Article Text

Download PDFPDF

The relationship between socioeconomic status and ‘hardcore’ smoking over time – greater accumulation of hardened smokers in low-SES than high-SES smokers
  1. Philip Clare1,
  2. Deborah Bradford1,
  3. Ryan J Courtney1,
  4. Kristy Martire2,
  5. Richard P Mattick1
  1. 1National Drug and Alcohol Research Centre, University of NSW, Sydney, Australia
  2. 2School of Psychology, University of NSW, Sydney, Australia
  1. Correspondence to Philip Clare, NDARC, University of New South Wales, Sydney NSW 2052, Australia; p.clare{at}


Objectives This paper used national survey data to investigate ‘hardcore’ smoking as predicted by the ‘hardening hypothesis’, and to examine the relationship between ‘hardcore’ smoking and socioeconomic status (SES).

Methods Analyses were performed using data from four waves of the Australian National Drug Strategy Household Survey between 2001 and 2010, a large national survey with a sample size of approximately 24 000 participants per wave. The primary outcome variable was ‘hardcore’ smoking, comprised of the variables: ‘no quit attempt in past 12 months’; ‘no plan to quit’; and smoking more than 15 cigarettes per day. The main predictor variables used were SES assessed by the Socio-Economic Indexes for Areas (SEIFA), and survey wave. Other sociodemographic variables were also examined.

Results Overall, ‘hardcore’ smoking remained stable from 2001 to 2010. However, ‘hardcore’ smoking declined among high-SES smokers (from 1.8% to 1.0%), but not among low-SES smokers (around 3.4%). ‘Hardcore’ smoking was strongly associated with SEIFA quintile (p<0.001). There was a significant interaction effect between top and bottom SEIFA quintiles and wave (p=0.025), with a decline in ‘hardcore’ smoking measures over the four waves among those in the top two SEIFA quintiles, with odds in 2010 of 0.39 (95% CI 0.17 to 0.87; p=0.012), down from 0.64 (95% CI 0.50 to 0.82; p<0.001) in 2001, while ‘hardcore’ smoking remained stable among those in the bottom two SEIFA quintiles.

Conclusions The results from high SES smokers suggest ‘hardcore’ smokers are able to quit, but outcomes among low-SES smokers are less encouraging.

  • Socioeconomic status
  • Disparities
  • Priority/special populations

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Smoking is a leading cause of preventable death worldwide,1 projected to be responsible for eight to ten million deaths per year by 2030.2 ,3 It has been suggested that those who continue to smoke despite increasing pressures to quit may represent a particular subpopulation of smokers who are more resistant, or ‘hardened’,4 in their smoking behaviours. That is, those who continue to smoke may be either more resistant to the economic and social/legislative pressures to quit smoking or less able to give up.5 If the ‘hardening hypothesis’ is correct, it would be expected that non-‘hardcore’ smokers will gradually quit smoking while ‘hardcore’ smokers will be less likely to do so. Thus, as the overall rate of smoking decreases, the prevalence of ‘hardcore’ smoking would remain relatively stable.

The ‘hardening hypothesis’ is popular and intuitively appealing. Numerous attempts have been made to define this subpopulation of ‘hardened’ smokers.6 These definitions have usually involved a combination of characteristics including nicotine dependence, history of quit attempts, intention or desire to quit, and history of continuous smoking.7–11

The findings of extant research have been varied, in terms of the rate and the role or importance of ‘hardening’. Research in Western countries has found varying prevalence of ‘hardcore’ smoking from 5.4% to 30%,7–9 12–15 with differences in the prevalence of ‘hardcore’ smoking between and within countries. However, it is unclear how much of this variation is due to actual differences between countries and how much is due to changes over time, further compounded by inconsistencies in the definitions used to define ‘hardcore’ smoking across studies. Additionally, most previous studies of ‘hardcore’ smoking have focussed on the role of affective correlates, such as psychological distress,16 rather than examining an overall change over time in ‘hardcore’ smoking as it is directly defined.

A link has also been observed between smoking and socioeconomic status (SES), with persons from low socioeconomic backgrounds more likely to be smokers in Western countries, including Australia,17–21 the UK18 ,20 ,22 and the USA,18 ,20 ,22 ,23 although this relationship may not generalise to less developed countries.24 ,25 The exact mechanisms behind socioeconomic difference in smoking are unknown, but have been linked to impoverished environments and lack of access to positive activities and alternatives to drugs.26 Some evidence suggests those from low socioeconomic backgrounds are less likely to intend to quit,20 while other evidence suggests low-SES smokers are equally likely to make a quit attempt but less likely to succeed.27 ,28 It would thus be expected that low SES smokers would be over-represented among ‘hardened’ smokers, and that this could compound to leave a population of smokers who are highly disadvantaged and increasingly resistant to quitting.29

This research examined the prevalence and change in ‘hardcore’ smoking measures in Australia over four waves of data in a period of 10 years from 2001 to 2010, to investigate if there has been an increase in the proportion of smokers who are ‘hardcore’ smokers, as is predicted by the ‘hardening hypothesis’, and if there is any difference in the proportion of ‘hardcore’ smokers based on SES.


We used data from four waves—2001, 2004, 2007 and 2010—of the National Drug Strategy Household Survey (NDSHS), a large national survey of drug behaviours conducted in Australia by the Australian Institute of Health and Welfare (AIHW) with a sample of around 24 000 participants per wave and a usual response rate around 50%. More detailed descriptions of the NDSHS design and methodology for each wave are available elsewhere.30–33


Smoking status

Respondents were asked a number of questions about their ‘smoking status’, including ‘Have you ever smoked on a daily basis?’ and ‘How often do you now smoke cigarettes, pipes or other tobacco products?’ Answers to these questions were used to create a single binary variable (‘non-smoker’ or ‘smoker’). Smokers were defined as those participants who reported they were a current smoker (daily, weekly or less than weekly) on any of the relevant questions.

‘Hardcore’ smoking measures

Respondents who identified as smokers were asked how many cigarettes they smoked per day, week or month, depending on how often they smoked. Based on prior definitions of ‘hardcore’ smoking7 ,8 those who smoked more than 15 cigarettes a day on average were coded as ‘heavy’ smokers. Smokers were also asked whether they had made a quit attempt in the past 12 months, and whether they intend to quit, both of which were included as binary indicator variables in analysis (‘no quit attempt in past 12 months’ and ‘no plan to quit’).

In line with past definitions of ‘hardcore’ smoking,6 these variables were combined to create a binary indicator variable for ‘hardcore’ smoking. Hence, respondents were coded as ‘hardcore’ smokers if they smoked more than 15 cigarettes a day, had not made an attempt to quit in the last 12 months, and did not plan to quit.

Sociodemographic characteristics

A number of sociodemographic variables were included as predictors or covariates, including age, sex, educational attainment, marital status, employment status, main language spoken, country of birth, Aboriginal Torres Strait Islander status and being a single parent.

Educational attainment

Based on two questions about education, respondents were coded into a 3-level variable of educational attainment. Those who said they had not completed a qualification were coded as ‘High school or less’, while those who had completed any kind of diploma or certificate were coded as ‘Diploma or certificate’, and those who had completed at least a bachelor degree were coded as ‘University degree’.

Employment status

Based on responses to relevant questions in each wave, those who said they were employed full-time or part-time (including those who were self-employed) were combined into the category ‘Employed’. Those who said they were full-time or part-time students were combined into the category ‘Student’. Those who were unemployed but looking for work were included as ‘Not employed—in the workforce’, while those who reported their employment status as ‘home duties’, said they were on a pension or retired, or were unable to work were combined into the category ‘Not employed—not in workforce’.

Relative socioeconomic advantage and disadvantage measure

Socio-Economic Indexes for Areas (SEIFA) is an area-level index of relative advantage or disadvantage developed by the Australian Bureau of Statistics (ABS).34 Participants were assigned to SEIFA quintiles based on the census collection area of their place of residence, with the bottom quintile representing the most disadvantaged areas and the top quintile representing the most advantaged. In order to maximise difference between groups and to specifically examine differences between ‘low’ and ‘high’ SES, SEIFA quintiles were used to create a binary variable of SEIFA, with the bottom-two quintiles combined into the category ‘bottom-two’ and the top-two quintiles combined into ‘top-two’, while the middle quintile was excluded.

Statistical analysis

All variables except age were entered into the regression analyses as categorical variables. Age was entered in regression analyses as a continuous variable, but was coded into a 4-level categorical variable for descriptive tables, with the categories ‘18–24’, ‘25–39’, ‘40–54’ and ‘55+’. Respondents under 18 years of age were excluded from analysis.

Multiple logistic regression analyses were used to analyse the relationships between the sociodemographic variables and smoking status. Univariate logistic regression of smoking status was used to select variables, with only variables with p values less than 0.05 included in the final regression model. For the sake of comparison, the same variables were used in the regression analyses of the components of ‘hardcore’ smoking, even though not all variables were significant in those analyses. Additionally, one interaction effect was included for ‘hardcore’ smoking but not for the other analyses in which the interaction was not significant (p=0.220, p=0.648 and p=0.215, respectively).

Only responses with complete data for all variables were included in the analysis, for a total sample of n=71 650. The logistic regression models used to analyse the relationship between sociodemographic variables, and no reported quit attempts in the past year, no intention to quit smoking and heavy smoking, were limited to samples of smokers who responded to the relevant questions in the questionnaire, with sample sizes of n=14 094 for reported quit attempts, n=13 799 for intention to quit and n=13 740 for heavy smoking.

All analyses were performed using the ‘svy’ function of Stata V.11.2,35 which controls for complex sample designs and allows for the calculation of SEs and CIs.



Table 1 shows the sociodemographic characteristics of participants, broken down by smoking status. In total, 19.9% in the sample across all waves were smokers: 22.4% in 2001, 20.0% in 2004, 18.5% in 2007 and 19.0% in 2010.

Table 1

Weighted sample characteristics by smoking status

‘Hardcore’ smoking

Approximately 2% of the sample were ‘hardcore’ smokers. Changes over time in the rate of ‘hardcore’ and non-‘hardcore’ smoking are shown in figure 1. The rates of each of the aspects of ‘hardcore’ smoking are shown in table 2. The proportion of smokers who reported no quit attempt in the past 12 months and/or heavy smoking was consistent across the four waves. However, there was a significant association between wave and having no plan to quit (p=0.004), with the odds of not having plans to quit significantly lower in 2010 compared with previous years (OR: 0.87; 95% CI 0.77 to 0.98; p=0.026).

Table 2

Weighted proportion of smokers with ‘hardcore’ smoking characteristics

Figure 1

Prevalence of ‘hardcore’ and non-‘hardcore’ smoking in population, top-two SEIFA quintiles and bottom-two SEIFA quintiles.

‘Hardcore’ smoking was strongly associated with SEIFA (p<0.001), with 3.4% of those from the bottom-two SEIFA quintiles ‘hardcore’ smokers compared with 1.3% of those in the top-two SEIFA quintiles. There was a significant interaction effect between SEIFA quintiles and wave (p=0.025). As shown in figure 2, the odds of being a ‘hardcore’ smoker declined over the four waves among those in the top-two SEIFA quintiles, but not among those in the bottom-two SEIFA quintiles.

Figure 2

Odds and prevalence of ‘hardcore’ smoking by top-two/bottom-two SEIFA quintile and survey year. Note: Prevalence of hardcore smoking for each data point are shown in brackets.

‘Hardcore’ smoking was also related to educational attainment. The odds of ‘hardcore’ smoking were significantly (p<0.001) lower among those with diploma/certificate level qualifications and university degrees compared to those with lower education.

‘Hardcore’ smoking was also significantly associated with a number of other sociodemographic variables, as shown in table 3.

Table 3

Weighted, adjusted OR and 95% CIs for the effects of covariates on ‘hardcore’ smoking (n=71 650)

Components of ‘hardcore’ smoking

There was no significant difference by SEIFA quintile in not making a quit attempt in the last 12 months. However, those in the top-two quintiles showed significantly lower odds of having no plan to quit (OR: 0.78; 95% CI 0.71 to 0.86; p<0.001) and heavy smoking (OR: 0.67; 95% CI 0.62 to 0.74; p<0.001). A similar relationship was seen with employment status, with those unemployed but still in the workforce showing higher odds of having no plan to quit, (OR: 1.31; 95% CI 1.07 to 1.61; p=0.010) and of being heavy smokers, (OR: 1.46; 95% CI 1.21 to 1.77; p<0.001). There were also significant associations between educational attainment and all three component variables, with the odds decreasing as educational attainment increased.


The overall rate of ‘hardcore’ smoking remained stable over the 10 years of the analyses. The results show a strong differential in ‘hardcore’ smoking based on SES, with those from low SES more likely to be ‘hardcore’ smokers in 2001 than those from a high SES, a difference which increased over the 10 years to 2010. Specifically, while the rate of ‘hardcore’ smoking in the population decreased among those from high SES, it did not decease among low SES smokers. There was a similarly strong relationship between ‘hardcore’ smoking and educational attainment, with those with higher education less likely to be ‘hardcore’ smokers than those with lower levels of education, unsurprising, given education is commonly used as an indicator of SES.18 ,36–38 This is contrary to the ‘hardcore’ hypothesis, which states that as non-‘hardcore’ smokers quit, the proportion of smokers who are ‘hardcore’ should increase, which did not occur. Rather, it appears that ‘hardcore’ smokers are able to quit, but have more difficulty when combined with low SES.

As expected, similar patterns were observed in having no plan to quit and heavy smoking, although not with past quit attempts. Having no plan to quit was significantly associated with wave, SEIFA, employment status and education. Similarly, heavy smoking was associated with SEIFA, employment status and education. Having no quit attempts in the past 12 months, however, remained stable across the four waves, and did not show significant associations with SEIFA or employment status, although it was significantly associated with education, with those with higher education more likely to have made a quit attempt in the past 12 months. This result suggests the socioeconomic drivers of ‘hardcore’ smoking are linked more strongly to quit intention and measures of nicotine dependence, but not to actual quit attempts, with low socioeconomic smokers equally likely to have made a quit attempt, but more likely to have no plan to quit, and to smoke more heavily than higher SES smokers. This is consistent with past research that has shown low SES smokers are just as likely to have made a quit attempt,27 ,28 but are less likely to plan to quit.20

The nature of the data makes it impossible to determine if changes in the rate of ‘hardcore’ smoking are due to quitting, or reductions in smoking severity (ie, ‘hardcore’ smokers may continue to smoke but no longer be classed as ‘hardcore’). Regardless, the decline in ‘hardcore’ smoking in high SES smokers is encouraging, suggesting that it is possible to target this group, but that focused work needs to be conducted to further help low SES ‘hardcore’ smokers. The lower rate of decline among those from low socioeconomic backgrounds does suggest the problem of ‘hardcore’ smoking may be due to the resources available rather than to an inherent barrier to quitting among the ‘hardcore’ smoking population. Notably, access to quitting aids, such as nicotine replacement therapy (NRT) has been lower among resource-poor, low-SES smokers. These smokers may also be less likely to hear and understand health messages to quit.39 In addition, research has demonstrated a link between financial stress and smoking,25 ,40–42 which could further contribute to differences associated with area-level SES, and also individual SES indicators such as employment status.

However, it is not known whether access to existing resources explains poorer cessation rates, or if the currently available resources are simply less effective among ‘hardcore’ smokers from low socioeconomic backgrounds. Because ‘hardcore’ smokers are, by definition, more resistant to the possibility of quitting, it is likely that existing quit programmes would have lower engagement among this group. However, it is not clear what kinds of support or services lead to ‘hardcore’ smokers being able to successfully quit. Past evidence has been inconclusive, with some evidence suggesting those from low socioeconomic backgrounds are less likely to access quit services and support, such as ‘Quitlines’ or NRT, while other research has shown relatively equal use of services, but leading to lower outcomes. Focussed research to explore the types of programmes that are successful among the ‘hardcore’ smoking population, and particularly those from more disadvantaged backgrounds, could assist. Additionally, the risk of further marginalising low SES smokers who are resistant to quitting by increasing prices of tobacco (such as via taxation increases) needs to be considered.29


The major limitation of the analyses presented in this paper is due to the nature of the questions asked in the NDSHS. Although the NDSHS questionnaire contains a number of questions about smoking and related behaviours, it does not contain a robust measure of nicotine dependence, which is arguably central to the definition of ‘hardcore’ smoking. While the number of cigarettes smoked is related to dependence, it is not a particularly reliable estimator, and thus, the estimates of the prevalence of ‘hardcore’ smoking may vary from the actual prevalence.5 A further limitation of this analysis is the lack of individual-based SES indicators in the NDSHS datasets. Because of the limited data available, only distal indicators of SES, such as education and area-level indicators, could be used. This limits the generalisability of the results.

It should also be noted that these findings are cross-sectional, and thus limited in their ability to imply causation. For similar reasons, it is not possible to examine individual outcomes. Further studies using longitudinal methods are required to address these issues. Lastly, it is also important to remember that while measures are taken to make the survey samples representative of the Australian population, it is impossible to guarantee the sample is entirely representative, particularly of minority groups and disadvantaged populations. This is compounded by a relatively low response rate of around 50% in each wave, which may have resulted in selection bias in the sample. That is, as social disapproval of smoking has increased in recent years, it may be possible that smokers are less likely to respond to surveys about their tobacco use.


The findings of this paper suggest there may be a population of smokers that is more resistant to quitting smoking, even beyond the existing difficulties faced by those from low socioeconomic backgrounds. However, this population is relatively small, and the positive outcome among higher SES smokers suggests ‘hardcore’ smokers are able to quit, and that the unfavourable outcome among those from low socioeconomic backgrounds may be overcome given the right conditions and support. Unfortunately, it is not clear exactly which factors are responsible for the more favourable outcome among those from higher socioeconomic backgrounds. More research is required to investigate what leads to positive outcomes among ‘hardcore’ smokers, and the reasons those factors are more prevalent among higher SES smokers.

What this paper adds

  • The ‘hardcore’ hypothesis is a potential barrier to continuing reductions in smoking prevalence.

  • A central axiom of the ‘hardcore’ hypothesis is temporal stability, however, there has been little research examining changes in ‘hardcore’ smoking over time.

  • This paper showed that ‘hardcore’ smoking remained stable between 2000 and 2010 in Australia, as predicted by the ‘hardcore’ hypothesis.

  • ‘Hardcore’ smoking is a real problem for low socioeconomic status (SES) smokers, however, ‘hardcore’ smoking declined among those from a higher SES, which is encouraging.

  • Work needs to be done to further investigate the mechanisms underlying socioeconomic difference in smoking, in order to more effectively target low SES smokers.


The National Drug and Alcohol Research Centre at the University of NSW is supported by funding from the Australian Government under the Substance Misuse Prevention and Service Improvements Grants Fund. We would also like to acknowledge The AIHW who conducted the NDSHS surveys on behalf of the Department of Health, and The Australian Social Science Data Archive for providing access to the NDSHS data.



  • Contributors PC, DB, RC, KM and RPM have participated in study design, analysis and interpretation of the data and preparation or approval of the manuscript.

  • Funding This work was supported by funding received by the National Health and Medical Research Council (grant number APP1021862).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement All data used in the analyses in this paper is available to researchers through The Australian Social Science Data Archive.