Smoking in New Zealand: a Census Investigation

Australian Journal of Public Health, Vol 19, No 2, April 1995, p.125-129.

Abstract: New Zealand may well be unique in that in the 1976 and the 1981 Census of Population and Dwellings each person over the age of 15 was asked about their cigarette smoking habits. The data is available on the basis of age and ethnicity, enabling an examination of the prevalence of ever-smoking by cohort born some 80 years before the censuses was taken, at the end of the last century. Thus the impact of ever-smoking on mortality can be calculated.  Ever regularly smoking reduced the life expectation of males by 11.7 years and females by 15.6 years. The corresponding reductions for Maori were 19.3 years and 23.8 years. The effect of smoking on health has been of epidemic proportions, notably for the Maori.
 Keywords: Health; Maori; Statistics;



The New Zealand Population Census
Every five years the Department of Statistics (now Statistics New Zealand) surveys every person (and dwelling) in New Zealand on a particular Tuesday night (usually) in March. A wide range of objective questions are asked, including age, gender, ethnicity, and other socio-economic characteristics. Respondents self-categorize.
            In 1976 and 1981 census asked the following question:
“Cigarette Smoking: Tick the box which best describes your current cigarette smoking
G Never smoked cigarettes at all, or never smoked them regularly.
G Do not smoke cigarettes now, but used to smoke them regularly (1 or more a day).
G Currently smoke cigarettes regularly (1 or more a day).”
            (In addition respondents were asked to “specify number smoked yesterday …cigarettes (if none write `0′).” An evaluation indicated “the data elicited on number of cigarettes smoked was extremely unreliable. The level of under-reporting of cigarettes smoked per day for the total population was estimated to be 26 percent in 1976, and 28 percent in 1981.” (1,2)  However Statistics New Zealand is satisfied with the information relating to cigarette use.)
            While it might be assumed that there would be a reasonably accurate recall on this question, some respondents may misrepresent their smoking experience. The group most likely to under-report are teenagers. (This study, however, focuses on the over 25s.) Unfortunately the question treats those who smoke, and have only smoked, cigars and pipe tobacco as non-smokers of cigarettes. (Chewing tobacco and snuff is uncommon in New Zealand.) Sales of loose tobacco exceeded manufactured cigarettes (by weight) until 1955 (3). But loose tobacco is used for roll-your-owns as well as pipes, and many pipe smokers may also have used cigarettes on a regular basis at some stage.

New Zealand’s indigenous people (tangata whenua) are the Maori, who are of Polynesian origin. The 1976 and 1981 Censuses defined a Maori as anyone who reported being at least one half Maori descent. In 1976 6.7 percent of the population over 15 were Maori by this definition, and in 1981 it was 7.3 percent. The Maori population is a somewhat younger than the whole population. Since their numbers in older age groups are small, statistics derived from the earlier cohorts are less reliable.
            As we shall see Maori smoking behaviour is different to the population as a whole. There was no smoking of tobacco or other substances by the Maori before the arrival of the European at the end of the eighteenth century, but they took up the habit with alacrity. One observer commented that in the early 1840s “smoking is universal among New Zealanders [Maori] of both sexes” (4). At the turn of the century Maori women are frequently depicted smoking pipes, although smoking among European (pakeha) women was uncommon. Today rates among Maori remain high. (5)

Method: The Prevalence of Smoking in the 1976 Population
Table 1 shows the prevalence of cigarette smoking by gender and birth date in 1976 for the whole population. The data is given for those who have “ever-smoked”, and also the “quit rate” (quit rate = (ever-smoked – now-smoke)/(ever-smoke))  Table 2 is the parallel data for the Maori. (The data for 1981 is available from the author.)
            It is not useful to compare the levels in the various cohorts, since smoking has a higher rate of mortality. Differences in levels may reflect the greater death rate of smokers and ex-smokers. We can look at differences between social groups of the same age. In the population as a whole men were more likely to have smoked at any age than women, except for the 1956/61 cohort who were between 15 and 20 years old at the time of the 1976 census. Except for women under 25 in 1976, the quit rates are higher for men than women, although in every age cohort except adolescents there were more men than women still smoking.
            Surprisingly, given popular perception, older living Maori males are less likely to have smoked than all males, although we shall see this is a mortality effect. Maori male quit rates are also lower than all men (and all women), so the proportion of Maori men smoking in 1976 was higher than for the whole population, which probably generates the impression of the widespread Maori prevalence of smoking.
            Proportionally more Maori women have been smoking than all women. For Maori women under the age of 45 in 1976, the prevalence of ever-smoking is similar to Maori men. Their quit rates are the lowest of the four groups, so those under 50 in 1976 were more likely to be smoking than other groups.

Method: The Smoking Mortality Differential
One of the reasons why the ever-smoker proportion falls off for older people is because smokers experience earlier mortality than non-smokers. The existence of the same question in the 1976 and 1981 population censuses makes it possible to estimate the (census) survivorship differential.
            (Suppose the population in a cohort is x in year t, and x’ in year t+5. Suppose that the rate of out-migration in the five year period is m, and the survivorship rate is S. In which case the remaining population is x(1-m)S = x’. Hence S =  x’/(x(1-m)). Suppose there are two population groups with the same migration rate, but initial populations x and y, final populations x’ and y’, and survivorship rates S and T. Then S/T = (x’/x)/(y’/y), which gives the survivorship differential.)
            This requires the following assumptions: there is no change in reportage between the two censuses; the smoking habits of migrants are the same as the population in the same age group; never-smokers do not take up smoking in the quinquennium, which is why differential survivorship rates were not calculated for those under the age of 25.)
            The resulting survivorship differential is shown in Table 3. The percentage should be interpreted as follows. Suppose it is 10 percent. If the survivorship rate for never-smokers in a particular age group over the next five years is 99,000 per 100,000 then the mortality rate for ever-smokers is 90,000 per 100,000 (i.e 99/1.1).
            The differentials in the rates are not implausible, tending to show rising rates with age for men. Women seem to suffer higher mortality rates at younger and middle ages, with a lower rate later perhaps because of the early deaths.
            The rates reflect a mix of complicated smoking behaviour – since different cohorts smoked at different intensities, and for different lengths of time – and from the way the smoking related diseases affect individuals. There is an erraticness in the rates which may reflect these, measurement error, or a departure from the assumptions.

Method: The Prevalence of Ever-Smoking By Cohort
We use the differential ratios to infer back the relative death rate of ever-smokers and never-smokers, and hence estimate the prevalence of smoking of each generation, including the dead as well as the alive. This requires the further assumption that the age specific differential survivorship rates which applied between 1976 and 1981, applied for earlier years as well. The estimates are shown in Table 4.
            Around 80 percent of All Men born before 1926 went through a regular cigarette smoking phase. There is a drop off among those younger, so the prevalence for those born after the Second World War was below 60 percent. We can but conjecture why there was the change in trend, but cigarettes were distributed as a part of a soldier’s rations. A very high proportion of those born before 1926 would have had service experience which few born after 1926 would have experienced.
            The pattern for All Women is that ever-smoking was low for those born at the turn of the century, consistent with anecdote. However the prevalence rose among those who would have been “flappers” in the 1920s. For the generation born in the 1920s the prevalence was about 50 percent, a level which continued on, at least, until for those born in the early 1950s.
            The ever-smoking prevalence among Maori Men has been consistently in the 75 percent to 80 percent for the whole period. The lower prevalence we observed in the actual data of the 1976 census reflected higher mortality. The prevalence among Maori Women has been rising – from 65 percent at the turn of the century to above the male Maori rate for those born in the early 1950s. Whatever the forces which have reduced the prevalence of ever-smoking in the population as a whole, they have not been as effective on the Maori, a gloomy conclusion reinforced by the lower reported quit rates for the Maori.
            For the non-Maori (mainly European or “Pakeha”) New Zealander, the census data, adjusted for differential mortality, confirms the impression from history, especially the prevalence of smoking among women increased markedly in the inter-war period; the differential prevalence between men and women has been falling; and this has been due to reduced prevalence among men, with little change for women.

Discussion: The Effect of Ever-Smoking on Life Expectancy
The effect of ever-smoking upon life expectancy can be calculated by adjusting the official estimates of average survivorship for the 1975-77 period (6), to obtain actual survivorship rates for each cohort.

 Table 5: LIFE EXPECTANCY: Years at Age 25

<> <> <>  <>Never Smoker <>Ever Smoker
Table 5 shows that the average life expectancy for a 25 year old who never smoked would be 55.4 years for a male and 60.2 years for a female. On the other hand the expectancies for an ever-smoker are 43.7 years and 44.6 years respectively. The smoking experience of ever-smokers (and any other factors which are associated with ever-smoking) is to reduce male life expectancy for a 25 year old male by 11.7 years and a female by 15.6 years.
            The Maori differences are even more startling. Table 5 shows that the average life expectancy for a 25 year old Maori who never smoked would be 58.7 years for a male and 65.3 years for a female. On the other hand the life expectancies for an ever-smoker are 39.4 years and 41.5 years respectively. The smoking experience of ever-smokers (and any other factors which are associated with ever-smoking) is to reduce life expectancy for a 25 year old male by 19.3 years and a female by 23.8 years.
            Whether the never-smoking Maori is in fact more robust than the non-Maori counterpart, or whether this is the consequences of a measurement error and/or the unreality of the assumption of constant ever-smoking mortality effects by cohort can only be speculated. (Implicitly the calculations assume that the ever-smoking intensities and life experience are the same through generations.) It should be noted however that this conclusion differs from Smith & Pearce (7), insofar it seems to suggest that almost all of the Maori mortality differential can be explained by their higher smoking rates. The effect arises from more smoking and, apparently, their smoking more intensively (as evidenced by their lower quit rates).
            We can estimate the aggregate differential mortality effects of ever-smoking by asking what would have been the population size if there had been no smoking at all. This is done by applying the never-smoking survivorship rates to the whole population. The calculation involves the additional assumption that smoking, including mortality from smoking, does not affect the birth rate, and there are no smoking induced deaths from passive smoking, or for those under the of age 25.
            Table 6 shows that the total population over the age of 25 of males would be 22.1 percent larger if there had ben no smoking, and 13.9 percent larger by females. The Maori figures are, not unexpectedly, bigger: the population would be 28.7 percent larger for Maori males and 28.3 percent for Maori females. We also looked at the impact on the age group from 45 to 64, from whence much of social leadership is drawn. While the total population would have been 23 percent larger, the Maori population would have been 52 percent larger. Thus the Maori leadership age groups have suffered severely, in a period of major social transformation.
            Mortality rates of 181 per 1000 (i.e. 221/1221) of all males over 25 and 122 per 1000 of females, smoking takes on epidemic proportions. By comparison the notorious influenza epidemic of 1918 had a mortality rate of 5.5 per thousand (for the whole population) for Europeans, and 42.3 per thousand of Maori (8). As devastating as the influenza was, smoking has been much more insidious albeit the deaths are over a longer period.
            It should be noted that all these figures apply for the period from 1976 to 1981, and reflect the smoking behaviour which preceded that. Insofar as there has been a reduction in smoking in the 1980s, and that quitting (and reducing) smoking reduces mortality, the population impact would not be as severe today.

Discussion: Potential Research Extensions
The smoking question was not asked in the 1986 or 1991 Population Censuses although, following active lobbying by the anti-tobacco lobby, a smoking question may be asked again in the 1996 census (without the question of the number of cigarettes being smoked).  If it is, it may be possible to estimate differential survivorship rates from smoking behaviour more accurately, and identify trends over time.
            It would also allow investigation of what has happened to the post-1951 birth cohorts. There is survey data since 1981 on smoking prevalence, which shows further declines in smoking in New Zealand (3). However it is not usually on the same basis as the census question so it is difficult to blend the data.
            Using the existing data, the analysis could be extended to other socio-economic groups (e.g. income, occupation, labour force status, other ethnicities, location), including those too small to be picked up by a sample survey. It might also be worth improving the differential survivorship rates, by using a finer level of data and statistical smoothing techniques.
            A useful adjunct would be a supplementary survey of pipe smokers, to learn whether they were ever regular cigarette smokers.

Discussion: Policy Implications
The census data had a major effect on tobacco policy in New Zealand, when it gave the anti-tobacco lobbies some idea of the magnitude of the phenomenon. Almost everyone recounting the successes of the movement in the last two decades mentions the census as a major stimulant in the campaign (9).
            Because the aggregate levels were so important, detailed analysis was not carried out at the time. This paper is perhaps a pioneer in the microanalysis of the census data. But it describes a situation almost two decades ago. While of historical interest it perhaps has little to say for current policy issues, except it confirms the build up of today’s tobacco related health problems that the smoking then was promising.
            The case for a reintroducing the question in the 1996 New Zealand Population Census is that the entire enumeration of the population is that it provided detailed information on small groups in the community, which samples are unable to provide data. Such groups are increasingly focuses for the anti-smoking campaign.
            Thus the New Zealand anti-tobacco lobbies have had a strong case for their campaign to re-include the question in the 1996 census. Lobbies elsewhere in the world have an even stronger case for pursuing the inclusion of a similar question in their next national censuses.
            In the interim we may be sure that this data confirms the picture of the smoking in the past having a major mortality effect, especially on Maori women.
The author is grateful for helpful comments on earlier versions of this paper from Robert Bowie, Neil Pearce, and Eru Pomare (Wellington Medical School), Brian Cox (Otago Medical School), Murray Laugesen (Public Health Commission), Ian Poole (Population Studies Centre, Waikato University), Sarah Thomson (New Zealand Cancer Society), Statistics New Zealand (the Department of Statistics), and two referees.

1. Statistics New Zealand. 1996 Census of Population and Dwellings: Preliminary Views on Content. Wellington, 1993:94-95.
2. Jackson R, and Beaglehole R. Secular Trends in Underreporting of Cigarette Consumption. Am J Epidemiol, 1985, 122:341-344.
3. Departments of Statistics and Health. Tobacco Statistics. Wellington, 1992:12-13.
4. `Journal’ of a surgeon Henry Weeks, reported in Rutherford J. and Skinner W.H. The Establishment of the New Plymouth Settlement. New Plymouth, 1940:93-94.
5. Reid P. and Robert R. Te-Taonga-mai-Tawhiti (the-gift-from-a-distant-place). Auckland: Niho Taniwha.
6. Department of Statistics. Demographic Trends 1992, Wellington, 1993:85-6.
7. Smith A.H, & N.E. Pearce, Determinants of Difference in Mortality Between New Zealand Maoris and Non-Maoris Aged 15-64, NZ Med J, Feb 22, 1984, Vol 97, No 750, p.101-8.
8. Rice, G. Black November: The 1918 Influenza Epidemic in New Zealand. Wellington, 1988:141-145.
9.Thomson, S. Stubbing Out the Social Cigarette. M.A. thesis, University of Auckland, 1992.


Table 1: 1976 PREVALENCE OF ALL SMOKING: Living Population (Percentage)
 Source: 1976 Population Census

Cohort Men Men Men Women Women Women
March Years
Before 1896 61.3 20.4 66.8 13.3 05.6 58.0
1897-1901 69.9 26.9 61.5 21.3 10.7 49.7
1902-1906 72.6 30.7 57.7 26.9 14.4 46.3
1907-1911 74.5 34.7 53.3 34.8 20.2 42.0
1912-1916 74.4 38.3 48.5 40.7 26.0 36.2
1917-1921 75.1 40.6 46.0 45.2 30.6 32.3
1922-1926 75.6 44.0 41.8 47.8 34.1 28.6
1927-1931 71.8 45.0 37.3 45.8 35.8 26.2
1932-1936 67.0 43.8 34.7 47.4 35.3 25.5
1937-1941 63.4 43.3 31.8 46.9 35.0 25.3
1942-1946 62.0 44.0 28.9 50.7 37.8 25.4
1947-1951 57.3 42.5 25.8 50.3 37.8 24.9
1952-1956 52.2 41.8 20.0 49.3 38.9 21.0
1957-1961 35.3 29.8 15.8 36.2 30.4 16.1



Table 2: 1976 PREVALENCE OF MAORI SMOKING: Living Population (Percentage)
 Source: 1976 Population Census



Source: Based on 1976 to 1981 Censuses

 Source: (based on 1976 Population Census, and differential survivorship rates)


Table 6: INCREASED POPULATION IF THERE WAS NO EVER-SMOKING (Percentage above Actual 1981 Population)