Article Text
Abstract
Background The main goal of this study is to estimate the crowding out impact of tobacco expenditures on the household budget allocation to other mutually exclusive commodity groups in Montenegro.
Methodology The analysis uses the Household Budget Survey data from 2005 to 2017 to estimate a system of Engel curves using a three-stage least squares approach. As the tobacco expenditure variable is endogenous to budget shares on other consumption items, instrumental variables were included to obtain consistent estimates.
Results Overall, the results confirm the existence of the crowding out effect of tobacco spending on various commodities, such as some food items (eg, cereals, fruits and vegetables and dairy products), clothing, housing and utilities, education and recreation while a positive effect of tobacco consumption was estimated on budget shares on bars and restaurants, alcohol, coffee and sugary drinks. These results are consistent throughout the income groups of households. The estimates indicate that an increase in tobacco expenditures leads to reduction in budget shares on essential goods, which is likely to have negative impacts on the household living standard.
Conclusions Tobacco expenditure crowds out household spending on necessities, especially in case of the poorest households, thus increasing inequality, hampering human capital development and potentially causing long-term adverse effects on the households in Montenegro. Our results are similar to evidence from other low and middle-income countries. This paper contributes to the analysis of the crowding out effect of tobacco consumption, which was conducted for the first time in Montenegro.
- Low/Middle income country
- Price
- Taxation
- Socioeconomic status
Data availability statement
Data may be obtained from a third party and are not publicly available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Global evidence suggests that tobacco consumption constitutes a sizeable portion of household consumption expenditure, which consequently reduces the resources spent on other basic commodities. Tobacco taxation policy is one of the most effective tools to reduce tobacco use, specifically among the poor.
Despite the existing empirical evidence on tobacco economics in Montenegro, there is a lack of scientific research on the impact of tobacco spending on intrahousehold resource allocation.
WHAT THIS STUDY ADDS
The estimates confirm the existence of the crowding out effect of tobacco spending on various commodity groups. Considering the budget constraints, poorest households as the most vulnerable group would experience long-term adverse effect of tobacco use.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Montenegro could benefit from the first scientific results related to the crowding out effect of tobacco spending, as a crucial input for evidence-based policymaking towards the increase of living standards and population welfare.
Introduction
Expenditure on tobacco represents a large share of the household budget in many countries. With limited resources, spending on tobacco crowds out other spending, including on clothing, housing, education, furniture and recreation. Therefore, tobacco spending can worsen a household’s living standard and impact the development of children and the future earning potential of household members. This is especially concerning for the poor households who already have insufficient resources to support basic spending needs.
Since the first studies in the early 2000s,1 2 research especially from low and middle-income countries (LMIC) analysed the crowding out effect of spending on tobacco and consistently found evidence that this effect negatively affected spending on mainly basic necessities.3–22
For example, in 2001, Efroymson et al 1 conducted the first study considering the topic, by providing a simple comparison of consumption patterns between smoking and non-smoking households. Results confirmed the economic impact of tobacco consumption in Bangladesh and showed that tobacco use represents a large burden on the budget of households, especially those of the poor. They found that the poorest households spent twice as much on tobacco as the wealthiest households; and male smokers spent more than twice as much money on cigarettes as on clothing, housing, health and education combined, in relation to females. Similar research conducted in China4 finds that spending on tobacco negatively affects investments in human capital, productivity and financial security. Tobacco spending also has important distributional effects within the family as the costs of smoking can affect other family members by reducing the expenditure on basic needs of all members. Results in this study were obtained by estimating the almost ideal demand system that controls for sociodemographic variables. Following this research, the next generation of studies addressed the issue of endogeneity of tobacco use, with John pioneering the use of instrumental variables (IV) technique in the analysis in 2008.5 This study finds that tobacco use crowds out certain products such as food, education, clean fuels and entertainment.
The empirical evidence from the LMICs shows that the poor households are the most affected.5 6 8 9 14 15 For example, in rural Indonesia,8 households with at least one smoker tend to divert a significant amount of their already scarce budget to tobacco products. In such manner, spending on tobacco products impacts children’s nutrition caused by reduced food consumption. In Cambodia,9 tobacco spending crowds out education and clothing expenditures, as well as food expenditures for the low-income and middle-income households. Lower education levels imply higher likelihood of smoking, which in turn results in insufficient resources for investment in education. The case of Chile15 shows that potential health and education disparities can occur as a consequence of tobacco consumption. The study shows that tobacco expenditures are associated with a reduction in the budget share related to healthcare, education and housing costs, especially for poorer households. Non-smoking households had up to 32% and 16% higher budget shares on health and education, respectively, in comparison with households without smokers.
To conclude, spending on tobacco has a significant impact on household consumption patterns because the cost of purchasing tobacco products represents a significant percentage of their budget. The crowding out effect has a greater impact on lower income households, as spending on tobacco sacrifices expenditures on other goods, such as housing, health, food, education and other. Similar trends of a negative impact on the structure of total consumption are also noticeable for middle-income and high-income households, but the crowding out effect is relatively lower due to their larger incomes.
The evidence6 13 14 17 shows the inconsistent association between food and tobacco consumption, as the relationship between them is less straightforward. This inconsistency may be due to the heterogeneity of food spending since it comprises both genuine necessity (for sustenance) and more discretionary spending (for relative luxury). As spending on basic necessities represents a major share of food spending for low-income households, it is not a surprise to obtain a positive relationship between tobacco spending and share of food spending in the remaining budget.5
Montenegro experienced a rise in smoking prevalence in recent years, by 5.3 percentage points between 201723 and 2019.24 According to a survey of smokers in 2019, the prevalence of adult tobacco use was 40.7% in Montenegro.24 Based on the Montenegro Household Budget Survey (HBS), spending on tobacco in households with smokers accounted for, on average, between 3.7% and 5.4% of their budget during 2005–2017. In 2020, 22.6% of total population in Montenegro was at risk of poverty, while 13.5% of population lived in households that cannot afford at least four out of nine material deprivation items and the child poverty rate was ten percentage points higher than the national poverty rate.25
The data suggest that tobacco represents a commodity that influences the expenditure decision in a large number of households. Thus, the main goal of this study is to examine the impact of tobacco spending on household expenditure patterns in Montenegro. This paper contributes to the existing empirical evidence by providing the first results of the impact of tobacco use on intrahousehold resource allocation. Research estimates can be used to support national efforts to prevent smoking and frame the issue of adequate tobacco control policies. To the best of our knowledge, this is the first such study in Montenegro.
Theoretical framework and empirical approach
The theoretical framework for the analysis is the consumption theory (Engel curves), according to which a household maximises a utility which is a function of a set of commodities. A household’s utility is a function of n commodities, including tobacco. Following Pollak,26 we assume that a household’s demand for tobacco ( ) is predetermined at level ( ), so the household maximises the following utility function:
subject to the budget constraint , where M represents the remaining budget after deducting expenditure on tobacco ( ), while a represents a vector of household characteristics.
Since the demand for tobacco is predetermined, the demand for other commodities is conditional on the consumption of tobacco , the prices of all commodities except tobacco , the remaining budget ( ) and a set of household characteristics. Hence, we estimate the following model:
(2)
where, for each household j , represents a share of spending on a commodity i in the remaining budget M after deducting spending on tobacco ( ), is a binary variable which equals to 1 if a household has a smoker, is the expenditures on tobacco ( ) and is a vector of household characteristics.
The binary variable is included in the model to account for a difference in preferences between households with and without smokers. In other words, this variable explains whether the households with reported zero expense on tobacco do not consume tobacco because they cannot afford it (ie, corner solution) or because they have no tobacco in their utility function (ie, abstention).
Testing the null hypothesis that coefficients associated with the binary variable in equation (2) are jointly significant ( ) is done using the Wald test. Joint significance of the coefficients indicates that the households with and without smokers have different preferences. This means that the utility functions of households with reported positive tobacco spending are significantly different from the utility functions of households with zero spending.
The literature has identified a few econometric problems in estimating equation (2). First, and are likely endogenous. Second, there is likely contemporaneous correlation as the shares of spending on different commodities may affect each other. Finally, the errors may be heteroscedastic. To address these issues, it is recommended to apply the generalised method of moments three-stage least squares (GMM 3SLS) method as a more efficient estimator of a system of Engel curves. However, as the GMM 3SLS did not converge, we estimated a traditional 3SLS model, which is effectively a combination of seemingly unrelated regressions and the IV approach. We tested the null hypothesis for the presence of heteroscedasticity in the IV regression using the Pagan-Hall statistic, which was confirmed.27 As 3SLS is less efficient with heteroscedastic SEs, the estimation included 1000 bootstrap replications to account for heteroscedasticity.
The C or GMM distance test was used to test the endogeneity of the regressors. A valid instrument needs to satisfy the following two strong assumptions for the IV estimation to provide a consistent estimator: (1) instrument is partially correlated with the endogenous regressors (ie, inclusion restriction); and (2) instrument affects the dependent variable only through the regressors and not directly (ie, exclusion restriction). To test the inclusion restrictions, the LM (Lagrange Multiplier) test statistic is applied for under identification (Kleibergen-Paap rk LM test). The exclusion restriction was tested using the Hansen J statistic (test of overidentifying restrictions), since a larger number of instruments than the number of endogenous variables were used. IVs used in the analysis are described in the following section.
Data and descriptive statistics
This study uses the HBS data for Montenegro from 2005 to 2017 (except 2016, when HBS was not conducted) to estimate the crowding out effect of tobacco expenditure. HBS is conducted annually by the Statistical Office of Montenegro (Monstat) in 21 municipalities across three regions: north, central and south. The 2005–2017 total sample comprises 15 068 households, with an average number of households per year of 1256. HBS provides information on average household consumption, expenditure by commodity, household size and structure, as well as detailed information on their demographic characteristics. As HBS does not provide information on household income, we use total reported spending as a proxy for income. The households were divided into three income groups—low income, middle income and high income—based on income per household member.
HBS contains data on household expenditures in 12 broad commodity groups according to the Classification of Individual Consumption According to Purpose,28 developed by the United Nations Statistics Division. As spending on tobacco may differently impact expenditures on different food categories,5 we separated expenditures on food and non-alcoholic beverages into 11 subcategories (cereals, meat, fish, milk, other dairy products, oils and fats, fruits and vegetables, desserts, ready-made food, coffee and tea and other non-alcoholic beverages). Additionally, we disaggregated tobacco and alcohol consumption into two separate items, resulting in a total of 23 groups used in this analysis (more details in online supplemental table S1). Table 1 presents the average monthly household expenditures and budget shares on tobacco in households with smokers, by year and income group.
Supplemental material
Expenditures on tobacco have increased over the observed period both in terms of euros and shares in total household spending for all household income groups. The low-income households allocate the highest share of their budgets on tobacco in comparison with their wealthier counterparts, which suggests that they could benefit the most from the reallocation of funds to more beneficial spending.
Due to budget constraints, households with smokers may not spend as much as households without smokers on basic necessities. This may be especially the case with low-income households. Table 2 shows the budget allocation on different commodity groups for households with zero versus households with positive tobacco expenditures, as well as the Student’s t-test of the difference in shares. Smoking households spend a slightly higher shares of their budgets on clothes, transportation, bars, restaurants and hotels, alcohol and sugary drinks, while lower shares on housing and utilities, health, total dairy products, fruits and vegetables and oils and fats. The statistically significant difference in expenditures between the two types of households suggests differences in taste and preferences, which presumably means that tobacco spending may have an effect on household expenditure.
For different income groups of households with smokers (results provided in online supplemental table S2) there are differences in the budget shares for different commodities. The low-income households with smokers spend the largest share of their budget on food (47.7%) in comparison to the other two income groups (34.5% for middle-income and 27.0% for high-income households with smokers). Regarding the food commodity group, all three income groups of households spend the most on cereals, meat, dairy products and fruits and vegetables. On the other hand, the wealthiest households allocate the largest share of their budget on housing and utilities (30.0%, compared with the poorest counterparts which on this item allocate only 19.5% of their budget).
As for the budget allocation among different types of households in the low-income group, it is noticed that the households with smokers spend a relatively smaller share of their budget on essential commodities such as food (statistically significant difference is evident in the case of milk, other dairy products and oils and fats), health, education and housing and utilities, compared with households without smokers. Food spending decomposition showed that the middle-income and high-income households with tobacco spenders allocate a lower share of their budget on fruits and vegetables compared with households without tobacco users, but spend relatively more on sugary drinks. As expected, in all income groups smoking households allocate relatively larger shares on alcohol and bars, restaurants and hotels than non-smoking households.
Descriptive statistics analysis does not account and control for the impact of sociodemographic characteristics, which is why a more complex econometric modelling is needed to estimate the crowding out effect. Due to that fact, our research includes the following sociodemographic variables: household size, average age of the household members, maximum education (defined as years of education attained by the highest educated member in the household), number of children 0–2 and 3–6 years of age, number of household members 65 or more years of age, household type defined by the economic activity—unemployed (if all members are unemployed), pensioners (if at least one member is pensioner, and other members are unemployed) and employed (if at least one member is employed), region (north, south and central) and year fixed effects.
In the Results section we estimate a system of quadratic Engel curves using HBS data between 2005 and 2017. From the categories of 23 commodities, the ‘other commodities’ group was excluded to ensure the adding up restriction in the system of equations. Based on the previous studies on this topic,29 we use the following instruments for tobacco expenditure: adult sex ratio (adult male to female ratio), the per cent of adults and the per cent of male adults per household. Adults are persons 18 years old and above. Generally, as smoking prevalence among males is higher compared with females, both the adult sex ratio and the adult ratio are assumed to be uncorrelated with budget shares on other products, while correlated with tobacco expenditure.
In addition, smoking prevalence and smoking intensity7 (measured by an average number of cigarettes consumed by household) by year and municipality are constructed as an instrument of household smoking characteristics. For total expenditure without tobacco we use total expenditures per household as an instrument. The test results (presented in online supplemental table S3) show the relevance and suitability of the used IVs.
Results
As the results of the Wald test suggest, households with smokers and those without smokers have different preferences. In other words, due to abstention households without smokers report zero expenditure on tobacco because tobacco is not in their utility function, no matter the price.
The results for all households and by income group (table 3) show the evidence of crowding out. The extent of the crowding out effect declines in magnitude with rise in income level for clothing and housing. The share of spending on recreation and culture was found to be significant in the case of low-income and middle-income groups, as lower consumption of these items could particularly affect children’s health, future development and earning potential. Smokers allocate fewer financial resources to education among all income groups, which again negatively impacts their human development and future productivity, and of their entire household. Spending on tobacco impacts the likelihood of having healthier nutrition, where the gap is more prominent among the poorest households and mostly visible in case of the consumption of cereals.
The results also indicate that tobacco consumption among all income groups crowds in alcohol consumption and provides evidence confirming a strong complementarity of these two categories, as found in the literature.30–33 Spending on restaurants showed to be positively associated with tobacco expenditure, with the highest magnitude of the impact in the wealthiest group. This result is as expected, considering a relatively larger available budget among the high-income households to be spent on commodities which are not considered as essential.
It should be emphasised that a positive relationship does not mean that expenditure on a certain food category or a commodity group would necessarily increase, but rather only that its share in the remaining budget would be higher. As the variable on the left-hand side is the share of spending on food in the remaining budget after deducting the tobacco expenditure ( M ), with a constant total budget, as tobacco spending increases, M becomes smaller. For example, this means that as spending on tobacco increases, the share of expenditure on alcohol, coffee and sugary drinks in the remaining budget (after deducting tobacco expenditures) increases, but in absolute terms, these types of spending may decrease, or increase, or remain unchanged. On the other hand, for categories for which we find evidence of the crowding out effect, such as cereals, and fruits and vegetables, the share of spending on the remaining budget decreases as tobacco spending increases. As total budget remains constant, this means that the absolute amount of spending on these categories also decreases.
Discussion and conclusion
The problem of tobacco use and its negative relationship with the living standards of the population has been broadly recognised in the scientific research.34–36 The evidence on crowding out of tobacco spending shows a high burden especially in LMICs, many of which continue to have high smoking prevalence. As a result, the households with smokers sacrifice spending on other commodities, including those essential for human capital development of all household members and especially children. This issue is specifically concerning in case of the poor households with constrained budgets as it exacerbates the long-term risk of falling into the poverty trap.
This study analysed which commodity groups are displaced by tobacco in the household budget in Montenegro. The analysis was conducted using the HBS data to estimate a 3SLS model. The results confirm that tobacco consumption sacrifices the resources on necessities, while benefiting other non-healthy consumption. Similar to the previous studies in LMICs, this study shows that tobacco spending crowds out the resources on clothing, housing and education.3 6 7 10 13 15 18 Considering food categories, households with smokers compared with those without smokers spend less on cereals, dairy products (other than milk) and fruits and vegetables. These items are very important for healthy nutrition and are sacrificed due to tobacco consumption. Tobacco spending in households impacts the level of food quantity, which consequently affects the health of all household members, especially children.5 We also find that spending on tobacco crowds out spending on recreation for low-income and middle-income households. The estimated effects on spending on health, transport and communication are not statistically and economically significant. Tobacco expenditure positively affects the budget shares on bars and restaurants and alcohol, as well as spending on coffee and sugary drinks, in line with previous research,30–33 37–39 creating a negative effect on nutrition balance and healthy lifestyles. The results by income groups mostly reflect those for the full sample.
One of the limitations of this study is a lack of a more recent HBS data which prevents us from analysing the crowding out effect in the last 5 years during which certain relevant tobacco control policies have been passed in Montenegro. Moreover, the HBS consumption data are self-reported information, so it may include measurement errors. In addition, as we are using the household-level data, we were not able to analyse the impact of the intrahousehold resource allocation on individual household members.
Despite the above limitations, this study adds to the empirical evidence on the adverse effect of tobacco use on household welfare. The crowding out effect of tobacco use negatively impacts the economy as a whole, implying reduced investment in human capital development. The estimated household budgetary effects indicate that strengthening and accelerating tobacco control policies is necessary to reduce the consumption and spending on tobacco since it would enhance household-level and population-level well-being, especially of the most financially vulnerable groups of the society. Moreover, the estimated positive effect of tobacco spending on budget shares on bars and restaurants might indicate the issue of the smoke-free policy not being effectively implemented.
Data availability statement
Data may be obtained from a third party and are not publicly available.
Ethics statements
Patient consent for publication
Ethics approval
The study does not directly involve human participants but uses individual data from the Household Budget Survey. There is no need for ethics approval when using Household Budget Survey data in Montenegro because those data are already deidentified or fully anonymised and do not contain personal information
Acknowledgments
We thank Rijo John, PhD, Frank J Chaloupka, PhD, and Jeffrey Drope, PhD, for very helpful comments.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @IseaMne, @v_vulovic
Contributors AM (guarantor) and MC contributed to developing ideas, data collection and analysis, and wrote the first draft. VV reviewed the existing empirical and theoretical evidence, interpreted the results and made revisions.
Funding This paper is funded by the University of Illinois at Chicago’s Institute for Health Research and Policy through its partnership with Bloomberg Philanthropies.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.