Objectives To estimate the impact of tobacco cessation on household spending on non-tobacco goods in the USA.
Methods Using 2006–2015 Consumer Expenditure Survey data, 9130 tobacco-consuming households were followed for four quarters. Households were categorised during the fourth quarter as having: (1) recent tobacco cessation, (2) long-term cessation, (3) relapsed cessation or (4) no cessation. Generalised linear models were used to compare fourth quarter expenditures on alcohol, food at home, food away from home, housing, healthcare, transportation, entertainment and other goods between the no-cessation households and those with recent, long-term or relapsed cessation. The full sample was analysed, and then analysed by income quartile.
Results In the full sample, households with long-term and recent cessation had lower spending on alcohol, food, entertainment and transportation (p<0.001). Recent cessation was further associated with reduced spending on food at home (p<0.001), whereas relapsed cessation was associated with higher spending on healthcare and food away from home (p<0.001). In the highest income quartile, long-term and recent cessations were associated with reduced alcohol spending only (p<0.001), whereas in the lowest income quartile, long-term and recent cessations were associated with lower spending on alcohol, food at home, transportation and entertainment (p<0.001).
Conclusions Households that quit tobacco spend less in areas that enable or complement their tobacco cessation, most of which may be motivated by financial strain. The most robust association between tobacco cessation and spending was the significantly lower spending on alcohol.
- Public policy
Statistics from Altmetric.com
Approximately 15% of adults in the USA smoke cigarettes and 21% use any form of tobacco (eg, cigars and waterpipe).1–4 The financial costs of tobacco use are often thought to be restricted to its indirect effects on increasing healthcare utilisation and decreasing productivity,2 but spending on tobacco may generate immediate opportunity costs through the use of funds that could otherwise be spent on other goods and services.5–8 In the USA, 7% of all smokers (12% of those with low income) report smoking-induced deprivation (or the inability to purchase household essentials because money was spent on cigarettes).9 10 Outside the USA, tobacco expenditures are associated with lower spending on healthy food, housing, education, transportation and healthcare.5 7 11 12 Within the USA, spending on tobacco is associated with reduced spending on housing and clothing,13–15 and smoking households are more likely to experience food or housing insecurity.14 16 17
Tobacco cessation may ameliorate household financial insecurity by increasing discretionary income to spend on household essentials, but there are literature gaps regarding the relationship between tobacco cessation and household spending on other goods. In particular, no longitudinal studies have examined changes in expenditure patterns following the cessation of tobacco use. Therefore, it is currently unclear whether the spending patterns found in the existing cross-sectional research reflect tobacco-induced expenditure restriction that would be eliminated with the cessation of tobacco use or whether these patterns reflect the preferences of smokers that would remain in absence of tobacco use.
Additionally, the association between tobacco cessation and spending may extend beyond a monetary relationship predicted by an increase in discretionary income. Tobacco use often occurs and changes with other health-related behaviours, including alcohol and food intake. In the USA, smokers are four times more likely to be dependent on alcohol, and this relationship has neurobiological, social and environmental mechanisms.18–22 Conversely, prior research has identified a negative relationship between smoking and body weight, the mechanisms of which include metabolic effects of nicotine and differences in eating patterns (eg, short-term increases in eating after cessation) and dietary composition between smokers and non-smokers.23 24 Thus, quitting smoking may affect the consumption of alcohol and food independent of what one would predict from the change in discretionary income that occurs after tobacco cessation.23 25–30
In summary, quitting smoking may impact household finances and other health behaviours, but there have been no longitudinal studies examining change in expenditures following cessation. The current study was designed to estimate the relationships between long-term, short-term and relapsed tobacco cessations on household dollars spent on alcohol, food, housing, transportation, healthcare, entertainment and other goods among US households.
The study was guided by the theory of rational addiction,31 which assumes that consumers seek to maximise their utility over the lifetime, where utility is defined as a function of the current consumption of an addictive good (eg, tobacco), the past consumption of the addictive good and the current consumption of all other goods. The theory further assumes that consumers are forward looking and spend within a budget constraint, where the primary determinants of the consumption of an addictive good are its price (monetary, health and social), past consumption (addiction), anticipated future consumption, income, consumption of all other goods and preferences.
The theory leads to several predictions relevant to the current analysis. First, when the consumption of an addictive good such as tobacco ends, a household will experience an increase in discretionary income and a decrease in utility. The increase in income eases the budget constraint and allows individuals to increase consumption of other goods. The utility deficit motivates individuals to increase consumption of substitute goods. Second, the higher the level of past consumption, the greater the utility deficit on cessation and the greater the motivation to consume substitute goods and decrease consumption of complementary goods. Third, as individuals sustain abstinence from an addictive good over time, their level of addiction (past consumption) declines and their motivation to consume substitute goods declines while their discretionary income for other goods increases.
Guided by this theory and prior literature, we predicted that recent cessation would be associated with lower spending on alcohol (a tobacco complement) but higher spending on food (a tobacco substitute, especially during nicotine withdrawals). We predicted that the increase in food expenditures would be pronounced in low-income households, who may be experiencing food insecurity during tobacco use.15 We further predicted that long-term cessation (after nicotine withdrawal subsides) would be associated with increased spending on additional types of goods, especially among very low-income households that may be experiencing smoking-induced deprivation. Lastly, we studied relapsed households as a unique group that were motivated to quit tobacco but failed. Their motivation (but failure) to quit provides insight into how spending changes may contribute to relapse and identifies spending changes that could not be due to the increase in discretionary income that comes from reducing tobacco spending.
Study design and data sources
The study used a cohort design, using de-identified data from the quarterly interview survey of the Consumer Expenditure Survey (CES) conducted by the US Bureau of Labor Statistics (BLS) and the US Census Bureau.32 CES participants are recruited monthly throughout the year and complete five in-person interviews over four quarters (12 months). A household respondent completes the five interviews approximately 3 months apart. Interview 1 captures descriptive information about the household and a household reference person. Interviews 2 to 5 update household characteristics data and capture detailed expenditures in the prior 3 months. Households are not reimbursed or incentivised for participating. For the current study, data from 2006 to 2015 were combined to achieve sufficient sample size. CES response rates ranged from 70% to 76% during these years.33 Data from the BLS’s Local Area Unemployment Statistics programme served as a control for state-level quarterly unemployment rates.34
The study’s cohort comprised households that reported any tobacco spending during the first study quarter. Households were excluded for having missing data, negative income or negative spending in any category. There were no significant differences between smoking households that were included or excluded from the current study in the average age, gender or race of the household reference person or in the average household size or income (p>0.05). The final sample comprised 9130 households with complete data.
Primary independent variable
The primary independent variable classified the presence and length of tobacco cessation in each household based on their tobacco spending during quarters 2, 3 and 4. Households were categorised as: (1) recent cessation, (2) long-term cessation, (3) relapsed cessation or (4) no cessation. No cessation was categorised for households with positive tobacco spending during all four quarters. Recent cessation was categorised for households that had positive tobacco spending during the first three quarters of study participation and zero spending on tobacco during the fourth quarter (ie, 3 months of cessation).35 Long-term cessation was categorised for households with zero tobacco spending in the second and/or third quarter through the fourth quarter (ie, 6–9 months of cessation).35 Relapse was categorised for households with zero tobacco spending during the second and/or third quarters with a resumption of tobacco spending in the fourth quarter.
Dependent variables included fourth quarter dollars spent on alcohol, food (at home and away), housing, transportation, healthcare, entertainment and other goods. Because we combined data from 2006 to 2015, we used the US BLS’s all-items urban consumers Consumer Price Index to adjust expenditures to 2015 dollars prior to analysis.36 In addition, the raw expenditure data were right skewed, so inverse hyperbolic sine (IHS) transformation was conducted.37 Following transformation, skewness was between −0.5 and 0.5 for all variables, except for health (skewness=−0.62) and other (skewness=−0.82). For three expenditure categories in which greater than 15% of the sample had zero spending (alcohol, food away from home and healthcare), we also examined the proportion of households with any spending within the category during the fourth quarter.
Other explanatory variables
Additional variables were included in all models that may explain fourth-quarter spending, including: first quarter dollars spent, characteristics of the household reference person (age,38 39 gender,38 40–42 race,43 44 marital status45) and characteristics of the household (income,42 size, state of residence and state-level unemployment rate). Survey year and month were included in the models to control for annual and seasonal variations in tobacco use and spending. We included confounders such as: an increase or decrease in household size, a decrease in income and a decrease in number of earners (an indicator for household job loss, which is associated with tobacco use and drinking46–48) between the first and fourth quarters.
Analyses were conducted using STATA V.14. Descriptive statistics were first calculated to summarise the study sample, and analysis of variance and χ2 analyses were used to compare study groups at baseline. The inferential analysis approach assumed that households maximise household utility through consumption and spend within a budget constrained by income. The following generalised linear model was employed:
where is the IHS-transformed dollars spent on good i by household j during the fourth quarter (Q4). was a binary indicator that household j had recent cessation, was a binary indicator that household j had long-term cessation and was a binary indicator that household j had relapsed cessation. The reference group for these three variables was a binary indicator that household j did not quit tobacco spending. The model included the IHS-transformed dollars spent on good i by household j during the first quarter ( ), fourth quarter pretax income ( ) and family size ( ). The model also included a binary indicator for a decrease in household income between the first and fourth quarters ( ), binary indicators for an increase or decrease in family size from the first to four quarters ), a binary indicator for an increase in state-level unemployment rates between the first and fourth quarters ) and a binary indicator for a decrease in household earners from the first to four quarters ( ). Lastly, dummies for the survey year and month ( ) were included to account for seasonality in spending and tobacco cessation as well as characteristics of the household head ( ) and the state in which households resided ( ).
As shown in table 1, over 15% of households reported zero spending in three categories (alcohol, food away from home and healthcare). For these categories, a two-part model49 50 was also employed to examine the relationship between cessation and any spending, then between cessation and spending among spenders. Probit regression was used to estimate the relationship between tobacco cessation and the probability of having non-zero expenditures on fourth quarter spending. The relationship between tobacco cessation and the amount of spending among spenders was then estimated using equation (1) on the previous page, restricting the sample to only those households with non-zero spending in the category .
To examine how results may vary by socioeconomic status, we conducted the above analyses stratified by household income quartile, comparing results of households at the lowest income quartile with those at the highest quartile. Lastly, to further explore the causal effects of cessation on spending, we conducted the above analyses in a subsample of households (n=4196) that did not experience an income decline or change in household size between the first and fourth study quarters. We lacked sufficient sample size to stratify this latter analysis by income.
Table 2 describes the study sample by household category. Most household reference persons were women (51.0%), white (84.9%) and married (51.9%). Reference persons were on average 49.8 (±14.5) years old, and households were composed of an average of 2.6 (±1.5) persons. Approximately 14.8% of households reported an annual income below the local poverty level. The four study groups were significantly different in several explanatory variables, including age, race and marital status of the reference person, average household size, annual income and whether the household experienced an income change or a decrease in earners between the first and fourth study quarters (p<0.05).
Table 1 presents first quarter spending, when all households were still consuming tobacco. Households that did not quit were significantly different than households that quit in several expenditure categories, including tobacco, alcohol, food and housing (p<0.05). Spending on entertainment, transportation and ‘other’ goods was not significantly different between no quit households and those with any type of cessation (p>0.05).
Spending on alcohol and food
Table 3 displays the marginal effects of tobacco cessation on fourth quarter spending on alcohol and food for the full sample (full model coefficients are located in the online supplementary file), by income quartile, and then among households without an income decline or change in family size during the study. Model 1 presents the relationship between cessation and dollars spent among all households. Model 2a provides the relationship between cessation and the probability of having any spending in the fourth quarter, and model 2b provides the relationship between cessation and dollars spent, limiting the sample to households with non-zero spending.
Results show that in the full sample, households with long-term and recent cessations had lower spending on alcohol (p<0.01), but relapsed households were not significantly different than those that did not quit in alcohol spending (p=0.28). In model 2a, long-term and recent cessation, but not relapsed cessation, were associated with a lower probability of having any spending on alcohol (p<0.01). When examining dollars spent on alcohol among spenders (model 2b), households with long-term and relapsed cessations were not significantly different than households that did not quit (p>0.05), but households with recent cessation had significantly lower spending (p<0.05). These alcohol findings were replicated in households at the lowest income quartile and in households without an income decline or change in family size during the study. Among households at the highest income quartile, long-term cessation was associated with lower alcohol spending and lower probability of having any spending (p<0.001), but recent cessation was associated with lower spending only (p<0.05).
The food models showed that among all households, those with recent cessation had lower spending on total food (p<0.01), and this was due to lower spending on food at home (p<0.05). Households with recent or long-term cessation were not different in their probability of having any spending on food away from home or in their spending among spenders compared with households that did not quit (p>0.05). Households with relapsed cessation were not significantly different than those that did not quit on their total food or food at home spending (p>0.05). However, among households that were spending on food away from home, relapse was associated with higher spending (model 2b, p<0.05). These food findings were replicated in households at the lowest income quartile, but there were no group differences in food spending among households that did not experience an income decline or change in family size during the study. Additionally, there were no group differences in total or food at home spending among households at the highest income level, but relapsed households at the highest income level had a higher probably of having any spending on food away from home (p<0.01).
Spending on healthcare, housing, entertainment, transportation and other
Table 4 displays the relationship between cessation and the remaining expenditure categories (full model coefficients are located in the online supplementary file), stratified by income and among households without an income decline or change in family size. When examining all households together, long-term and recent cessations were not associated with spending on healthcare (p>0.05), but relapse was associated with higher spending on healthcare among spenders (p<0.01). These results were replicated in the subsample of households without an income decline or change in family size; in this group long-term cessation was also associated with lower probability of having any spending on healthcare (p<0.05). The low-income and high-income samples displayed a pattern of results similar to that of the full sample, but none of the group differences were statistically significant at p<0.05. There were no significant relationships between any type of cessation and spending on housing or ‘other’ goods in the full sample or any of the subsamples (p>0.05). However, in the full sample, long-term and recent cessations were associated with lower spending on transportation (p<0.05) and entertainment (p<0.01), and relapsed households had lower spending on entertainment (p<0.01). This pattern was similar in the low-income group, except recent cessation was not associated with transportation spending (p>0.05) and relapse was not associated with entertainment spending (p>0.05). In the high-income group, cessation was not associated with spending on entertainment or transportation. Lastly, in the full sample, compared with households that did not quit, households with long-term or recent cessation had lower total spending on non-tobacco goods (p<0.01), whereas relapsed households were not different in their
total non-tobacco spending than households that quit (p>0.05); this pattern of total spending was found in the low-income group only.
This study was designed to estimate the impact of tobacco cessation on household spending by comparing non-tobacco expenditures of households that had recent, long-term or relapsed cessation to the expenditures of households that did not quit, controlling for expenditure differences prior to cessation and potential covariates. The study did not find evidence that households that cease tobacco use spend more on non-tobacco goods after quitting compared with households that do not quit. Thus, tobacco spending may not have been crowding out other expenditures or households were quitting in the midst of financial strain (eg, poverty and income decline) that precluded the ability to increase spending in other areas.
Overall, results showed that households that quit tobacco spend less in areas of the budget that may have enabled or complemented their tobacco cessation (alcohol, food, entertainment and transportation). The reduced alcohol spending among long-term and recent quitters was the most robust finding across subsamples and is consistent with prior research finding that tobacco cessation is associated with a reduction in alcohol intake and increased odds of alcohol abstinence in smokers with problem drinking.30 The current study extends these prior findings by showing that the reduction in drinking during tobacco cessation occurs in non-clinical populations, occurs as an expenditure change and occurs at the level of the household. The alcohol results were similar in households at the lowest and highest income levels and were present among households without an income decline or change in family size during the study, suggesting that they are not due to an income shock or change in family composition. Nicotine has been found to increase alcohol reinforcement and cravings,51 so tobacco cessation may have operated through the elimination of nicotine as a neurobiological reinforcer of alcohol use. Alternatively, tobacco users who quit may have avoided environmental smoking triggers (eg, bars and restaurants), thus removing opportunities and cues to drink.52
The reduced spending on food among households that had recently quit tobacco use was surprising, given prior literature finding that tobacco and food are substitutes and that smoking households are more likely to experience food insecurity.15 Even in the low-income group, households that recently ceased tobacco use had significantly lower spending on food at home suggesting that they did not—or could not—use tobacco funds to alleviate pre-existing food insecurity. The lower food spending was not found in the long-term cessation group and is thus temporary. Financial strain may have jointly motivated tobacco cessation and lower food spending,53 but if the motivation for lower spending was purely financial, one would expect to see lower spending on food away from home,53–55 which was not found. Prior research has found that former smokers often turn to foods with higher sugar and fat content,56 and on average, healthy foods can cost more per calorie than unhealthy foods.57–59 In the current study, households may have switched to unhealthy, yet lower cost, foods after quitting, thereby modifying the composition or nutrient content of their food25 while reducing their food spending.
Relapsed households displayed a unique pattern of spending, such that their only differences with households that did not quit included higher healthcare spending (possibly due a serious medical condition that motivated a quit attempt) and higher spending on food away from home. Prior research has found that binge eating and weight gain during tobacco cessation reduce odds of long-term abstinence.60 61 In the current sample, a maladaptive increase in the consumption of food away from home and associated weight gain may have placed households at risk for relapse. Alternatively, as discussed previously in relation to alcohol consumption, tobacco use is vulnerable to conditional reinforcement and environmental cues.62 Individuals who continued to drink and eat away from home may have been more likely to be exposed to environmental cues to smoke, leading to relapse.
Given that in the current study tobacco cessation was only associated with lower spending (even among households that did not have an income decline), it is not clear where they reallocated their tobacco money after quitting. Households may have saved the money or used it to pay debts, which should be explored in future research.
CES participation is voluntary, so there is a risk of selection or non-response bias,63–65 and the CES does not capture the behaviour of unhoused smokers. Therefore, the results may not generalise to populations not represented by the CES. In addition, data were derived by self-report and are at risk for recall or reporting bias. However, the CES interview survey has been shown to perform well compared with national income account data,65 and the prevalence of alcohol and tobacco use captured by the CES is similar to the proportions of adults reporting past-month drinking or tobacco use on the National Survey of Drug Use and Health66 and the National Health Interview Survey.1 As an observational study, there may be unmeasured variables affecting the relationships between tobacco cessation and spending (eg, health shock).
Overall, findings suggest households that quit tobacco execute expenditure reductions that enable or complement their tobacco cessation, which many may be motivated by concurrent financial strain (eg, poverty and income decline). In the absence of observable financial strain, households that quit have limited differences in their spending compared with households that do not quit (except for lower alcohol spending and short-term lower entertainment spending). The relapsed household findings suggest one of a struggle between a health priority (eg, a quit attempt and higher healthcare spending) and an increase in eating away from home. Future research should seek to understand the nature of household relapse and how spending patterns in the postquit period place households at risk for relapse. Future research should also examine how cessation may impact household savings or debt payments.
What this paper adds
This study builds on the existing cross-sectional research on the crowd-out effects of tobacco expenditures by conducting a longitudinal analysis of the relationship between household tobacco cessation and spending on non-tobacco goods in the USA.
Results showed that long-term and recent cessations were associated with lower household spending on alcohol, food, entertainment and transportation. Tobacco relapse was associated with higher spending on healthcare and food away from home. When segmenting the sample to remove potential financial motivations to reduce spending (eg, poverty and income decline), cessation was only associated with lower spending on alcohol and (among recent quitters) lower spending on entertainment.
Households that quit tobacco spend less in areas that may enable or complement their tobacco cessation, most of which may be motivated by financial strain. The most robust association between tobacco cessation and spending was the significantly lower spending on alcohol.
Contributors ESR led the planning, conduct and reporting of the work described in this article, including analysis of data and drafting of the manuscript. She is the guarantor. DMD, AP, MF and WTG contributed to the planning, conduct and reporting of the work described in this article, including contributing to the statistical analysis plan and reviewing the manuscript.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
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.