Dangerous But Consumed: Drivers of Tobacco and Alcohol Expenditure and Their Implications on Healthcare Costs and Food Poverty in Tanzania

Abstract

The escalating use of tobacco and alcohol has become a pressing concern for policymakers worldwide due to its severe impact on public health. However, there is a paucity of research focusing on the determinants of tobacco and alcohol spending, particularly in Tanzania. The aim of this study was to examine the factors influencing tobacco and alcohol expenditures with their implications on food poverty and healthcare costs in Tanzania. The analysis relied on secondary data obtained from the 2017/2018 Household Budget Survey, while Tobit, logit and generalised linear models were employed for data analysis. Being male, unmarried and residing in rural areas exhibited positive associations while age had an inverted U-shaped relationship with tobacco and alcohol expenditures. Increased spending on these substances amplified the incidence of food poverty by 0.0002%, particularly among married couples. Tobacco and alcohol expenditures were found to jointly contribute to a 0.01% rise in healthcare costs. Results underscore the importance of targeting interventions towards mitigating the adverse effects of tobacco and alcohol consumption, with a particular focus on youth, rural areas, males and unmarried household heads. Such interventions have the potential to alleviate the incidence of food poverty and reduce the financial burden on healthcare systems.

Keywords

Drivers tobacco healthcare costs alcohol expenditure food poverty

Introduction

Tobacco and alcohol spending has attracted significant attention from policymakers due to their profound impact on public health worldwide. These substances are major contributors to cardiovascular diseases and accidents, including lung cancer (James et al., 2022). Annually, more than 8 million deaths globally are attributed to tobacco and alcohol consumption (Carney et al., 2021; WHO-AFRO, 2021). Alcohol consumption is also associated with 6% and 1.6% of the disability-adjusted life years among males and females, respectively (Gray, 2018). Over the course of the last 30 years, these commodities have caused approximately 200 million deaths and incurred a staggering economic cost of around 1 trillion US dollars (Goodchild et al., 2018). Low- and middle-income countries, including those in Africa, account for approximately 80% of the 1.3 billion users of these substances (WHO, 2019).

According to a report by the World Bank (2020), the prevalence of adult smoking in Africa, as a percentage of the total population, was estimated at 9.6%, while per capita alcohol consumption stood at approximately 5.8 L/year. These figures are lower than the global averages of 23% for adult smoking prevalence and 6.2 L per capita for alcohol consumption. However, it is projected that these figures will double in Africa by 2030 due to factors such as an increase in the youth population with varying preferences, urbanisation, Westernisation and the presence of unregulated tobacco and alcohol industries (Carney et al., 2021; WHO, 2019).

The burden of tobacco and alcohol expenditures on health and the economy is particularly significant in Africa, where consumption rates are relatively high, per capita income is low, household sizes are larger, effective prevention measures are lacking and there are limited human and financial resources (İğdeli, 2021; Kidane et al., 2015). This burden is further compounded by the fact that tobacco and alcohol not only impact health but also strain household budgets by reducing the available funds for other essential goods, including food, leading to food poverty. Additionally, it results in increased healthcare expenditures for treating tobacco- and alcohol-related diseases and accidents (Aksoy et al., 2019; Jumrani & Birthal, 2017).

Tanzania, as one of the developing nations in Africa, adversely faces the impacts of tobacco and alcohol consumption. The estimated per capita alcohol consumption is 11.7 L, and the prevalence of adult smoking as a percentage of the total population is approximately 9.2% (World Bank, 2020). These figures surpass the African average of 9.6% and 5.8 L per capita, leading to roughly 17,200 deaths annually (NBS, 2018). Alcohol consumption in Tanzania has been increasing since 2000, rising from 8.3 L per capita to 11.27 L in 2018. Conversely, the prevalence of smoking among adults as a percentage of the total population has declined from 27.8% to 9.2% during the same period (World Bank, 2020). Smoking has consequences such as reduced calorie intake, which particularly affects low-income households and those with larger household sizes (Kidane et al., 2015).

Given the concerns regarding the economic and health issues associated with tobacco and alcohol, as well as food poverty, and the ongoing anti-tobacco and alcohol campaigns, it is crucial to analyse the determinants of tobacco and alcohol expenditures in Tanzania. Such analysis will provide valuable insights for policymakers to develop effective interventions to address these issues.

While some studies have examined the factors influencing tobacco and alcohol use, they often focus on either alcohol or tobacco and primarily concentrate on developed countries (Boachie & Ross, 2020; Cetinkaya & Marquez, 2017; Dikmen, 2021; Michas et al., 2022). This limited scope of research fails to capture the interplay between tobacco and alcohol, despite their reciprocal influence on each other. As substitutes, restrictions on tobacco expenditure tend to result in higher alcohol expenditure, whereas policies aimed at reducing alcohol expenditure have minimal impact on tobacco expenditure when they are viewed as complementary substances (Aksoy et al., 2019; Anderson et al., 2020). Consequently, there is a need for more comprehensive research that explores the combined effects of tobacco and alcohol and their interdependence.

To address this gap, it is crucial to examine the determinants of tobacco and alcohol expenditures jointly, which have not been extensively studied before. As far as we know, only one study by Mwaisakila and Adrison (2021) in Tanzania has estimated the determinants of tobacco and alcohol expenditures, but it did not consider their effects on food poverty, which is the focus of this study. Another study by Kidane et al. (2015) explored the effects of smoking on nutrition and food poverty in Tanzania using household budget survey data from 2007/2008 and employed OLS regression.

The current study differs from this previous research by using the more recent household budget dataset from 2017/2018 and employing a Tobit model, which explicitly accounts for limited dependent variables. Therefore, the objective of this article is to estimate the determinants of tobacco and alcohol expenditures and examine their effects on food poverty and healthcare costs in Tanzania. This research will assist in designing policies to address tobacco and alcohol expenditures and mitigate their negative consequences. The subsequent sections are organised as follows: the second section presents the literature review, the third section describes the methodology, the fourth section presents the findings and the last section concludes with policy implications.

Literature Review

Determinants of Household Tobacco and Alcohol Expenditure

In this section, a literature review on the factors impacting tobacco and alcohol expenditures, yielding diverse and inconclusive results is presented. Studies conducted in developing countries, particularly those in Sub-Saharan Africa are first presented followed by those conducted in developed countries which is then followed by a synthesis of the distinctions and similarities of the findings on the drivers of tobacco and alcohol expenditures. Boachie and Ross (2020) conducted research in South Africa, investigating smoking intensity determinants in township communities. Their findings revealed an inverse correlation between smoking intensity and cigarette prices, consistent with the law of demand. Furthermore, age, wealth status and education of respondents negatively affected cigarette expenditure. As income rose, the demand for tobacco products decreased, suggesting tobacco’s classification as an inferior good with a negative income elasticity of −0.504. Educated individuals, equipped with information on tobacco-related health risks, displayed reduced tobacco expenditure. The study also unveiled that males smoked approximately 27%–36% more than females, a trend influenced by African culture, which affords males greater social and cultural freedom.

These findings are in line with those presented by Chelwa and van Walbeek (2019) study in Uganda using national representative expenditure data. The findings show that cigarette use is price inelastic and a 10% increase in the price of cigarettes leads to less than proportionate decline in cigarette demand by 2.6%–3.3% ceteris paribus. This is attributed to the reason that cigarette smoking is mainly influenced by habit as compared to other factors. In addition to these findings, a study conducted in 24 African countries using panel data spanning from 2010 to 2016 revealed that an increase in tobacco price as well as tobacco tax as tools to reduce tobacco consumption were found to reduce smoking rates of between 0.11% and 0.14% and between 0.25% and 0.36%, respectively (Immurana et al., 2021).

Other factors associated with the use of tobacco include residing in rural areas and being poor (Cham et al., 2019). In most parts of Africa, poor people are found in rural areas and being poor increases the likelihood of smoking as a way to relieve distress associated with being poor. Similarly, the existence of stable and enforceable tobacco control policies helps in reducing the incidence of smoking (Nyagwachi et al., 2020). Furthermore, Mwaisakila and Adrison (2021) investigated the factors influencing tobacco and alcohol expenditures in Tanzania, employing a Tobit model. Their findings revealed that a 1-Tsh increase in household income corresponded to an approximate 0.001-Tsh increase in tobacco and alcohol expenditures. Additionally, residing in rural areas was associated with an expenditure that was 10,793.7 Tsh higher than that of urban dwellers. Older individuals tended to spend less than younger individuals, suggesting a non-linear relationship between age and expenditure on tobacco and alcohol.

On the determinants of alcohol consumption, numerous studies have examined the determinants of alcohol expenditure in households. Sommer et al. (2020) conducted a qualitative study in Tanzania, exploring the social and structural factors influencing alcohol use among youths. Their findings revealed that male youths exhibited higher vulnerability to alcohol expenditure compared to their female counterparts, attributed to greater freedom and fewer restrictions during their younger age. Conversely, female youths were more protected by their parents, making them more susceptible to risky practices, including alcohol consumption.

Alcohol consumption incidence is further influenced by psychological and sociological factors including being lonely, bullied, truancy, suicidal attempts and ideation as well as peer pressure, particularly among adolescents (Asante & Kugbey, 2019). Similarly, lack or seldom attendance to religious gatherings and rituals, being male and being adult is associated with an increase in the incidence of alcohol consumption, particularly among youth attending universities (Ajayi et al., 2019). However, Ajayi et al. (2019) further found that residing in middle-class families leads to more likelihood of consuming alcohol which contradicts the findings by Cham et al. (2019) in Gambia which found that coming from a poor family increased the likelihood of alcohol consumption.

Apart from developing countries of Africa, Michas et al. (2022) undertook research in Greece, utilising a representative sample of adults to explore smoking prevalence and influencing factors. The findings indicated an overall smoking prevalence of approximately 33.1%, with a higher prevalence among males (39.1%) compared to females (29.4%). Factors such as financial stress, chronic pain and sleep problems were positively linked to smoking. Similarly, Çebi-Karaaslan (2022) investigated sociodemographic indicators of tobacco expenditure in Turkey, revealing that individuals below the age of 30, those with a low level of education, male household heads and single individuals were more likely to smoke and had higher tobacco expenditure. However, income level was associated with a reduction in tobacco expenditure.

Additionally, Polanska (2022) conducted a study in five Central and Eastern European countries, examining tobacco use susceptibility among youths. The study revealed that a low knowledge of the health effects of tobacco use increased susceptibility to tobacco expenditure. In contrast to previous studies, females exhibited higher susceptibility to tobacco expenditure in the Czech Republic and Slovenia, while the pattern was reversed in Romania. Surprisingly, increased income among youths was associated with higher susceptibility to tobacco expenditure in all examined countries except Lithuania, contrary to initial expectations.

In contrast, Leung et al. (2019) investigated socio-economic factors, alcohol availability and the built environment’s impact on alcohol expenditure in Toronto, yielding contrasting results. The study indicated that an increase in years of education and the percentage of individuals in management occupations was associated with a significant rise in area-level alcohol expenditure, while holding other factors constant. Some studies examined both tobacco and alcohol expenditure determinants. Aksoy et al. (2019) explored the determinants of tobacco and alcohol expenditures in Turkey using a multivariate sampling system of models. The study found that sociodemographic factors such as education, age, marital status, household size and income significantly affected both tobacco and alcohol expenditures.

Furthermore, Tan et al. (2016) conducted a study in Malaysia to explore the determinants of household expenditure on cigarettes and alcohol. The research identified influential factors, including income, household size, age, education, gender and location. In contrast to earlier studies, this research found that an increase in income led to higher expenditure on both cigarettes and alcohol, with an approximate increase of RM 3.29. The negative coefficient for household size suggested that alcohol and cigarette expenditure decreased as family size increased, likely due to larger families allocating more of their budget to essential goods rather than luxury items. Additionally, education had a positive effect, while age had a negative effect on alcohol and cigarette expenditure.

In a study conducted by Alkan and Abar (2019) on tobacco consumption in Turkey, employing logistic and probit regression analysis, the findings indicated that individuals aged 24–44 years, secondary school graduates, divorced and widowed individuals, and those exposed to cigarette smoke were more likely to engage in tobacco expenditure compared to other groups. Similarly, Amalia et al. (2019) investigated social inequalities and factors influencing cigarette smoking in Indonesia from 2007 to 2014. The results revealed that sex, age and education played significant roles as determinants of cigarette smoking. Specifically, males, individuals under the age of 55, and those with lower levels of education had a higher likelihood of smoking compared to their counterparts.

Synthesis between studies conducted in Africa and those conducted outside of Africa particularly in relatively developed countries on the determinants of tobacco and alcohol expenditures show more similarities than differences with the exception of some few variables. For example, in most of the African countries examined, people with lower education levels and being poor were more likely to smoke or consume alcohol than those with higher education levels. In some countries in Europe, high incidences of smoking and alcohol use are inherent in well-off families and those with higher education levels. This implies that there are still mixed and contradictory results which could not be generalised to all countries.

Effect of Tobacco and Alcohol Expenditure on Food Poverty and Health Care Costs

Although tobacco and alcohol contribute to the economic growth of Tanzania, their spending negatively impacts food poverty and healthcare expenses. The consumption of these items is associated with higher occurrences of lung cancer and cardiovascular diseases, leading to increased healthcare costs and a shift in household budget allocation away from food consumption towards healthcare. Kidane et al. (2015) emphasise the additional financial burdens these diseases place on households. Moreover, the use of tobacco and alcohol can result in increased work absenteeism, leading to a loss of wage income. Nyakutsikwa et al. (2021) and Ozughalu (2016) support this idea and suggest that reduced appetite and poor nutrition further worsen household food poverty. Tobacco consumption further leads to a reduction in the household budget allocated to necessary goods including food items, healthcare, education and communication (Nyagwachi et al., 2020).

However, some existing studies found that, at the aggregate national level, the reduction in cigarette use is associated with increased poverty level through a decline in employment rates generated by tobacco industries. For example, Jha et al. (2018) examined the effect of reducing smoking prevalence on the economy and environment in Tanzania. This study concluded that with a 30% decline in smoking prevalence, employment rates in tobacco and tobacco products would also decline by 20.8% and 7.8%, respectively.

Theoretical Model

This study is based on the theoretical foundation of random utility maximisation theory, initially introduced by McFadden (1980). According to this theory, households are presumed to optimise their utility through choices influenced by random elements, while considering their budget limitations. These choices ultimately impact their expenditure on tobacco and alcohol. The utility function can be mathematically expressed as follows:

v (q, z, x) = d U (q, z, x) + (1 - d) U^{*} (z, x)

(1)

In this context, the random utility function (v) is determined by the quantity of the commodity (q) represented in terms of its price. The composite variable (z) denotes other goods consumed by the household, which are standardised to a value of 1. The utility resulting from spending on tobacco and alcohol (U) is compared to the utility obtained from not spending on them (U*). A binary variable (d) assumes a value of 1 if the household spends on these commodities and 0 otherwise. The vector of explanatory variables (x) encompasses social, economic and environmental factors that influence household expenditure on tobacco and alcohol.

Tobacco and alcohol consumption are perceived as addictive behaviours guided by rational expectations. Household decisions are influenced by past, present and future consumption patterns, all aimed at maximising lifetime utility. The utility at a specific time (T), denoted as U, is shaped by the current level of addictive consumption (C), consumption of other goods (Z) and the accumulation of prior consumption (S).

U^{t} = U^{i} [C (t), Z (t), S (t)]

(2)

The utility function described in Equation (2) is concave in relation to Z, C and S. It assumes that the lifecycle utility can be separated into Z, S and C, rather than just Z and C.

The variables utilised in the analysis are drawn from existing literature and economic theories. Previous research has identified several key factors influencing household expenditure on tobacco and alcohol, including household income, age, sex, education level of the household head, marital status, location (urban/rural), household size and ethnicity (Aksoy et al., 2019; Amalia et al., 2019; Boachie & Ross, 2020; Cetinkaya & Marquez, 2017; Dikmen, 2021; İğdeli, 2021; Michas et al., 2022). For instance, older age has been linked to reduced tobacco and alcohol expenditures, possibly due to higher health risks, financial stability and greater awareness of the negative effects (Aksoy et al., 2019; Cetinkaya & Marquez, 2017). In a study by Aksoy et al. (2019) conducted in Turkey, it was observed that older age negatively impacted expenditure on tobacco and alcohol. The researchers proposed several contributing factors. First, older individuals are more susceptible to health issues, leading them to reduce spending on tobacco and alcohol. Additionally, with age, people tend to have greater financial stability and accumulated life experience, increasing their awareness of the negative consequences of such expenditures, which, in turn, influences them to decrease spending. This finding aligns with another study by Boachie and Ross (2020) conducted in South Africa, which also found that individuals younger than 30 years old spent more on tobacco and alcohol compared to those aged 55 years and above.

According to research conducted by Tan et al. (2016) and James et al. (2022), there exists a negative association between household size and expenditure on tobacco and alcohol. This implies that as the size of the household increases, the spending on tobacco and alcohol tends to decrease. The underlying reason is that larger households demand more essential items, leading to a reduction in the budget allocated for luxury goods like tobacco and alcohol.

Consistent with multiple studies (Boachie & Ross, 2020; Perera et al., 2017), the level of education of the household head negatively affects expenditure on tobacco and alcohol. Individuals with higher education levels have better access to information about the health risks associated with tobacco and alcohol consumption, leading them to reduce their spending on these products.

Regarding income’s influence on household expenditure on tobacco and alcohol, the existing literature provides mixed findings. Some studies suggest a positive effect (Polanska, 2022), indicating that higher income, especially among young individuals, leads to increased spending on luxury goods due to peer pressure. On the contrary, financial difficulties and stress have been linked to higher tobacco expenditure (James et al., 2022). From a negative perspective, rising income allows individuals to access health information, causing cigarettes and alcohol to be considered inferior goods as income increases. When it comes to the gender of the household head, it is expected to have a positive–negative effect on tobacco and alcohol expenditures, as seen in previous studies (Carney et al., 2021; Perera et al., 2022). This can be attributed to the fact that females are generally more cautious about risks to the household compared to males, and cultural norms often favour males over females. For example, in many African cultural settings, females are prohibited from drinking alcohol or smoking cigarettes.

Moreover, it is hypothesised that marital status negatively affects expenditure on tobacco and alcohol. Married individuals are inclined to allocate a smaller portion of their budget to tobacco and alcohol due to increased responsibilities that prioritise essential goods benefiting the family unit, unlike unmarried individuals. This idea is corroborated by earlier research conducted by Aksoy et al. (2019) in Turkey, which demonstrated that married individuals spend approximately 7 TL less per month on tobacco and alcohol compared to their unmarried counterparts.

Methodology

Data

This research utilised secondary data sourced from the Tanzania National Household Budget Survey, which was conducted by the Tanzania National Bureau of Statistics between November 2017 and December 2018. The dataset employed in this study is nationally representative, covering 26 regions in mainland Tanzania, and involves a total sample size of 9,463 households. The data collection followed a two-stage cluster sampling design. In the first stage, enumeration areas (primary sampling units or PSUs) were selected from the 2012 population and housing census. From the 796 PSUs chosen, 69 were from Dar es Salaam (Tanzania’s largest city), 167 were from other urban areas and the remaining 560 PSUs were from rural areas. In the second stage, a systematic sampling method was used to select a sample of 12 households from the updated list of PSUs, resulting in a final sample size of 9,463 households. However, due to some missing values, the sample size was reduced to 7,335 households for further econometric analysis.

The variables extracted from the dataset included monthly household expenditure on tobacco and alcohol, expressed in Tanzanian shillings (TZS), which served as the dependent variable. Additionally, several independent variables were extracted, including age, education, location, marital status, income level and sex of the household head. Table 1 provides the descriptive statistics of these variables. The results indicate that the majority of respondents (72.73%) were males, while only 27.27% constituted female household heads. On average, the typical household head was approximately 47 years old, indicating that most respondents were young adults. Among them, 80.32% had primary education, while the remaining had higher levels of education, including 14.58% with secondary education and 5.10% with tertiary education.

Table 1.

Definition of Variables and Descriptive Statistics.

Variable	Description	Mean	Std Dev.
Continuous variables
Expenditure on tobacco and alcohol	Dependent variable (TZS/month)	1,990.56	10,361.94
Age of the household head	Age of household head in years	47.07	15.74
Household size	Number of members in a household	4.85	2.91
Annual household income (TZS)	Annual household income in TZS	946,863.4	2,343,936
Categorical variables		Frequency	%
Sex of household head	Dummy (1 = male, 0 = female)
Male		6,882	72.73
Female		2,581	27.27
Education level of household head	(1 = primary, 2 = secondary, 3 = tertiary)
Primary education		5,972	80.32
Secondary education		1,084	14.58
Tertiary education		379	5.1
Marital status of household head	Dummy (1 = married, 0 = not married)
Married		6,411	67.8
Not married		3,045	32.2
Location of household head	Dummy(1 = rural, 0 = urban)
Rural		6,675	70.54
Urban		2,788	29.46
Alcohol and smoking status	Dummy (1 = yes, 0 = no)	1,272	13.44
Food poor household	Dummy (1 = yes, 0 = no)	375	3.96

Note: N = 9,463.

Furthermore, most respondents (70.54%) resided in rural areas, while only 29.46% were urban dwellers. Approximately 67.8% of the respondents were married, either in monogamous or polygamous families, while 32.2% were not married, which included those who were divorced, widowed, never married, separated or living together in a cohabitation relationship. About 13.4% of the sample had expenditure on tobacco and alcohol, with an average monthly household expenditure of approximately 1,990.6 TZS. The average household size was 4.8 people, and the annual household income was around 946,863 TZS. Food-poor households constituted about 3.96% of the total sample.

Similarly, Table 2 displays the distribution of household expenditures for individual goods and services as a percentage of the total household expenditure. The findings reveal that, on average, the monthly total household expenditure amounted to 423,692.1 TZS, with 62.5% allocated to food expenditure and 37.5% allocated to non-food expenditure. The proportion of expenditure allocated to tobacco and alcohol was remarkably low, at approximately 0.6%, indicating that only a small portion of the monthly household budget is spent on these items. Furthermore, the allocation for education was relatively modest, accounting for 0.9% of the total expenditure. This can be attributed to the government’s implementation of free education for primary and secondary schools at the time of the survey. Among the non-food expenditure categories, the largest share was allocated to housing, water, electricity and gas, accounting for 15.96%. This was followed by clothing, which accounted for 4.56%, and transport, which accounted for 4.41% of the total expenditure.

Table 2.

Monthly Percentage Share of Household Total Expenditure.

Variable	Mean (TZS)	Std Dev.	% Share
Tobacco and alcohol	1,990.564	10,361.94	0.595
Clothing and footwear	18,928.373	34,993.866	4.557
Housing, water, electricity and gas	52,797.481	67,083.737	15.956
Furnishings, equipment and maintenance	9,410.949	27,235.279	2.206
Health	7,970.33	46,666.161	1.837
Transport	27,306.465	120,045.75	4.414
Communication	10,344.275	19,830.942	2.757
Recreation and culture	12,648.48	370,714.21	1.362
Education	9,382.303	87,197.729	0.9
Miscellaneous goods and services	15,537.534	51,814.811	2.934
Total non-food expenditure	159,072.77	141,664.73	37.518
Total food expenditure	264,619.34	1,673,248.4	62.482
Total household expenditure	423,692.1	1,693,704.2	100.00

Measuring Food Poverty

As defined by Zhang et al. (2018), food poverty occurs when a household’s food consumption falls below the food poverty threshold, representing the minimum daily food requirement of 2,200 kcal per adult equivalent. Food poverty indices, as highlighted by the National Bureau of Statistics (NBS, 2020), are utilised to assess the prevalence, depth and severity of food poverty within the population by determining the percentage of households that fall below the established poverty threshold. In this study, the measurement of food poverty utilised the same approach as Zhang et al. (2018) and the National Bureau of Statistics (NBS) in 2020. The chosen method was the Foster–Greer–Thorbecke (FGT) index, which belongs to a group of indices that provide a comprehensive assessment of poverty distribution within a population, distinguishing it from other measurement approaches. The FGT index is calculated using the following formula:

F G T = \frac{1}{N} \sum_{i = 1}^{n p} {[\frac{Z - y_{i}}{Z}]}^{α}

(3)

The FGT index, denoted as FGT, is employed in this study to measure food poverty. The index takes into account the food poverty line (Z), the food expenditure below this line ( $y_{i}$ ), and the sensitivity parameter (α) that determines the type of poverty index used. Specifically, when α equals 0, the index represents the headcount ratio, which measures the proportion of households classified as food poor. For α = 1, the index becomes the poverty gap index, which quantifies the average expenditure shortfall of households compared to the food poverty line. Finally, when α is set to 2, the index transforms into the poverty severity index (PSI), which assesses the relative distance between poor households and the food poverty line, as well as the degree of inequality among the poor (NBS, 2020).

Figure 1 illustrates the trajectory of food poverty in Tanzania based on the 2017/2018 household budget survey, comparing it to the previous surveys conducted in 2007/2008 and 2011/2012. The graph indicates a decline in food poverty rates across both rural and urban areas of Tanzania, as well as at the national level. From 2007/2008 to 2011/2012, the prevalence of food poverty decreased from 11.7% to 9.8%, and it further declined to 8% in the 2017/2018 survey for the entire population. However, the rate of food poverty remained higher in rural areas, decreasing from 13.3% in 2007/2008 to 11.3% in 2011/2012 and reaching 9.7% in 2017/2018. In contrast, urban areas experienced a lower rate of food poverty, which decreased from 7.1% to 6.1% between 2007/2008 and 2011/2012, and dropped even further to 4.4% during the same period (NBS, 2020).

Econometric Model Specification

Considering that a substantial number of households (8,191 out of 9,463) reported zero expenditure on tobacco and alcohol, it becomes evident that the data in this study are censored. The prevalence of numerous zeros in the dependent variable violates the assumption of normality in linear regression, making it necessary to adopt a more suitable model (Thibaut et al., 2020). The presence of a large number of zeros can be attributed to various factors. First, infrequent spending within the short data collection duration may result in sample selection. Second, some individuals may be absent from the market due to different reasons. Lastly, budget constraints could also contribute to the occurrence of zeros (Fu & Florkowski, 2016).

When handling a significant number of zeros in the dependent variable, previous research has utilised two models: the Heckman selection model and the Tobit model. The Heckman model is applied when the zeros arise from sample selection (Ba et al., 2019), whereas the Tobit model is preferred when the zeros result from budgetary constraints within households (Thibaut et al., 2020). In this study, the Tobit model was chosen since the dependent variable (tobacco and alcohol expenditures) is continuous, the data are censored, and the zeros are a consequence of constrained budgets. In line with previous studies (Tan et al., 2016; Thibaut et al., 2020), the model is specified according to the following equation:.

y_{i}^{*} = α X_{i} + μ_{i}, \{\begin{matrix} {y_{i} = y}_{i}^{*}, i f {y_{i}^{*}}_{i} > 0 \\ y_{i} = 0, i f y_{i}^{*} \leq 0 \end{matrix}, i = 1, 2, \dots, n

(4)

In the context of this model, the latent explained variable is represented as $y_{i}^{*}$ , while the observed explained variable is denoted as $y_{i}$ . The vector $X_{i}$ comprises the explanatory variables and α represents the corresponding coefficients. The random error term µ is assumed to follow a normal distribution, independently and with a mean of zero and a constant variance $μ$ ~ N(0, σ²). To facilitate a meaningful interpretation of the results, the marginal effects were calculated after applying the Tobit model. This was done by utilising the following formula, where ɸ represents the cumulative normal distribution.

\frac{\partial E (y | x)}{\partial x_{i}} = Φ (\frac{x^{'} α}{σ}) α_{i}

(5)

The marginal effects were calculated using the STATA command ‘mfx, predict (ystar(0,.))’.

Furthermore, a binary logit model was utilised to analyse the impact of tobacco and alcohol expenditures on the occurrence of food poverty. This model was chosen since the dependent variable was binary, taking a value of 1 to indicate food poverty (when the household falls below the food poverty line), and a value of zero to indicate non-food poverty (when the household’s expenditure is above the food poverty line). The logit model was preferred over the probit model due to its better fit in cases where extreme explanatory variables are present in multivariate link models, as indicated by Hahn and Soyer (2005). Following the approach of Gujarat and Porter (2009), the logit model was specified as shown in the following equation:

l n [\frac{P}{1 - P}] = X_{i}^{'} β + ε

(6)

In the above equation, X represents a vector of explanatory variables (as shown in Table 5) that influence the occurrence of food poverty in households. P represents the probability that a household falls below the food poverty line, while 1 − P represents the probability that it does not fall below the food poverty line. ε denotes the disturbance term in the model. The effect of tobacco and alcohol expenditures on healthcare costs was estimated by using a generalised linear model with a log-link function following Isaranuwatchai et al. (2019)

H E = β_{0} + β_{1} {A T + X}_{i}^{'} β + ε

(7)

where HE is healthcare expenditure, AT is the tobacco and alcohol expenditures, X represents a vector of socio-economic characteristics while $ε$ is a disturbance term.

Results

Descriptive Results

The findings from Table 3 illustrate the outcomes of the analysis, specifically focusing on the breakdown of monthly expenditure on tobacco and alcohol based on variables such as sex, household size and location of the household head, education level, marital status, household income and age of the household head. The results demonstrate significant variations in tobacco and alcohol expenditures across all the examined variables. It was observed that male-headed households allocate a larger portion of their budget towards tobacco and alcohol compared to female-headed households (p < .01). This disparity can be attributed to cultural norms prevalent in many African societies, where women are typically prohibited from smoking and consuming alcohol due to their responsibilities in caring for children and their heightened awareness of health risks compared to men (Boachie & Ross, 2020; Perera et al., 2017).

Table 3.

Monthly Prevalence of Household Tobacco and Alcohol Expenditure (TZS).

Variable	Categories	N	Mean	Std Dev.	χ ²	P > χ²
Sex of household head	Male	6,882	2,278.30	10,761.78	91.44	.000*
	Female	2,581	1,223.35	9,169.27
Household size	1–5 members	6,242	1,848.98	9,064.99	545.90	.000*
	6–10 members	2,865	2,120.95	11,872.47
	11 and above members	356	3,423.71	16,655.87
Location of household	Rural	6,675	2,174.65	10,559.88	18.14	.000*
	Urban	2,788	1,549.83	9,859.81
Education level	Primary education	5,972	2,182.39	10,871.01	239.36	.000*
	Secondary education	1,084	1,654.02	8,111.30
	Tertiary education	379	2,581.92	14,747.26
Marital status	Married	6,411	2,090.64	10,821.73	87.28	.000*
	Not married	3,045	1,784.01	9,331.11
Household income (TZS)	Less/equal to 1,000,000	6,904	1,953.11	10,492.11	9.44	.009*
	1,000,001–5,000,000	2,378	2,106.77	10,078.34
	Greater than 5,000,000	92	2,822.04	11,933.20
Age of the household head	13–35 years	2,509	1,423.58	9,512.83	478.90	.000*
	36–58 years	4,750	2,347.40	11,719.34
	59 years and above	2,204	1,866.97	7,836.52

Note: *Denotes significance at 0.01.

The analysis reveals notable variations in tobacco and alcohol expenditures across different age groups (p < .01). Specifically, young adults between the ages of 36 and 58 display higher levels of expenditure on these items compared to both younger individuals aged 13–35 and older individuals aged 59.7 years and above. The higher expenditure among young adults can be attributed to factors such as financial stress and peer pressure, as they may view these products as a means of alleviating the burdens associated with financial difficulties. As individuals advance in age, there is a decline in smoking and drinking behaviours. This decline can be attributed to older individuals being more susceptible to the health risks associated with tobacco and alcohol consumption, having accumulated a wealth of knowledge regarding the detrimental effects of these harmful substances (Aksoy et al., 2019; Tan et al., 2016).

Expenditure on tobacco and alcohol demonstrates significant variations across different income groups (p < .01). The findings suggest that higher income is associated with increased spending on these items, indicating that tobacco and alcohol are perceived as luxury goods. Households with greater income have the financial capacity to afford both essential necessities and indulgent commodities, while economically disadvantaged households struggle to meet basic needs, let alone allocate funds for luxury items. Nevertheless, the existing literature on the income elasticities of demand for tobacco and alcohol presents mixed results. Some studies report positive elasticities of demand (Tan et al., 2016), whereas others find negative income elasticities (Mwaisakila & Adrison, 2021).

Additionally, the results indicate that rural households spend approximately 624.82 units more on tobacco and alcohol compared to their urban counterparts. In rural areas, alcohol consumption is intertwined with cultural practices, especially during local ceremonies and informal gatherings where individuals convene in local clubs to discuss various economic and sociocultural matters. Moreover, the majority of alcohol consumed in rural areas consists of locally brewed beverages, which are relatively more affordable than commercially produced alcoholic drinks. These findings align with the study conducted by Mwaisakila and Adrison (2021), but they contradict the findings of Carney et al. (2021), which suggest higher tobacco and alcohol consumption in urban areas compared to rural areas.

Furthermore, the extent of tobacco and alcohol expenditures varies depending on the education level of the household head, household size and marital status. Household heads with tertiary education tend to spend more on these commodities, followed by those with primary education, while those with secondary education exhibit lower expenditure. Larger households tend to allocate more funds to tobacco and alcohol compared to smaller households. Similarly, married household heads demonstrate higher expenditure on these products compared to unmarried household heads.

The estimates from Table 4 reveal the incidence of food poverty, food poverty gap and food poverty severity indices to be 3.96%, 1.16% and 0.57%, respectively. Female-headed households, on average, experienced higher levels of food poverty compared to male-headed households. This disparity can be attributed to prevailing patriarchal cultural norms and traditions in African societies, where males typically have greater access to productive and financial resources. Additionally, food poverty was more prevalent in rural areas than in urban areas, but it decreased with higher levels of education, consistent with previous studies (Kidane et al., 2015; NBS, 2020).

Table 4.

Monthly Household Food Poverty Prevalence.

Variable	Categories	Poverty Incidence	Poverty Gap	Poverty Severity
Overall		0.0396	0.0116	0.0057
Tobacco and alcohol expenditures	Yes	0.0314	0.0068	0.0025
	No	0.0409	0.0123	0.0062
Sex of household head	Male	0.0325	0.0090	0.0042
	Female	0.0585	0.0185	0.0098
Household size	1–5 members	0.0495	0.0146	0.0072
	6–10 members	0.0206	0.0059	0.0031
	11 and above members	0.0197	0.0039	0.0010
Location of household	Rural	0.0511	0.0148	0.0071
	Urban	0.0122	0.0038	0.0024
Education level	Primary education	0.0355	0.0096	0.0045
	Secondary education	0.0129	0.0042	0.0021
	Tertiary education	0.0053	0.0053	0.0053
Marital status	Married	0.0328	0.0096	0.0048
	Not married	0.0542	0.0158	0.0078
Household income (TZS)	Less/equal to 1,000,000	0.0427	0.0124	0.0060
	1,000,001–5,000,000	0.0324	0.0098	0.0054
	Greater than 5,000,000	0.0217	0.0038	0.0010
Age of the household head	13–35 years	0.0359	0.0117	0.0064
	36–58 years	0.0318	0.0087	0.0043
	59 years and above	0.0608	0.0176	0.0081

Household income also played a significant role in reducing the incidence of food poverty. The incidence of food poverty was lower among households with higher incomes (2.17%) compared to those with lower incomes (4.27%). This can be attributed to the fact that higher income levels enable households to better cope with fluctuations in food supply and demand (Kidane et al., 2015). Interestingly, contrary to expectations, households headed by individuals who did not spend on tobacco and alcohol had higher levels of food poverty compared to those who did spend on these items. The results indicate that the incidence of food poverty was lower among smokers and alcohol drinkers (3.14%) than among non-smokers and non-alcohol drinkers (4.1%). Furthermore, the incidence of food poverty was higher among households headed by older individuals (6.1%) compared to households headed by young adults (3.2%) or youth (3.6%). This can be attributed to the increased financial constraints faced by elderly individuals as they age and become less productive (Zhang et al., 2018).

Similarly, households with unmarried heads experienced a higher incidence of food poverty (5.4%) compared to households headed by married individuals (3.3%). Additionally, households with larger sizes had lower incidences of food poverty than those with smaller sizes.

Econometric Results

Determinants of Tobacco and Alcohol Expenditure

The empirical results pertaining to the determinants of household tobacco and alcohol expenditures in Tanzania are presented in Table 5. The likelihood ratio test yielded a significant outcome at the 1% level, indicating a well-estimated model. Among the variables included in the model, the age of the household head, sex, location and marital status emerged as influential factors affecting household tobacco and alcohol expenditures in Tanzania. Specifically, age, sex and location had a positive and significant impact, while marital status had a negative and significant effect on tobacco and alcohol expenditures.

Table 5.

Determinants of Household Expenditure on Tobacco and Alcohol.

Variable	Coefficient	Marginal Effect
Household size	106	13.01495
	(267.7)
Household head age	1,294.0***	158.8533
	(293.4)
Household age squared	−10.83***	−1.330027
	(2.882)
Total household income (TZS)	−0.000693	−0.000085
	(0.000694)
Sex of household head (1 = male, 0 = female)	15,077.4***	1,565.616
	(2,243.3)
Location of household head (1 = rural, 0 = urban)	11,652.6***	1,329.6
	(1,695.5)
Marital status (1 = married, 0 = not married)	−8,426.3***	−1,117.191
	(1,961.1)
Education of household head (1 = secondary, 0 = tertiary	993	119.6324
	(3,680.6)
Constant	−94,077.5***
	(8,366.3)
Var. (e. expenditure on tobacco and alcohol)	1.55578e + 09***
	(79,567,178.2)
Log-likelihood	−13,804.905
LR χ² (8)	139.5
Prob. > χ²	0.0000
Left censored observations	6,333
Uncensored observations	1,010
N	7,343

Notes: Standard errors in parentheses.

***p < .001.

The findings suggest that initially, for each additional year in the age of the household head, tobacco and alcohol expenditures increase by approximately 158.85 TZS, holding other factors constant. However, the relationship between age and tobacco and alcohol expenditures is not linear, as evidenced by the negative and significant coefficient for the squared term of age. Beyond a certain point, known as the turning point (at 59.7 years), and with a maximum expenditure of 4,743.1 TZS on these commodities, expenditure on tobacco and alcohol starts to decline by approximately 1.33 TZS. This implies that expenditure on tobacco and alcohol tends to increase with age until the household head reaches around 59.7 years, after which it starts to decrease.

The coefficient related to the sex of the household head demonstrates that male-headed households allocate more funds to tobacco and alcohol compared to female-headed households. Specifically, male-headed households spend approximately 1,565.6 TZS more on these commodities than their female-headed counterparts. Additionally, the geographic location of the household is a significant factor influencing tobacco and alcohol expenditures in Tanzania. Particularly, rural households spend more on tobacco and alcohol than urban households, with rural households spending approximately 1,329.6 TZS more than those in urban areas. Moreover, the results indicate that married household heads tend to spend less on tobacco and alcohol compared to unmarried household heads, with an expenditure difference of approximately 1,117.2 TZS. However, the study finds no significant effects of the education level, household size and income of the household head on tobacco and alcohol expenditures.

Effects of Tobacco and Alcohol Expenditure on Food Poverty

Moving on to the effects of tobacco and alcohol expenditures on food poverty, the logit results presented in Table 6 demonstrate that although expenditure on tobacco and alcohol positively affects food poverty, it initially appears to be insignificant. However, the significance becomes apparent when interacting with marital status. The interaction term between married household heads and expenditure on tobacco and alcohol is found to be positively significant (p < .001). This implies that all else being equal, households headed by married individuals who spend on tobacco and alcohol have a slightly higher probability of experiencing food poverty compared to unmarried heads who spend on these commodities. Similarly, the coefficient for household size is positive and significant (p < .001), indicating that as household size increases by one member, the probability of food poverty increases by approximately 0.36% when holding other factors constant.

Table 6.

Effects of Tobacco and Alcohol Expenditure on Food Poverty.

Dependent Variable: Food Poverty
Dependent Variables	Coefficient	Marginal Effect
Expenditure on tobacco and alcohol (TZS)	0.000250	4.25E − 06
	(0.000410)
Marital status × Expenditure on tobacco and alcohol (TZS)	0.000107*	1.81E − 06
	(0.0000444)
Age of the household head	0.00674	0.0001145
	(0.0233)
Age of the household head squared	−0.000149	−2.53E − 06
	(0.000224)
Sex of the household head	0.284	0.005221
	(0.182)
Household size	0.211***	0.003591
	(0.0350)
Education level (1 = secondary, 0 = tertiary)	−1.393*	−0.013838
	(0.718)
Location of household (1 = rural, 0 = urban)	−1.605***	−0.023322
	(0.226)
Marital status (1 = married, 0 = not married)	−0.106	−0.001757
	(0.181)
Household income (TZS)	−2.95e − 09	−5.01E − 11
	(5.40e − 08)
Location × Expenditure on tobacco and alcohol (TZS)	−0.000251	−4.26E − 06
	(0.000409)
Household head sex × Expenditure on tobacco and alcohol (TZS)	−0.0000129	−2.20E − 07
	(0.0000197)
Constant	5.014***
	(0.927)
LR χ² (13)	156.42
Prob. > χ²	0.0000
Log-likelihood	−934.42201
Pseudo R²	0.0772
N	7,343

Notes: Standard errors in parentheses.

*p < .05, ***p < .001.

Furthermore, the findings suggest that the education level of the household head has a negative and significant effect (p < .05) on the level of food poverty. As the household head’s level of education increases by one level, the probability of food poverty decreases by approximately 1.38%, assuming other factors remain constant. In other words, higher levels of education are associated with lower incidences of household food poverty. Similarly, the location of the household is found to have a negative and significant (p < .001) impact on the likelihood of food poverty. Households residing in rural areas have approximately 2.3% lower probability of experiencing food poverty compared to those in urban areas, while controlling for other variables. The remaining variables in the model, including age, sex and income of the household head, were not found to be statistically significant, although income and age displayed the expected signs while the coefficient for sex did not align with the prior expectations.

Effect of Tobacco and Alcohol Expenditure on Healthcare Expenditure

Table 7 presents the effect of tobacco and alcohol expenditures on household’s health expenditure. Results show that expenditure on tobacco and alcohol has positive effect on household’s health expenditure level. An increase in tobacco and alcohol expenditures by 1% increase household’s healthcare expenditures by 0.01%. This is consistent with expected prior. Apart from tobacco and alcohol expenditures, some control variables included in the model were also significant factors affecting household’s healthcare expenditure. The coefficient of household size was positive and significant suggesting that an increase in household size leads to an increase in household’s healthcare expenditure. An increase in the household size by one member leads to an increase in the level of household’s health expenditure by about 0.7% holding other factors affecting health expenditure constant. Similarly, a household headed with a married household head had 3.78% more health expenditure than those headed by unmarried heads. Furthermore, the coefficient of location of household head had a negative and significant effect on healthcare expenditure. Households residing in rural areas had 7.2% lower health expenditure than their urban counterparts.

Table 7.

Effect of Tobacco and Alcohol Expenditure on Healthcare Expenditure.

Dependent Variable: LN (Monthly Health Expenditure)
Variable	Coefficient
LN (tobacco and alcohol expenditures)	0.00964*
	(0.00477)
Age of household head	0.00213
	(0.00221)
Age of household head squared	−0.0000138
	(0.0000203)
Household size	0.00677***
	(0.00192)
Sex of household head (1 = male, 0 = female)	0.0279
	(0.0195)
Location of household (1 = rural, 0 = urban)	−0.0744***
	(0.0170)
Marital status (1 = married, 0 = not married)	0.0371*
	(0.0176)
Education level (1 = secondary, 0 = tertiary)	−0.00949
	(0.0135)
Insurance cover (1 = yes, 0 = no)	−0.0164
	(0.0182)
Ln(household income)	0.00558
	(0.00912)
Constant	1.861***
	(0.141)
AIC	6.243
BIC	−4,319.75
Log-likelihood	−2,517.6

Notes: Standard errors in parentheses.

*p < .05, ***p < .001.

Discussion of Findings

The findings displayed in Table 5 illustrate a positive correlation between age and expenditure on tobacco and alcohol, suggesting that as individual’s age, their spending on these products increases. However, this trend reverses as individuals reach older age, and expenditure on tobacco and alcohol starts to decline. This result is in line with prior expectations and is consistent with earlier research. Younger individuals, especially youths aged 15–35 years and young adults below 60 years, tend to have higher expenditure on tobacco and alcohol due to factors such as peer pressure, limited knowledge about the harmful effects of drinking and smoking, financial stress and exposure to smoking environments. As individuals age, they accumulate more knowledge about the health consequences associated with tobacco and alcohol consumption, become more susceptible to cardiovascular diseases including lung cancer and are less influenced by peer pressure. Consequently, their expenditure on tobacco and alcohol decreases. This finding is supported by several previous studies, including Mwaisakila and Adrison (2021), Aksoy et al. (2019) and Boachie and Ross (2020). It is also consistent with the findings of the National Bureau of Statistics (NBS, 2018) on the global adult tobacco survey, which revealed that smoking in Tanzania is more prevalent among individuals aged 20–34 years, accounting for approximately 31.1% of all smokers.

Moreover, the study outcomes indicate that male-headed households tend to allocate higher spending on tobacco and alcohol compared to female-headed households, aligning with existing literature from James et al. (2022), Perera et al. (2017), İğdeli (2021) and Boachie and Ross (2020). For instance, Boachie and Ross (2020) found in a township community study in South Africa that the rate of cigarette smoking among males was 27–36% higher than among females. This can be attributed to cultural norms in African societies, which grant men more freedom to engage in practices involving alcohol and cigarettes, while females are discouraged from such behaviours as they often bear the responsibility for caring for their family members. However, Polanska (2022) conducted a study in five Eastern European countries and found contrasting results. In the Czech Republic and Slovenia, females showed higher susceptibility to tobacco expenditure than males, while the results in Romania supported the present study’s findings.

Furthermore, marital status significantly influences tobacco and alcohol expenditures in Tanzania. Married heads of households tend to spend less on tobacco compared to unmarried individuals. This is likely due to the increased responsibilities of married household heads in caring for their families, leading to a larger allocation of their budget towards necessary goods and services rather than luxury items like alcohol. In contrast, unmarried individuals may have fewer responsibilities and perceive lower health risks associated with tobacco and alcohol consumption. These findings align with studies conducted by Aksoy et al. (2019) and İğdeli (2021) in Turkey, which reported similar results. This observation is also consistent with the random utility theory, where households aim to maximise their utility within the constraints of their budget.

Additionally, the results reveal that rural households in Tanzania tend to exhibit higher expenditure on tobacco and alcohol compared to urban households. This can be explained by the persistence of strong cultural and traditional practices in rural areas, including the prevalence of local brews during various ceremonies. Limited interaction with other communities and lower economic activity levels may also contribute to higher tobacco and alcohol expenditures in rural areas. This finding is supported by Carney et al. (2021), who identified idleness as a contributing factor to harmful alcohol and cigarette consumption among youth in Tanzania. Furthermore, it corroborates with the findings of Mwaisakila and Adrison (2021), who reported a higher prevalence of tobacco and alcohol expenditures in rural areas compared to urban areas.

The results presented in Table 6 demonstrate the effects of tobacco and alcohol expenditures on food poverty. Married household heads who spend on tobacco and alcohol have a higher probability of experiencing food poverty compared to unmarried heads who engage in such expenditures. This can be attributed to the fact that married heads bear more responsibilities in providing essential necessities, including food, for their households. The allocation of household budget towards tobacco and alcohol reduces the resources available for purchasing food, thereby increasing the likelihood of food poverty (Bjerregaard et al., 2021; Nyakutsikwa et al., 2021).

Similarly, an increase in household size contributes to a higher vulnerability to food poverty. Larger households imply fewer resources available to be shared among household members, particularly when there are fewer working-age members compared to dependents. This is consistent with the population structure in Tanzania, where a large portion of the population consists of dependents (MoFP et al., 2020). This finding aligns with the results of Kidane et al. (2015) in Tanzania, who found that the impact of smoking on poverty was greater in larger households compared to households with fewer members.

Furthermore, the results indicate that the incidence of food poverty decreases as the education level of the household head increases. Education enables households to make informed decisions regarding resource allocation, helping them cope with food demand and supply shocks resulting from market dynamics (Ward, 2016). Education also enhances household earnings through increased productivity and facilitates the development of coping strategies for potential future uncertainties (Zhang et al., 2018). Interestingly, the findings suggest that rural households have a lower probability of experiencing food poverty compared to urban households, which contrasts with the findings of the Tanzania National Bureau of Statistics (MoFP et al., 2020).

Results presented in Table 7 show the effect of tobacco and alcohol expenditures on a household’s healthcare expenditure. Expenditure on tobacco and alcohol significantly increases a household’s healthcare expenditure. This is explained by the reason that these goods are the major cause of cardiovascular diseases including lung cancer and heart attack which in turn increases medication expenses in search for a cure (Goodchild et al., 2018; James et al., 2022). Expenditure on tobacco and alcohol also reduces the income that would be allocated in promoting a healthy diet leading to the occurrence of malnutrition, malnutrition-related diseases and thereby leading to an increase in healthcare expenditure. Similarly, an increase in household size increases healthcare expenditure since more resources are needed to meet the increasing demand for health improvement. Rural dwellers had lower healthcare expenditures since they lacked access to information on healthcare improvement and healthcare services due to insufficient health facilities and the use of local medicines that are not accounted for in the calculation of the total household’s healthcare expenditure.

Conclusion and Policy Implications

The objective of this study was to analyse the factors influencing tobacco and alcohol expenditures and their impact on food poverty and healthcare costs in Tanzania, utilising data from the 2017/2018 household budget survey (HBS). Descriptive statistics and the Tobit model were employed for data analysis. The descriptive statistics revealed that tobacco and alcohol expenditures accounted for approximately 0.6% of total household expenditure, with a significant portion allocated to food (62.5%) and non-food items (37.5%). Several determinants were identified as significant factors affecting tobacco and alcohol expenditures in Tanzania, including the age of the household head, gender, marital status and the rural or urban location of the household. Specifically, younger household heads below 59.7 years of age, males, unmarried individuals and those residing in rural areas exhibited a positive association with higher tobacco and alcohol expenditures. Furthermore, the study found that both tobacco and alcohol expenditures and household size had a positive effect on food poverty, while education had a negative effect. Specifically, married household heads who spent on tobacco and alcohol had a slightly higher probability of experiencing food poverty compared to unmarried household heads, all else being equal. Tobacco and alcohol expenditures increase household’s healthcare expenditure by 0.01%.

The results from this study have three implications. First, the policy interventions targeted at addressing tobacco and alcohol consumption should put more emphasis on youths aged 15–35 years and young adults below 59.7 years who are particularly more susceptible to smoking and alcohol consumption. Additionally, interventions should focus on rural areas, males and unmarried individuals to tackle the issue of tobacco and alcohol expenditures. Second, to mitigate the impact of tobacco and alcohol expenditures on food poverty, strategies should involve married household heads, family planning initiatives and education on effective food budget planning and forecasting to alleviate food poverty. Third, to mitigate the negative effect of tobacco and alcohol expenditures on healthcare expenditure, interventions that aim at reducing/combating expenditure on these goods including the provision of health education and family planning are crucial. However, due to data limitations, this study did not disentangle the effect of tobacco from that of alcohol on healthcare costs and food poverty and thus leaving gaps for further research.

Appendix A

Appendix A provides a mathematical derivations of how the turning point of age of household head (59.7 years) presented in the first paragraph of page 19 was obtained.

The turning point of the household age (59.7 years) was calculated as follows:

h h exp A T = 158.85 Age - 1.33 Age 2

(A1)

First order condition (F.O.C),

\frac{\partial h h e x p A T}{\partial Age} = 158.85 - 2 (1.33) Age

(A2)

= 158.85 − 2.66 Age

But, the first-order condition implies that $\frac{\partial h h e x p A T}{\partial A g e}$ = 0

2.66 Age = 158.85

Age = 59.718 ≈ 59.7 years, and thus hh exp AT = 158.85(59.7) − 1.33(59.7)2 = 4,743.11 TZS

The second-order condition for a maximum point $\frac{\partial h h e x p A T}{\partial A g e}$ < 0

From dhhexpAT/dAge = 158.85 − 2(1.33) Age,

(d² hhexpAT)/(d[(Age)]²) = −2(1.33) = −2.66 < 0

Hence, 59.7 years is the maximum turning point with a maximum monthly household expenditure on tobacco and alcohol of about 4,743.11 TZS.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon request from https://www.nbs.go.tz/index.php/en/

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

ORCID iD

Furaha N. Rashid

References

Ajayi

A. I.

, Owolabi

E. O.

, & Olajire

O. O.

(2019). Alcohol use among Nigerian university students: Prevalence, correlates and frequency of use. BMC Public Health , 19, 1–11.

Aksoy

, Bilgic

, Yen

S. T.

, & Urak

(2019). Determinants of household tobacco and alcohol expenditures in Turkey. Journal of Family and Economic Issues , 40(4), 609–622.

Alkan

Ö.

, & Abar

(2020). Determination of factors influencing tobacco consumption in Turkey using categorical data analyses. Archives of Environmental & Occupational Health , 75(1), 27–35.

Amalia

, Cadogan

S. L.

, Prabandari

Y. S.

, & Filippidis

F. T.

(2019). Socio-demographic inequalities in cigarette smoking in Indonesia, 2007 to 2014. Preventive Medicine , 123, 27–33.

Anderson

C. L.

, Mons

, & Winkler

(2020). Global progress in tobacco control: The question of policy compliance. Global Health Action , 13(1), 1844977.

Asante

K. O.

, & Kugbey

(2019). Alcohol use by school-going adolescents in Ghana: Prevalence and correlates. Mental Health & Prevention , 13, 75–81.

H. A.

, de Mey

, Thoron

, & Demont

(2019). Inclusiveness of contract farming along the vertical coordination continuum: Evidence from the Vietnamese rice sector. Land Use Policy , 87, 104050.

Bjerregaard

, Olesen

, & Larsen

C. V. L.

(2021). Association of food insecurity with dietary patterns and expenditure on food, tobacco and alcohol amongst indigenous Inuit in Greenland: Results from a population health survey. Land Use Policy , 87, 104050.

Boachie

M. K.

, & Ross

(2020). Determinants of smoking intensity in South Africa: Evidence from township communities. Preventive Medicine Reports , 19, 101099.

10.

Carney

, Kaaya

, Kajula

, Ibitoye

, Marwerwe

, & Sommer

(2020). ‘Most of the youth are drinking because they have nothing to do’: How idle time facilitates adolescent alcohol use in urban Tanzania. Journal of Child & Adolescent Substance Abuse , 29(2), 129–142.

11.

Çebi-Karaaslan

(2022). Assessment of sociodemographic indicators of tobacco expenditure: An application of the censored regression model. Journal of Substance Use , 28(3), 441–446.

12.

Cetinkaya

, & Marquez

P. V.

(2017). Tobacco taxation in Turkey . Retrieved, 20 June 2022, from https://www.worldbank.org

13.

Cham

, Scholes

, Groce

N. E.

, & Mindell

J. S.

(2019). Prevalence and predictors of smoking among Gambian men: A cross-sectional national who step survey. International Journal of Environmental Research and Public Health , 16(23), 4719.

14.

Chelwa

, & van Walbeek

(2019). Does cigarette demand respond to price increases in Uganda? Price elasticity estimates using the Uganda National Panel Survey and Deaton’s method. BMJ Open , 9(3), e026150.

15.

Dikmen

F. H.

(2021). A micro-data analysis of factors affecting alcohol consumption in Turkey. In

Tarıq

, Dikmen

F. H.

& Köktaş

A. M.

(Eds.), Economic, social and business issues evidence from developing world . IJOPEC Publication.

16.

, & Florkowski

(2016). Polish household consumption of tobacco and alcohol: A censored system (No. 1376-2016-109715). World Health Organization.

17.

Goodchild

, Nargis

, & d’Espaignet

E. T.

(2018). Global economic cost of smoking-attributable diseases. Tobacco Control , 27(1), 58–64.

18.

Gray

(2018). Mixed up: Alcohol and society. The Lancet Psychiatry , 5(12), 970.

19.

Gujarati

D. N.

, & Porter

D. C.

(2009). Basic econometrics (Fifth edition). McGraw-hill.

20.

Hahn

E. D.

& Soyer

(2005). Probit and logit models: Differences in the multivariate realm. The Journal of the Royal Statistical Society, Series B , 67, 1–12.

21.

İğdeli

(2021). Regional determinants of tobacco and alcohol expenditures. In Economic, social and business issues evidence from developing world , (1st ed, pp. 247–268). IJOPEC Publication Limited (No.2021/02).

22.

Immurana

, Boachie

M. K.

, & Iddrisu

A. A.

(2021). The effects of tobacco taxation and pricing on the prevalence of smoking in Africa. Global Health Research and Policy , 6(1), 14.

23.

Isaranuwatchai

, Archer

R. A.

, Teerawattananon

, & Culyer

(2019). Non-communicable disease prevention: Best buys, wasted buys and contestable buys (p. 220). Open Book Publishers.

24.

James

P. B.

, Bah

A. J.

, Kabba

J. A.

, Kassim

S. A.

, & Dalinjong

P. A.

(2022). Prevalence and correlates of current tobacco use and non-user susceptibility to using tobacco products among school-going adolescents in 22 African countries: A secondary analysis of the 2013–2018 global youth tobacco surveys. Archives of Public Health , 80(1), 1–15.

25.

Jha

, Narayanan

B. G.

, Wadhwa

, & Tesche

(2020). Economic and environmental effects of reduction in smoking prevalence in Tanzania. Tobacco Control , 29(1), 24–28.

26.

Jumrani

, & Birthal

P. S.

(2017). Does consumption of tobacco and alcohol affect household food security? Evidence from rural India. Food Security , 9(2), 255–279.

27.

Kidane

, Mduma

, Alexis

N. A. H. O.

, & Hu

T. W.

(2015). Impact of smoking on nutrition and the food poverty level in Tanzania. Journal of Poverty Alleviation and International Development , 6(1), 131.

28.

Leung

, Law

, Cooke

, & Leatherdale

(2019). Exploring and visualizing the small-area-level socioeconomic factors, alcohol availability and built environment influences of alcohol expenditure for the city of Toronto: A spatial analysis approach. Chronic Diseases and Injuries in Canada , 39(1). https://doi.org/10.24095/hpcdp.39.1.02

29.

McFadden

(2019). Econometric models of probabilistic choice, structural analysis of discrete data with econometric applications (pp. 198–272). The MIT Press.

30.

Michas

, Magriplis

, Micha

, Chourdakis

, Koutelidakis

, Dimitriadis

, & Roma

(2022). Prevalence and factors associated with smoking in a nationally representative sample of Greek adults: The Hellenic National Nutrition and Health Survey (HNNHS). Hellenic Journal of Cardiology , 67, 19–27.

31.

Ministry of Finance and Planning (MoFP), National Bureau of Statistics (NBS), & World Bank. (2020). Tanzania mainland household Budget Survey 2017/18 [Final report]. Dodoma, Tanzania MOFP-PED, NBS and Washington DC USA, and the WB.

32.

Mwaisakila

S. R.

, & Adrison

(2021). The determinants of household spending on tobacco and alcohol in Tanzania. East African Journal of Business and Economics , 3(1), 173–186.

33.

NBS. (2018). 2018 Tanzania global adult tobacco survey. Country report . Retrieved, 18 June 2022, from https://www.nbs.go.tz/index.php/en/census-surveys/health-statistics/tobacco-survey-report/517-2018-tanzania-global-adult-tobacco-survey-key-findings-report

34.

National Bureau of Statistics (NBS). (2020). Tanzania in figures.. Retrieved from https://www.nbs.go.tz

35.

Nyagwachi

A. O.

, Chelwa

, & van Walbeek

(2020). The effect of tobacco- and alcohol-control policies on household spending patterns in Kenya: An approach using matched difference in differences. Social Science & Medicine , 256, 113029.

36.

Nyakutsikwa

, Britton

, & Langley

(2021). The effect of tobacco and alcohol consumption on poverty in the United Kingdom. Addiction , 116(1), 150–158.

37.

Ozughalu

U. M.

(2016). Relationship between household food poverty and vulnerability to food poverty: Evidence from Nigeria. Social Indicators Research , 125(2), 567–587.

38.

Perera

K. M. N.

, Guruge

G. D.

, & Jayawardana

P. L.

(2017). Household expenditure on tobacco consumption in a poverty-stricken rural district in Sri Lanka. Asia-Pacific Journal of Public Health , 29(2), 140–148.

39.

Polanska

, Znyk

, & Kaleta

(2022). Susceptibility to tobacco use and associated factors among youth in five central and eastern European countries. BMC Public Health , 22(1), 1–11.

40.

Sommer

, Kaaya

, Kajula

, Marwerwe

, Hamisi

, & Parker

(2021). Social and structural determinants of youth alcohol use in Tanzania: The role of gender, social vulnerability and stigma. Global Public Health , 16(1), 75–87.

41.

Tan

A. K.

, Yen

S. T.

, & Hasan

A. R.

(2016). Cigarette and alcohol expenditures in Malaysia: Implications for anti-smoking and drinking policies. Asian Economic Journal , 30(4), 401–421.

42.

Thibaut

, Eakins

, Willem

, & Scheerder

(2020). Financial barriers for sports consumption: The dynamics of the income–expenditure relation. Sport, Business and Management: An International Journal , 10(3), 245–261.

43.

Ward

P. S.

(2016). Transient poverty, poverty dynamics, and vulnerability to poverty: An empirical analysis using a balanced panel from rural China. World Development , 78, 541–553.

44.

WHO-AFRO. (2021). Tobacco control fact sheet 2021 . Retrieved, 20 June 2022, from World Wide Web: https://www.afro.who.int/health-topics/tobacco-control

45.

World Bank. (2020). Prevalence of current tobacco use (% adults) . Retrieved, 17 June 2022, from World Wide Web: https://data.worldbank.org/indicator/SH.PRV.SMOK

46.

World Health Organization (WHO). (2019). Tobacco . https://www.who.int/news-room/fact-sheet/detail/tobacco

47.

Zhang

, Luo

, & Robinson

(2018). Reducing food poverty and vulnerability among the rural elderly with chronic diseases: The role of the new rural pension scheme in China. International Journal of Environmental Research and Public Health , 15(6), 1253.