1. Introduction
About 20 percent of the world′s adult population smoke cigarettes [
1]. Tobacco caused 100 million deaths over the course of the 20th century [
2]. China is the largest producer and consumer of cigarettes [
3]. In 2013, China consumed more than 30 percent of the world′s cigarettes and smoking accounts for 1.36 million deaths annually there [
4]. As a leading risk factor for global disease burden, smoking consumption is of particular interest [
5].
The area-level unemployment rate reflects the macro-economic context/environment, where an individual resides, that may influence her or his smoking behavior [
6]. A number of studies, based on data from developed countries, have examined the relationship between area unemployment rate and smoking behavior. The exact nature of the relationship between area unemployment rate and smoking behavior remains unclear due to the mixed results observed in the literature review [
6]. Some studies reported that the area-level unemployment rate was associated with an increase in smoking [
7,
8,
9,
10], whereas some were null or decreased tobacco consumption [
6,
11,
12,
13,
14,
15,
16]. Charles and DeCicca found an increase in smoking among those who are least likely to be employed; however, smoking declined for those in the highest employment decile in U.S. [
17]. The spillover effects of high unemployment rates are less well understood [
18].
This study, examining the impacts of unemployment rate on smoking in China contributed to the literature in three aspects. First, this study contributed to a better understanding of the relationship in an under-studied Chinese population. No previous studies have focused on the impacts of macroeconomic environments in China [
19]. The relationship between smoking and unemployment rate may depend on whether use is examined in developing or developed countries. For example, in developed countries, smokers might consume more in a downturn. This positive relationship found for smoking does not hold for developing countries where disposable income is low [
20]. As a developing country, the Chinese macroeconomic policies and environments are drastically different from those of the U. S. and European countries where similar comparisons and implications may not appropriately be made between China and those countries. Second, we attempted to examine the effects of the area-based unemployment rate on smoking of individuals who were either employed or unemployed. Contextual effects of unemployment may exhibit different signs on alternative social groups. Previous studies have calculated an “average” effect of unemployment on smoking behaviors, research is lacking on the effects of unemployment rate on smoking of workforce and non-workforce populations, respectively [
18]. In addition, considering cigarette and alcohol are related in consumption, comparative analysis on smoking influence of unemployment rate between current drinkers and non-drinkers was carried out [
21,
22]. The cross-relation between smoking and alcohol consumption was another point that the previous literature on smoking effects of unemployment rate failed to highlight. Findings showed that the unemployment rate was positively and significantly associated with smoking in China, and the effects were stronger for those who were employment than the entire sample. Drinkers were less likely to smoke when unemployment rate was on the rise.
2. Conceptual Framework
Economic factors such as unemployment can lead individuals to change their behaviors without much attention to the potential effects on their personal health [
23]. The canonical economic model of health-related decisions derived by Grossman was applied to examine the association between the unemployment rate and smoking [
24]. Individuals would consider and make trade-offs between current benefits obtained from risky and unhealthy behaviors and future costs of such behaviors including potentially undesirable consequences related to their health. The decision to smoke can be seen as divestment in the stock of health capital in return for a short-run increase in welfare [
25].
Mounting evidence has shown substantial costs being associated with smoking in the long-run. Smokers have higher risks of developing debilitating and often serious illnesses such as cancer, cardiovascular and circulatory diseases, and chronic respiratory diseases. The chemicals in tobacco damage the lungs, blood vessels, and other tissues and keep them inflamed; disrupting the way body heals itself [
26]. In short, smoking decreases the individual′s health capital and productivity in the long-run.
However, smokers may experience short-term benefits from smoking, such as relieves in anxiety or appetite suppression [
27,
28], both of which may help people to ease stresses resulting from possible budget constraints given an environment where the unemployment rate is increasing. For Chinese people, job insecurity due to increased unemployment usually results in stress, anxiety, and psychological hardship [
29], one outlet for coping with this stress is to take up or increase smoking [
30,
31,
32]. Besides, smokers may attribute the smoking behavior to unemployment rate, which make smokers be easy to accept behavioral consequences resulting from smoking [
33].
Unemployment often provokes a shift from a preference towards the future to a preference leaning towards the present time [
34]. Many studies have found that unemployment leads to an erosion of biographical plans, an abandonment of the future and a loss of hope [
35,
36]. A short-term perspective not only increases the likelihood of seeking short-term rewards that may result in unhealthy behavior such as smoking, but also leads to underestimating the long-term consequences of the unhealthy behavior (e.g., health risks associated with these behaviors).
On the other hand, budget constraints due to increased unemployment may also have opposite effects on smoking [
37]. Because unemployment leaves individuals with less discretionary income, individuals may simply decide to abstain or reduce tobacco consumption in order to save money [
38]. For example, in China between 1990 and 2005, urban poor families used 7.7% of their reported expenditures on smoking and the rate among rural family was 11.2% [
39]. Given that smoking represents a significant share of an individual′s disposable income, higher rates of unemployment may force individuals to reduce or quit smoking.
In addition, the unemployment rate may influence smoking not only on those who have lost their jobs but also on those who remain employed [
6]. High unemployment induces a negative externality on the employed. If workers attach their future prospects to their local unemployment rate, anxiety that one might be the next on the list of fired persons will grow with rising unemployment rates.
4. Results
The average city-level unemployment rate was 3.19% with 3.4% in the wave of 2008 and 2.95% in the wave of 2012. Variations in the unemployment rate existed across cities, with the lowest being 2.50% in Wenzhou of Zhejiang province and the highest being 3.91% in Linxia of Gansu province in 2008; with the lowest being 1.63% in Hangzhou of Zhejiang, the highest being 3. 68% in Dingxi of Gansu in 2012.
Table 1 shows descriptive statistics for the variables. The average age of current smokers was 61 years. Thirty-one percent of the respondents identified themselves as smokers: 6 percent smoked fewer than 10 cigarettes per day, 8 percent smoked more than 10 but fewer than 20 cigarettes, and 17 percent smoked more than 20 cigarettes daily. The distribution of smokers in the CHARLS data was very similar to the distribution in China report on health hazards of smoking [
47].
Table 1 also shows socioeconomic and lifestyle characteristics of individuals who were employed. After excluding the unemployed, the sample size was down to 3908, among them 35 percent identified themselves as smokers: 7 percent who smoked fewer than 10 cigarettes a day, 9 percent who smoked more than 10 but fewer than 20 cigarettes, and 19 percent who smoked more than 20 cigarettes a day. The smoking prevalence was higher among those who were employed than that of the entire sample.
Table 2 shows the multivariable regression results of the relationship between the area unemployment rate and individual smoking behavior after controlling for socio-economic status, alcohol consumption and other demographic characteristics. The positive association supported the hypothesis that the unemployment rate had positive impacts on smoking behavior in China.
Table 3 presents the marginal effects of unemployment rate on smoking behavior in China. In the Logit model, the unemployment rate increased the likelihood of smoking by 10 percent. In the Ordered Logit model, where the dependent variable was an ordinal variable representing the decision to smoke or not and how much, the unemployment rate increased the likelihood of smoking by 9.1 percent. The unemployment rate increased the likelihood of smoking 1–10 cigarettes daily by 2.9 percent; the likelihood of smoking 10–20 cigarettes daily by 2.8 percent; and the likelihood of smoking more than 20 cigarettes by 3.4 percent.
Table 1.
Individual socioeconomic and life style characteristics.
Table 1.
Individual socioeconomic and life style characteristics.
Demographics | All Sample | Employed People |
---|
Mean or N | Std. Dev. or % | Mean or N | Std. Dev. or % |
---|
Age | 61.1 | 10.2 | 61.8 | 10.6 |
Male | 2247 | 49 | 2071 | 53 |
Unmarried | 779 | 17 | 704 | 18 |
Educational attainment | | | | |
Primary school and below | 3622 | 79 | 1602 | 41 |
Middle school | 550 | 12 | 1251 | 32 |
High school and above | 413 | 9 | 1055 | 27 |
Economic status | | | | |
Individual income | 12,333.93 | 16,285 | 12,490 | 15,513 |
Employed | 3908 | 85 | 3908 | 100 |
Urban | 963 | 21 | 782 | 20 |
Tobacco price | 5.54 | 2.33 | 5.59 | 2.33 |
Life style | | | | |
Current drinker | 1559 | 34 | 1211 | 31 |
Drink frequently | 734 | 16 | 625 | 16 |
Smoking status | 1421 | 31 | 1368 | 35 |
1-10 a day | 275 | 6 | 273 | 7 |
10-20 a day | 367 | 8 | 352 | 9 |
More than 20 a day | 779 | 17 | 743 | 19 |
Obs | 4585 | 3908 |
Table 2.
Effects of unemployment rate on smoking behavior.
Table 2.
Effects of unemployment rate on smoking behavior.
Variables | Smoking | Number of Cigarettes Smoked Daily |
---|
Cof. Std. Error | Cof. Std. Error |
---|
Unemployment rate | 0.66 *** | 0.23 | 0.69 *** | 0.22 |
Tobacco price | −0.30 ** | 0.14 | −0.61 *** | 0.18 |
Age | −0.58 *** | 0.31 | −1.19 *** | 0.34 |
Male | 1.61 *** | 0.35 | 4.48 *** | 0.19 |
Unmarried | 0.08 | 0.11 | 0.16 | 0.11 |
Educational attainment | | | | |
Middle school | −0.03 | 0.10 | −0.28 ** | 0.12 |
High school and above | −0.13 | 0.13 | −0.32 ** | 0.14 |
Household income | 1.13 × 10−7 | 1.86 × 10−7 | 1.47 × 10−7 | 9.90 × 10−8 |
Employed | 0.40 ** | 0.16 | 0.40 *** | 0.15 |
Living in urban area | −0.15 * | 0.08 | −0.08 | 0.14 |
Drink frequently | 0.55 *** | 0.12 | 0.37 *** | 0.12 |
Obs | 4585 | 4585 |
Prob > chi2 | p < 0.01 | p < 0.01 |
Table 3.
Marginal effects of unemployment rate on smoking behavior.
Table 3.
Marginal effects of unemployment rate on smoking behavior.
Variables | Marginal Effect |
---|
Smoking (n = 4585) | 0.10 *** |
Number of Cigarettes Smoked Daily (n = 4585) | |
Non-smoker | −0.091 *** |
1–10 a day | 0.029 *** |
10–20 a day | 0.028 *** |
More than 20 a day | 0.034 *** |
Table 4 shows effects of unemployment rate on smoking behavior of employed individuals. The area unemployment rate was positively associated with smoking behavior, consistent with results obtained from the entire sample.
Table 5 presents the marginal effects of the unemployment rate on smoking behavior among employed individuals. In the Logit model, the unemployment rate increased the likelihood of smoking by 13.6 percent. In the ordered logic model, one percent increase in the unemployment rate resulted in the increase in the likelihood of smoking about 10.8 percent comprised of compounded increases in those smoking 1–10 cigarettes per day by 3.3 percent; those smoking 11–20 cigarettes per day by 3.3 percent; and those smoking 20 cigarettes or more by 4.2 percent.
Table 4.
Effects of unemployment rate on smoking behavior of employed individuals.
Table 4.
Effects of unemployment rate on smoking behavior of employed individuals.
Variables | Smoking | Number of Cigarettes Smoked Daily |
---|
Cof. Std. Error | Cof. Std. Error |
---|
Unemployment rate | 0.74 *** | 0.22 | 0.72 *** | 0.23 |
Tobacco price | −0.12 | 0.13 | −0.6 *** | 0.18 |
Age | −0.94 *** | 0.31 | −1.18 *** | 0.35 |
Male | 1.56 *** | 0.35 | 4.40 *** | 0.18 |
Unmarried | 0.06 | 0.11 | 0.12 | 0.11 |
Educational attainment | | | | |
Middle school | −0.07 | 0.11 | −0.31 ** | 0.12 |
High school and above | −0.12 | 0.13 | −0.31 ** | 0.14 |
Household income | 9.35 × 10−8 | 1.95 × 10−7 | 1.56 × 10−7 | 9.88 × 10−7 |
Living in urban area | −0.13 | 0.08 | −0.11 | 0.14 |
Drink frequently | 0.56 *** | 0.12 | 0.37 *** | 0.12 |
Obs | 3908 | 3908 |
Prob > chi2 | p < 0.01 | p < 0.01 |
Table 5.
Marginal effects of unemployment rate on smoking behavior of employed individuals.
Table 5.
Marginal effects of unemployment rate on smoking behavior of employed individuals.
Variables | Marginal Effect |
---|
Smoking (n = 3908) | 0.136 *** |
Number of Cigarettes Smoked Daily (n = 3908) | |
Non-smoker | −0.108 *** |
1–10 a day | 0.033 *** |
10–20 a day | 0.033 *** |
More than 20 a day | 0.042 *** |
Table 6 shows effects of unemployment rate on smoking of drinkers and non-drinkers, respectively. The area unemployment rate was positively associated with smoking behavior among individuals who were either drinkers or non-drinkers. Compared to drinkers, non-drinkers were more likely to smoke.
Table 7 presents the marginal effects. In the ordered logic model, one percent increase in the unemployment rate increased the likelihood of smoking by 20.6 percent among non-drinkers and 4.6 percent among drinkers.
Table 6.
Effects of unemployment rate on smoking behavior of drinkers and non-drinkers.
Table 6.
Effects of unemployment rate on smoking behavior of drinkers and non-drinkers.
Current Drinkers |
Variables | Smoking | Number of Cigarettes (Smoked Daily) |
| Cof. | Std. Error | Cof. | Std. Error |
Unemployment rate | 0.59 *** | 0.20 | 0.52 * | 0.27 |
Obs | 1559 | 1559 |
Prob > chi2 | p < 0.01 | p < 0.01 |
Non- Drinkers |
Variables | Smoking | Number of Cigarettes (Smoked Daily) |
| Cof. | Std. Error | Cof. | Std. Error |
Unemployment rate | 0.84 * | 0.47 | 1.11 *** | 0.41 |
Obs | 3026 | 3026 |
Prob > chi2 | p < 0.01 | p < 0.01 |
Table 7.
Marginal effects of unemployment rate on smoking behavior of drinkers and non-drinkers.
Table 7.
Marginal effects of unemployment rate on smoking behavior of drinkers and non-drinkers.
Variables | Non-Drinkers (n = 3026) | Current Drinkers (n = 1559) |
---|
Smoking | 0.206 *** | 0.046 *** |
Number of Cigarettes Smoked Daily | | |
Non-smoker | −0.243 *** | −0.048 *** |
1–10 a day | 0.049 ** | 0.017 *** |
10–20 a day | 0.081 *** | 0.014 * |
More than 20 a day | 0.113 ** | 0.017 * |
Table 8 shows effects of unemployment rate on smoking by 2SLS. Since Hausman test statistics were not significant, we cannot reject the hypothesis that unemployment rate exogenous when dependent variable was smoking status. However, Hausman test indicated that unemployment rate was endogenous in the regression of number of cigarette smoked daily (Hausman test statistics are available at request). When estimated with 2SLS, the effects of unemployment rate on smoking were consistent. However, these effects were weaker after correcting for endogenous.
Table 8.
Effects of unemployment rate on smoking behavior by 2SLS.
Table 8.
Effects of unemployment rate on smoking behavior by 2SLS.
Variables | Smoking | Number of Cigarettes Smoked Daily |
---|
| Cof. Std. Error | Cof. Std. Error |
---|
Unemployment rate | 0.03 | 0.07 | 0.32 * | 0.17 |
Tobacco price | −0.04 * | 0.02 | −0.2 *** | 0.05 |
Age | −0.16 *** | 0.04 | −0.45 *** | 0.1 |
Male | 0.28 *** | 0.01 | 1.43 *** | 0.03 |
Unmarried | 0.02 | 0.02 | 0.06 | 0.04 |
Educational attainment | | | | |
Middle school | −0.006 | 0.02 | −0.10 ** | 0.04 |
High school and above | −0.02 | 0.02 | −0.13 ** | 0.05 |
Household income | 2.33 × 10−8 | 3.81 × 10−7 | 9.37 × 10−8 | 8.54 × 10−8 |
Living in urban area | −0.02 | 0.01 | −0.04 | 0.03 |
Drink frequently | 0.12 *** | 0.02 | 0.24 *** | 0.04 |
Obs | 4585 | 4585 |
Prob > chi2 | p < 0.01 | p < 0.01 |
5. Discussion
A strong positive association of the area-level unemployment rate with individual smoking was observed in China, even after controlling for demographic, social economic status, and alcohol consumption. In China, the increased unemployment rate provoked a shift from a preference towards the future to a preference leaning towards the present time, which resulted in increased smoking. Changes in smoking behavior resulting from financial burdens were less substantial. The growth of consumer incomes in China has been much faster than the increase in consumer expenditures on cigarettes in recent years. From 1990 to 2005, the former, on average, was about 2.8 times as much as the latter. The ability to pay for cigarettes has doubled as well [
39].
According to most studies, smoking patterns appear to be procyclical in developed countries [
6,
11,
12,
13,
14,
15,
16]. A decrease in smoking was followed when the unemployment rate increases. In contrast, a countercyclical trend was observed in China, which was consistent with those findings among individuals with lower social-economic status (e.g., construction workers and African Americans), old people or women in developed countries [
7,
8,
9,
10]. Combing our findings with existing literature, the positive association may be explained by the social supports that individuals enjoy [
48]. Since the social safety net in China was not yet well developed, fluctuations in the unemployment rate tended to easily affect individual physical living conditions and increased the psychological stress [
49]. It was possible that, without the protection from a sophisticated social welfare safety net, the high unemployment rate created anxiety for individuals, and as a result, trigged the increase in their smoking consumption. According to the economics theory, this was the case that individuals changed their time preference toward the present
versus the future and coped the stress along with rising unemployment rate with smoking [
24].
The effects of area-level unemployment rate on smoking behavior were stronger for those who were employed than the entire population, which is consistent with the literature that observes the negative influence of individual unemployment status on smoking behavior and that employed respondents were more likely to smoke than unemployed respondents with budget constraints [
6,
7].
Relative to their drinking counterparts, non-drinkers were more likely to smoke following the increased unemployment rate. Smoking and drinking were interdependent. When smokers also drank alcohol, they were likely to deal with the stress resulting from unemployment rate by increasing their alcohol consumption, whereas, given that drinking cannot be used to alleviate the pressure for smokers who do not drink, it is very plausible to expect that these smokers would increase their tobacco consumption more than their drinking counterparts [
21,
22].
Hausman test assisted that unemployment rate was not an endogenous variable when the dependent variable was smoking status, but the endogeneity problem existed between unemployment rate and number of cigarette smoked daily. Smoking behavior does not hurt labor market performance until the number of cigarette smoked daily reaches a certain limit [
50]. In order to correct for the endogeneity of unemployment rate, we employed an instrumental variable (area-level unemployed insurance gold) and two-stage least squares (2SLS) estimation. A number of tests confirmed that the instrument was valid, and area-level unemployed insurance gold was negatively and significantly associated with unemployment rate. The results of the two-stage least squares (2SLS) estimation indicated the effects of unemployment rate on number of cigarette smoked daily were positive, but smaller than the results without correcting for endogeneity. A possible explanation for the difference between the endogeneity-corrected and uncorrected results was that individuals might get smoking earn penalty in labor markets [
44,
45].
Our finding highlights the need for multi-level approaches aimed at smoking reduction and cessation. Policy interventions for smoking-control should not only focus on the individual characteristics but also take the macro-economic environment into consideration. One intervention option is to implement different tobacco control policies based on the economic situations in areas or regions. The findings of positive impact of the area-level unemployment rate on smoking from our study indicate potentially effective policy strategies consisting of improving the social welfare safety net and increasing the cigarette tax in China, especially during economic recessions. The government should increase recourse allocation in health to strengthen the social welfare safety net to help ease economic hardship and mental stress. To achieve this, one option is to increase cigarette tax to make up for social welfare expenses and meantime adjust cigarette demand [
51]. Other policies may include offering more smoking reduction and cessation services during economic recessions and strengthening anti-smoking education campaigns, especially during economic booms.
Our findings also have global public health implications. The vast majority of the world′s smokers live in low- and middle-income countries, which results in nearly 80 percent of deaths in middle- and low- income countries [
1]. At present, global health is affected by economic fluctuation. Compared to developed countries, the social welfare safety net in developing countries is weaker [
49]. Smoking behavior in developing countries is more easily affected by economic downturn. The gap between the numbers of smokers in developing countries
versus developed countries is expected to widen [
1]. As a result, special attention should be paid to smoking control in developing countries, especially during the economic recessions.
Our study has some limitations. The data sample only included those who were 45 years old or older; the findings may be generalized to the entire population with caution. Cigarette consumption may be higher among older populations who may not see the time preference of the future to the present as significant as younger people do. Young people may decrease their consumption for this reason. In addition, given their relatively low earnings, younger people may decrease their consumption simply because of the increase in cost of cigarettes. On the other hand, young people may not see the urgency of trading current smoking for the future healthy life because they are still healthy at a young age or their health stocks are high. There could be some wash-out effects between the two possibilities. In fact, we did run the analysis by dividing the sample into the working age group and retired group (the legal retirement age for women is 50 and the retirement age for men is 60) and did not find smoking behavior was significantly different between young people and old people. Further, the positive association of smoking and unemployment rate among retired people was comparable to that among working age people (in view of the article length, such results are not show in the article). Therefore, we, to some extent, believe that our findings are quite generalizable.