Economic crisis and smoking behaviour: prospective cohort study in Iceland.

OBJECTIVE: To examine the associations between the 2008 economic collapse in Iceland and smoking behaviour at the national and individual levels.
DESIGN: A population-based, prospective cohort study based on a mail survey (Health and Wellbeing in Iceland) assessed in 2007 and 2009.
SETTING: National mail survey. PARTICIPANTS: Representative cohort (n=3755) of Icelandic adults. MAIN OUTCOME MEASURE: Smoking status.
RESULTS: A significant reduction in the prevalence of smoking was observed from 2007 (pre-economic collapse) to 2009 (postcollapse) in both males (17.4-14.8%; p 0.01) and females (20.0-17.5%; p 0.01) in the cohort (n=3755). At the individual level of analysis, male former smokers experiencing a reduction in income during the same period were less likely to relapse (OR 0.37; 95% CI 0.16 to 0.85). Female smokers were less likely to quit over time compared to males (OR 0.65; 95% CI 0.45 to 0.93). Among male former smokers who experienced an increase in income between 2007 and 2009, we observed an elevated risk of smoking relapse (OR 4.02; 95% CI 1.15 to 14.00).
CONCLUSIONS: The national prevalence of smoking in Iceland declined following the 2008 economic crisis. This could be due to the procyclical relationship between macro-economic conditions and smoking behaviour (ie, hard times lead to less smoking because of lower affordability), or it may simply reflect a continuation of trends already in place prior to the crisis. In individual-level analysis, we find that former smokers who experienced a decline in income were less likely to relapse; and conversely, an increase in income raises the risk. However, caution is warranted since these findings are based on small numbers.

An examination on the association between economic crises and smoking behaviours, that is, is change in income (at both the national and individual levels) related to a change in smoking status?

National smoking prevalence declined in Iceland following the 2008 economic crisis.

In individual-level analysis, male former smokers whose incomes declined experienced a reduced risk of smoking relapse.

Conversely, an increase in income from 2007 to 2009 was associated with increased risk of relapse.

Our findings are consistent with the hypothesis that economic downturns may result in decreased tobacco use (procyclical effect).

A representative prospective cohort study assessed at two time points, which straddle the start of a severe economic crisis.

Owing to the low number of individuals that change their smoking behaviours in a short period, we were unable to assess the effects of a change in employment on smoking habits.

Findings are based on a low number of subjects and must be taken with caution.

Introduction

The Icelandic economy was severely affected by the global economic collapse of 2008. After a decade-long period of financial prosperity the nation was plunged into a recession of great severity, resulting in a severe currency crisis, as well as a drastic increase in national and household debts, runaway unemployment rates and decreased per capita purchasing power.1 2

Previous research on the health consequences of the Icelandic economic collapse has suggested adverse impacts on cardiovascular and mental health among women.3 4 In the broader literature on economic crises and population health, however, it has been debated whether health moves in a procyclical or counter-cyclical direction to macroeconomic conditions. The work of Brenner5 beginning in the 1970s suggested that mortality is counter-cyclical, that is, when the economy is down, death rates—in particular, suicides—rise. However, in more recent years, a series of econometric studies have suggested that mortality is procyclical, that is, during economic contractions death rates decline.6–9 There are plausible reasons for this unexpected finding—for instance, during the 1998 Korean financial crisis, economic activity was so depressed that there was a detectable decline in traffic-related mortality.10 Others have speculated—without direct evidence—that people are more likely to be over-worked and ‘stressed’ during economic booms than during busts, having less time flexibility to engage in health promoting behaviours.11 12

Few studies, however, have used individual-level data to test the association between recession and health, especially smoking. Most of the evidence to date has been at the ecological level, though not all.13 For instance, Shaw et al14 found a direct association between economic hardship and a propensity to smoke. Using US data, Ruhm12 previously reported that economic recession was associated with a decline in the prevalence of cigarette smoking. A recent report from Gallus et al15 found that the recent economic contraction in Italy has given rise to an increase in the percentage of current smokers—primarily for women. In the present study, we took advantage of the natural experiment afforded by the Icelandic crisis to examine the relationship between changes in economic conditions and smoking behaviour. Utilising a prospective cohort of Icelandic adults assessed before (in 2007) and after the start of the collapse (in 2009), we sought to examine the risk of relapse among precollapse former smokers, as well as quitting behaviour among current smokers in terms of economic changes. Furthermore, because of the important role of perceived stress on smoking status, we sought to examine the potential influence of stress on the studied associations.16 17

Methods

Design and samples

Cohort

Our cohort is based on the Health and Wellbeing in Iceland health survey. Data were collected by a questionnaire in two waves: (1) from October to December of 2007 (10–12 months precollapse), then again (2) between November and December of 2009 (13–14 months postcollapse). The cohort was based on a stratified random sample of the Icelandic population (n=9807), which was selected from 12 strata: consisting of two geographic regions further stratified by six age groups. Of the initial 9807, a total of 5918 responded to the initial 2007 assessment (response rate of 60.3%), with 4092 responding again to the modified version of the survey in 2007 (response rate of 82.8% of those who responded to the precollapse baseline survey). Because of the importance of stress as a potential predictor of smoking behaviour, we excluded individuals who did not have complete responses to the Perceived Stress Scale (PSS-4) in both 2007 and 2009. This left a final analytical sample of n=3755. Figure 1 shows the cohort attrition over questionnaire waves.

Figure 1

The cohort of the ‘Health and well-being’ study.

Measures

Smoking status and behaviour

In the questionnaire, we inquired about smoking status, that is, whether respondents were current smokers, had quit smoking or had never smoked. In order to examine the likelihood of relapsing or quitting following an economic collapse, respondents were stratified according to their smoking status: non-smoker, relapsed and quit smoking.

Non-smoker: an individual was classified as a non-smoker if they responded that they did not currently smoke on both the 2007 and 2009 assessments.

Relapsed smoker: an individual was identified as relapsed if they indicated that they (a) were a former smoker on the 2007 questionnaire, but indicated they had (b) smoked in any frequency in 2009. In our analyses estimating the ORs of relapse, the base population was restricted to individuals who were former smokers at baseline.

Quit smoking: a respondent who had quit smoking must have indicated that they were (a) currently smoking in 2007, yet had (b) quit smoking by 2009. In our analyses estimating the ORs of quitting, our base population was restricted to individuals who were current smokers at baseline.

Change in economic status

Additional socio-economic questions pertained to employment and income status. Household income was classified into income ranges of (in terms of Icelandic currency; ISK) (1) low (≤3.4 million ISK), (2) middle (3.5–9.4 million ISK) and (3) high (≥9.5 million ISK); corresponding approximately to (1) ≤28 000 US$, (2) 28 000–77 000 US$ and (3) ≥77 000 US$. For analysis of income change, household income was further dichotomised into either high or ‘low’ (which combined the middle-income and low-income categories). We examined two types of income change: (a) drop in income between 2007 and 2009 from high to low and (b) a rise in income between 2007 and 2009 from low to high.

Change in perceived stress

Psychological stress was measured in both 2007 and 2009 using the four-item PSS-4.18 The PSS-4 is a shortened, validated and acceptable substitute for the original scale,19 with scores ranging from 0 to 16; the higher the score, the higher the perceived stress. An increase in stress was classified as any increase from baseline to follow-up; conversely, a decrease was classified as any decrease from baseline to follow-up. For example, an individual with a score of 5 in 2007 and a score of 10 in 2009 would be classified as having an increase in stress.

Explanatory variables and demographics

Our regression models controlled for the following socio-demographic covariates: age, sex, marital status and education. Education was categorised as (1) basic (completed primary school or less), (2) middle (completed high school or equivalent) and (3) university (a completed university degree). Employment status was categorised as (1) employed, (2) unemployed, (3) student (4) homemaker/paternal leave, (5) retired and (6) disabled.

Statistical analyses

Table 1 presents the distribution of socio-demographic characteristics according to change in smoking status between 2007 and 2009.

Table 1

Baseline characteristics (in 2007) of the cohort and among differential smoking status

	Cohort	Relapsed in 2009	Quit smoking in 2009
n	3755	56	160
Age mean±SD	52.3±16.0	45.7±14.2	47.4±15.5
Sex	n (% of category)
Male	1763 (47.0)	31 (55.4)	82 (51.3)
Female	1992 (53.0)	25 (44.6)	78 (48.8)
Marital status
Single/divorced	556 (14.9)	7 (13.0)	31 (19.9)
Committed, not cohabiting	131 (3.5)	2 (3.7)	9 (5.8)
Married, cohabiting	2871 (76.7)	45 (83.3)	116 (74.4)
Education
Basic	1688 (47.1)	22 (40.7)	65 (42.5)
Middle	971 (27.1)	15 (27.8)	51 (33.3)
University	928 (25.9)	17 (31.5)	37 (24.2)
Employment status
Employed	2019 (58.4)	37 (71.2)	98 (64.5)
Unemployed	169 (4.9)	3 (5.8)	10 (6.6)
Student	122 (3.5)	1 (1.9)	5 (3.3)
Homemaker/paternal leave	159 (4.6)	2 (3.8)	9 (5.9)
Retired	872 (25.2)	4 (7.7)	24 (15.8)
Disabled	119 (3.4)	5 (9.6)	6 (3.9)
Household income
Low	621 (20.6)	8 (17.0)	22 (16.5)
Middle	1855 (61.4)	25 (53.2)	80 (60.2)
High	543 (18.0)	14 (29.8)	31 (23.3)

Binary logistic regression was used to estimate OR (corresponding 95% CIs) of relapse in 2009 (table 2), and the odds of quitting smoking in 2009 (table 3) by background characteristics, change in income and stress levels. Analyses were also stratified by gender. Models were adjusted for age and sex; models for household income and income change were additionally adjusted for baseline income levels. As previous research supports the role of stress as a mediator of an individual's propensity to change smoking status,15 16 20 we also ran models of relapse and cessation with and without the inclusion of (1) changes in stress levels between 2007 and 2009 and (2) baseline stress levels.

Table 2

The OR of relapsing in 2009 among those who had quit smoking at the baseline (2007)

	OR (95% CI)*
			Male	Female
2009 status	Overall		Ref	0.67 (0.38 to 1.18)
Household income in 2009†	n‡
Low	8	0.66 (0.26 to 1.70)	1.13 (0.24 to 5.36)	0.56 (0.15 to 2.08)
Middle	25	1.57 (0.48 to 5.17)	2.28 (0.38 to 13.55)	1.31 (0.21 to 8.32)
High	17	Ref	Ref	Ref
Household income in 2009 (among high income at baseline)†
High income in 2009	5	Ref	Ref	Ref
Lower income in 2009	9	0.53 (0.28 to 1.01)	0.37 (0.16 to 0.85)	0.92 (0.29 to 2.88)
Household income in 2009 (among low incomes at baseline)†
High income in 2009	23	3.14 (1.27 to 7.72)	4.02 (1.15 to 14.00)	2.43 (0.64 to 9.19)
Lower income in 2009	7	Ref	Ref	Ref
Change in stress from 2007 to 2009§
Same	7	Ref	Ref	Ref
Decreased	15	0.91 (0.35to 2.36)	0.83 (0.23 to 2.99)	1.03 (0.25 to 4.28)
Increased	34	1.71 (0.86 to 3.37)	1.75 (0.68 to 4.53)	1.64 (0.61 to 4.39)

*OR adjusted for statuses in 2009: age and sex.

†OR adjusted for statuses in 2009: age, sex and baseline income in 2007.

‡Totals do not include missing values from 2009.

§OR additionally adjusted for baseline stress (2007).

Table 3

The OR of smoking cessation in 2009 among those who were smokers at the baseline (2007)

	OR (95% CI)*
			Male	Female
2009 status	Overall		Ref	0.65 (0.45 to  0.93)
Household income in 2009†	n‡
Low	22	0.89 (0.49 to 1.60)	0.75 (0.33 to 1.74)	1.01 (0.43 to 2.36)
Middle	80	0.98 (0.45 to 2.13)	0.80 (0.27 to 2.38)	1.12 (0.36 to 3.46)
High	31	Ref	Ref	Ref
Household income in 2009 (among high income at baseline)†
High income in 2009	19	Ref	Ref	Ref
Lower income in 2009	6	0.75 (0.46 to 1.22)	0.82 (0.41to 1.62)	0.68 (0.34 to 1.37)
Household income in 2009 (among low incomes at baseline)†
High income in 2009	85	0.68 (0.30 to 1.55)	0.61 (0.19 to 1.97)	0.77 (0.24 to 2.41)
Lower income in 2009	8	Ref	Ref	Ref
Change in stress from 2007 to 2009§
Same	22	Ref	Ref	Ref
Decreased	62	0.84 (0.47 to1.48)	0.73 (0.34 to 1.56)	0.98 (0.41 to 2.31)
Increased	76	0.98 (0.64 to 1.51)	0.66 (0.36 to 1.22)	1.38 (0.74 to 2.58)

*OR adjusted for statuses in 2009: age and sex.

† OR adjusted for statuses in 2009: age, sex and baseline income in 2007.

‡Totals do not include missing values from 2009.

§OR additionally adjusted for baseline stress (2007).

Repeated measures analysis of variance (p values, F statistic) was used to examine overall and gender-specific mean differences in stress levels from 2007 to 2009 (table 4). Statistical analyses were conducted with IBM SPSS Statistics V.19.0 (SPSS Inc, Chicago, Illinois, USA). Statistical significance was set at the 0.05 level, and all tests were two-tailed.

Table 4

Average stress levels according to smoking status—among waves (2007 and 2009)

	2007	2009
	Stress mean (SD)	Stress mean (SD)	p Value (F)*
Never smoker
Male	3.70 (2.75)	3.83 (2.69)	0.31 (1.02)
Female	4.18 (2.70)	4.40 (2.90)	0.44 (0.60)
Relapsed
Male	3.52 (2.28)	4.94 (2.80)	0.28 (1.20)
Female	3.96 (2.52)	5.24 (3.46)	0.01 (7.67)
Quit smoking
Male	4.21 (2.71)	4.16 (2.78)	0.91 (0.01)
Female	4.38 (3.49)	5.03 (3.35)	0.13 (2.31)

*Repeated measures analysis of variance (p values, F statistic) used to examine overall and gender-specific mean differences in stress levels from 2007 to 2009; adjusted for age in 2009.

Results

Baseline characteristics

Table 1 describes the baseline characteristics of the cohort in 2007 (n=3755), which was 53.0% female, 76.7% married/cohabiting and with a mean (SD) age of 52.3 (16.0). Table 1 also describes the characteristics of those that had relapsed and quit: 72.2% (n=2711) of the cohort were non-smokers, 4.0% (n=56) of the former smokers at baseline had relapsed in 2009 and 22.2% (n=149) of smokers at baseline had quit smoking in 2009. A significant reduction (p<0.01) in the prevalence of smokers was observed from 2007 to 2009 in both males (17.4–14.8%) and females (20.0–17.5%).

Relapse smoking

Among individuals who were former smokers at baseline (table 2), decreased odds of relapsing in 2009 (after the collapse) were observed in the older age groups (compared to those aged 18–39), regardless of gender (age of 40–59: OR 0.38; 95% CI 0.21, 0.69 | age ≥60: 0.10; 0.04, 0.23).

While an individual's employment status was not involved in their risk of relapsing, retired women showed a significant increased risk of relapsing (4.12; 1.11, 15.29), compared to the employed.

Among men in the lower-income groups at baseline (ie, low and middle), those who moved into the high-income group in 2009 experienced an increased risk of relapse (4.02; 1.15, 14.00)—while among those in the high-income group at baseline, those whose incomes dropped had a decreased risk of relapsing (0.37; 0.16, 0.85). Further adjustments for a change in stress levels from 2007 to 2009, showed limited attenuation in the coefficients, suggesting some mediation by perceived stress—that is, former smokers whose incomes increased between 2007 and 2009 may have relapsed in part because of an increase in stress.

Smoking cessation

Women were less likely to quit smoking in 2009 (0.65; 0.45, 0.93), compared to men. An increased likelihood of quitting in 2009 was observed among the following female groups: those with middle (2.78; 1.48, 5.21) or university-level (2.73; 1.38, 5.40) education compared to a basic, and the disabled (3.42; 1.23, 9.52) compared to the employed. Compared to women aged 18–29, those in the middle-aged group (0.46; 0.26, 0.83) were less likely to quit. Additional adjustments for a change in stress levels from baseline to follow-up in the cessation models revealed no diminished significance in effect sizes.

Stress and smoking

Though stress change (increase versus Stable and decrease versus stable) did not predict a relapse in women in aforementioned analyses, further examination of changes in stress levels among smoking status displayed a significant change in mean stress levels (SD) among women that had relapsed, with a significant increase in stress scores from 3.96 (2.52) in 2007 to 5.24 (3.46) (p 0.01; F=7.67).

Discussion

In response to the severe economic collapse in Iceland, we found that the prevalence of smoking continued to decrease for both genders in the short period after. This drop in smoking may be attributed to background secular trends,21 while other factors, such as changes in the price of cigarettes, and changing norms about the acceptability of smoking, may also have played a role. The strength of our study is that we were able to document changes in individual economic status straddling the economic downturn and link these exposures to individual changes in smoking habits. Additionally, in comparison to national smoking rates (2007: 23.0% of population; 2009: 19.0%) the prevalence rates from 2007 to 2009 of this sample are relatively analogous—offering support for the generalisability of the sample.

Our findings partially corroborate previous research on the procyclical nature of the association between economic downturns and smoking habit, that is, during recessions, smoking habits may be dampened. Among male former smokers, those who experienced a decline in income during the economic recession had a significantly lower risk of relapse two years later. Conversely, among men whose incomes increased during the period of recession, their risk of relapse was considerably higher compared to those whose incomes stayed the same. Although the direction of associations was similar among women, none of the estimates were statistically significant.

Taken together, the main significant finding of our analyses is that male former smokers whose incomes fell during the period of the economic collapse experienced a reduced risk of relapse. Ruhm22 hypothesised that this risk reduction is possibly driven by a tendency to adopt healthier behaviours during periods of reduced income—driven by an increase in positive health behaviours (ie, exercise) that accompanies newly acquired increased leisure time during economic contractions. It could also be argued their behaviour change in a recession can be either intentional or inadvertent. When facing enforced economic inactivity—individuals may choose to fill their time by actively investing in positive personal health changes, which include stopping smoking or joining a fitness club. However, our results did not indicate an increased risk of quitting among those whose incomes fell—which is inconsistent with previous research by Siahpush and Carlin.23

It is possible that smoking cessation and smoking relapse are ‘asymmetric’ behaviours with different triggers. Thus, a former smoker who experiences a drop in income may be less tempted to start smoking again because of the reduced affordability of cigarettes. However, someone who is already smoking may be less sensitive to an income drop (higher income inelasticity)—that is, he is unable to quit his ongoing behavior because of the offsetting increase in stress (although our data on self-reported stress did not support this).

There is an apparent discrepancy between the national decline in smoking in Iceland and the fact that smokers whose incomes declined were not more likely to quit. This underscores the point that macrolevel data and individual-level patterns are often driven by a different set of causes. Thus, the overall decline in national smoking rates could be either due to the procyclical nature of smoking (ie, recessions are good for health), or it may simply reflect a continuation of trends already in place prior to the recession (ie, national antismoking campaigns, declining social acceptability of smoking, etc). In other words, national averages are driven by more than the group of smokers whose incomes decreased after the crisis.

Furthermore, we caution that our findings regarding recession, income change and smoking habits cannot be generalised to other health outcomes. For example, observational reports found a spike in female cardiac emergency visits during the week corresponding to the economic collapse in October 2008.3 In accordance with this, our previous analysis on changes in mental health revealed significant increases in stress for mainly women.4 This increase in stress for women, however, threatening to related health outcomes, did not prove to be associated with an increased likelihood of relapsing.

Our findings are also congruent with multiple models explaining the link between stress levels and smoking behaviour. Though much research shows stress as a cause of smoking,15 16 additional research actually points to cigarette smoking as a cause of stress and, furthermore, smoking cessation as leading to a reduction in stress.20 This is in line with our findings, as both male and female relapsed smokers had the lowest levels of stress before the collapse when they considered themselves as having quit smoking in 2007 (table 4), yet experienced an increase in stress postcollapse—significantly for women. This may also point to a vulnerability of this group to use smoking as a means of alleviating stress—explaining their relapse in smoking after the collapse.24 This vulnerability has been discussed and supported by previous research showing economic stress as a cause of adverse mental health.25 This increased stress may have also been amplified by a return to smoking, as Cohen and Lichtenstein26 have found. Caution is warranted in interpreting the findings on stress, however, since smokers may be citing an increase in perceived stress to justify their relapse or failure to quit. We cannot conclusively argue that stress did not play a mediating role in the association between income change and smoking behaviour because of measurement error.

Study limitations

Some limitations of our study should be noted. Relapsed smokers and quitters represent a small proportion of the population, and hence our ORs were estimated with imprecision and must be interpreted with caution. Similarly, we lacked statistical power to directly examine the effects of a change in employment status on change in smoking habits. In other words, though we were able to examine the effects of income change, we were not able to directly estimate the effects of unemployment as there were too few individuals in the sample who lost their jobs between 2007 and 2009. While our findings are based on the potential effects of an economic crisis on a change in smoking status, it is not clear whether these similar findings would hold true in normal scenarios and, thus, caution is warranted when generalising our findings to other normative scenarios. Finally, smoking status was based on self-report only, and not validated by biomarkers such as cotinine. This may have produced misclassification of the outcome, though it is not clear whether this misclassification was differential by exposure status (eg, income changes).

Conclusions

Our large population-based cohort with assessment points straddling the 2008 economic crisis in Iceland revealed a reduction in smoking rates from the short periods before and after the start of the crisis—though our study could not disentangle the direct effects of the crisis with other mechanisms, for example, secular trends and changing cigarette prices. Chiefly, this examination revealed a role of income change on the risk of relapse after the collapse among former male smokers.