2018 If the 2006 census smoking prevalence remains unchanged into the future (i.e. \u2018total in Table 2), we estimate the difference in 2040 between M ori and non-M ori life expectancy to range from 1.8 to 6.1 years across the six scenarios and two sexes (average 3.8 years; \u2018ethnic gap estimates in column (iii)). By comparing the life expectancy of \u2018never-smokers and \u2018total across scenarios A to F, we have estimates of the additional gains in projected life expectancy in 2040 if nobody smoked tobacco from 2020 compared to the 2006 smoking distribution continuing indefinitely (i.e. column (v) in Table 2). Accordingly, we estimate additional gains in life expectancy for M ori ranging from 2.5 to 7.9 years (average 4.7) and for non-M ori ranging from 1.2 to 5.4 years (average 2.9). That is, going smokefree as a nation will (we estimate) result in larger improvements in M ori life expectancy, compared to non-M ori, and therefore result in a closing in ethnic inequalities in life expectancy ranging from 0.3 to 4.6 years (average 1.8 years). The estimated closing of ethnic gaps in life expectancy was consistently greater for females, reflecting the particularly high 2006 smoking prevalence among M ori females. Discussion If the 2006 census smoking prevalence remains unchanged into the future, and background non-smoking related mortality continues to decrease more so for M ori than non-M ori, we estimate that the difference in 2040 between M ori and non-M ori life expectancy will be about three and a half years (averaged across sex). However, if nobody smokes tobacco from 2020 onwards, we estimate about 5 years of additional gain in life expectancy for M ori (range across six scenarios and sexes 2.5 to 7.9 years) andabout 3 years for non-M ori (range 1.2 to 5.4 years), therefore contributing about 2 years (range 0.3 to 4.6 years) of closing in ethnic inequalities in life expectancy. We emphasise that our exact estimates are quantitatively uncertain. But we do conclude that if nobody smokes tobacco by 2020 in New Zealand there will be substantive improvements in overall population life expectancy. And it will be an important, if not necessary, step towards the ending of M ori:non-M ori inequalities in mortality by 2040200 years after the signing of the Treaty of Waitangi. A number of subsidiary findings also arise out of our projections. Perhaps surprisingly at first glance, we find that the gap between current- and never-smoker life expectancy in 1996-99 was less among M ori (4.3 and 3.9 for males and females respectively) than among non-M ori (7.4 and 6.2). The reason for this is that the absolute additional mortality burden from smoking (i.e. the rate difference in mortality between current- smokers and never-smokers) is about the same across ethnic groups, sexes and time.8 The corollary of this is that the relative risk comparing the mortality of smokers and never-smokers is lessamong M ori. This is due to the much higher background mortality among M ori never-smokers compared to non-M ori never-smokers, which in turn is due to all the other non-tobacco determinants of mortality that vary between M ori and non-M ori.9 11 However, if as we assumed in this paper, M ori mortality rates fall faster than non-M ori mortality rates in the next few decades (i.e. if mortality rate reductions return to their long-run trends of the last century, as opposed to a reversed pattern in the 1980s and 1990s3 4 13 associated with structural changes in the economy), then the relative impact of smoking on mortality among M ori will increase faster than among non-M ori. That is, the M ori life-table will move into a state where smoking has a larger impact on life expectancy gains than it does now. Thus, by 2040 we estimate that th

The public health case for concern and action on tobacco use in Aotearoa-New Zealand is overwhelming; 4500 to 5000 deaths per year in New Zealand can be attributed to tobacco use.1The long-run trends in life expectancy show continual improvement in non-M ori life expectancy and a substantial increase in M ori life expectancy since the turn of the 19th Century.2-4 Over the last century, average annual reductions in mortality have been approximately 3.5% per annum for M ori and 2.0% per annum for non-M ori (Woodward and Blakely, work in progress). Such reductions do not occur by chance, but reflect concerted policy and public health efforts, in addition to general improvements in health services and standard of living. Furthermore, closing of ethnic inequalities in life expectancy are far from guaranteed, as evidenced by a widening of M ori:non-M ori life expectancy gaps in the 1980s and 1990s associated with structural changes in New Zealand society.3,5-7The gap in life expectancy between M ori and non-M ori c.2006 remains large at 7 to 8 years, but has narrowed from the 9 to 10 year gap c.1996.3,4,7 Such a pattern suggests the possibility of a return to long-run trends of closing ethnic gaps in mortality in New Zealand.One of the greatest obstacles to extending the long-run improvements in life expectancy into the future is tobacco use. We have previously quantified the impact of smoking on ethnic and socioeconomic inequalities in New Zealand during the 1990s.8-11 A Tupeka Kore Vision has been developed which seeks the end of tobacco use in New Zealand by 2020.12This paper asks the question: if New Zealand did end tobacco smoking by 2020, what would the effect be on life expectancy by 2040, and in particular the gap between M ori and non-M ori?To attempt to answer this question, we present estimates of life expectancy for M ori and non-M ori, and current- and never-smokers, in 1996-99 and then project these life expectancy estimates out to 2040. Assuming a substantive upgrade in tobacco control such that smoking prevalence is negligible by 2020, and allowing a 20 year wash-out period for the majority of tobaccos effect on excess mortality, we assume that life expectancy in 2040 will be that projected for never-smokers.Methods In an accompanying paper in this issue of the Journal,7 we present life-tables and life expectancies for multiple combinations of time, ethnic group, income tertiles and smoking status, using mortality rates from linked census-mortality data.7 Here we use life-tables and life expectancy by sex/smoking/ethnicity, estimates of future mortality decline among never-smokers, and estimates of future rate differences for current- versus never-smokers (and ex- compared to never-smokers in one analysis), to estimate life expectancy in 2040. Life-tables and life expectancy Briefly, the method used to create the life-tables brings together three pieces of information: The official Statistics New Zealand life-tables by year and sex; The distribution of the total New Zealand population by the variables of interest (e.g. the proportion who smoke and ethnicity); and Estimates of the differences in subpopulation mortality rates (from the New Zealand Census-Mortality Study [NZCMS]). These three inputs were combined to produce mortality rates by single year of age for each subpopulation, and complete life-tables including central death rates (mx) and probabilities of death (qx), calculated for each age (x:0-100), which were then used to derive other life-table functions such as life expectancy for each subpopulation. The life-tables used in this paper were generated for males and females by ethnicity (M ori and non-M ori), and smoking status (never-, current- and ex-smoker), for the 1996-99 census-mortality cohort. For smoking life-tables, mortality rates up to age 14 were assumed to be those of the sex by ethnic group (i.e. not stratified by smoking status). (The 2006 Census, which also includes a smoking variable, is not yet linked to mortality data.) Projections We focus first on projections for never- and current- smokers. The method of estimation to 2040 involved three steps: Estimating mortality rates by single year of age for never-smokers (by sex and ethnic group) in 2040; Estimating mortality rate differences (by single year of age) between current- and never-smokers in 2040, and adding this to the never-smoker mortality rates to get the smoker mortality rates; and Deriving life-tables and life expectancy in 2040 using these estimated mortality rates (mx) to calculate mortality risk (qx) and remaining life-table parameters. These steps are outlined in more detail below. Step 1We set an initial estimate of the annual percentage reduction in never-smoker mortality rates mx of 2.5% for M ori and 1.5% for non-M ori. These figures are consistent with Statistics New Zealand projections for low, medium and high mortality scenarios of 2.1%, 1.6% and 1.0% percent per year reductions in mortality rates for males (ethnic groups combined), and 2.4%, 1.8% and 1.3% percent per year reductions for females (http://www.stats.govt.nz/methods_and_services/TableBuilder/population-projections-tables.aspx). We modified these estimates to allow for ethnic variation in mortality rate reductions (the long-run trends suggest mortality rates are falling faster for M ori), and assumed that mortality reductions in the future will be the same for males and females. Extrapolating to 2040, we multiplied the non-M ori mx in 1996, for every single year of age, by 0.98544= 0.514, where 0.985 is one minus the annual percentage reduction of 1.5% and 44 is 2040 minus 1996. The M ori mx was multiplied by 0.97544 = 0.328. We also projected more optimistic annual reductions in mortality rates of 3.5% per annum for M ori and 2.0% per annum for non-M ori using the long-run trends of improving life expectancy over the last 100 years. Step 2We have previously found that the mortality rate difference (not the rate ratio) comparing smokers to never-smokers, is consistent across time and ethnic group.11 Therefore, we set one option for the smoking:non-smoking mortality rate differences in 2040 as being the same as that observed in 1996-99. However, constant rate differences over time mean increasing rate ratios if the mortality rate among never-smokers is reducing. For example, the rate ratio comparing current- to never-smokers within M ori is roughly 1.5 in 1996-99. Under the assumption of 2.5% per annum reduction in M ori never-smoker mortality rates to 2040, a constant rate difference would see the rate ratio increase to 1+ (1.5-1)/0.328 = 2.5. We also explored the effects of alternative assumptions of 1% and 2% per annum reductions in the rate difference, applied similarly within sex by ethnic groups. Under the 2% per annum reduction in the rate difference (and the 2.5% reduction in M ori never-smoker mortality rates), the rate ratio of 1.5 in 1996 would be about 1.6 in 2040. Step 3With mortality rates by single year of age for all sex/ethnicity/smoking status groups, life-tables were easily calculated for 2040. Because of the relatively simple nature of our projections and the differential mortality rate declines by ethnic group, it was possible for estimated M ori never-smoker mortality rates to be less than non-M ori never-smoker mortality rates in 2040 for some single years of age, and likewise for estimates of M ori current-smoker mortality rates to be less than non-M ori current- smoker rates in 2040. It is possible that M ori mortality will fall below that of non-M ori at some time in the future, if the long-run trend since 1900 continues. However, for the purposes of this analysis, we assumed no more than convergence and therefore, where necessary, forced the M ori rate to equal the projected non-M ori rate in 2040. Ex-smokersThe ex-smoker compared to never-smoker mortality rate differences and rate ratios obtained from the NZCMS should be treated with caution. This is because we do not have data on the time since quitting, which makes ex-smoker mortality experience in 1996-99 difficult to interpret and potentially unreliable as the basis for future projections. Nevertheless, to provide a point of comparison, we also calculated life expectancy in 2040 for sex by ethnic groups assuming the 2006 census distribution of smoking behaviour (never-, current-, and ex-smoker). A parallel method to that described above for current-smoker mortality rate projections was used for ex-smokers. Sensitivity analyses In addition to varying the percentage annual decline in never-smoker mortality rates, and percentage decline in the smoking:non-smokiing mortality rate difference, we tested two other assumptions. First, life expectancy in the future will be influenced by mortality over the age of 100, as the proportion of centenarians in the population increases. Thus, we extended the life-tables out to age 120 by simply assuming that the mortality rate increased by 6% per year of age for every year of age over 100, where 6% is approximately the change in mortality by year of age from 90 to 100. (The methods and life-tables in the accompanying paper in this Journal7 apply to age 100, the top end of current \u2018official Statistics New Zealand life-tables.) Altering this 6% percent increase down to 4% and up to 10% had only a negligible influence of the results presented in this paper, and is therefore not discussed further here. Second, we had to estimate mortality rate ratios for smokers compared to never-smokers beyond the age of 80 in the accompanying paper,7 because the 1996-99 census-mortality cohort only included deaths up to age 77. Our assumption was that the predicted mortality rate ratio for current-smokers compared to never-smokers at age 80 reduced linearly to 1.0 by age 100. For the estimates to 2040 in this paper, we investigated setting a minimum rate ratio (and hence rate difference) at all ages - essentially ages above 80 years. Setting such a minimum at 1.2, or even 1.5, had negligible impact on the estimations in this paper, so is not presented further. A copy of the Microsoft Excel spreadsheet used to generate all estimates in this current paper is provided at the NZCMS website (www.uow.otago.ac.nz/nzcms-info.html), and allows interested users to alter any of the input assumptions specified above. Results Life epectancy 1996-9 Figure 1 shows estimates of life expectancy in 1996-99, the most recent cohort for which we have smoking data. The estimates are reproduced in Table 1, with the addition of gaps in years of life expectancy between M ori and non-M ori within smoking status groups, and conversely gaps in life expectancy between smoking status groups within M ori and non-M ori populations. Table 1. Life expectancy estimates for 1996-99 (observed) Life expectancy by smoking status Smoking gap 1996-99 (i.e. Figure 1) Never Current Total Never: Current Never: Total Males M ori 68.1 63.8 66.4 4.3 1.7 Non-M ori 78.3 70.9 75.4 7.4 2.9 Ethnic gap 10.2 7.2 9.0 Females M ori 73.4 69.5 71.4 3.9 2.0 Non-M ori 82.2 76.0 80.5 6.2 1.7 Ethnic gap 8.8 6.5 9.1 Figure 1. Life expectancy in years for 1996-99, by sex, ethnicity and smoking status Two patterns are evident. First, current-smokers have lower life expectancy among males and females for M ori and non-M ori strata. However, the gap between current- and never-smokers is less among M ori than among non-M ori (4.3 and 3.9 years among M ori males and females respectively, compared to 7.4 and 6.2 years among non-M ori). Second, M ori have lower life expectancy in all smoking strata. However, the gap between M ori and non-M ori is greater among never-smokers than among current-smokers (10.2 and 8.8 years for male and female never-smokers respectively, compared to 7.2 and 6.5 years for current-smokers). Projections to 2040 Table 2 shows our projected 2040 life expectancy estimates for never- and current-smokers for the six scenarios. Assuming New Zealand is smokefree by 2020, and allowing for a wash-out period of 20 years for past smoking-related mortality risk, then we might assume that 2040 life expectancies are approximated by the never-smokers. For all scenarios, and all strata of ethnicity by smoking, there are substantial improvements in life expectancy compared to 1996-99. Regardless of the scenario, ethnic gaps within strata of smoking are also reduced. Gaps in life expectancy between current- and never-smokers (column iv, Table 2) range from 5 to 6 years in Scenario C (2.5/1.5% annual reduction in M ori/non-M ori never-smoker mortality; 2% per annum reduction in mortality rate difference between current- and never-smokers) up to 11 to 13 years in Scenario D (3.5/2.5% annual reduction in M ori/non-M ori never-smoker mortality; 0% per annum reduction in mortality rate difference between current- and never-smokers). Regardless of the scenario, the impact of smoking on life expectancy is more similar for M ori and non-M ori than that observed in 1996-99. \u2018 If the 2006 census smoking prevalence remains unchanged into the future (i.e. \u2018total in Table 2), we estimate the difference in 2040 between M ori and non-M ori life expectancy to range from 1.8 to 6.1 years across the six scenarios and two sexes (average 3.8 years; \u2018ethnic gap estimates in column (iii)). By comparing the life expectancy of \u2018never-smokers and \u2018total across scenarios A to F, we have estimates of the additional gains in projected life expectancy in 2040 if nobody smoked tobacco from 2020 compared to the 2006 smoking distribution continuing indefinitely (i.e. column (v) in Table 2). Accordingly, we estimate additional gains in life expectancy for M ori ranging from 2.5 to 7.9 years (average 4.7) and for non-M ori ranging from 1.2 to 5.4 years (average 2.9). That is, going smokefree as a nation will (we estimate) result in larger improvements in M ori life expectancy, compared to non-M ori, and therefore result in a closing in ethnic inequalities in life expectancy ranging from 0.3 to 4.6 years (average 1.8 years). The estimated closing of ethnic gaps in life expectancy was consistently greater for females, reflecting the particularly high 2006 smoking prevalence among M ori females. Discussion If the 2006 census smoking prevalence remains unchanged into the future, and background non-smoking related mortality continues to decrease more so for M ori than non-M ori, we estimate that the difference in 2040 between M ori and non-M ori life expectancy will be about three and a half years (averaged across sex). However, if nobody smokes tobacco from 2020 onwards, we estimate about 5 years of additional gain in life expectancy for M ori (range across six scenarios and sexes 2.5 to 7.9 years) andabout 3 years for non-M ori (range 1.2 to 5.4 years), therefore contributing about 2 years (range 0.3 to 4.6 years) of closing in ethnic inequalities in life expectancy. We emphasise that our exact estimates are quantitatively uncertain. But we do conclude that if nobody smokes tobacco by 2020 in New Zealand there will be substantive improvements in overall population life expectancy. And it will be an important, if not necessary, step towards the ending of M ori:non-M ori inequalities in mortality by 2040200 years after the signing of the Treaty of Waitangi. A number of subsidiary findings also arise out of our projections. Perhaps surprisingly at first glance, we find that the gap between current- and never-smoker life expectancy in 1996-99 was less among M ori (4.3 and 3.9 for males and females respectively) than among non-M ori (7.4 and 6.2). The reason for this is that the absolute additional mortality burden from smoking (i.e. the rate difference in mortality between current- smokers and never-smokers) is about the same across ethnic groups, sexes and time.8 The corollary of this is that the relative risk comparing the mortality of smokers and never-smokers is lessamong M ori. This is due to the much higher background mortality among M ori never-smokers compared to non-M ori never-smokers, which in turn is due to all the other non-tobacco determinants of mortality that vary between M ori and non-M ori.9 11 However, if as we assumed in this paper, M ori mortality rates fall faster than non-M ori mortality rates in the next few decades (i.e. if mortality rate reductions return to their long-run trends of the last century, as opposed to a reversed pattern in the 1980s and 1990s3 4 13 associated with structural changes in the economy), then the relative impact of smoking on mortality among M ori will increase faster than among non-M ori. That is, the M ori life-table will move into a state where smoking has a larger impact on life expectancy gains than it does now. Thus, by 2040 we estimate that th