Food expenditure shares and income elasticities in Zimbabwe: Accounting for gender and poverty differences

Abstract

This study analysed food expenditure patterns and income elasticities differentiated by gender of household head and income levels. Data came from 32, 256 households drawn from the Zimbabwe Poverty, Income, Consumption and Expenditure Survey of 2017. Female headed households had higher vegetable and fruit expenditures while male headed households had higher expenditure shares on animal protein sources. Poorer households had poor dietary choices characterized by higher consumption of starchy foods and lower consumption of fruits, eggs, and milk. All income elasticities of all food items and groups were positive, indicating they were normal goods and increasing income will contribute to more diversified diets. Richer groups spend a lower share of their income on food compared to their poorer counterparts. The income elasticities for vegetables, bread, maize, and poultry were high for the poorest households, implying an increase in income would substantially increase their consumption. Policies aimed at increasing household income are vital to improve food consumption. Poor households and male headed households allocated higher proportion of income to food budget than the non-poor and female headed households. Food aid and social protection schemes, for example, income transfers should target very poor households to help them get access to better diets and cushion them from food price increase. Food consumption patterns and income elasticities varied by gender of household head and income group. Therefore, targeted food policies should be formulated based on specific food demand patterns for each group of households.

Keywords:

• expenditure patterns
• income elasticity
• socio-economic groups
• food policy
• Zimbabwe

PUBLIC INTEREST STATEMENT

Understanding gender and poverty differentiated food expenditure patterns and income elasticities are important for guiding specific food policies and interventions for different groups of households. Poor households and male headed households allocated higher proportion of income to food budget than the non-poor and female headed households. Food aid and social protection schemes, for example, income transfers should target very poor households to help them get access to better diets and cushion them from food price increase. Food consumption patterns and income elasticities varied by gender of household head and income group. Therefore, targeted food policies and interventions should be formulated based on specific food demand patterns for each group of households.

1. Introduction

Zimbabwean households rely on a variety of food commodities to meet their dietary and nutritional requirements. In the country, major food commodities include cereals, vegetables, legumes, fruits, meat and poultry, eggs, and milk. For healthy and normal growth, nutritionists recommend certain levels of these food groups to provide balanced diets. However and oftentimes, many households fail to meet these dietary requirements due to limited purchasing power and food shortages emanating from structural constraints (Ansah et al., 2020). On the other hand, a rise in employment from agricultural productivity growth and industrial transformation coupled with the associated improvement in living standards changes food consumption patterns as well as the demand for food quality and safety (Béné et al., 2020; Timmer, 2017). Due to this, demand for the various food groups varies across socioeconomic groups and locations (Ansah et al., 2020; Colen et al., 2018; Headey & Alderman, 2019). In 2017, the mean per capita consumption expenditure for female headed households was US$69.74 and this was statistically higher than the observed US$67.41 for male headed households in Zimbabwe (ZIMSTAT, 2018). Using these estimates, the female headed were wealthier compared to male headed households. Consequently, heterogeneity in food expenditure as a share of disposable income among male and female headed households (Zhang et al., 2018) is nontrivial. This implies that variations in food prices and disposable incomes could affect the various social groupings differently with varying implications for food and nutrition insecurity.

According to the Global Nutrition Report (2020), there is high malnutrition among under-five population in Zimbabwe. The report revealed that the national prevalence of overweight among the under-five was 5.6% in 2015. Furthermore, the prevalence of stunting among the under-five stood at 27.1%, and was greater than the average of 25% in developing countries (Global Nutrition Report, 2020). The Zimbabwean adult population also faces high malnutrition levels with 28.8% of women of reproductive age having anaemia, while 7.6% of adult women had diabetes. In addition, 25.3% of women and 4.7% of men were obese (Global Nutrition Report, 2020). These negative statistics are consequences of the limited purchasing power of the people in society. In Zimbabwe, both the young and old, male and females, urbanites and rural residents as well as poor and rich are faced with the same policy settings. This means that higher food prices affect all the categories of individuals, and sometimes the poor and females are the most affected due to structural rigidities and their confinement to rural areas.

Zimbabwe is one of those countries in Africa for which no or hardly any food demand evidence exists. Hence, periodically, there is a need to update food consumption characteristics data to guide various policy decisions and intervention strategies (Ansah et al., 2020; Nzuma & Sarker, 2010), for example, in the agricultural, health, and food supply chains. People differ in food purchasing behaviour based on socioeconomic and location characteristics. Therefore, it becomes imperative to better understand how gender and various socioeconomic groups differ in their food expenditure patterns (Ansah et al., 2020; Cirera & Masset, 2010).

This article analysed (i) the gender and income differentiated distribution of expenditure shares on various food items and groups and (ii) the distributions of income elasticities of the various food items and groups disaggregated by gender of the head, and poverty status. Our study builds on a few emerging studies on income elasticities associated with different types of food and nutrients in Southern Africa (Chisanga & Zulu-Mbata, 2018; Harris et al., 2019). This study is unique as it used recent evidence to provide gender and poverty differentiated income elasticity estimates vital to improve food demand projections and to design effective agricultural, food and nutrition policies in the country. In particular, such analysis is essential for food policy decisions, since estimated elasticities could help inform and design targeted food and tax policies suitable for the various socioeconomic groups including the poor and vulnerable.

2. Data and methods

2.1. Data and data sources

The quantitative data analysed are from the Zimbabwe Poverty, Income, Consumption, and Expenditure Survey (PICES) of 2017 (ZIMSTAT, 2018), collected by Zimbabwe National Statistics Agency (ZIMSTAT). The survey targeted all private households in all ten provinces in Zimbabwe, excluding hospitals, prisons or military barracks. The survey employed a stratified two-stage sampling design with the first level of stratification corresponding to 62 administrative districts (ZIMSTAT, 2018).

At the first sampling stage, the sample Enumeration Areas (EAs) were selected within each stratum (administrative district) using random systematic sampling with Probability Proportional to Size (PPS) from the ordered list of EAs in the sampling frame. The measure of size for each EA is based on the total number of households identified in the 2012 Population Census. At the second sampling stage, 36 EAs were selected in each district 14 households were selected using systematic random sampling, giving a total of 504 households per district. A total of 2,304 primary sampling units were selected. The final sample consists of 32,256 households. The sampling weights for the data were constructed so that the sample is representative of all household heads in Zimbabwe (ZIMSTAT, 2018).

2.2. Methods

2.2.1. Household consumption expenditure

Household consumption expenditure consisted of the cost incurred by households on goods and services, and this included all expenditure on all food and non-food items.

2.2.2. Food expenditure shares

The expenditure shares of each food item and group (expressed as a percentage) were used as an indicator of consumption (Ansah et al., 2020; Chisanga & Zulu-Mbata, 2018). To address variations in household size, we use per capita household expenditure to standardize food expenditure across households with more or fewer people. A total of 33 food items were identified, as shown in Table 1.

Table 1. Main food items

1	Bread
2	Maize (grain and meal)
3	Small grains (e.g., millet, sorghum)
4	Wheat
5	Rice
6	Other cereals (e.g., biscuits, breakfast cereals, spaghetti, macaroni etc)
7	Beef
8	Poultry
9	Goat and game meat
10	Pork and crocodile
11	Other meat (e.g., mutton, madora, tinned meat, offals)
12	Other meat2 (e.g., purchased outside country, own produce)
13	Fish
14	Milk
15	Other milk products (e.g., skimmed, powdered milk, cheese, cream)
16	Eggs
17	Oils
18	Fruit
19	Other fruits (e.g., tinned and dried fruits)
20	Vegetables
21	Nuts
22	Cassava
23	Potatoes
24	Other tubers (e.g., sweet potatoes, manioc, own produce)
25	Sugar
26	Coffee tea
27	Condiments
28	Takeaway (e.g chips, chicken and chips, hamburger, sandwiches, pie, samosa, pizza)
29	Sit down (e.g., sit-down meals in country and outside)
30	Beverages
31	Alcohol
32	Other beverages (e.g., beverages purchased outside the country, gifts)
33	Tobacco

Each food item was appropriately placed in one of the 14 groups: a) cereals; b) tubers; c) pulses/nuts such as groundnuts; d) vegetables; e) fruits; f) meats; g) fish; h) eggs; i) milk; j) sugars; k) oils; l) condiments; m) beverages such as tea and coffee; n) food away from home and o) alcohol and tobacco.

2.2.3. Income elasticities

To analyze the income-food consumption relationship, the income elasticity of food expenditure was computed. The income elasticity of demand measures the percentage change in food expenditure in response to a one-percentage increase in income at any given set of prices (Babu et al. 2016a). The Engel curve represents the relationship between household expenditure $f{e}_i$ on item and household income, $y_i$. In this function, price is assumed to be independent of $y_i$ and the relationship between $f{e}_i$ and $y_i$ reflects changes in the quantity purchased in response to a change in while holding prices fixed (Babu et al. 2016a; Chisanga & Zulu-Mbata, 2018). The following relationship between expenditure on food items was estimated using Ordinary Least Squares (OLS):

(1) $f{e}_i=n_i{y}_i+{\epsilon }_i$(1)

Where: $f{e}_i$ is the expenditure on food commodity item $i$; $y_i$ is the income for the household measured by the total expenditure on all food and non-food items as a proxy; and is the income elasticity of expenditure on food item $i$. Using the PICES dataset of 2017, Eq. (1)(1) $f{e}_i=n_i{y}_i+{\epsilon }_i$(1) was estimated for 32 food items and 14 food groups. From Eq. (1(1) $f{e}_i=n_i{y}_i+{\epsilon }_i$(1) ), income elasticity of food expenditure was estimated as follows (Babu et al. 2016a; Chisanga & Zulu-Mbata, 2018):

(2) $n_i=\frac{\mathrm{\partial }lnf{e}_i}{\mathrm{\partial }ln{y}_i}$(2)

Where is as defined already, is the natural logarithm of expenditure on food item $i$ and $ln{y}_i$ is the natural logarithm of income for the household using total expenditure and used as the proxy for income.

The comparisons of food expenditure shares and income elasticities were calculated for each of the food items and groups by gender of the head and income groups. Households were grouped into quintiles to examine the variation in food expenditure shares and income elasticities across the different income groups. Per capita consumption expenditure was used to compute the income quintiles (Chisanga & Zulu-Mbata, 2018). The average per capita consumption expenditures were US$20.46 for the poorest income group (quintile 1), US$31.24 for the poor income group (quintile 2), US$44.87 for the moderate-income group (quintile 3), US$69.71 for the rich income group (quintile 4) and US$175.20 for richest income group (quintile 5).

3. Results and discussion

3.1. Socioeconomic characteristics

Table 2 shows the per capita consumption expenditure and food expenditure shares. On average per capita consumption expenditure for female headed households is US$70.33 and US$2.40 higher than for male headed households. Female headed households were better off compared to male headed households.

Table 2. Per capita consumption and food expenditure shares

	Overall	Male	Female	Differences
Per capita consumption expenditure	68.84	67.93	70.33	2.40**
Food expenditure share	43.03	43.05	43.00	−0.05
Observation	29,087	18,073	11,014

^{*, **, ***. Statistically significant at the 10%, 5%, and 1% level, respectively. Differences based on t-test.}

3.2. Food expenditure shares by food items

The proportion of food budgets spent on each food item is shown in Tables 3 and indicates that vegetables and bread dominated the monthly food expenditure shares. Vegetables accounted for the largest share of the food expenditures (27.3%) for an average household in Zimbabwe. Vegetables are a key food group providing essential vitamins and minerals (Choudhury et al., 2020), and their intake is particularly important in settings where micronutrient deficiencies are widespread, such as in Zimbabwe. However, the higher vegetable expenditure share can also be linked to expensive meat beyond the reach of many Zimbabwean families with low income levels. Maize, which is the major staple food of the country, accounted for 7% of the food expenditure share and tops other cereals and rice which constituted 2.9% and 3.5%, respectively. Within the cereal food group, small grains only constitute 0.1% of the food expenditure, yet their production and consumption should be vigorously promoted given that they are drought tolerant and have high vitamin and micronutrients. An analysis of the production and consumption constraints of small grains coupled with identification of entry points to upscale the small grains value chain is needed. Beef and other meats were the most important source of animal protein in Zimbabwe. The beef expenditure share was 4.8% while poultry consumption constituted 3.1% of the monthly food expenditure share. Other meats besides beef and poultry took up a higher share in the meat food group with 5.5% share and this could be due to high beef and poultry prices compared to other alternative meat sources. These results taken together may imply that households were substituting animal protein sources with plant sources, for example, beans or soya chucks to satisfy their protein requirements. Other milk products and milk constituted 2.4% and 1%, respectively, while fruits accounted for 1.4% of the food expenditure. The expenditure shares of these three food items were quite low and comparable to Chisanga and Zulu-Mbata (2018) who found that expenditure shares for dairy products and fruits were 1.4 and 1.5, respectively, in Zambia.

Table 3. Proportion of food budgets spent on each food item for full sample and by gender of head

Food item	Overall sample	Male	Female	Gender differences
Bread	14.1	13.73	14.71	0.98***
Maize	6.99	7.05	6.89	−0.16
Small grain	0.12	0.12	0.12	−0.00
Wheat	0.03	0.03	0.02	−0.01
Rice	3.5	3.55	3.42	−0.13**
Other cereals	2.87	2.94	2.77	−0.16***
Beef	4.75	5.06	4.24	−0.82***
Poultry	3.07	3.22	2.81	−0.41***
Goat and game	0.41	0.47	0.33	−0.14***
Pork and crocodile	0.50	0.55	0.42	−0.13***
Other meat	0.60	0.61	0.58	−0.03
Other meat2	5.48	5.51	5.44	−0.08
Fish	5.9	6.26	5.32	−0.94***
Milk	1.01	1.03	0.97	−0.07**
Other milk products	2.33	2.41	2.21	−0.20***
Eggs	0.71	0.73	0.67	−0.06***
Oils	6.0	6.10	5.84	−0.26***
Fruit	1.35	1.26	1.48	0.22***
Other fruits	0.05	0.05	0.05	0.00
Vegetables	27.26	26.01	29.30	3.29***
Nuts	1.94	1.75	2.25	0.50***
Cassava	0.01	0.01	0.01	−0.00
Potatoes	0.69	0.71	0.66	−0.05*
Other tubers	2.02	1.95	2.12	0.16***
Sugar	5.18	5.26	5.03	−0.23***
Coffee tea	0.50	0.48	0.53	0.05***
Condiments	1.62	1.64	1.60	−0.04
Takeaway	0.78	0.87	0.63	−0.23***
Sit down	0.33	0.34	0.30	−0.04
Beverages	1.71	1.77	1.62	−0.16***
Alcohol	0.74	1.08	0.18	−0.90***
Other beverages	0.23	0.22	0.24	0.02
Tobacco	0.18	0.25	0.05	−0.20***
N	29805	18,503	11,302

^{*, **, ***. Statistically significant at the 10%, 5%, and 1% level, respectively.}

An understanding of the gender differences in food expenditure patterns is important for policy and targeting of interventions. Table 3, column 3 to 5 shows the food expenditure shares by the gender of the household head. Male headed households had significantly higher expenditures on the majority of food items while their female counterparts had significantly higher shares on bread, fruits, and vegetables only. If the fruit and vegetables expenditures translate to intake, study findings are quite interesting for female headed households given that adequate consumption of fruit and vegetables is key to improved diet-related health and associated with a lowered risk of cancer and cardiovascular disease (Choudhury et al., 2020; Aune et al. 2017). Male headed households had significantly higher food expenditure shares than their female counterparts with significantly larger differences observed for meat commodities. Male headed households tend to devote a higher budget share for meat products, beverages, and alcohol.

3.3. Food expenditure shares by food groups

The analysis of food budgets on each food group shows that cereals (28%), vegetables (27%), and meats (15%) accounted for relatively large shares of the food expenditure followed by oils, sugars, tubers, pulses, condiments, beverages and fruits (Table 4, column 2). The gender differences in food expenditure shares are shown in columns 3 to 5. Female headed households seem to have significantly higher food expenditure shares on cereals, vegetables, tubers while male headed households dominated on meats, milk, eggs, sugars, and oil food expenditure shares. Fruits and vegetables are important as they lower the risk of cancer and many non-communicable diseases (Aune et al. 2017). It is also of interest to note that male headed households depend more on food away from home, beverages, alcohol, and tobacco than the female headed households. The risks associated with unhealthy diets and sedentary lifestyles, for example, obesity and non-communicable diseases may be more pronounced in male headed households (Ssewanyana et al., 2018; Wilde et al., 2012). This evidence demonstrates that food policies in Zimbabwe need to be gender inclusive to ensure food and nutrition security for all, rather than one-size fits all.

Table 4. Proportion of food budgets spent on each food group for full sample by and gender of head

Food group	Full sample	Male	Female	Gender differences
Cereals	27.61	27.41	27.93	0.52***
Tubers	2.71	2.67	2.78	0.11*
Pulses	1.94	1.75	2.25	0.50***
Vegetables	27.26	26.01	29.30	3.29***
Fruits	1.4	1.31	1.54	0.22***
Meats	14.81	15.42	13.82	−1.60***
Fish	5.9	6.26	5.32	−0.94***
Eggs	0.71	0.73	0.67	−0.06***
Milk	3.34	3.44	3.17	−0.26***
Sugars	5.18	5.26	5.03	−0.23***
Oils	6.0	6.10	5.84	−0.26***
Condiments	2.12	2.12	2.13	0.01
Beverages	1.94	2.00	1.86	−0.14***
Food away from home	1.1	1.21	0.94	−0.27***
Alcohol & tobacco	0.92	1.34	0.24	−1.10***
N	29805	18,503	11,302

^{*, **, ***. Statistically significant at the 10%, 5%, and 1% level, respectively.}

3.4. Proportion of food budget spent on food groups by income levels

In this section, we are interested in the link between income and food consumption. To understand this, food shares by household income category are analysed. Results in Table 5 show that across all quintiles the larger proportion of the food budget is devoted to cereals and vegetables and followed by meat. The cereal food expenditure share was 28% and 27% for the poorest and richest groups, respectively. For four types of food groups (cereals, pulses, vegetables, and sugars), food expenditure shares are larger for poorer than for richer people. The rich households tend to have balanced diets characterized by consumption of starchy vegetables, fruits, eggs, milk, and meat. The food expenditure shares for fruits, meat, eggs, and milk were lower for the poorer than for the richer households. Milk consumption provides all the necessary energy and nutrients vital for growth and bone mass formation (Pereira, 2014). Eggs are dense in high-quality protein and a wide range of micronutrients (Iannotti et al., 2017) while fruits supply a wide range of micronutrients (Choudhury et al., 2020). The lower expenditures on fruits, eggs, and milk among the poor may be a sign of inadequate consumption. Low intake of these foods is associated with a risk of wasting, stunting, and micro-nutrient deficiencies and also tends to slow cognitive development in children (Choudhury et al., 2020; Headey & Alderman, 2019; Pereira, 2014). Poorer households had poor dietary choices characterized by higher consumption of starchy foods and lower consumption of fruits, eggs, and milk.

Table 5. Food expenditure shares for food groups by income quintiles

Food group	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5
Cereals	28.15	27.58	27.78	28.01	26.53
Tubers	2.56	2.68	2.89	2.71	2.71
Pulses	1.81	2.33	2.65	1.87	1.07
Vegetables	38.01	33.08	27.91	22.58	14.9
Fruits	1.10	1.37	1.52	1.49	1.49
Meats	9.90	12.08	13.68	16.90	21.51
Fish	5.88	6.05	6.17	5.97	5.45
Eggs	0.23	0.33	0.47	0.86	1.65
Milk	2.25	2.93	3.51	3.76	4.23
Sugars	5.37	5.59	5.54	5.07	4.32
Oils	5.79	5.96	5.97	6.18	6.11
Condiments	1.78	1.82	2.07	2.26	2.64
Beverages	0.54	0.90	1.46	2.47	4.35
Food away from home	0.27	0.41	0.61	1.21	3.03
Alcohol & tobacco	0.37	0.45	0.69	1.11	1.98

3.5. Income elasticities for food items

Food-income elasticities show the responsiveness of expenditure on each food item to an increase in income. Table 6 provides the summaries of income elasticities of food items for the full sample (column 2) and by gender of head (column 3 to 5). The positive signs of the expenditure elasticities indicate that all the food items are normal goods in the full sample and the gender categories. The positive sign also implies that an increase in household income will cause an increase in the expenses on all the food items, ceteris paribus. In the full sample, all food items are necessary goods with inelastic expenditure elasticities ranging from 0.10% to 0.66%. The pooled results show that household bread budget share increases at an increasing rate of 6.6% for every 10% increase in the income, other things equal. For every 1% increase in the household income, expenditure on vegetable increases by 0.83% in the full sample, 0.82% in male headed households, and 0.84% in female headed households.

Table 6. Income elasticities for food items for full sample and by gender of head

Food item	Full sample	Male	Female	Gender differences
Bread	0.66	0.66	0.66	0.00
Maize	0.45	0.45	0.44	−0.01*
Small grain	0.44	0.45	0.41	−0.04
Wheat	0.42	0.45	0.37	−0.08*
Rice	0.41	0.41	0.41	−0.00
Other cereals	0.36	0.36	0.36	−0.00
Beef	0.49	0.50	0.47	−0.03***
Poultry	0.49	0.49	0.48	−0.01
Goat and game	0.44	0.45	0.43	−0.02
Pork and crocodile	0.29	0.30	0.27	−0.02***
Other meat	0.22	0.22	0.22	−0.01
Other meat2	0.60	0.62	0.59	−0.03***
Fish	0.46	0.48	0.44	−0.03***
Milk	0.22	0.22	0.22	−0.00
Other milk products	0.34	0.35	0.33	−0.02***
Eggs	0.21	0.21	0.21	−0.00
Oils	0.50	0.50	0.49	−0.01***
Fruit	0.26	0.26	0.28	0.02***
Other fruits	0.24	0.23	0.26	0.03
Vegetables	0.83	0.82	0.84	0.02***
Nuts	0.41	0.40	0.42	0.02***
Cassava	0.22	0.23	0.20	−0.03
Potatoes	0.28	0.29	0.27	−0.02***
Other tubers	0.44	0.44	0.43	−0.00
Sugar	0.45	0.45	0.44	−0.01***
Coffee tea	0.10	0.09	0.12	0.02***
Condiments	0.15	0.15	0.15	0.00
Takeaway	0.10	0.11	0.09	−0.02**
Sit down	0.41	0.43	0.39	−0.04
Beverages	0.29	0.29	0.28	−0.01*
Alcohol	0.40	0.42	0.26	−0.16***
Other beverages	0.23	0.24	0.23	−0.01
Tobacco	0.22	0.23	0.20	−0.02

^{*, **, ***. Statistically significant at the 10%, 5%, and 1% level, respectively.}

Beef in all categories had expenditure elasticities less than one and positive, indicating that they are normal goods, where an increase in household income increases the quantities demanded. For the pooled data, a 10% increase in household income guarantees a 4.9% increase in the household beef budget in Zimbabwe, ceteris paribus. For every 1% increase in the household income, there will be an increase in the expenditure on beef by 0.50% in male headed households, and 0.47% in female headed households. The expenditure elasticities for beef are significantly higher in male headed households compared to female headed households. Findings conform to other studies that as the household income increase, families tend to shift towards protein-related foods (e.g., beef) with significantly higher effects among male headed households (Ansah et al., 2020; Chisanga & Zulu-Mbata, 2018; Harris et al., 2019).

3.6. Income elasticities for food items by income levels

In Table 7, we analysed the income elasticities for food items by different income levels. All the food items were normal goods given their elasticities are positive (Ansah et al., 2020; Babu et al., 2016b). For all types of food items except food away from home, income elasticities were larger for the poorer than the richer households. For most food items, the income elasticities for the poorest households are two-fold larger than those in the richest quintile. As discussed earlier, and noted by Qaim (2019) poor people increase the quantity of food consumption more with rising income than rich people who tend to increase their demand for food quality. The income elasticities for vegetables, bread, maize, and poultry were quite high and closer to one for the poorest households, implying an increase in income would substantially increase consumption of these foods. Study findings demonstrate that as income increases, not only does the quantity of food increase less than proportionally, but the composition of the food basket also changes. In line with “Bennet’s law”, the consumption of starchy cereals in Zimbabwe declines with an increase in income (Cirera & Masset, 2010; Colen et al., 2018; Qaim, 2019).

Table 7. Income elasticities for food items by income levels

Food item	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5
Bread	0.91	0.74	0.64	0.57	0.47
Maize	0.58	0.48	0.42	0.40	0.35
Small grain	0.69	0.54	0.41	0.32	0.22
Wheat	0.67	0.45	0.38	0.43	0.26
Rice	0.55	0.46	0.42	0.38	0.30
Other cereals	0.50	0.41	0.36	0.32	0.28
Beef	0.63	0.52	0.47	0.47	0.44
Poultry	0.70	0.58	0.50	0.46	0.41
Goat and game	0.57	0.49	0.43	0.40	0.37
Pork and crocodile	0.47	0.37	0.29	0.29	0.26
Other meat	0.35	0.26	0.21	0.20	0.18
Other meat2	0.76	0.66	0.57	0.51	0.45
Fish	0.62	0.51	0.45	0.40	0.33
Milk	0.35	0.26	0.21	0.19	0.18
Other milk products	0.46	0.39	0.35	0.30	0.25
Eggs	0.30	0.22	0.19	0.19	0.20
Oils	0.68	0.56	0.49	0.44	0.37
Fruit	0.43	0.35	0.29	0.22	0.16
Other fruits	0.43	0.23	0.14	0.14	0.07
Vegetables	1.18	0.98	0.83	0.68	0.49
Nuts	0.53	0.46	0.42	0.34	0.25
Cassava	0.39	0.26	0.22	0.15	0.18
Potatoes	0.43	0.32	0.28	0.27	0.26
Other tubers	0.61	0.51	0.43	0.35	0.26
Sugar	0.61	0.51	0.45	0.38	0.29
Coffee tea	0.18	0.12	0.09	0.08	0.09
Condiments	0.22	0.16	0.14	0.13	0.12
Takeaway	0.0	0.02	0.02	0.09	0.19
Sit down	0.53	0.36	0.38	0.42	0.39
Beverages	0.30	0.26	0.26	0.28	0.30
Alcohol	0.49	0.39	0.38	0.38	0.40
Other beverages	0.37	0.25	0.23	0.20	0.21
Tobacco	0.27	0.23	0.21	0.22	0.22

3.7. Income elasticities for food groups by gender of head

Results in Table 8 and column 2 for the overall sample confirm large variations in income elasticities according to the food group. As expected, the values of the income elasticities for food groups were positive and less than one (inelastic), even though they typically vary between different types of food groups. As such positive elasticities show that these are normal goods (Qaim, 2019) and resonate with the Engel’s law that food demand increases less than proportionally with income (Babu et al., 2016b; Cirera & Masset, 2010; Qaim, 2019) is fulfilled. Elasticities for cereals, vegetables, meat, and oils are significantly higher than demand for other food groups, i.e. demand for these types of foods is most responsive to income changes. A 1% rise in income would increase cereal, vegetable, meat, and oil consumption by 0.9%, 0.8%, 0.7% and 0.5%, respectively. Demand for fruits, eggs, milk, and beverages is thus less responsive to income changes. Such results tend to be contrary to the meta-analysis by Colen et al. (2018), and (Chisanga & Zulu-Mbata, 2018) for Zambia. Part of the explanation may relate to the fact of different country contexts.

Table 8. Income elasticities for food groups for full sample and by gender of head

Food group	Full sample	Male	Female	Gender differences
Cereals	0.85	0.85	0.84	−0.01***
Tubers	0.42	0.42	0.41	−0.01
Pulses	0.41	0.40	0.42	0.02***
Vegetables	0.83	0.82	0.84	0.02***
Fruits	0.27	0.26	0.28	0.02***
Meats	0.67	0.69	0.65	−0.04***
Fish	0.46	0.48	0.44	−0.03***
Eggs	0.21	0.21	0.21	−0.00
Milk	0.35	0.35	0.34	−0.01***
Sugars	0.45	0.45	0.44	−0.01***
Oils	0.50	0.50	0.49	−0.01***
Condiments	0.20	0.20	0.21	0.01*
Beverages	0.30	0.30	0.29	−0.01**
Food away from home	0.14	0.15	0.13	−0.02**
Alcohol & tobacco	0.39	0.41	0.27	−0.15***

^{*, **, ***. Statistically significant at the 10%, 5%, and 1% level, respectively.}

Table 8, (column 3 to 5) shows the income elasticities for food groups by gender of household head. We find significant evidence that the magnitude of food-income elasticities for female headed households was slightly lower for most of the food groups (cereals, meat, milk, oils, and sugars) compared to male counterparts. This conforms to Engel’s law that richer households (female headed households in our case—see, Table 2) spend a lower share of their income on food compared to their male counterparts (Cirera & Masset, 2010).

3.8. Income elasticities for food groups by income levels

In this subsection, we investigated whether income elasticities for food groups vary between people at different income levels (Table 9). For all types of food groups except food away from home, income elasticities were larger for poorer than for richer people. The main explanation is that poor people increase their food consumption more with rising income than rich people who may further increase their demand for food quality, but not necessarily food quantity (Qaim, 2019). The income elasticities for cereals and vegetables were quite high and greater than one for the poorest households, implying an increase in income would substantially increase cereal and vegetable consumption. As income increases, not only does the quantity of food increase less than proportionally, but the composition of the food basket also changes. Cereals are the main starchy staple food in the country. In particular, these findings fulfill the “Bennet’s law” that documents that the consumption of starchy staple food declines with an increase in income compared to other foods (Babu et al. 2016a; Cirera & Masset, 2010; Colen et al., 2018; Qaim, 2019).

Table 9. Income elasticities for food groups by income quintiles

Food group	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5
Cereals	1.08	0.93	0.84	0.76	0.64
Tubers	0.60	0.50	0.42	0.35	0.29
Pulses	0.53	0.46	0.42	0.34	0.25
Vegetables	1.18	0.98	0.83	0.68	0.49
Fruits	0.44	0.35	0.29	0.22	0.16
Meats	0.81	0.72	0.66	0.64	0.58
Fish	0.62	0.51	0.45	0.40	0.33
Eggs	0.30	0.22	0.19	0.19	0.20
Milk	0.46	0.39	0.36	0.31	0.28
Sugars	0.61	0.51	0.45	0.38	0.29
Oils	0.68	0.56	0.49	0.44	0.37
Condiments	0.26	0.21	0.19	0.19	0.18
Beverages	0.33	0.29	0.28	0.29	0.31
Food away from home	0.13	0.08	0.07	0.13	0.22
Alcohol & tobacco	0.46	0.38	0.37	0.037	0.40

4. Conclusion

The study analysed the expenditure patterns and income elasticities differentiated by gender of head and income levels in Zimbabwe. Vegetables and bread dominated the monthly food expenditure shares. Vegetables accounted for the largest share of the food expenditures for an average household in Zimbabwe. Male headed households had significantly higher expenditures on the majority of food items while their female counterparts had significantly higher shares on bread, fruits, and vegetables only.

An analysis of each food group shows that cereals (28%), vegetables (27%), and meats (15%) accounted for relatively large shares on the food expenditure. Female headed households had higher food expenditure shares on cereals, vegetables, and tubers while male headed households dominated on meats, milk, eggs, sugars, and oil food expenditure shares. Male headed households were relying heavily on animal protein sources. With regard to poverty, results show that poor people had poor dietary choices characterized by higher consumption of starchy foods and lower consumption of fruits, eggs, and milk.

All income elasticities for all food items and groups were positive, an indication of normal goods. These findings suggest that policies aimed at increasing household income can still provide an important contribution to increasing food consumption. Our sample show heterogeneity in income elasticities for food items and food groups. Vegetables, bread, other meats, and oils had elasticities higher than 0.5 in the full sample. The expenditure elasticities for fruits and vegetables were significantly higher in female headed households. On the other hand, income elasticities for beef were significantly higher in male headed households. For most food items, the income elasticities for the poorest households are two-fold larger than those in the richest quintile.

The income elasticities for food groups were positive and less than one, even though they vary between different types of food groups. Elasticities for cereals, vegetables, meat, and oils are significantly higher and most responsive to income changes than demand for other food groups. We find evidence that the magnitude of food-income elasticities for richer groups and female headed were slightly lower for most of the food groups. This conforms to Engel’s law that richer segments of the society spend a lower share of their income on food compared to poorer counterparts. For all food groups except food away from home, income elasticities were larger for poorer than for richer people. As discussed earlier, poor people increase the quantity of food consumption more with rising income than rich people who tend to emphasize the demand for food quality as their income rises. The income elasticities for vegetables, bread, maize, and poultry were quite high and closer to one for the poorest households, implying an increase in income would substantially increase consumption of these. This demonstrates that as income increases, not only does the quantity of food increase less than proportionally, but the composition of the food basket also changes.

5. Policy implications

The study findings indicate that food consumption patterns in Zimbabwe are affected by the gender of household head, and income. All food items had positive expenditure elasticities, and this suggests that increased income will contribute to more diversified diets. Hence, policies and interventions that improve income streams among households are crucial for food consumption. However, it is not clear whether more diverse diets will necessarily be healthier. A number of public policy approaches have been found to promote the consumption of healthier diets, for example, nutrition education aimed at promoting consumption of healthy foods like fruits and vegetables (Rustad & Smith, 2013), food fortification or biofortification (Guo et al., 2019); and subsidies for healthy foods (Donovan & Gelli, 2019).

We found that very poor households, and male headed households allocated the highest proportion of income to the food budget than the non-poor and female headed households. Poor households do not have access to proper nutrition because of insufficient income that limits them to meet their dietary needs. This suggests that food and social protection policies in Zimbabwe should be designed to protect the very poor from escalating food prices. Hence, it might be prudent for food aid and social protection schemes such as social assistance in the form of income transfer to target very poor households to help them get access to better diets. Overall, we find that demand functions for gender and income groups were different. Therefore, targeted food policies should be formulated based on specific food demand patterns in the groups.

6. Limitations

This study is based on cross-sectional data and should be interpreted with caution when considering food expenditure and income growth over time. Although this article relies on OLS to sufficiently answer the research objectives (Chisanga & Zulu-Mbata, 2018), the use of OLS and other estimation techniques such as the almost-ideal demand system and quadratic almost-ideal demand system do not account for the endogeneity of income and other right-hand variables (Ansah et al., 2020; Lecocq & Robin, 2015). Future research articles may need to consider using the aidsills STATA command by Lecocq and Robin (2015) for estimating almost-ideal demand systems and accounts for endogeneity.

Highlights

• Analyse food expenditure patterns and income elasticities

• Differential analysis by gender of household head and income levels

• Female headed households had higher vegetable and fruit expenditures

• Male headed households had higher expenditure shares on animal protein sources

• Income elasticities positive indicating normal goods

• Poor and male headed households spend higher proportion of income on food

• Food policy interventions to target the poor and promote healthy diets

Data availability

The data are publicly available from Zimbabwe National Statistics Agency. Permission to use the data set is obtained from this institution.

Acknowledgements

We are grateful to Zimbabwe National Statistics Agency for providing the dataset used in this article. The anonymous reviewers are acknowledged for their constructive comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Conrad Murendo

Conrad Murendo is a Collaboration, Learning and Adaptation Lead at CARE International in Zimbabwe. He holds a PhD in Agricultural Economics from the University of Goettingen, Germany. His research experiences are in food and nutrition security, gender, health economics, consumer finance, and technology adoption.

Grown Chirongwe

Grown Chirongwe is an Economist at Zimbabwe National Statistics Agency. He holds a MSc Degree in Economics from the University of Zimbabwe. His research experiences are in poverty analysis, financial economics and international development.

Givious Sisito

Givious Sisito is a Principal Research Officer in the Ministry of Lands, Agriculture, Water and Rural Resettlement, Zimbabwe. He holds an MSc in Operations Research from National University of Science and Technology, Zimbabwe. His research experience is on livestock production systems, climate and economic modelling, agricultural development, and community stakeholder engagements.

This study contributes to the literature on food expenditure, income elasticities, and socio-economic disparities in a developing country context.