Three essays on changing food consumption patterns in Indonesia

Abstract

Food consumption differs significantly across households over time. Economists tend to explain the differences in food consumption in terms of traditional economic variables such as prices and income. While these factors have been observed to explain the differences in consumption in a greater extent, other factors such as migration and education are assumed to substantially alter patterns of food consumption. In this thesis, I study the differences in food consumption, mainly focusing on non-traditional economic factors that have been perceived to be important drivers of changing food consumption patterns. Previous research suggests that households’ resources (i.e. total expenditure), demographics, migration, and education, are the key determining factors influencing household welfare. Because households have different levels expenditure and expenditure is related to household’s welfare, factors that may affect expenditure such as household size, natural disasters, levels of education, out-migration and so on may also affect welfare. In this dissertation, I explore the impact that the above factors have on the patterns of food consumption, and hence welfare, of households in Indonesia using rich and comprehensive longitudinal Indonesian Family Life Survey (IFLS) data. An Engel curve depicts the mean budget share of a particular food group at each level of household expenditure while prices of goods are held constant. Employing the correct specification of an Engel curve in food demand analysis plays a key role in estimating precise food consumption parameters. The first essay analyses the food consumption patterns by applying Lewbel and Pendakur’s (2009) Exact Affine Stone Index (EASI) to the IFLS dataset. The EASI demand system is a powerful framework for analysing consumer food choices and policy evaluation, as it can be applied to any higher order polynomials of per capita food expenditure as a main explanatory variable when estimating the Engel curve. To my knowledge, my first essay titled “Changing food consumption patterns: An application of the Exact Affine Stone Index demand system in Indonesia” is the first to apply the EASI method to investigate consumer food demand functions in the context of a developing economy, namely Indonesia. I find that the estimated food Engel curves have a variety of shapes with sufficient curvatures, and that the rank of the food demand functions (i.e. Engel curves) can be approximated by up to 3rd order polynomial functions of real household expenditure. Furthermore, poorer and richer households have statistically significantly different food consumption choices. The most striking and somewhat surprising finding is that the wealthier households do not appear to diversify their food consumption further when their income rises, whereas poorer households tend to diversify their food consumption significantly when their wealth increases. The second essay addresses the impact of internal migration on food consumption patterns. This issue is pertinent to Indonesia as it has a large number of internal and interprovincial migrations throughout its history. I use distance to the migrant’s destination and propensity score-matching to generate plausibly exogenous variations in migration to identify the effect of migration on food consumption. Overall, I find that on average, the migrant-sending household’s per capita food consumption is larger (13.4%) than that of non-migrant sending households living in the same neighbourhood (10.7%). To claim that this finding is not driven by other unobserved variables, this study has employed both fixed effects (FE) and instrumental variable (IV) regressions and these estimates consistently support OLS findings. Moreover, migrant households appear to make a substantial shift from the consumption of rice, corn, and wheat towards the consumption of vegetables and fruits, dairy products, and ‘meat and animal’ foods. The results have a suggestive evidence about the value of internal migration for improving welfare in terms of changing food consumption patterns of migrant-sending households. Since the 1970s, Indonesia has had an impressive record of educational extension, including six years of compulsory schooling (effective from 1984) and nine years of compulsory schooling (effective from 1994). Enrolment rates in primary schools are close to universal and about 75% for secondary education. There is an ongoing effort to expand secondary school attainment at the universal level. Whereas an overwhelming portion of the literature has focused on the labour market (monetary) returns to education in both the developed and developing economies, to date, only a few studies have investigated the impact of education on food consumption. The third essay attempts to fill in this gap by exploring the relationship between the household head’s educational attainments and household consumption patterns in Indonesia. To obtain consistent and causal estimates, I employ a quasi-parametric selection model and instrumental variable approach to address the endogeneity of education (i.e. schooling). I use distance from the household to the institutions in which the household has attained education in Indonesia as a source of plausibly exogenous variation in schooling. The first-stage result shows that distance to the school is a strong predictor to education. The ordinary least squares (OLS) estimates suggest that households with graduates from the higher secondary schools tend to consume 31.5% more healthy foods and 22.8% less unhealthy foods than households with graduates from lower secondary schools. These findings have been confirmed by IV estimation and imply that the OLS estimates are not driven by other unobserved characteristics in the households. The results also demonstrate that households have heterogeneous food consumption returns due to different educational attainments. Taken together, these three essays is an attempt to provide an empirical investigation into how household welfare, measured in terms of food consumption, could be influenced by important socio-economic variables such as household’s resources, migration, and education. The findings from the three essays of the dissertation suggest that non-traditional economic variables (as opposed to tractional economic variable such as prices and income) market access, natural disasters, migration and education such as also influence food consumption significantly across households. The findings may have policy implications that the government may perhaps undertake in relation to migration and education so as to enhance welfare within a household.