Professional Documents
Culture Documents
_9 Bopape Document
_9 Bopape Document
†
Lesiba Bopape and Robert Myers
Abstract
This study analyzes food expenditure patterns in South Africa, taking into account
differences in demand behavior across rural and urban households, as well as across
income groups. The analysis is carried out using the QUAIDS model, accounting for
demographic effects, structural change, and seasonality effects. Expenditure endogeneity
is also tested and controlled for. The study makes use of household food consumption
data, collected as part of the KwaZulu-Natal Income Dynamics Study. Demand behavior
differs significantly between rural and urban households, as well as across income
groups, implying that an accurate analysis of expenditure patterns in South Africa
requires a disaggregated analysis that takes into account these differences in demand
behavior.
Selected paper prepared for presentation at the African Econometrics Society Annual
Conference, Cape Town, South Africa, July 4-6, 2007
†
Lesiba Bopape is an economist, Economic Statistics Cluster, Statistics South Africa, Pretoria, South
Africa; and Robert Myers is University Distinguished Professor, Department of Agricultural Economics,
East Lansing, Michigan, U.S.A.
This work is part of Bopape’s Ph.D. dissertation at Michigan State University. We thank Jeffrey
Wooldridge, John Staatz, and Kelly Raper for their helpful comments. All remaining errors are our own.
Analysis of Household Demand for Food in South Africa: Model Selection,
Expenditure Endogeneity, and the Influence of Socio-Demographic Effects
1. Introduction
South Africa has the highest per capita income in Sub-Saharan Africa and is
data suggest that South Africa is food secure in almost all basic foodstuffs (IFSS, 2002).
These facts suggest that hunger and food security should not be major policy issues in the
country. However, these aggregate data mask a highly unequal distribution of income and
a huge divide between relatively affluent urban areas and destitute conditions in many
rural communities. For instance, a study to assess the household level food security status
indicated that over 30% of the population is vulnerable to food insecurity and over 20%
of the children are estimated to be stunted and vitamin A deficient (HSRC, 2004).
an important part of the social and economic reforms introduced by the democratic
government in 1994. But income inequality and household food insecurity remain. One
of the problems is that little is known about how food expenditure patterns differ across
different income groups, and across different geographic regions. Without a thorough
these patterns are changing over time, it will continue to be difficult to design policies
that improve food security effectively over a broad range of heterogeneous low-income
households.
patterns in South Africa, taking into account differences in demand behavior across rural
1
and urban households, as well as across income groups. The study makes use of an
unusually rich dataset on household food consumption, collected as part of the KwaZulu-
Natal Income Dynamics Study (KIDS). The KIDS dataset contains detailed information
characteristics. The KIDS interviewed households over a ten-year period, with surveys in
This paper uses the KIDS data to estimate demand functions for seven food
groups grains, meat and fish, fruits and vegetables, dairy, oils and fats, sugar, and all
other foods. The Quadratic Almost Ideal Demand System (QUAIDS) of Banks et al.
(1997) is used to estimate price and expenditure elasticities, taking into account the
Previous studies on food demand in South Africa have typically been based on
highly aggregate data and have either been limited to examining only one commodity
(e.g., Taljaard, 2003; Nieuwoudt, 1998) or ignored any impact of demographic factors on
food demand (Bowmaker and Nieuwoudt, 1990; Liebenberg and Groenewald, 1997). The
exception is the study by Agbola (2003) which is based on micro data and incorporates
household demographics. However, Agbola uses cross-sectional data that was collected
in 1993, one year prior to the major reforms introduced by the democratic government.
System (LA/AIDS) model, which does not allow for adequate curvature in the Engel
curves. In a related study using the KIDS dataset, Bopape and Myers (2006) explicitly
test for whether the demand model should be specified with a quadratic (QUAIDS) or a
linear (AIDS) expenditure term and found evidence against AIDS. This study also tests
2
for expenditure endogeneity and controls for it where necessary. The problem of
2. Theoretical Framework
The almost ideal demand system (AIDS) of Deaton and Muellbauer (1980) has
been a popular functional form to model demand behavior during the past two decades.
The AIDS model has budget shares that are linear functions of log total expenditure.
demand models (Muellbauer, 1976), which are derived from indirect utility functions that
are themselves linear in log total expenditure. However, there is a growing body of
literature providing evidence on the importance of allowing for nonlinearity in the budget
share equations (Lewbel, 1991; Banks et al., 1997). The quadratic almost ideal demand
system (QUAIDS) model developed by Banks et al. (1997), which has budget shares that
are quadratic in log total expenditure, is an example of the empirical demand systems that
−1
ln x − ln a (p) −1
ln V = + λ (p ) (1)
b(p )
of degree one in prices, and b(p) and (p) are functions that are homogeneous of degree
3
zero in prices. As in the original AIDS model, ln a(p) and ln b(p) are specified as the
K
1 K K
ln a(p ) = α 0 + ∑α i ln pi + ∑∑ γ ij ln pi ln p j (2)
i =1 2 i =1 j =1
K
b(p ) ∏p
βi
= i (3)
i =1
K K
λ( p) = ∑ λi ln pi where ∑λ = 0 .i (4)
i =1 i=1
Application of Roy’s identity to (1) gives the QUAIDS budget share equations. To
incorporate demographic variables (z) into the QUAIDS model through the linear
demographic translating method (Pollak and Wales, 1981). This leads to the following
2
K
x λi x L
wi = α i + ∑ γ ij ln p j + β i ln +
ln ∑δ is z s
+ (5)
j =1 a(p ) b(p ) a (p ) s =1
expenditure and price elasticities are derived by differentiating the budget share equations
with respect to ln x and ln pj, respectively. Following Banks et al. (1997), we simplify the
∂wi 2λi x
µi ≡ = βi + ln
b(p ) a(p )
(6)
∂ ln x
2
∂wi K
λ β x
µ ij ≡ = γ ij − µi α j + ∑ γ jl ln pl − i j ln . (7)
∂ ln p j l =1 b(p) a (p )
4
In terms of the µi , the formula for expenditure elasticities can be written as:
µi
ei = 1 + . (8)
wi
The expression for the Marshallian or uncompensated price elasticities can be written as:
µ ij
eiju = − δ ij (9)
wi
Hicksian or compensated price elasticities are obtained from the Slutsky equation:
The theoretical restrictions of adding-up, homogeneity, and symmetry are imposed on the
parameters to ensure integrability of the demand system (Moro and Sckokai, 2000).
Addding-up simply requires that the household does not spend more than its total budget
K K K K
∑αi = 1
i=1
∑ βi = 0
i =1
∑ λi = 0
i =1
∑γ
i=1
ij =0 ∀ j. (11)
Marshallian demands are homogenous of degree zero in (p, m ) , and this property is
∑γ
j =1
ij =0 ∀ j. (12)
∑γ
j =1
ij =0 ∀ j. (13)
5
2.2. Expenditure Endogeneity
Most empirical demand analyses do not cover all products and services that households
purchase. Data limitations, finite computer memory, and the increased complexity and
time required for estimating large models make it necessary to abstract from a completely
specified demand system containing a different equation for each of the myriad goods
available in the market (LaFrance, 1991). The practice is typically to assume that
preferences are separable and estimate a set of conditional demands for the goods of
interest as functions of prices and total expenditure on these goods (Pollak and Wales,
1969). However, such a practice raises questions regarding the possibility of simultaneity
bias in the budget share equations of the demand model. Total expenditure may be
determined jointly with the expenditure shares of the individual commodities that enter
the demand model, making it endogenous in the expenditure share equations. Estimation
In this study, we follow Bundell and Robin (1999) and control for endogeneity using an
Hausman (1978).
To illustrate how the augmented regression technique works, consider the regression
variable, z2, exists for y2. Correction for the endogeneity of y2 using the augmented
regression technique proceeds in two steps. The first step involves estimating a reduced
form regression of the endogenous variable on a set of instrumental variables, where the
set of instrumental variables include all the other exogenous explanatory variables (i.e.,
6
regress y2 on z and z2). The residuals, v , from this first-stage regression are then included
the parameters and in this augmented regression is identical to the Two-Stage Least
Squares (2SLS) estimator (Blundell and Robin, 1999). Moreover, testing for the
significance of the coefficient on v is a test for the exogeneity of y2. We use total
For income to be a good instrumental variable for expenditure, it must meet two
conditions: the relevance condition, which requires that income be sufficiently correlated
with expenditure (the endogenous variable), and the exogeneity condition, which requires
that income must not be correlated with the error term in the demand model. The former
condition is testable while the latter condition is maintained. A test for the relevance
We conduct this test following Wooldridge (2002, pp. 118-122) by determining whether
not unreasonable to assume that it (total household income) satisfies the exogeneity
condition.
3. Data
Data used in this study comes from the KwaZulu-Natal Income Dynamics Study
(KIDS). KIDS is a panel dataset comprising three surveys: the 1993 Project for Statistics
on Living Standards and Development (PSLSD) survey, and the 1998 and 2004 surveys
7
which interviewed households from the 1993 PSLSD survey who reside in KwaZulu-
Natal Province.
consortium of South African survey groups and universities, including the South African
Labor and Development Research Unit of the University of Cape Town and the World
Bank. The main instrument was a comprehensive household survey collecting data on a
broad array of socio-economic conditions of households, including their food and non-
was administered in each cluster to collect data on prices for a detailed list of food
KwaZulu-Natal Province were reinterviewed by the KIDS survey. The third KIDS survey
was undertaken in 2004. KwaZulu-Natal is the most populous province in South Africa,
constituting approximately 20% of South Africa’s population. The economic, social, and
racial stratification of KwaZulu-Natal mirrors that of the country as a whole: the province
includes a wealthy metropolitan area, Durban, poor townships surrounding it and a poor
and largely rural former homeland, KwaZulu. Also, poverty and inequality in the
province are relatively similar to those at the national level (Woolard et al., 2002), so that
results of the analysis should provide important insights about the conditions in other
provinces.
The recall period in all the three surveys is one month. The 1998 and 2004
surveys re-interviewed the original 1993 households, and tracked and interviewed
households that had moved, as well as new households that split from the original 1993
households. After removing observations deemed unusable for the current purpose (such
8
as households reporting zero consumption on all food items, or those missing critical
the three surveys is used. The problem of reported zero expenditures is not significant in
this dataset, mainly because of the broad commodity group definitions used here. Table 1
presents a summary of the food expenditure shares of the sampled households, including
the differences across income groups, and rural and urban areas. ‡ Grains constitute the
largest share of household total food expenditure, ranging from about 26 and 24% among
the high income and urban households to 37% among the low income and rural
households. The share of grains in the households’ budgets is lower at higher income
levels. The budget share of meat products, a more expensive source of calories, is higher
among the high income and relatively affluent urban households. The mean monthly
income of high income households is seven times more than that of low income
‡
A detailed listing of the food items contained in the food categories are
9
Table 1. Average Expenditure Shares by Income Group and Region
Table 2. On average, urban households are of smaller size, headed by younger males with
households, except the latter have larger family sizes. The rural and low-income groups
10
4. Demand Model Results
estimation. Table 3 presents the estimated coefficients for the QUAIDS model. The
equations. It is in the expenditure share equations for meat and fish, fruits and vegetables,
and oils and fats that the null hypothesis of expenditure linearity is not rejected. However,
as the results to be presented below indicate (Table 4), the hypothesis that the quadratic
expenditure term is zero across all equations is strongly rejected. Only 9 of the 28 price
effects are significantly different from zero at the 10% significance level, suggesting that
there is not much quantity response to movements in relative prices, possibly due to the
level of aggregation in the commodity groups. Most (34 out of 49) of the coefficient
estimates on the demographic variables are statistically different from zero. Households
with large sizes consume more grains and dairy products while their small-sized
counterparts consume more meat and fish and fruits and vegetables. These results are as
expected, given that grains provide a relatively cheap source of calories compared to such
foods as meat and fish, and that household size is negatively correlated with income.
Also, large households are likely to have more children who consume milk and other
dairy products.
11
Table 3 Estimated Price, Income, and Demographic Effects
Grains Meat/fish Fruits/Veg Dairy Oils/butter/ Sugar Other
products fats foods
Constant 0.1185 0.2965 0.2328 0.0404 0.0571 0.0151 0.2395
(0.050) (0.0234) (0.0167) (0.0141) (0.0084) (0.0071) (0.0214)
Race (1= black) 0.0814 0.0257 -0.0189 -0.0293 -0.0096 0.004 -0.0534
(0.0093) (0.0089) (0.0065) (0.0053) (0.0031) (0.0026) (0.0080)
Rural (1= rural) 0.0499 -0.0482 0.0193 -0.0159 -0.0019 0.0061 -0.0093
(0.0066) (0.0062) (0.0045) (0.0036) (0.0021) (0.0018) (0.0055)
12
Table A1 in the appendix presents the parameter estimates of the reduced form
regression for ln x. The results of the F test for the joint significance of ln m and (ln m)2
are presented in the bottom row of Table A1. These results provide evidence of a strong
partial correlation between ln m and (ln m)2 and ln x, and hence, provide evidence that
income is a relevant instrument for expenditure. Results of the test for expenditure
exogeneity are presented at the bottom row of Table 3. The t values for the significance
of the v residuals in the individual budget share equations are clearly low, implying that
The middle column of Table A1 in the appendix presents the parameter estimates of the
reduced form regression for ln x. The results of the F test for the joint significance of ln m
and (ln m)2 are presented in the bottom row of Table A1. These results provide evidence
of a strong partial correlation between ln m and (ln m)2 and ln x, and hence, provide
evidence that income is a relevant instrument for expenditure. Results of the formal test
of the hypothesis that log expenditure is exogenous across all the budget share equations
are presented in Table 4. As these results clearly show, expenditure exogeneity cannot be
rejected. Hence, in estimating the demand system using this sample of households, there
variables clearly influence demand behavior. Aggregate time effects, captured by year
dummy variables, are significant, which provides some evidence of structural change
during the sample period. The month of the survey is also significant, which indicates the
13
importance of seasonality in the food purchase and consumption patterns of these
households.
Table 4 Results of the Wald Tests for AIDS specification, Demographic effects, Structural
Change, and Seasonality
2
Degrees of freedom p-value
of expenditure elasticities ( i’s) are presented in Table 5. The first column presents
expenditure elasticity estimates for the entire sample, while the other columns reports
those for the rural, urban, and income groups sub-samples. All expenditure elasticity
estimates are positive, as would be expected for broadly defined food aggregates like the
ones considered here. Most (39 out of 42) of the estimates are statistically different from
zero at less than 1% significance level. Grains are expenditure inelastic across all
household groups, with substantial rural-urban and income group differences. These
estimates for expenditure elasticity for grains differ significantly from the earlier
estimates ( i = 1.250) reported by Agbola (2003) using the 1993 nationwide cross-
sectional data. Meat and fish are luxuries across all household groups; expenditure on
meat and fish is more elastic among rural and low income households than among urban
14
and high income households. The expenditure elasticity estimates for meat and fish are
close to the 1.027 reported by Agbola. Expenditures on fruits and vegetables and oils,
butter and fats are relatively more responsive to increases in total household expenditure
among urban and high income households than among rural and low income households.
Expenditure elasticities ( i )1
Table 6 presents estimates of the Marshallian and Hicksian own price elasticities.
Own price elasticities are all negative as expected. Based on the uncompensated price
elasticity estimates, fruits and vegetables and dairy products are price inelastic across all
household groups, and all other food groups are price elastic. Hence, households respond
more than proportionately to changes in the prices of these foods. However, when the
substitution effects are considered, grains and meat and fish become price elastic, with
compensated own-price elasticity estimates of less than unity. It is only in the case of
high income households that grains remain price elastic when both the uncompensated
15
and compensated elasticity estimates are considered. This indicates the greater
16
Agbola’s uncompensated and compensated price elasticity estimates for grains are –1.730
and –1.394, respectively. Grains comprise mainly of items that are staple foods in South
The expenditure and price elasticities estimated in this study correct for many of
the problems associated with previous estimates in South Africa. Our estimates are based
on the flexible QUAIDS model that allows for more curvature in the Engel curves, with
account taken for the influence of demographic, seasonality and aggregate time effects.
Hence, the price and expenditure elasticity estimates presented in this study are more
5. Conclusions
South Africa. Various tests are performed to decide on the most appropriate model
specification. The traditional AIDS model is rejected in favor of the more flexible
QUAIDS model. This is the first study to apply the QUAIDS model to analyze food
demand in South Africa; previous studies were based on restrictive functional forms
LA/AIDS and log-linear models. The QUAIDS model is estimated accounting for
and fish are luxuries across all household groups. Expenditure on meat and fish is more
elastic among rural and low income households than among urban and high income
households. Sugar products and oils, butter and fats are own-price elastic based on both
the compensated and uncompensated price elasticity estimates. Grains, which comprise
17
items that are mainly staple foods in South Africa, are price elastic among high income
households.
There are substantial differences in the consumption patterns of rural and urban
enhancement programs needs to be region-specific and take into account these behavioral
18
Appendix
Variable
ln x std. Err.
Constant 7.0905 (0.3453)
Household size 0.0465 (0.0031)
Race (1 if black) -0.3944 (0.0427)
Rural (1 if rural) 0.0129 (0.0308)
Education 0.0118 (0.0026)
Dummy for 1998 -0.4460 (0.0904)
Dummy for 2004 -1.0427 (0.1411)
Survey month -0.0050 (0.0094)
Total household income -0.0409 (0.0706)
Total household income2 0.0152 (0.0049)
lnprices Jointly significant
R2 0.4364
19
References
Agbola, F.W. “Estimation of Food Demand Patterns in South Africa Based on a Survey
of Households.” Journal of Agricultural and Applied Economics 35(3)(December
2003):662-670
Banks, J.; Blundell, R., and A. Lewbel. “Quadratic Engel Curves and Consumer
Demand.” The Review of Economics and Statistics 79(4)(November 1997):527-
539.
Blundell, R.W. and J.M. Robin. “Estimation in Large and Disaggregated Demand
Systems: An Estimator for Conditionally Linear Systems.” Journal of Applied
Econometrics 14(1999):209-232.
Bowmaker, P.A., and W.L Nieuwoudt. “Demand equations for selected South African
agricultural commodities.” Agrekon 29(1)(1990):46–50.
Deaton, A., and J. Muellbauer. “An Almost Ideal Demand System.” American Economic
Review 70(3)(June 1980b):312-336.
Dhar, T., J.P. Chavas, and B.W. Gould. “An Empirical Assessment of Endogeneity Issues
in Demand Analysis for Differentiated Products.” American Journal of
Agricultural Economics 85(3)(August 2003):605-617.
Döckel, J.A., and J.A. Groenewald. “The demand for food in South Africa.” Agrekon
9(4)(1970):15-20.
Elliot, M.B., and J. van Zyl. “The effect of income on the market for carbohydrates in
South Africa.” Agrekon 30(4)(1991):269-71.
Human Science Research Council (January 2004). Food Security in South Africa: Key
Policies Issues for the Medium Term (Online). Available from:
www.sarpn.org.za/documents/d0000685/Food_security_SA_January2004.pdf
20
Integrated Food Security Strategy (IFSS). National Department of Agriculture Policy
Document. Pretoria, South Africa, 2002.
Loubser, M. (1990). “Income Elasticities of the demand for consumer goods and
services.” Bureau of Market Research No. 175, University of South Africa,
Pretoria.
Lubbe, W.F. “Decomposition of price time series components of the red meat industry for
efficient policy and marketing strategies.” Agrekon 29(4)(1990): 296 - 304.
May, J., and I. Woolard. “Measurement and experience”. In J. May (Ed.), Poverty and
inequality in South Africa: report prepared for the office of the executive deputy
president and the inter-ministerial committee on poverty and inequality. Durban,
Praxis Publishing, 1998.
Nieuwoudt, W.L. “The demand for livestock products in South Africa for 2000, 2010 and
2020: Part 1." Agrekon 37(2)(1998).
Poi, B. P. “From the Help Desk: Demand System Estimation.” Stata Journal
2(4)(2002):403-410.
Pollak, R. A., and T.J. Wales. “Estimation of the Linear Expenditure System.”
Econometrica 37(4)(October 1969): 611-627.
Poonyth, D., R. Hassan, and J.F. Kirsten. “Random Coefficients Analysis of Changes in
Meat Consumption Preferences in South Africa.” Agrekon 40 (2001):426-437
Taljaard, P.R., Z.G. Alemu, and H.D. van Schalkwyk. “The Demand for Meat in South
Africa: An Almost Ideal Estimation.” Agrekon 43(4)(December 2004):430-443.
Woolard, I., S. Klasen, and M. Leibbrandt. “Income mobility and household dynamics in
South Africa: The case of KwaZulu-Natal.” Labour Markets and Social Frontiers
2(2002):5-11.
Wooldridge, J.M. Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: MIT Press, 2002.
World Bank. GNI Per Capita 2004, Atlas Method and PPP. At: www.worldbank.org
21