Household Head Smoking Behavior and Household Food Insecurity in South Africa: Evidence from National Income Dynamics Study Survey

Abstract

This study examines how the smoking behavior of household head impacts the food insecurity status of household in South Africa, with a focus on smoking behavior transition. The study used the fifth wave of the 2017 National Income Dynamics Study survey. Additionally, the smoking transition variable was computed using smoking status in waves 3 and 5 of the survey. Foster-Greer-Thorbecke food insecurity index, descriptive statistics, and logit regression techniques were deployed in the study. Surprisingly, the analysis shows that households headed by current smokers have a lower prevalence of food insecurity (compared to households whose heads are nonsmokers). Similarly, household food insecurity prevalence is lesser for homes whose head initiated smoking behavior or remained a smoker compared to those whose head remained a nonsmoker or transition to nonsmoking. Nevertheless, the findings from the applied logit regression suggested that household food insecurity significantly increases when a household head is a current smoker and a consistent smoker. The finding indicates that household head smoking behavior does contribute to household food insecurity in South Africa. The need to address food insecurity at the household level is a compelling argument for tobacco smoking cessation or control in addition to health consequence concerns.

1. Introduction

The importance of food for people justifies the inclusion of food security and the necessitation of the pursuit of same as a development outcome by governments of various countries and several development partners. It is known that food alongside clothing and shelter are the three basic needs of mankind [1, 2, 3]. Food plays an essential role in human sustenance and survival [4]. It is needed to build, rejuvenate, and where necessary replace worn-out body cells and tissues, to produce needed energy for warmth and physical activities and to promote the prevention of diseases, support the recovery from sickness, and aid good health in general [5, 6]. According to FAO [7], food security is the state in which everyone consistently has the means to access enough safe and nutritious food to meet their dietary requirements and preferences, ensuring a hale, hearty, and healthful life. Without a doubt, the absence of hunger and availability and at least ease of accessing nutritionally adequate, acceptable, and safe foods without theft, scavenging, use of coping strategies, or depending on crisis-related food supplies are the several dynamic reflections of a food secure situation.

Statistics from FAO et al. [8] shows that 830.2 million people (i.e., about 10.7% of the world population) experienced severe level of food insecurity between 2019 and 2021. The number of persons that were moderately as well as severely food insecure rose from about 1609.1 million people between 2014 and 2016 to about 2187.4 people between 2019 and 2021. This number also varies by region with Asia (1109.5 million) and Africa (743.5 million) accounting for the largest proportion in 2020. In South Africa, various statistical reports have provided evidence of the reality of food insecurity in the country. For instance, FAO et al. [8] reported that about 6.9% of the South African population are considered undernourished between 2019 and 2021. Likewise, the occurrence of moderate or chronic food insecurity within the entire population is 19%. Equally, with a Global Index score of 12.9 in 2022, South Africa was ranked to be having moderate hunger severity [9]. According to Statistics South Africa (Stat SA) [10], the proportion of the population who were classified as severely and moderately food insecure were estimated at 17.3% (i.e., 10.1 million South Africans) in South Africa in 2019. These international and national statistics underscore the undeniability of the reality of food insecurity in the country and justify the inclusion of food insecurity goal in the sustainable development agenda 2030. It is therefore a priority to conduct studies that identify the factors that determine household food insecurity. An important instrument to consider is one related to lifestyle risk factors.

Smoking as an unhealthy behavior or lifestyle is one of the identified direct or indirect contributors to food insecurity. Conceptually, tobacco smoking is linked with household food stock, food access, and utilization through crowding out of food expenditure, unequal spending of household income on cigarettes, and financial strain. Furthermore, the contribution of smoking to depreciated health leads to limited labor force participation, lowering of productivity at work due to ill health, and reduction of wage income, which eventually inhibits the affordability and availability of needed quality and quantity of food at the household and individual levels, [11]. WHO [12] estimated that about 1.3 billion persons are estimated to be users of tobacco products, out of which 80% live in countries that are categorized as middle- and low-income. According to Global Adult Tobacco Survey (GATS) [13], 12.7 million South African adults (i.e., 29.4% of adult population) are current users of both smoked and smokeless tobacco product. Of these statistics, 41.7% are men, while 17.9% are women. GATS [13] also reported that the median monthly expenditure on manufactured cigarettes was R263 (US$ 17.79).¹ Apparently, this smoking behavior prevalence raises concern of it being a potential inhibitor of the achievement of household food security in South Africa. Therefore, given that food insecurity and smoking are prevalent in South Africa, this study’s objective is to empirically investigate if and to what extent smoking behavior of household head can predict household food insecurity. The objective takes the current smoking status and smoking behavior transition patterns into consideration. The other sections of the article are organized as follows. Sections 2 and 3 provide a brief review of related previous works and methodology, respectively. Section 4 reports the outcome, discussion, and inferences regarding the analyses that were conducted. Conclusion and suggested policy directions are presented lastly in Section 5.

2. Brief literature review: empirics

Empirically, studies that have incorporated smoking behavior as an independent variable while investigating household food insecurity are very few. Moreover, available evidence on the relationships between the two issues from the literature is mixed. For example, Kang and Cho [14] demonstrated that individuals who had ever used any tobacco products, whether current or former users, had increased odds of experiencing food insecurity in comparison to those who had never used any tobacco products. Likewise, the findings from different studies [15, 16, 17, 18, 19, 20, 21, 22, 23] are consistent with Kang and Cho [14] in terms of association between smoking behavior and food security. Regarding health behavior transition, only Berry et al. [24] have investigated the impact of smoking transition on food security. Evidence from the modified Poisson regression utilized by the study showed that compared to those who have smoked and recently quit smoking, those who maintain smoking habit have increased risk (that is about twice as much of recent quitters) of household food insecurity.

In South Africa, a set of different research articles [25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35] have identified several demographic and socioeconomic factors (such as household head characteristics [gender, age, relationship or marital status, status in the labor market, education], number of household members, total income and income category of the household, geographical location or place of residence) that influence well-being in terms of food security. However, the potential link between health behavior and food insecurity have been unaccounted for in these studies, and it is observed as a gap in the literature in the case of South Africa. Specifically, the quantification of smoking behavior to welfare (food insecurity) impacts has been overlooked.

3. Research and methodology

3.1 Data

The study is a quantitative study that utilizes a cross-sectional research design. It uses the fifth wave of the 2017 National Income Dynamics Study (NIDS) survey, which was accessed and acquired through [36, 37] website. NIDS is a panel survey of individuals and households that reside in South Africa, conducted every 2 years from 2008 to 2017 [36, 37]. NIDS provided information on demographic and socioeconomic characteristics of individual and household including gender, level of education attained, food security, health behavior, and household size [38]. This study uses information of household heads and households from the round 5 of NIDS survey that was conducted in 2017/2018. The fifth wave of the NIDS survey was used because it is the most recent data available in South Africa that captures the key variables of interest. Furthermore, to construct a transition matrix of the targeted variable, NIDS wave 3 survey data of the targeted variable was included. The selection criteria used in this study are household heads and households in wave 5 [36, 37] whose heads were interviewed in wave 3 [36, 37] of the NIDS surveys. Consequently, a sample of 7676 households was used in this study. A detailed narrative of the dataset that was used in this article is obtainable at www.nids.uct.ac.za.

3.2 Analytical techniques

3.2.1 Foster-Greer and Thorbecke food insecurity (FGT-FI) index

FGT-FI index was used in profiling household food insecurity in this study. The FGT-FI index was adapted from Foster et al. [39], an index that is primarily used in the computation of the incidence, depth, and severity of poverty but has been extended to profile food insecurity situation in geographical areas (see [4, 40, 41]). The original FGT index is a generalized poverty index that utilizes a poverty threshold and considers the inequality among the poor (or food insecure) and allows one to vary the amount of weight on income (or expenditure) levels when analyzing poverty situation [42]. As a decomposable measure of poverty, FGT index links overall poverty with the poverty levels in population subgroups. In the context of this study, household food (in) security threshold was based on the two-thirds of the mean per capita household food expenditure (MPCHFE) [4, 40, 41]. Households whose MPCHFE is below the food security threshold were categorized as food insecure, while those whose MPCHFE is above the food security line were considered as food secure households. FGT-FI index for this study is represented by the following Eq. (1):

$$\mathrm{FGT}-\mathrm{FI}\mathrm{\alpha }\mathbf{=}\frac{\mathbf{1}}{\mathit{N}}\sum _{\mathit{i}\mathbf{=}\mathbf{1}}^{\mathit{q}}{\left(\frac{\mathit{z}\mathbf{-}{\mathit{y}}_{\mathit{i}}}{\mathit{z}}\right)}^{\mathbf{\alpha }}$$E1

where N = total households’ number under consideration, z = food security line, q = number of households under the food security threshold, ${\mathit{y}}_{\mathit{i}}$ = per capita monthly food expenditure of the ith household, and $\mathit{z}\mathit{-}{\mathit{y}}_{\mathit{i}}$ is the food expenditure shortfall of the ith household. The FGT-FI measure “α” take the values of 0,1,2, representing food insecurity headcount, food insecurity depth, and food insecurity severity, respectively.

3.2.2 Logit regression

The study used binary logistic regression in estimating the relationship between the variables of interest. Logit regression estimation technique was used because the dependent variable has a binary outcome. To avoid bias related to bivariate modeling, other possible determinants of household food insecurity were included as control variables in the model. Hence, the specification of binary logistic regression model for this study is stated in Eq. (2):

$${\mathit{FI}}_{\mathit{i}}={\varnothing }_{\mathbf{0}}+{\varnothing }_{\mathbf{1}}{\mathit{SB}}_{\mathbf{1}}+\sum {\varnothing }_{\mathit{i}}{\mathit{X}}_{\mathit{i}}+{\mathit{e}}_{\mathit{t}}$$E2

where the dependent variable FI is household food insecurity status, ${\mathit{\varnothing }}_{\mathbf{0}}$ = the constant term, SB represents household head smoking behavior (captured by both current smoking status and smoking transition pattern “S”), ${\mathit{\varnothing }}_{\mathbf{1}}$ = the vector of coefficients, ${\mathit{X}}_{\mathit{i}}$ =vector of explanatory variables, and ${\mathit{e}}_{\mathit{t}}$is the error term.

All the analyses in this study were conducted using STATA 15 software. Table 1 presents the description of the variables in the model, both the dependent variable and the independent variables.

Variable	Level	Measurement	Reference
Dependent Variable
FI	Food Insecurity Status	1 if Food Insecure; 0 otherwise
	Independent Variables
SB	a) Current Smoking Status	1 if a current smoker; 0 otherwise	Amrullah et al. [17]
	b) Smoking Transition Pattern (S)	1 if S(0,0)= Not a smoker in both 2012 and 2017; 0 otherwise; 1 if S(0,1) = Not a smoker in 2012 but a smoker in 2017; 0 otherwise; 1 if S(1,0)= Was a smoker in 2012 but not a smoker in 2017; 0 otherwise; 1 if S(1,1)= Was a smoker in both 2012 and 2017; 0 otherwise	[14, 24]
X1	Gender of head	1 if male, 0 otherwise
X2	Age of head	Age of household head	Azwardi et al. [43]
X3	Marital Status of Head	1 if married, 0 otherwise
X4	Education of Head	1 if No education, 0 Otherwise; 1 if Primary, 0 Otherwise; 1 if Secondary, 0 Otherwise; 1 if Matric, 0 Otherwise; 1 if Certificate and diploma, 0 Otherwise; 1 if higher degree, 0 Otherwise	[44, 45]
X5	Employment Status	1 if employed, 0 otherwise	[45, 46]
X6	Household Size	Number of people in the household	[47, 48]
X7	Health Insurance Coverage	1 if yes, 0 otherwise	[21, 49]
X8	Household Income Quantile	1 if Quantile 1, 0 Otherwise; 1 if Quantile 2, 0 Otherwise; 1 if Quantile 3, 0 Otherwise; 1 if Quantile 4, 0 Otherwise	[18, 21]
X9	Population Group	1 if African, 0 Otherwise	Jonah and May [29]
X10	Place of Residence	1 if Urban, 0 Otherwise	[17, 50]
X11	Province	1 if WC, 0 Otherwise; 1 if EC, 0 Otherwise; 1 if NC, 0 Otherwise; 1 if FS, 0 Otherwise; 1 if KZN, 0 Otherwise; 1 if NW, 0 Otherwise; 1 if G, 0 Otherwise; 1 if Mp, 0 Otherwise; 1 if L, 0 Otherwise

Table 1.

Summary description of variables in the model.

Note 1: S_(0,0) = not a smoker in both 2012 and 2017, 0 otherwise (consistent nonsmoker); S_(0,1) = not a smoker in 2012 but a smoker in 2017, 0 otherwise (initiate); S_(1,0) = was a smoker in 2012 but not a smoker in 2017, 0 otherwise (quitters); S_(1,1) = was a smoker in both 2012 and 2017, 0 otherwise (consistent smoker).

Note 2: WC, EC, NC, FS, KZN, NW, G, Mp, and L stand for Western cape, eastern cape, northern cape, Free State, KwaZulu Natal, north-west, Gauteng, Mpumalanga, and Limpopo, respectively.

4. Results and discussion

4.1 Analysis of household food insecurity

4.1.1 FGT-FI index

As earlier mentioned, it is needed to determine the MPCHFE and the 2/3 of MPCHFE when profiling household food security. In this study, the computed MPCHFE is R498.61 (US$ 37.69) as shown in the Table 2. Also, the calculated 2/3 of MPCHFE, which is the food insecurity (FI) threshold used in this study, is R332.41 (US$ 25.13). The result of the FGT-FI index analysis that was used to assess the food insecurity situation in the study area is presented in Table 2. The food insecurity index analysis indicated that household food insecurity incidence and gap were 66.1 and 30%, respectively. The results imply that 66.1% of the households can be categorized as food insecure and that for food-insecure households to be food-secure, they will need to increase their current food spending by about 30%.

FGT Food insecurity (FGT-FI) indices	Value
FGT-FI0 (incidence of food insecurity)	0.661
FGT-FI1 (depth of food insecurity)	0.303
FGT-FI2 (severity of food insecurity)	0.170
MPCHFE	R498.61 (US$ 37.69)*
Food insecurity (FI) Threshold (i.e., 2/3 of MPCHFE)	R332.41 (US$ 25.13)*

Table 2.

FGT-FI index on extent of household food insecurity.

^*

Exchange Rate in 2018: US$ 1 = Rands 13.23.

Note: “R” means Rand.

Source: Computed by Author.

4.1.2 Distribution of food insecurity indices based on household head smoking behavior

In line with evidence from literature suggesting that socioeconomic features of the head of a household influence the well-being of households [51, 52], the disaggregation of the FGT-FI index by the smoking status of household head and household’s and household head’s socioeconomic and demographic characteristics are analyzed, and they are presented in Table 3.

Variable	Level	Food Insecurity Indices
		FGT-FI0	FGT-FI1	FGT-FI2
SB: Current Smoking Status	No	0.676	0.311	0.173
	Yes	0.584	0.267	0.154
SB: Smoking Transition	S(0,0)	0.677	0.311	0.173
	S(0,1)	0.576	0.270	0.159
	S(1,0)	0.666	0.296	0.163
	S(1,1)	0.589	0.265	0.151

Table 3.

Decomposition of FGT-FI indices by household head smoking behavior.

It needs to be known that this profiling does not imply a cause or effect between variables of interest. Table 3 reveals that in terms of smoking status, the reported food insecurity incidence of 58.4% is lower among households whose heads have a current smoking status than the 67.6% found among households whose heads do not smoke currently. Also, the proportion of food expenditure required to lift food insecure households whose head does not smoke currently out of food insecurity situation is higher at 31% than that of their counterpart that requires only a 22% increase in food expenditure. It is surprising to observe that households whose heads maintained a nonsmoking behavior status or transitioned out of smoking had a higher food insecurity incidence compared to those whose heads initiated a smoking behavior or remained a smoker.

4.1.3 Distribution of food insecurity indices based on households’ socioeconomic and demographic characteristics

Table 4 shows that households headed by a female had a higher food insecurity incidence at 73% compared to male-headed households (53%). Also, the proportion of food expenditure required to lift female-headed food insecure households out of the food insecurity situation is higher at 34% than that of male-headed households that require only a 23% increase in food expenditure. As seen in Table 3, the incidence of food insecurity increased with the age of the household head. It was highest (75%) for households headed by persons of 61 years and greater. Food insecurity depth and severity were highest at 35.1 and 19.7%, respectively, among households headed by individuals greater than 60 years in age, which corresponds with [4, 53]. Contrarily, households headed by people under the age of 35 years were the least affected by food insecurity.

Variable	Level	Food Insecurity Indices
		FGT-FI0	FGT-FI1	FGT-FI2
Gender	Female	0.730	0.344	0.195
	Male	0.533	0.228	0.124
Age of Head (Year)	22–34 years	0.522	0.224	0.120
	35–64 years	0.675	0.312	0.176
	> 64 years	0.748	0.351	0.197
Marital Status	Married	0.618	0.267	0.146
	Otherwise	0.688	0.325	0.184
Education Attainment	No Schooling	0.851	0.431	0.254
	Primary	0.802	0.391	0.226
	Secondary	0.652	0.285	0.154
	Matric	0.447	0.170	0.085
	Certificate and Diploma	0.426	0.164	0.085
	Degree	0.199	0.061	0.024
Place of Residence	Rural	0.774	0.369	0.210
	Urban	0.544	0.235	0.128
Province	WC	0.473	0.185	0.098
	EC	0.705	0.323	0.181
	NC	0.589	0.272	0.157
	FS	0.607	0.266	0.142
	KZN	0.775	0.369	0.209
	NW	0.679	0.352	0.211
	Gauteng	0.492	0.204	0.109
	Mpumalanga	0.685	0.294	0.155
	Limpopo	0.722	0.345	0.198
Population Group	African	0.704	0.327	0.184
	Non-African	0.446	0.183	0.099
Health Insurance Coverage	No	0.700	0.325	0.183
	Yes	0.254	0.087	0.032
Employment Status	No	0.759	0.365	0.209
	Yes	0.254	0.221	0.117
Household Size	1–5	0.479	0.177	0.086
	6–10	0.857	0.425	0.245
	>10	0.987	0.589	0.384
Household Income Group	Quartile 1	0.742	0.363	0.211
	Quartile 2	0.721	0.340	0.192
	Quartile 3	0.705	0.328	0.185
	Quartile 4	0.520	0.212	0.111

Table 4.

Decomposition of FGT-FI indices by household socioeconomic and demographic characteristics.

Source: Computed by Author.

Table 4 further shows that the prevalence, depth, and inequality of food insecurity are higher for households residing in the Eastern Cape, KwaZulu Natal, and Limpopo provinces; rural areas; and households with a size of more than five. Likewise, in terms of food insecurity depth particularly in household socioeconomic distributions where household food insecurity incidence is prevalent, for a household to be food secure, there is a need for about 33, 43, 37, 35, and 33% raise in their monthly per capita food expenditure for households headed by a person from the African population group, with no or primary education, not employed, and without medical insurance.

4.2 Logit regression result on the impact of household head smoking behavior on household food insecurity

The estimated logit regression function’s findings (i.e., estimated coefficients, standard error, and probability values) on the relationship between household head smoking behavior and household food insecurity are shown in Table 5. The validity of the estimation results in Table 5 is supported by the diagnostic Pearson χ-2 goodness-of-fit test that was performed. Fit testing helps determine whether a model fits well and can be accepted or otherwise discarded. A statistically insignificant Pearson χ-2 indicates a well-fit model. In this study, the estimated goodness of fit of the models (model 1: Pearson χ2 = 7032.64, p = 0.8368; model 2: Pearson χ2 = 7072.56, p = 0.8235) corresponds with the desirable criteria. The validity of the goodness of fit of the models is further buttressed by the Hosmer-Lemeshow [H-L] χ2 (8) results (model 1: H-L χ2 = 12.03, p = 0.1497; model 2: H-L χ2 = 9.46, p = 0.3047), reported in Table 5. Additionally, to further verify that the estimated models were unbiased and consistent, a check for possible problem of multicollinearity among the explanatory variables was performed using the variance inflation factor (VIF). The mean VIF test value of 2.26 and 2.18 obtained for models 1 and 2 is less than 10; therefore, it can be concluded that there is no multicollinearity problem. The results in Table 5 show that household head smoking behavior, marital status, household size, and household head population group (African) increase the odds of household food insecurity. However, the odds of household food insecurity incidence are decreased by the gender and age of the head, health insurance coverage, income, and living in an urban area, Western Cape, and Gauteng provinces.

Variable	Level	Model 1 (Current Smoking Only)			Model 2 (Smoking Transition Pattern)
Column 1	Column 2	Column 3	Column 4	Column 5	Column 6	Column 7	Column 8
		Odds Ratio	P > |z|	95% CI	Odds Ratio	P > |z|	[95% CI]
SB: Current Smoking	Yes	1.161	0.061***	0.993 – 1.356	—	—	—
	No (Reference)
SB: Smoking Transition	S(0,0) (Reference)	—	—	—	—	—	—
	S(0,1)	—	—	—	0.985	0.896	0.786–1.235
	S(1,0)	—	—	—	1.271	0.093***	0.961 – 1.681
	S(1,1)	—	—	—	1.339	0.002*	1.113 – 1.611
Male	Male	0.484	0.000*	0.427 – 0.549	0.472	0.000*	0.415 – 0.537
	Female (Reference)
Age	Age	0.996	0.094***	0.992 – 1.001	0.996	0.076***	0.992 – 1.000
Marital Status	Marital Status	1.346	0.000*	1.185 – 1.530	1.351	0.000*	1.188 – 1.535
Employed	Employed	0.775	0.000*	0.686 – 0.876	0.778	0.000*	0.689 – 0.880
Health Insurance Coverage	Yes	0.456	0.000*	0.355 – 0.584	0.456	0.000*	0.356 – 0.586
Education Attainment	No Education	Ref
	Primary	1.009	0.933	0.828–1.229	1.005	0.964	0.824–1.224
	Secondary	0.642	0.000*	0.521 – 0.790	0.644	0.000*	0.523 – 0.793
	Matric	0.384	0.000*	0.297 – 0.498	0.386	0.000*	0.298 – 0.500
	Certificate & Diploma	0.396	0.000*	0.310 – 0.505	0.398	0.000*	0.312 – 0.508
	Degree	0.195	0.000*	0.119 – 0.321	0.197	0.000*	0.119 – 0.324
Household Size	NA	9.259	0.000*	7.844 – 10.929	9.231	0.000*	7.821 – 10.896
Income Quartile	1	Ref.	—		—	—
	2	0.775	0.001*	0.670 – 0.896	0.777	0.001*	0.672 – 0.896
	3	0.733	0.000*	0.628 – 0.855	0.732	0.000*	0.627 – 0.854
	4	0.442	0.000*	0.364 – 0.538	0.442	0.000*	0.364 – 0.538
Population Group	African	1.588	0.000*	1.294 – 1.949	1.614	0.000*	1.314 – 1.983
	Non-African (Reference)
Place of Residence	Urban	0.836	0.008*	0.732 – 0.955	0.835	0.008*	0.731 – 0.954
	Rural (Reference)
Province	Western Cape	0.729	0.030**	0.548 – 0.971	0.718	0.023*	0.540 – 0.956
	Eastern Cape	0.897	0.375	0.706–1.140	0.891	0.348	0.701–1.133
	Northern Cape	0.991	0.952	0.7737–1.332	0.991	0.954	0.737–1.333
	Free State	0.950	0.730	0.710–1.272	0.943	0.696	0.705–1.263
	Kwa-Zulu Natal	0.848	0.128	0.686–1.049	0.850	0.133	0.688–1.051
	North-West	0.930	0.612	0.702–1.232	0.932	0.626	0.704–1.236
	Gauteng	0.716	0.009*	0.556 – 0.921	0.716	0.009*	0.556 – 0.921
	Mpumalanga	0.962	0.777	0.734–1.260	0.963	0.787	0.735–1.262
	Limpopo (Reference)	—	—		—	—
	Constant	0.202	0.000*	0.126 – 0.325	0.201	0.000	0.125–0.323
Number of observations		7676			7676
LR chi^2 (29)		2622.79			2631.17
Prob > chi^2		0.0000			0.0000
Log likelihood		−3991.9815			−3987.7919
Pearson chi^2		7032.64			7072.56
Prob > chi^2		0.8368			0.8235
Hosmer-Lemeshow chi^2(8)		12.03			9.46
Prob > chi^2		0.1497			0.3047

Table 5.

Impact of household head smoking behavior on household food insecurity: Binary logistic regression.

^*

= p < 1%.

^**

= p < 5%.

^***

= p < 10%.

Source: Computed by Author.

First, Table 5 shows the odds ratio of this study’s variable of interest; household head smoking behavior in terms of current status (in column 3) is statistically significant (OR = 1.161; p < 0.01). Similarly, the smoking behavior transition (in column 6) is statistically significant for smoking cessation (OR = 1.271; p < 0.10) and smoking maintenance (OR = 1.339; p < 0.01). These results indicate that the odds of households being food insecure are 1.16, 1.271, and 1.34 times higher when the household head is a smoker, quitted smoking, and has a consistent smoking behavior, respectively. This implies that smoking habit of household heads exerts a significant and increasing influence on the likelihood of household food insecurity. That is, having a head that smokes (especially when such head can be said to be a tobacco-smoking addict [continuous smoker]) increases the probability of a household being food insecure. This finding is consistent with several studies [14, 15, 17, 18, 19, 23] that have established an influencing effect of household head smoking behavior on household food security status. As noted earlier in this study, tobacco spending crowds out spendings on food and other necessities. Similarly, possible financial constraints are plausible reasons for the established effect of being a continuous smoker on household food insecurity probability. For the quit smoking result, it can be inferred that when stopping or trying to stop smoking, people may run into problems like nicotine withdrawal or the need for cessation assistance, which can put more financial strain on households. The expenses linked to quitting smoking, such as medication, treatment, or nicotine replacement therapy, could take money away from food purchases, which could worsen food poverty temporarily.

The results of the regression model reveal that the odds ratio for gender is less than one and statistically significant (OR = 0.484; p < 0.01). This result, which is consistent with [19, 54, 55], implies that odds of household food insecurity are predicted to be about 0.427 to 0.484 times lower for male-headed household than those headed by women, when all other variables are kept constant. A plausible reason for this result that is in favor of men is the difference in economic opportunities and assets (e.g., land and productive economic resources). Gender differences in the type of employment that is in favor of men could translate into differences in income needed to meet household expenditures. In addition, woman-led households are susceptible to food insecurity when the dependency ratio in such households is high [56].

Table 5 shows that household head’s age is statistically significant (OR = 0.996; p < 0.10). According to this result, there is a decreasing likelihood of food insecurity in households as household head get older, provided that all other factors remain unchanged. The calculated odds ratio estimates specifically suggest that the probabilities of a home experiencing food insecurity are around 0.996 times lower for every unit or year that the household head’s age increases. This aligns with the a priori statement of this study and with Azwardi et al. [43], which argued that the maturity of the household heads in terms of age affects psychological and financial stability. In a related perspective, Mthethwa and Wale [44] noted that as people grow older, they accumulate assets, experience, and strategies that could reduce their household vulnerability to food insecurity. According to Megbowon [27], the high possibility of a head in urban areas of securing a job, accumulating income over a long period time, and a robust social security program for older people in the country are plausible reasons for this finding. Also, increase in age, which is related to farming experience for rural farming household heads, leads to more output resulting in better profitability and household food security [54]. This result is however contrary to the perspective that aging of a household head leads to limited participation in economic activities and a decrease in income earning potential, thus increasing household food insecurity tendency.

The odd ratio estimated for the employment status (employed) is statistically significant (OR = 0.778; p < 0.01). This means that the odds of household having a food insecure status are predicted to be about 0.78 times lower among households having employed heads than those whose heads are unemployed. This finding is in line with the theoretical expectation of this study where it was argued that being employed is an indicator of economic empowerment, such that it can inhibit or promote access to needed food nutrients. Having an employed household head enables households to maintain higher levels of expenditure that can aid household welfare. This finding is consistent with the results of other studies [45, 46, 57], who all established that a reduced household food insecure status likelihood is linked with employed status (being employed) of the head is, the greater the chances that his or her household will be less food insecure. While unemployment increases resource and welfare deprivations, economic access to needed food for a household and availability of food would be largely guaranteed when there is a source of income, which comes when there is an economic engagement or labor market participation.

Table 5 also shows that the odds ratio of health insurance coverage is statistically significant (OR = 0.456; p < 0.01). This means that the odds of households being food insecure is 0.46 times lower when the household head is covered by medical insurance compared to those whose head is not covered by any medical insurance. This result is clearly in line with the expectations of this study, and it is also consistent with previous evidence [21, 49, 58]. Indeed, medical aid would enable individuals and households to reallocate income away from medical care, thereby leaving more funds toward food consumption. The odd ratios estimated for the education attainment categories of household head are less than one (i.e., secondary: OR = 0.644; Martic: OR = 0.386; Certificate and Diploma: OR = 0.398; Degree: OR = 0.1957) and statistically significant except for primary schooling attainment. The results imply that compared to households whose head has no education, households headed by educated people are less likely to experience food insecurity. It is also striking to observe that the odds of a household being food insecure reduces further as the education of the household head increases. This result is consistent with theory, a priori expectation, and the results found by other authors [44, 45]. This finding supports the importance of education and especially higher education in household well-being. Education provides opportunities for labor supply and employment chances and acquisition of skill and application, which could translate to better income to access household necessities and improve household well-being.

It is also shown in Table 5 that household size’s odd ratio is statistically significant (OR = 9.25; p < 0.01). The results indicate that the odds of household food insecurity are about 9.2 times higher when household size increases by one unit increase. This implies that an increase in households’ size would further exacerbates households’ food insecurity. This result is consistent with a priori expectation of the study and argument put forward by previous studies [47, 48, 59], that a large household is an extra burden on household resources and strain on income because of higher food consumption expenditure. According to Mota et al. [48], per capita food expenditure and availability decline as the size of the family increases, hence increasing the likelihood of food insecurity. This result is further plausible in a situation where the dependency ratio is high in a household; that is, when the number of economically active members of the household is lower than the number of non-economically active due to age, health, and unemployment.

The odds ratios of household income quartiles are less than 1 and statistically significant (i.e., quartile two: (OR = 0.775; p < 0.01), quartile three: (OR = 0.733; p < 0.01), quartile four: (OR = 0.442; p < 0.01)). This implies that compared to households in the quartile one income group, households in quartile two, quartile three, and quartile four are 0.775, 0.733, 0.442 less likely to experience food insecurity. However, as can be seen, the probability of being food insecure is highest for households in the quartile 4 group. This finding is related to a previous study [18]. Furthermore, in this study, the odds ratios for African are 1.588 for model 1 and 1.614 for model 2, and both are statistically significant at p < 0.05. It implies that households whose heads are from African population have a higher likelihood of being food insecure. The finding of this study is consistent with Jonah and May [29], where it was established that Africans had the lowest dietary diversity, and the coloreds were vulnerable when considering the Household Food Insecurity Access Scale (HFIAS) measure of food insecurity [29]. According to Jonah and May [29], the distribution of population group and food insecurity in South Africa reflects socioeconomic status where previously disadvantaged groups (Colored and Africans) have remained vulnerable. High levels of economic and social inequalities in economic opportunities, employment, purchasing power, and education access that are not in favor of Africa are plausible factors that impact the well-being including food insecurity probability of African households in the country.

The odd ratios estimated for the place of residence (urban) are statistically significant (model 1: OR = 0.835; p < 0.01 and model 2: OR = 0.836; p < 0.01). The result proposes that the odds of a household being food insecure are 0.84 times lower for urban-based households than rural area-residing households, when other variables are held constant. It can also be said that rural households and populations are more likely to be food insecure or more vulnerable to food insecurity than urban households. This finding is consistent with [19, 50, 56]. Szabo et al. [50] established that urban households are less likely to experience insufficient dietary variety and insufficient energy intake compared with rural households that are agricultural households. The result explains the advantages of residing in urban areas. In an alternative interpretation, the remoteness of rural areas, poor economic activity, poor agricultural sector performance, limited livelihood opportunities, as well as limited sources of income are plausible reasons for the negative influence of urban-based household variables on household food insecurity. These identified constraints in the rural areas limit the ability of households in the area to purchase needed food for their well-being. Bulawayo et al. [55] also argued that households in urban areas are less likely to be food insecure because they have more households’ assets, and hence, more sustainable coping strategies are possible than in rural areas. Moreover, the result is unsurprising considering the higher prevalence of poverty in rural areas than in urban area in the South African context. Finally, for the provincial parameter, the odds of household food insecurity are lower for all provinces but only statistically significant at p < 0.05 and p < 0.01 for the Western Cape and Gauteng provinces, respectively. This implies that taking all other variables constant, the odds of households’ food insecurity are predicted to be about 0.73 and 0.72 lower for household residing in the Western Cape and Gauteng provinces, respectively.

5. Conclusion

South Africa is one of the countries struggling with the public health challenges of food insecurity and smoking, which have been argued to be interwoven. However, while several confounders have been linked with household food insecurity, the potential relative influence of household head smoking behavior on household food insecurity has been less examined and reported globally and in South Africa. Consequently, this study provides a better understanding of the potential contribution of the smoking behavior of household heads to household food insecurity in South Africa. These findings contribute to literature understudying the dynamic relationship between smoking behavior and household food insecurity. Findings from this study indicated that while food insecurity incidence is highest for households whose head does not smoke currently or remained a nonsmoker, nevertheless, household food insecurity was found to be associated with and can be predicted by current smoking and continual smoking behavior of the household head. There are other factors that influence household food insecurity irrespective of the smoking behavior of the household head as demonstrated in this study. However, considering the interest of this study, the finding indicates that household head smoking behavior does contribute to household food insecurity in South Africa and that the case for tobacco smoking quitting or control goes beyond health outcome reasons to include the need for household food security attainment. Measures to promote quitting smoking behavior need to be promoted and intensified by the government. Therefore, given the confirmed impact of smoking behavior, concurrent (rather than isolated) interventions that can discourage smoking and smoking uptake, encourage early discontinuation of smoking addiction, and promote food security are imperative. Furthermore, intensified food insecurity reduction interventions and policies, through empowerment programs for female-headed and rural households, and toward increasing literacy and employment opportunities especially in the formal sector, are imperative. The impact of being a member of health insurance scheme on food insecurity reduction as found in this study strengthens the urgency of implementation of the national health insurance scheme in South Africa.

Conflict of interest

The authors declare no conflict of interest.