The Effect of Household Characteristics on Food Expenditure in the Philippines (Research Paper Sample)

The Effect of Household Characteristics on Food Expenditure in the Philippines
Group 6
Introduction, Results, Conclusion
Abstract, Methodology and Data Discussion, Weaknesses of the Model and Remedies
Review of Related Literature, Weaknesses of the Model and Remedies
Economic Theories, Econometric Model, Results
Abstract
This study aims to provide information regarding the effect of household characteristics such as total number of household members, location of household (urban or rural area), total family income, age of household head and marital status of household head on total family food expenditure in the Philippines. The data was from the Family Income and Expenditure Survey (FIES) 2009. Three econometric models were constructed and Ordinary Least Squares (OLS) regression was utilized to estimate these. Three tests were performed to the models: Regression Specification Error Test (RESET), Variance Inflation Factor (VIF) and White Test. The test results revealed that functional form misspecification, omitted variable bias and heteroskedasticity are present among the regression models. One remedy we executed is to use robust standard errors. Estimation results suggest that both the number of household members and total household income have positive relationship with total food expenditure. The household located in an urban area has higher level of food expenditure than that of the household located in a rural area. Married household heads also have greater food expenditure compared to those household heads that have marital status other than “married”. Meanwhile, household head age negatively affects food spending.
Table of Contents
Introduction1
Review of Related Literature1
Economic Theory6
Econometric Model8
Weaknesses of the Model and Remedies10
Methodology and Data Discussion13
Results15
Conclusion18
List of Tables and Figures
Table 2.1 Studies on The Effect of Different Household Characteristics
on Household Expenditure
Table 5.1 Variance Inflation Factor of the First Model
Table 5.2 Variance Inflation Factor of the Second Model
Table 5.3 Variance Inflation Factor of the Third Model
Table 7.1 Results from regressing the three models
Chapter 1: Introduction
Food consumption is one of the major chunks of expenditures households have in any country in the world. Because eating is needed in order for humans to survive, it is already a given that households would allot a certain portion of their income for food expenditures. However, there are differences when it comes to the pattern of food spending across different households. This is caused by the variations in the household characteristics of each household. Each one has different set of characteristics, and these affect how the pattern of food spending would go. Factors such as level of income, location, number of household members, as well as characteristics of the household head, all affect food expenditure in certain ways. Thus, it is interesting to study the effects of these factors in order to assess, on average, their individual impacts on the food expenditures across different households.
This study aims to investigate the case of households in the entire Philippines, with regard to the patterns of their food consumption expenditure in relation to their household characteristics. We hope to determine whether these household characteristics really have an effect, and if they do, in what way do they influence the patterns? To do that, we will identify which household characteristics are possible influencers of food expenditure, and through regression analysis, determine what specific effects they cause to the variable being examined.
Chapter 2: Review of Related Literature
Various studies show that household characteristics are important factors in determining household expenditures. Each of these characteristics showcases different effect on the total expenditures of a household. Ritonga (1994,1-3) stresses the importance of incorporating household characteristics in observing consumption behavior of a household in Central Java, Indonesia. Generally, he investigated patterns of expenditures by estimating a complete household demand system for sixteen commodity groups. Ritonga (1994, 4-27) also showed that, through adding household characteristics, there is a huge changes found in household expenditure patterns. The Almost Ideal Demand System (AIDS) model was used in this study. This model is derived from the necessary sufficient conditions for the existence of a representative budget level with defined cost function representing the price-independent generalized logarithmic (PIGLOG) class of preferences. It is visually represented as: wᵢ = αᵢ + βᵢ (log Y/P) + ∑j ϒᵢj log pj where wᵢ is the household budget share on the ith commodity; αᵢ, βᵢ, ϒᵢj are the parameters of the system; Y denotes total household income; pj denotes price of jth commodity; and P is the price index. Using the Stone’s Price Index given by log p = ∑ᵢ wᵢ log pᵢ, the model will turn into Linear Approximation of the Almost Ideal Demand Systems (LA-AIDS). With Stone’s Price Index, the LA-AIDS can be written as: wᵢ = αᵢ + βᵢ (log Y- log ∑ᵢ wᵢ log pᵢ) + ∑j ϒᵢj log pj. However, this model does not portray household characteristics and to do so, Ritonga (1994, 44-47) used methods for incorporating household characteristics to extend the AIDS model. This can be written as wᵢ = α*ᵢ + βᵢ (log Y/P) + ∑j ϒᵢj log pj + ∑j δᵢj log hj where like the AIDS model derived above, wᵢ is the household budget share on the ith commodity; αᵢ, α*ᵢ, βᵢ, ϒᵢj, and δᵢj are the parameters of the system that will be estimated; Y denotes total household income; pj denotes price of jth commodity group; P is the price index and hj is the jth household characteristic.
The results show that through the test of model specifications using Chi-square test, AIDS model that includes household characteristics is considered as the best model because its improvement over the first model was significant at 1 percent level thus, this model was used to observe the expenditure patterns of a household given prices of commodity, household income plus other household characteristics. It must be noted that the household demand system estimated in this study comprised of sixteen commodity groups (See table 2.1). Within the sixteen commodity groups, budget shares on housing and on rice and grains appeared to be the two largest for households, higher income (holding prices constant), reduces a percentage spent in food, the higher the age of the household head, the more likely the household is to consume food prepared at home than spend for processed food & drinks, tobacco & betel nut, the more educated the household head, the more likely the household was to consume foods with high protein, the findings also indicate that the impact of each household characteristic used in the study were inelastic for all budget shares, homeowners spent significantly more on housing than renters, households whose main income from agricultural sector (either employed or self-employed) spent significantly more on rice and grains, cassava, and fish than those whose main income was from being employed in another sector, childless couples, single persons, and single-parent households spent significantly more on processed food and drinks than did two-parent households and budget shares of elderly households were less than that of nonelderly households for all expenditure groups except rice and grains, fish, vegetables, and clothing (Ritonga 1994, 80-124).
Burney and Khan (1992, 57-72) emphasized the importance of observing functional form transformations and comparing the differences after estimation. Using three different models varying in functional form, they analyzed household consumption patterns in Pakistan by estimating expenditure of different commodities and examined the effect of household size and other household composition on consumption pattern. They first employ the linear model of the Engel curve represented by the equation: Eij = αj + βj Ei + ϒj + HSi +∑ᵟ ᶥj Zᶥj where, i = 1,2…..,n households; j= 1,2….,m commodity groups; l= 1,2…,h demographic groups; Eij = expenditure of ith household (Rupees per month); Ei = total expenditure; HSi = number of members in ith household; Zᶥj = number of lth type of member in ith household. The next step was to transform the Engel curve equation into its double-logarithmic form and this can be written as: ln Eij = ϴj + λj ln Ei + ηj ln HSi + ∑ wᶥj Zᶥj and lastly, using the Working-Lester form which relates budget shares linearly to the logarithm of per capita total expenditure, logarithm of household-size and other variables for demographic characteristics of the households. It is represented by: sᵢj= φj+ λj ln Ei + ϩj ln HSi+ ∑ rᶥj Zᶥj where sᵢj= the expenditure share of the jth commodity in total expenditure of the ith household. The Engel’s curve estimation pays no attention to other compositions within households therefore Burney and Khan (1992, 57-72) incorporated other demographic compositions and studied their influence on consumption pattern.
Using the Household Income and Expenditure Survey (HIES) for the year 1984 and 1985, Burney and Khan (1992, 57-72) aggregated the list of commodities included in twelve major...