An investigation into food-away-from-home consumption in South Africa

ABSTRACT

Food policy that ignores food-away-from-home (FAFH) in a developing country like South Africa will be misleading given changes in demand for food over time. This study contributes to our understanding of the factors that influence the demand for FAFH in South Africa. Using panel data from the Income and Expenditure Survey, this study analyses the effects of income and socio-demographic variables on FAFH expenditure using a double-hurdle model. The results show that small-sized households headed by younger white females/males and living in an urban settlement are most likely to purchase FAFH while male-headed households spend more than female-headed households. Furthermore, income of the household head is an important determinant of household FAFH expenditures. The income elasticity of expenditure on FAFH is inelastic and a normal good. The small size of the participation elasticities means that growth in the FAFH sector will be driven by households with existing expenditure.

KEYWORDS:

• Double-hurdle model
• food-away-from-home
• South Africa

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The Income and Expenditure Survey classifies FAFH as prepared meals purchased from restaurants, cafés, canteens and the like (Statistics South Africa, 2008).

2. Source: Food Expenditures data product, Economic Research Service, United States Department of Agriculture (USDA) and National Bureau of Economic Research.

3. Traditional food demand models use data based on household expenditure on food items from the market and typically do not include FAFH consumption. As the share of FAFH expenditure in total food consumption increases, food demand models that ignore FAFH as a category of food demand will not reflect true expenditure and trade-offs.

4. The higher percentage of calories from saturated fat in fast foods was especially noteworthy at 13.5%, compared with 11.9% in restaurant foods, 12.3% in school foods and 10.7% in FAH.

5. This may be because of data availability.

6. The dependent double-hurdle model is used given it allows correlation between the participation decision and expenditure decision (García, 2013).

7. Ma et al. (2006) used both univariate and multivariate tobit models to conduct their analyses of FAFH expenditure for urban China. Angulo et al. (2002) used the two-stage process suggested by Chamberlain (1984) to address the problem of including censored dependent variables in the panel data framework. Stewart et al. (2004) used the method developed by Shonkwiler and Yen (1999) to estimate multiple equations simultaneously while accounting for zero-censoring because the usual methods of estimation are likely to be biased in the presence of zero expenditure observations (zero-censoring) for FAFH. The Shonkwiler and Yen method consists of two steps to correct for the problem of zero-censoring.

8. Household size can be expected to have both positive and negative impact on FAFH. In a country like China with restrictions on number of children, larger household size can actually signify more adults in a household than a large size with more kids. A household with more adults including grand parents may reduce the likelihood of demand for FAFH.

9. Stewart et al. (2004) stated that:

To test hypotheses about how a household’s demand for food away from home is affected by its structure and other characteristics, we need a data set with information on households, their characteristics, and how much they spend in each market segment. The ideal set of data for this study would include information on at least several thousand households, the characteristics of each household, and how much each household spent in each market segment. Moreover, it would follow this sample of households over 20 to 50 years, and report on how each household’s characteristics and expenditures have changed … Unfortunately, these data are not available.

10. A detailed overview of the sector is presented in the supplementary material to this paper.

11. Other data exist that can be used in South Africa such as the National Income and Dynamics Study. The IES is used given that the first panel comparable to the 2011 data is in 2005 versus the first panel for the National Income Dynamics Study (NIDS) starting in 2008 and there are issues of comparability between the first and second fieldwork phases of the NIDS wave 2 (Finn et al., 2014). Other studies can compare the results in the two datasets for robustness. The IESs are primarily dedicated to collecting information on households’ expenditures and, as such, the questionnaires are extensive and detailed with six visits by enumerators in 2005/6 and four visits in 2010/11.

12. Some households were dropped because of missing values in some or all of the variables of interest.

13. Variables such as hours worked and occupation are other variables that are included in the estimation.

14. Variyam (2005) argued that FAFH providers have different types of offerings, economies of scale and levels of recipe standardisation and that having a labelling policy will have varying impact on producers. The value of labelling for consumers may be endogenously determined – will consumers that already have good quality diets and healthy weights reap the benefits of labelling or those with poor diets and the overweight?