00074918.2015.1111997
Bulletin of Indonesian Economic Studies
ISSN: 0007-4918 (Print) 1472-7234 (Online) Journal homepage: http://www.tandfonline.com/loi/cbie20
Diet Transition and Supermarket Shopping
Behaviour: Is There a Link?
Hery Toiba, Wendy J. Umberger & Nicholas Minot
To cite this article: Hery Toiba, Wendy J. Umberger & Nicholas Minot (2015) Diet Transition and
Supermarket Shopping Behaviour: Is There a Link?, Bulletin of Indonesian Economic Studies,
51:3, 389-403, DOI: 10.1080/00074918.2015.1111997
To link to this article: http://dx.doi.org/10.1080/00074918.2015.1111997
Published online: 29 Nov 2015.
Submit your article to this journal
Article views: 50
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=cbie20
Download by: [Universitas Maritim Raja Ali Haji]
Date: 17 January 2016, At: 23:17
Bulletin of Indonesian Economic Studies, Vol. 51, No. 3, 2015: 389–403
DIET TRANSITION AND SUPERMARKET
SHOPPING BEHAVIOUR: IS THERE A LINK?
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Hery Toiba*
Brawijaya University
Wendy J. Umberger*
University of Adelaide
Nicholas Minot*
International Food Policy Research Institute
Supermarkets are leading the transformation of food markets in Asia, yet few studies have examined the impact of this so-called supermarket revolution on diet
transition and the related nutritional and health implications. We use data from
a sample of 1,180 urban households in Indonesia to explore the relation between
the increased use of modern food-retail outlets and the emergence of unhealthy
dietary patterns. The results of our ordinary least-squares and instrumental variables regressions suggest a negative and signiicant relation between the share of
food expenditure at modern food retailers and the healthiness of consumer food
purchases, even after we control for other characteristics that may inluence foodconsumption decisions.
Keywords: supermarket revolution, consumers, diet quality, Lewbel approach, diet transition
JEL classiication: I12, I15, F63, F68, Q13, Q18
INTRODUCTION
Economic growth, urbanisation, and foreign direct investment have all contributed to the rapid rise of domestic and international supermarket chains in developing and emerging economies, including Indonesia (Faiguenbaum, Berdegue,
and Reardon 2002; Reardon et al. 2003). The transformation of the food-retail
sector, or the so-called supermarket revolution, is having a profound effect on
the market conditions faced by both producers and consumers in many countries (Faiguenbaum, Berdegué, and Reardon 2002). Previous research on the
* We thank Wahida from the Indonesian Center for Agricultural Socio Economic and Policy Studies (ICASEPS) and Randy Stringer from Global Food Studies, at the University of
Adelaide, for their invaluable insight and support during the research-design and datacollection phases of this research. We also thank the ICASEPS research team for their assistance with all phases of the research project. The project was funded by the Australian
Centre for International Agricultural Research (grant ADP/2005/006). All views, interpretations, and conclusions are those of the authors and not necessarily those of the supporting or cooperating institutions.
ISSN 0007-4918 print/ISSN 1472-7234 online/15/000389-15
http://dx.doi.org/10.1080/00074918.2015.1111997
© 2015 Indonesia Project ANU
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
390
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
emergence of supermarkets in developing countries has focused on their impact
on smallholder farmers, rural development, small traders, and consumer purchasing habits,1 but only a handful of studies have considered their impact on foodconsumption patterns and the related health implications (such as Asfaw 2008;
Tessier et al. 2008; Banwell et al. 2013; Zhang, Van der Lans, and Dagevos 2012;
and Kelly et al. 2014). There is, however, increasing speculation that supermarket
penetration is one cause of the dramatic shift in Asian diets towards more westernised diets, typiied by the increased consumption of reined carbohydrates,
fats, and oils at the expense of grains, fresh fruits, and vegetables (Popkin 1999,
2006; Mendez and Popkin 2004; D’Haese and Van Huylenbroeck 2005; Asfaw
2008).
The World Health Organization is concerned about the impact of dietary
changes on the health status of consumers in many developing Asian countries, because the consumption of highly processed foods has been linked to an
increased risk of non-communicable chronic diseases such as cardiovascular disease and Type 2 diabetes (Asfaw 2008). The diffusion of supermarkets can have
important dietary implications, both positive (such as more diverse diets, lower
food prices, and increased food accessibility) and negative (such as greater inequalities in food accessibility and increased consumption of nutrient-poor, highly
processed foods) (Hawkes 2008; Banwell et al. 2013).
Previous empirical studies, using data from Tunisia, Guatemala, and Thailand,
have examined the relation between supermarket use and diet quality, but
their results have been mixed (Tessier et al. 2008; Asfaw 2008; Kelly et al. 2014).
Additional research is needed to understand the implications of supermarket-sector growth and diet transition in Indonesia. We use data from a survey of urban
households in Java to examine whether a higher share of food expenditures in
modern food-retail outlets (supermarkets) is associated with greater consumption
of speciic ‘unhealthy’ food products. We apply methods similar to those of Volpe,
Okrent, and Leibtag (2013) to develop household-level measures for, irst, shares
of expenditure on ‘healthy’ and ‘unhealthy’ foods, and, second, shares of food
expenditure in supermarkets, which include all types of modern food retailers—
hypermarkets, supermarkets, and convenience stores (mini-marts)—in Indonesia.
Indonesia, particularly Java, is an appropriate setting in which to explore the
relation between diet quality and supermarket use. Indonesia has been undergoing signiicant economic growth and urbanisation. Furthermore, its food-retail
sector has been transforming since 1998, when the Indonesian government began
to allow foreign direct investment in food retailing (World Bank 2007; Reardon
1. For the impact of supermarkets on smallholder farmers, see, for example, the work of
Bignebat, Koç, and Lemeilleur (2009); Hernández, Reardon, and Berdegué (2007); and
Michelson, Reardon, and Perez (2012). On rural development, see that of Dries, Reardon,
and Swinnen (2004); Hu et al. (2004); Neven et al. (2009); Rao and Qaim (2011); Reardon,
Stamoulis, and Pingali (2007); Reardon et al. (2003); and Weatherspoon and Reardon
2003). On small traders, see studies by Cadilhon et al. (2006); Faiguenbaum, Berdegué,
and Reardon (2002); Gorton, Sauer, and Supatpongkul (2011); Schipmann and Qaim (2011);
Suryadarma et al. (2010); and Zhang and Pan (2013). On consumer purchasing habits, see
the work of Bai, Wahl, and McCluskey (2008); D’Haese and Van Huylenbroeck (2005);
Neven et al. (2006); and Rodriguez et al. (2002).
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
391
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
and Timmer 2014). Between 1999 and 2009, the number of supermarkets increased
by 67%, with around three-quarters of these owned by multinational companies
(Dyck, Woolverton, and Rangkuti 2012). The growth was much more intense in
Java, particularly around the greater Jakarta metropolitan area, Jabotabek, which
has a population of around 28 million and includes the cities of Bogor, Depok,
Bekasi, and Tangerang.
DATA AND METHODS
Data
The analysis in this article uses data from our survey of urban consumers, which
includes a sample of 1,180 urban households in Surabaya, Bogor, and Surakarta,
representing large, medium, and small cities, respectively (Umberger et al. 2015).
These three cities also represent different stages of urbanisation and supermarket
penetration, with higher modern food-retail penetration in Bogor and Surabaya
than in Surakarta. In addition, it is likely that there are ethnic and cultural differences in the populations of these cities. Within each city, we used stratiied
random sampling to select households, oversampling higher-income neighbourhoods and areas close to supermarkets. We calculated sampling weights on the
basis of the inverse of the probability of selection, and used them to calculate all
results presented here.
The survey was implemented from November 2010 to February 2011 by trained
and experienced enumerators, who conducted the interviews at the homes of the
respondents. The questionnaire covered household composition, housing and
asset ownership, shopping behaviour at different types of outlets, food-expenditure patterns, and perceptions of each type of food retailer. The food-expenditure
module collected information on the amount that households spent on 67 food
items and the most important type of retailer for each item. We identiied and considered shopping behaviour at seven types of food-retail outlets: hypermarkets,
supermarkets, convenience stores, small shops (warung), semi-permanent stands,
traditional wet markets, and peddlers. In this study, we use the terms ‘modern
food retailers’ and ‘supermarkets’ to mean hypermarkets, supermarkets, and convenience stores.2
Econometric Approach
To explore the possible link between diet quality and supermarket use in
Indonesia, we estimate the household-level share of food expenditures spent on
healthy food products as a function of the household-level share of food expenditures at modern retail outlets. Similar to Volpe, Okrent and Leibtag (2013), we
use the United States Department of Agriculture’s Guidelines for Healthy Eating
(USDA 2010) to categorise food items as either healthy or unhealthy. According to
these guidelines, healthy items include grains, roots and tubers, fruit, vegetables,
2. Hypermarkets are very large modern stores that occupy more than 8,000 square metres,
have at least 10 or more cash registers, and sell both food and other groceries. Supermarkets
are medium to large modern stores that occupy between 300 and 8,000 square metres, have
3–9 cash registers, and sell both food and other groceries. Convenience stores, or minimarts, are small modern stores with 1–2 cash registers (Suryadarma et al. 2010).
392
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
unprocessed meat, milk, and eggs. Unhealthy items include sugar, sweetened
food and beverages, fats, oils, and processed foods such as snack foods, processed
meat, and ready-to-eat meals. The general form of the econometric model used in
this study is as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
HealthyShare = f (Pct_modern, Female, Age, Age2, Education, Hourjob, Label,
Income, Household_size, Children5, Domestic, Activity, FAFH, Surabaya, Bogor) (1)
Table 1 provides the deinitions, means, and standard errors of all variables.
Following the approach of Asfaw (2008) and Volpe, Okrent, and Leibtag (2013),
we use as the dependent variable a continuous variable, HealthyShare, that represents the share of household food expenditures on healthy food products.
Pct_modern is the explanatory variable of interest. It is a continuous variable
representing the household share of food expenditures in supermarkets. Excluded
are food expenditures made at traditional retail outlets, including small shops,
semi-permanent stands, traditional wet markets, and peddlers. As discussed earlier, consumers who buy a larger share of their food at modern retail outlets may
buy fewer healthy food products for their household.
The survey enumerators interviewed the primary person responsible for making food-purchasing decisions for each household. We include several variables
representing respondents’ socio-demographic characteristics, because these characteristics may inluence household food expenditures on certain foods (Huston
and Finke 2003; Schroeter, Anders, and Carlson 2013; Umberger et al. 2015).
Female is a dummy variable equal to one if the respondent was female. Age and
Age2 represent the age of the respondent and the squared age of the respondent.
The partial effect of age on the share of healthy food expenditures is expected
to be positive, because previous studies have shown that individuals become
more concerned about the healthfulness of their diet as they get older (Frazao and
Allshouse 2003).
Education is the number of years of education completed by the respondent.
Respondents with higher levels of education may be more aware of the relation
between diet, nutrition and health, and thus decide to buy healthier food for their
households (Turrell and Kavanagh 2006). More educated consumers may also
have a lower future discount rate, making them more likely to buy healthy food as
an investment in their family’s future (Huston and Finke 2003; Schroeter, Anders,
and Carlson 2013).
Hourjob is the number of hours per week that the respondent spends at work.
Respondents who work longer hours may have less time to spend on household
food shopping and food-purchasing decisions (Mancino and Kinsey 2004). Label
is a factor score, based on a Likert-scale response, that represents respondents’
attitudes to a series of questions about their use of nutrition labels and their nutrition information preferences. Respondents with a larger positive factor show a
preference for nutrition information and use nutrition labels more frequently.
Respondents who are more concerned about nutrition are expected to spend more
on healthy food (Huston and Finke 2003).
We include several household characteristics that may inluence the share of
total food expenditures on healthy food. We use Income as a proxy for per capita household expenditures. Households with higher incomes may face fewer
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
393
TABLE 1 Deinitions and Descriptive Statistics for Variables
Variable
Deinition
% of food expenditures classiied
as healthy
Pct_modern
% of food expenditures made at
modern retail outlets
Female
Gender (0 = male; 1 = female)
Age
Respondent’s age (years)
Respondent’s age (years)2
Age2
Education
Years of education completed
by respondent
Hourjob
Number of hours per week
spent at respondent's job
Label
Factor scores representing
respondents’ nutrition concerns,
food label use & perceived
nutrition knowledge
Income
Household income ($ ’000/year)
Household_size Number of family members
living in household
Children5
Dummy for household with
children under 5 years old
(1= yes; 0 = otherwise)
Domestic
Dummy for households with a
domestic employee
(1 = yes; 0 = otherwise)
Activity
Average hours per week adults
in the household exercise
FAFH
% of food expenditures spent
on meals and beverages eaten
outside of the home
Surabaya
Dummy for Surabaya
(1 = Surabaya; 0 = otherwise)
Bogor
Dummy for Bogor
(1 = Bogor; 0 = otherwise)
Time_modern Time needed to get to the
nearest modern food retailer
(mins)
Food_safety
Dummy variable (1 = respondent
indicated that a retail outlet
offering food-safety assurances
is important or extremely
important when deciding
where to purchase food;
0 = otherwise)
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
HealthyShare
Mean
SD
Min
Max
68.71
15.05
2.93
98.52
13.39
14.97
0
83.71
0.89
43.02
2,004
9.35
0.32
12.40
1,152
4.52
0
15
225
0
1
83
6,889
22
21.11
25.01
0
130
–0.07
1.06
–4.94
2.42
4.93
4.41
4.91
1.76
0.32
1
77.73
12
0.58
0.49
0
1
0.16
0.37
0
1
2.11
3.13
0
35
7.41
10.75
0
95.39
0.61
0.49
0
1
0.21
0.41
0
1
18.98
10.45
1.5
120.0
0.96
0.19
0
1
Note: SD = standard deviation
economic barriers to buying healthy food, particularly if healthy food is more
expensive than unhealthy food (Smith, Bogin, and Bishai 2005). We include
Household_size because households with more family members to support may
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
394
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
have less disposable income (Schroeter, Anders and Carlson 2013). Children5 is a
dummy variable equal to one if there are children under ive years of age in the
household. Households with young children may make different food-purchasing
decisions, owing to the speciic dietary needs of young children (Umberger et al.
2015). Domestic is a dummy variable equal to one if the household has a domestic
employee who helps with household duties, including shopping for food.
Activity is a continuous variable representing the average number of hours
per week that adults in the household spend exercising. Households with adults
who exercise more may place more value on foods that are healthy, and therefore may have a higher share of food expenditures on healthy food (Huston and
Finke 2003; Komlos, Smith, and Bogin 2004; Mancino and Kinsey 2004). FAFH is
an additional measure of the impact of the food-market environment on foodconsumption behaviour. It is the household’s share of total food expenditures
on food-away-from-home (FAFH), including meals and beverages prepared and
eaten outside the home. Previous research has found that individuals who consume more FAFH tend to have poorer diets as a result of other factors in the
food-market environment (Bezerra and Sichieri 2009; Drichoutis, Nayga, and
Lazaridis 2012).
Surabaya and Bogor are dummy variables representing city-level effects.
Compared with Surakarta, Surabaya and Bogor are more metropolitan and have
more mature modern food-retail sectors. There may also be unobservable citylevel factors (such as social norms, cultural traditions, and climatic conditions)
that have an effect on food-expenditure shares (Umberger et al. 2015).
Estimation and Instrumental Variables
We irst estimate equation (1) using ordinary least squares (OLS), controlling for
the set of observable factors that may inluence food-purchasing decisions. Under
the OLS approach, however, the coeficient on the modern-share variable (Pct_
modern) may not accurately measure the effect of shopping at modern retailers on
healthfulness of food-consumption patterns, for three reasons. First, there may
be reverse causation: a household that prefers to eat unhealthy food may decide
to shop at supermarkets, potentially resulting in negative bias of the Pct_modern
coeficient in equation (1). Second, an unobserved variable may be inluencing
both the modern share and the unhealthy share of food spending. For example,
if watching television exposes viewers to advertisements for supermarkets as
well as advertisements for unhealthy foods, the OLS coeficient estimate of the
Pct_modern variable will be negatively biased. Third, the estimates of food expenditures may be subject to measurement error (Gibson and Kim 2007) and therefore
the OLS estimation of the impact on food-consumption patterns of using modern
retailers will also be biased. In all three cases, the explanatory variable is correlated with the error term, thus violating one of the assumptions behind OLS and
giving biased and inconsistent estimates of the coeficient.
We can address these problems by using an instrumental variable (IV) technique, in which we replace the potentially endogenous explanatory variable, Pct_
modern, with an estimated value of the variable. The estimated value is based on a
supplementary regression of the potentially endogenous explanatory variable as
a function of one or more IVs (or instruments). Under the IV approach, the main
equation is as follows:
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
yi = α 0 + α 1 M i + βi Xi + ε i
395
(2)
where yi is the share of food expenditures of household i spent on healthy food; Mi
is the estimated share of food expenditures of household i purchased at modern
food-retail outlets; Xi is a vector of the independent variables; α0, α1, and βi are
parameters to estimate; and εi is the error term. We estimate Mi as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
M i = β i Xi + γ Z i + υ i
(3)
where Zi is a vector of instrumental variables, γ is an estimated parameter, and υi
is the error term.
One dificulty in implementing IV estimations is inding valid instruments.
Instruments must meet the relevance condition, meaning that they must be
strongly correlated with the potentially endogenous explanatory variable (Pct_
modern). They must also meet the exclusion restriction, meaning they cannot be
correlated with any omitted variable that may help to explain the dependent variable (that is, they must not be correlated with the error terms) (Stock and Watson
2012).
A standard IV used in studying the impact of supermarket use is the distance
from the household to the nearest retail outlet (Volpe, Okrent, and Leibtag 2013).
In Indonesia, however, distance is not a good measure of access to shopping outlets and food markets, owing to the absence of good public transportation and
related trafic congestion in urban areas. As suggested by Narayan, Rao, and
Sudhir (2012), the cost of buying a product from a particular retail outlet includes
the cost of travelling to the store. This suggests that travel time to a particular
type of outlet will be negatively related to the likelihood of using that outlet. In
this study, we use as an IV a variable representing the time it usually takes the
respondent to get to the nearest modern retail outlet, Time_modern, even though
using this instrument may raise other concerns. Some consumers, for example,
may use different transportation modes to access retail outlets. To address this
issue, we apply different identiication strategies to check the robustness of the
instruments.
We include, as an additional IV, the dummy variable Food_safety, which equals
one if the respondent indicated that food-safety assurances are an important or
extremely important attribute when deciding where to purchase food. We consider this to be a valid IV because previous studies suggest that consumer concerns
about food safety have increased the use of modern retail outlets in developing
countries (Gorton, Sauer, and Supatpongkul 2011; Minten and Reardon 2008).
Inappropriate IVs could potentially give estimations inferior to OLS results. To
mitigate this risk, we also use Lewbel’s (2012) two-stage, heteroskedasticity-based
estimator, which involves irst creating ‘higher moment’ internal instruments that
are the product of a set of selected demeaned regressors and the residuals obtained
from the irst-stage regression. According to Lewbel, these instruments should,
irst, satisfy the exclusion restriction if they are uncorrelated with the other regressors and, second, meet the relevance condition if hetroskedasticty exists in the
irst stage. Lewbel cautions that ‘higher moments can lead to noisier, less reliable
estimates than exclusion based on standard exclusion restrictions’ (78). They may
be more appropriate, however, if the usual instruments are weak.
396
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
We follow methods used by Emran and Shilpi (2012); Schroeter, Anders, and
Carlson (2013); Volpe, Okrent, and Leibtag (2013); and Umberger et al. (2015). We
estimate equation (4) as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
M i = π ′Z i + ξi
(4)
where Zi is a vector of internal instruments that, according to Lewbel, can be either
a subset of or identical to Xi (Zi = Xi). In this analysis, we consider three exogenous
variables (Education, Income, and Label) for Zi. We estimate equation (4) by regressing (Mi) on this subset of exogenous variables, obtaining the residual ξ i. We then
use this residual to generate the higher-moment instruments
(Z
i
− Zi ) ξ̂i
where Zi is the mean of Zi and ξ̂i is the residual from equation (4). We use Z! i to
denote the higher-moment instruments. We then use two-stage least squares and
re-estimate equation (3), substituting Z! i for Zi in equation (3), in the irst stage,
and then re-estimating equation (2). Finally, to test the robustness of the estimates,
we conduct a fourth estimation, using two-stage OLS: we estimate equation (3)
by using both the original IVs (Time_modern and Food_safety) and the highermoment instruments ( Z! i ) in the irst stage. We use the Breusch-Pagan test to test
for heteroskedasticity:3
(Cov ( Z ξ ) ≠ 0 )
2
i, i
RESULTS
Table 2 reports the results of the estimations of the healthy expenditure share
models. As discussed earlier, the Pct_modern variable may be endogenous, and
the OLS results provided in the irst column of table 2 may be biased. We therefore provide in the other columns the coeficients estimated using the three IV
methods. These results give an insight into potential biases that may result from
endogeneity.
The results are similar across all four models, particularly for our OLS, Lewbel,
and IV-Lewbel estimations. The coeficient for the main explanatory variable of
interest, Pct_modern, is consistently statistically signiicant and negative, which
suggests that an increase in the share of food expenditures in modern retail outlets
is associated with a decrease in the share of food expenditures on healthy food.
There are some notable differences between the results of the standard IV estimation and the others. The estimated coeficient for Pct_modern from the standard
IV regression is much larger than the OLS, Lewbel, and IV-Lewbel coeficients.
Additionally, four variables (Hourjob, Domestic, Surabaya, and Bogor) are signiicant only in the standard IV estimation. Our results are consistent with those of
other studies (such as Gao and Smyth 2015 and Sabia 2007) in that the Lewbeland OLS-estimated coeficients are similar and both much higher than those estimated using the standard IV approach.
Before discussing additional signiicant covariates, we consider the validity
of the results of the alternative IV approaches using several test statistics. The
3. We use the ivreg2 command developed by Baum, Schaffer, and Stillman (2010).
TABLE 2 Regression Results: Estimation of Shares of Expenditure on Healthy Food
Pct_modern
Female
Age
Age2
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Education
Hourjob
Activity
Surabaya
Bogor
FAFH
Label
Income
Household_size
Children5
Domestic
Constant
Observations
R2
First-stage regression
Time_modern
Food_safety
First stage (Stock & Yogo) F-test
Endogeneity test
Breusch-Pagan chi2
Hansen J-statistic
OLS
IV
Lewbel
IV-Lewbel
–0.20***
(0.03)
1.83
(1.35)
0.62***
(0.19)
–0.01***
(0.00)
0.04
(0.01)
–0.02
(0.01)
0.18*
(0.10)
0.74
(0.95)
1.05
(1.03)
–0.19***
(0.04)
0.12
(0.42)
0.00
(0.00)
–0.39*
(0.220)
1.19
(0.73)
1.69
(1.05)
60.77***
(4.66)
–0.70**
(0.31)
3.69*
(1.96)
0.59***
(0.22)
–0.01**
(0.00)
0.76
(0.47)
–0.04*
(0.02)
0.24*
(0.13)
3.08*
(1.85)
2.91*
(1.66)
–0.22***
(0.05)
0.70
(0.60)
0.00
(0.00)
–0.40*
(0.24)
1.25
(0.82)
3.19**
(1.52)
57.92***
(5.76)
–0.15**
(0.07)
1.64
(1.36)
0.62***
(0.19)
–0.01***
(0.00)
–0.03
(0.13)
–0.02
(0.01)
0.18*
(0.10)
0.50
(1.00)
0.86
(1.06)
–0.19***
(0.04)
0.06
(0.42)
–0.00
(0.00)
–0.39*
(0.22)
1.19
(0.73)
1.54
(1.07)
61.07***
(4.67)
–0.17**
(0.07)
1.72
(1.36)
0.62***
(0.19)
–0.01***
(0.00)
–0.00
(0.13)
–0.02
(0.01)
0.18*
(0.10)
0.59
(0.99)
0.94
(1.05)
–0.19***
(0.04)
0.08
(0.42)
0.00
(0.00)
–0.39*
(0.22)
1.19*
(0.72)
1.60
(1.07)
60.95***
(4.67)
1,180
0.13
1,180
–0.16
1,180
0.12
1,180
0.125
60.21***
–0.11
(0.06)*
5.03
(1.46)***
41.36***
–0.10
(0.05)*
4.71
(1.50)***
7.22
2.78
(0.09)
189.06
(0.00)
0.63
(0.73)
5.33
(0.26)
Note: The dependent variable is HealthyShare. Standard errors are in parentheses. IV uses two-stage
least squares with Time_modern and Food_safety as external instruments. Lewbel uses the heteroskedasticity-based estimator, which takes ( Zi − Zi ) ξi as instruments, where Zi includes the variables Education,
Income, and Label. IV-Lewbel uses both the external instruments and the Lewbel approach.
* p < 0.10; ** p < 0.05; *** p < 0.01.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
398
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
endogeneity test result (2.78) is signiicant, suggesting that the Pct_modern variable is endogenous. The two instruments, Time_Modern and Food_Safety, are both
statistically signiicant in explaining Pct_modern. As we expected, their signs are
based on previous literature and suggest that they satisfy the relevance condition.
However, the irst-stage Stock and Yogo (2005) F-test from the standard IV estimation (table 2) indicates that these variables are relatively weak instruments. In this
case, the endogeneity test is unreliable.
To support our use of the Lewbel and IV-Lewbel models, we consider the results
of two additional tests. First, the Breusch-Pagan chi-square statistic (189.06) is signiicant (table 2) and we can reject the null hypothesis of homoskedasticity of the
error term (ξ i). Second, the Hansen J statistics, used to test for overidentiication of
all instruments for the Lewbel and the IV-Lewbel models, are not signiicant, which
conirms that the sets of instruments included in these models are valid. These
diagnostic test results support the use of the Lewbel and IV-Lewbel approaches
as an alternative to the standard IV method.4 Thus, after controlling for various
respondent and household characteristics, we ind that a one-percentage-point
increase in the share of food purchased at a modern outlets (Pct_modern) results in
a decrease of 0.15 to 0.17 percentage points in the share of healthy food purchases.
This inding supports the notion that increased use of supermarkets for food shopping may lead to less healthy food-purchasing behaviour among households.
In both models, the coeficient on Age is positive and signiicant, while the coeficient on Age2 is negative and signiicant. This implies that the share of expenditures on healthy food rises and then falls with the age of the person responsible
for shopping for food for the household. More speciically, according to the coeficients for the Lewbel and IV-Lewbel models, the healthy share reaches its maximum at the age of approximately 57.5 Contrary to expectations, we found that
older people (above 57) spend less on healthy food for their household, other
factors being equal.
The coeficients for Activity are positive and signiicant, which is consistent
with previous studies and suggests that households with active adults are also
more likely to buy healthier food. As expected, FAFH was negative and highly
signiicant (at 1%) in both models, implying that households spending more on
food-away-from-home tend to buy less healthy food.
The coeficient on Income was not signiicant in explaining healthy expenditure
shares in either estimation. This result is surprising, but there are two possible
explanations. First, many low-income households receive food subsidies from the
government, especially for rice, allowing them to spend more of their income on
other healthy foods (such as unprocessed meats, fresh milk, and eggs).6 Second,
4. The coeficients estimated using the standard IV approach (table 2) are not reliable, owing to weak instruments, and we provide them only for comparison. As such, we focus
the remaining discussion on the results from the estimations of the Lewbel and IV-Lewbel
models.
5. Taking the derivative of the regression equation with respect to age, the maximum occurs when b + 2(a)(Age) = 0, where b is the linear coeficient and a is the quadratic coeficient. Rearranging terms, we get Age = –b/2a = –0.622/0.011 = 56.5.
6. Since 1998, the Indonesian government has provided food subsidies for low-income
households as part of its Raskin rice-subsidy program (Sumarto and Bazzi 2011)
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
399
high-income households are better able to afford processed foods and convenience foods, which are considered to be less healthy.
A related variable, Household_size, was negative and statistically signiicant in
both models, perhaps because, for a given income, larger households have less
disposable income to spend on more expensive, healthier foods, so household
members may have poorer diets. The coeficient on the Children5 variable in the
IV-Lewbel model (table 2) is positive and statistically signiicant, suggesting some
evidence of a positive relation between having young children in the household
and having healthier household food-expenditure shares.
There was no statistically signiicant difference in the share of healthy food
expenditures in Surabaya, Bogor, and Surakarta. Any differences across cities are
due to differences in other explanatory variables.
CONCLUSIONS AND IMPLICATIONS
There is increasing speculation that consumers’ dietary practices and health outcomes can be affected by the strategic decisions of modern retailers. It is unclear
from the literature, however, whether there is a connection between the use of
modern retail outlets and unhealthy dietary patterns. Tessier et al.’s (2008) study,
for example, which used data from Tunisia, determined that the diet quality of
frequent users of supermarkets was slightly higher than that of less frequent
supermarket users. On the other hand, studies using data from Guatemala (Asfaw
2008), Thailand (Kelly et al. 2014), and Kenya (Rischke et al. 2015) found higher
supermarket use to be associated with an increased consumption of highly processed foods, which are generally considered to be unhealthy.
We have attempted here to build on previous empirical studies and shed light
on the relation between supermarket shopping behaviour and diet transition in
Indonesia, a topic that has not previously been researched. Using cross-sectional
data from a survey of Indonesian urban households, we found a negative and signiicant relation between the share of food expenditures in modern food retailers
and the healthfulness of consumer food expenditures. Given, however, that the
coeficients on the supermarket variables (Pct_modern) estimated using the OLS,
Lewbel, and IV-Lewbel models are relatively small, so, too, is the actual magnitude or impact of changing food expenditures on diet quality. Speciically, after
controlling for other variables, we found that an increase of one percentage point
in the share of food expenditures in supermarkets is associated with a relatively
small decrease (of 0.15 to 0.17 percentage points) in the household’s share of food
expenditures on healthy food.
Nevertheless, our results support those of Asfaw (2008), Kelly et al. (2014),
and Rischke et al. (2015). As in Thailand, Guatemala, and Kenya, supermarket
penetration in urban areas of Indonesia is relatively high, and supermarkets are
generally accessible to both low-income and high-income households. As Kelly et
al. (2014) explain, modern food retailers, including hypermarkets, supermarkets,
and convenience stores, may be able to compete with traditional retailers by providing less expensive processed foods, which tend to be higher in fat, salt, and
sugar content, and fewer fresh foods, particularly fruits and vegetables.
We also found that diet quality is negatively associated with greater
food-expenditure shares on meals and beverages bought outside the home.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
400
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
Policymakers in Indonesia who are concerned about diet transition may want
to consider aspects of the food environment other than just supermarket penetration. The development of one-stop shopping centres (such as shopping malls
where supermarkets are typically co-located with food courts), for example, may
exacerbate the negative effects of supermarket penetration on diet quality.
This study has several limitations. We used cross-sectional data, so we were
unable to capture dynamic patterns of consumption; panel data would have
yielded more robust estimates of causal relations. We did not measure the quantity of food consumed by individuals in households, which made it dificult to
isolate the effect of food expenditure shares on the intake of calories and nutrients.
And we focused on urban households in Java, where modern food-retail penetration is much greater than in rural areas and in other provinces. The relation
between modern food-retail penetration and health outcomes, in Indonesia and
elsewhere, requires more research.
REFERENCES
Asfaw, Abay. 2008. ‘Does Supermarket Purchase Affect the Dietary Practices of Households? Some Empirical Evidence from Guatemala’. Development Policy Review 26 (2):
227–43.
Bai, Junfei, Thomas I. Wahl, and Jill J. McCluskey. 2008. ‘Consumer Choice of Retail Food
Store Formats in Qingdao, China’. Journal of International Food and Agribusiness Marketing 20 (2): 89–109.
Banwell, Cathy, Jane Dixon, Sam-ang Seubsman, Suttinan Pangsap, Matthew Kelly, and
Adrian Sleigh. 2013 ‘Evolving Food Retail Environments in Thailand and Implications
for the Health and Nutrition Transition’. Public Health Nutrition 16 (4): 608–15.
Baum, Christopher F., Mark E. Schaffer, and Steven Stillman. 2010. ‘IVREG2: Stata module
for extended instrumental variables/2SLS and GMM estimation’. http://ideas.repec.
org/c/boc/bocode/s425401.html.
Bezerra, Ilana Nogueira, and Rosely Sichieri. 2009. ‘Eating Out of Home and Obesity: A
Brazilian Nationwide Survey’. Public Health Nutrition 12 (11): 2037–43.
Bignebat, Céline, Ahmet Ali Koç, and Sylvaine Lemeilleur. 2009. ‘Small Producers, Supermarkets, and the Role of Intermediaries in Turkey’s Fresh Fruit and Vegetable Market’.
Agricultural Economics 40 (s1): 807–16.
Cadilhon, Jean-Joseph, Paule Moustier, Nigel D. Poole, Phan Thi Giac Tam, and Andrew P.
Fearne. 2006. ‘Traditional vs. Modern Food Systems? Insights from Vegetable Supply
Chains to Ho Chi Minh City (Vietnam)’. Development Policy Review 24 (1): 31–49.
D’Haese, Marijke, and Guido Van Huylenbroeck. 2005. ‘The Rise of Supermarkets and
Changing Expenditure Patterns of Poor Rural Households Case Study in the Transkei
Area, South Africa’. Food Policy 30 (1): 97–113.
Drichoutis, Andreas C., Rodolfo M. Nayga, Jr, and Panagiotis Lazaridis. 2012. ‘Food Away
from Home Expenditures and Obesity among Older Europeans: Are There Gender Differences?’. Empirical Economics 42 (3): 1051–78.
Dries, Liesbeth, Thomas Reardon, and Johan F. M. Swinnen. 2004. ‘The Rapid Rise of
Supermarkets in Central and Eastern Europe: Implications for the Agrifood Sector and
Rural Development’. Development Policy Review 22 (5): 525–56.
Dyck, John, Andrea E. Woolverton, and Fahwani Yuliati Rangkuti. 2012. Indonesia’s Modern Retail Food Sector: Interaction with Changing Food Consumption and Trade Patterns.
Economic Information Bulletin 97. Washington, DC: United States Department of
Agriculture.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
401
Emran, M. Shahe, and Forhad Shilpi. 2012. ‘The Extent of the Market and Stages of Agricultural Specialization’. Canadian Journal of Economics 45 (3): 1125–53.
Faiguenbaum, Sergio, Julio A. Berdegué, and Thomas Reardon. 2002. ‘The Rapid Rise of
Supermarkets in Chile: Effects on Dairy, Vegetable, and Beef Chains’. Development Policy Review 20 (4): 459–71.
Frazao, Elizabeth, and Jane Allshouse. 2003. ‘Strategies for Intervention: Commentary and
Debate’. Journal of Nutrition 133 (3): 844S–847S.
Gao, Wenshu, and Russell Smyth. 2015. ‘Education Expansion and Returns to Schooling in
Urban China, 2001–2010: Evidence from Three Waves of the China Urban Labor Survey’. Journal of the Asia Paciic Economy 20 (2): 178–201.
Gibson, John, and Bonggeun Kim. 2007. ‘Measurement Error in Recall Surveys and the
Relationship between Household Size and Food Demand’. American Journal of Agricultural Economics 89 (2): 473–89.
Gorton, Matthew, Johannes Sauer, and Pajaree Supatpongkul. 2011. ‘Wet Markets, Supermarkets and the “Big Middle” for Food Retailing in Developing Countries: Evidence
from Thailand’. World Development 39 (9): 1624–37.
Hawkes, Corinna. 2008. ‘Dietary Implications of Supermarket Development: A Global Perspective’. Development Policy Review 26 (6): 657–92.
Hernández, Ricardo, Thomas Reardon, and Julio Berdegué. 2007. ‘Supermarkets, Wholesalers, and Tomato Growers in Guatemala’. Agricultural Economics 36 (3): 281–90.
Hu, Dinghuan, Thomas Reardon, Scott Rozelle, Peter Timmer, and Honglin Wang. 2004.
‘The Emergence of Supermarkets with Chinese Characteristics: Challenges and
Opportunities for China’s Agricultural Development’. Development Policy Review 22
(5): 557–86.
Huston, Sandra J., and Michael S. Finke. 2003. ‘Diet Choice and the Role of Time Preference’. Journal of Consumer Affairs 37 (1): 143–60.
Kelly, Matthew, Sam-ang Seubsman, Cathy Banwell, Jane Dixon, and Adrian Sleigh. 2014,
‘Thailand’s Food Retail Transition: Supermarket and Fresh Market Effects on Diet
Quality and Health’. British Food Journal 116 (7): 1180–93.
Komlos, John, Patricia K. Smith, and Barry Bogin. 2004. ‘Obesity and the Rate of Time Preference: Is There a Connection?’. Journal of Biosocial Science 36 (2): 209–19.
Lewbel, Arthur. 2012. ‘Using Heteroskedasticity to Identify and Estimate Mismeasured
and Endogenous Regressor Models’. Journal of Business & Economic Statistics 30 (1):
67–80.
Mancino, Lisa, and Jean Kinsey. 2004. ‘Diet Quality and Calories Consumed: the Impact of
Being Hungrier, Busier and Eating Out’. Working paper 04-02, Food Industry Center,
University of Minnesota, St Paul, MN.
Mendez, Michelle A., and Barry M. Popkin. 2004. ‘Globalization, Urbanization and Nutritional Change in the Developing World’. Journal of Agricultural and Development Economics 1 (2): 55–80.
Michelson, Hope, Thomas Reardon, and Francisco Perez. 2012. ‘Small Farmers and Big
Retail: Trade-Offs of Supplying Supermarkets in Nicaragua’. World Development 40 (2):
342–54.
Minten, Bart, and Thomas Reardon. 2008. ‘Food Prices, Quality, and Quality’s Pricing in
Supermarkets versus Traditional Markets in Developing Countries’. Applied Economic
Perspectives and Policy 30 (3): 480–90.
Narayan, Vishal, Vithala R. Rao, and K. Sudhir. 2012. ‘Early Adoption of Modern Grocery
Retail in an Emerging Market: Evidence from India’. Johnson School Research Paper
27-2012, Samuel Curtis Johnson Graduate School of Management, Ithaca, NY.
Neven, David, Michael Makokha Odera, Thomas Reardon, and Honglin Wang. 2009. ‘Kenyan Supermarkets, Emerging Middle-Class Horticultural Farmers, and Employment
Impacts on the Rural Poor’. World Development 37 (11): 1802–11.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
402
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
Neven, David, Thomas Reardon, Jonathan Chege, and Honglin Wang. 2006. ‘Supermarkets
and Consumers in Africa: The Case of Nairobi, Kenya’. Journal of International Food &
Agribusiness Marketing 18 (1–2): 103–23.
Popkin, Barry M. 1999. ‘Urbanization, Lifestyle Changes and the Nutrition Transition’.
World Development 27 (11): 1905–16.
———. 2006. ‘Global Nutrition Dynamics: The World is Shifting Rapidly toward a Diet
Linked with Noncommunicable Diseases’. American Journal of Clinical Nutrition 84 (2):
289–98.
Rao, Elizaphan J. O., and Matin Qaim. 2011. ‘Supermarkets, Farm Household Income, and
Poverty: Insights from Kenya’. World Development 39 (5): 784–96.
Reardon, Thomas, Kostas Stamoulis, and Prabhu Pingali. 2007. ‘Rural Nonfarm Employment in Developing Countries in an Era of Globalization’. Agricultural Economics 37
(s1): 173–83.
Reardon, Thomas, and C. Peter Timmer. 2014. ‘Five Inter-Linked Transformations in
the Asian Agrifood Economy: Food Security Implications’. Global Food Security 3 (2):
108–17.
Reardon, Thomas, C. Peter Timmer, Christopher B. Barrett, and Julio Berdegué. 2003. ‘The
Rise of Supermarkets in Africa, Asia, and Latin America’. American Journal of Agricultural Economics 85 (5): 1140–46.
Rischke, Ramona, Simon C. Kimenju, Stephan Klasen, and Matin Qaim. 2015. ‘Supermarkets and Food Consumption Patterns: The Case of Small Towns in Kenya’. Food Policy
52: 9–21.
Rodriguez, Elsa, Miriam Berges, Karina Casellas, Rosángela Di Paola, Beatriz Lupín, Laura
Garrido, and Natacha Gentile. 2002. ‘Consumer Behaviour and Supermarkets in Argentina’. Development Policy Review 20 (4): 429–39.
Sabia, Joseph J. 2007. ‘The Effect of Body Weight on Adolescent Academic Performance’.
Southern Economic Journal 73 (4): 871–900.
Schipmann, Christin, and Matin Qaim. 2011. ‘Modern Food Retailers and Traditional Markets in Developing Countries: Comparing Quality, Prices, and Competition Strategies
in Thailand’. Applied Economic Perspectives and Policy 3 (3): 345–62.
Schroeter, Christiane, Sven Anders, and Andrea Carlson. 2013. ‘The Economics of Health
and Vitamin Consumption’. Applied Economic Perspectives and Policy 35 (1): 125–49.
Smith, Patricia K., Barry Bogin, and David Bishai. 2005. ‘Are Time Preference and Body
Mass Index Associated? Evidence from the National Longitudinal Survey of Youth’.
Economics & Human Biology 3 (2): 259–70.
Stock, James H., and Mark W. Watson. 2012. Introduction to Econometrics: International Edition, 3rd ed. Boston, MA: Pearson Education.
Stock, James H., and Motohiro Yogo. 2005. ‘Testing for Weak Instruments in Linear IV
Regression’. In Identiication and Inference for Econometric Models: Essays in Honor of
Thomas Rothenberg, edited by Donald W. K. Andrews and James H. Stock, 80–108. New
York: Cambridge University Press.
Sumarto, Sudarno, and Samuel Bazzi. 2011 ‘Social Protection in Indonesia: Past Experiences and Lessons for the Future’. Paper presented at the 2011 Annual Bank Conference
on Development Economics, Paris, 30 May–1 June.
Suryadarma, Daniel, Adri Poesoro, Sri Budiyati, Meuthia Rosfadhila, and Asep Suryahadi.
2010. ‘Traditional Food Traders in Developing Countries and Competition from Supermarkets: Evidence from Indonesia’. Food Policy 35 (1): 79–86.
Tessier, Sophie, Pierre Traissac, Bernard Maire, Nicolas Bricas, Sabrina Eymard-Duvernay,
Jalila El Ati, and Francis Delpeuch. 2008. ‘Regular Users of Supermarkets in Greater
Tunis Have a Slightly Improved Diet Quality’. Journal of Nutrition 138 (4): 768–74.
Turrell, Gavin, and Anne M. Kavanagh. 2006. ‘Socio-Economic Pathways to Diet: Modelling the Association between Socio-Economic Position and Food Purchasing Behaviour’. Public Health Nutrition 9 (3): 375–83.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
403
Umberger, Wendy J., Xiaobo He, Nicholas Minot, and Hery Toiba. 2015. ‘Examining the
Relationship between the Use of Supermarkets and Over-nutrition in Indonesia’. American Journal of Agricultural Economics 97 (2): 510–25.
USDA (United States Department of Agriculture). 2010. ‘Healthy Eating Index-2005
Development and Evaluation Technical Report Support Files’. Center for Nutrition
Policy and Promotion, USDA. http://origin.www.cnpp.usda.gov/HealthyEatingIndex2005report.htm.
Volpe, Richard, Abigail Okrent, and Ephraim Leibtag. 2013. ‘The Effect of Supercenter-Format Stores on the Healthfulness of Consumers’ Grocery Purchases’. American Journal of
Agricultural Economics 95 (3): 568–89.
Weatherspoon, Dave D., and Thomas Reardon. 2003. ‘The Rise of Supermarkets in Africa:
Implications for Agrifood Systems and the Rural Poor’. Development Policy Review 21
(3): 333–55.
World Bank, 2007. Horticultural Producers and Supermarket Development in Indonesia. Report
38543-ID. Washington, DC: World Bank.
Zhang, Qian Forrest, and Zi Pan. 2013. ‘The Transformation of Urban Vegetable Retail in
China: Wet Markets, Supermarkets and Informal Markets in Shanghai’. Journal of Contemporary Asia 43 (3): 497–518.
Zhang, Xiaoyong, Ivo van der Lans, and Hans Dagevos. 2012. ‘Impacts of Fast Food and
the Food Retail Environment on Overweight and Obesity in China: A Multilevel Latent
Class Cluster Approach’. Public Health Nutrition 15 (1): 88–96.
ISSN: 0007-4918 (Print) 1472-7234 (Online) Journal homepage: http://www.tandfonline.com/loi/cbie20
Diet Transition and Supermarket Shopping
Behaviour: Is There a Link?
Hery Toiba, Wendy J. Umberger & Nicholas Minot
To cite this article: Hery Toiba, Wendy J. Umberger & Nicholas Minot (2015) Diet Transition and
Supermarket Shopping Behaviour: Is There a Link?, Bulletin of Indonesian Economic Studies,
51:3, 389-403, DOI: 10.1080/00074918.2015.1111997
To link to this article: http://dx.doi.org/10.1080/00074918.2015.1111997
Published online: 29 Nov 2015.
Submit your article to this journal
Article views: 50
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=cbie20
Download by: [Universitas Maritim Raja Ali Haji]
Date: 17 January 2016, At: 23:17
Bulletin of Indonesian Economic Studies, Vol. 51, No. 3, 2015: 389–403
DIET TRANSITION AND SUPERMARKET
SHOPPING BEHAVIOUR: IS THERE A LINK?
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Hery Toiba*
Brawijaya University
Wendy J. Umberger*
University of Adelaide
Nicholas Minot*
International Food Policy Research Institute
Supermarkets are leading the transformation of food markets in Asia, yet few studies have examined the impact of this so-called supermarket revolution on diet
transition and the related nutritional and health implications. We use data from
a sample of 1,180 urban households in Indonesia to explore the relation between
the increased use of modern food-retail outlets and the emergence of unhealthy
dietary patterns. The results of our ordinary least-squares and instrumental variables regressions suggest a negative and signiicant relation between the share of
food expenditure at modern food retailers and the healthiness of consumer food
purchases, even after we control for other characteristics that may inluence foodconsumption decisions.
Keywords: supermarket revolution, consumers, diet quality, Lewbel approach, diet transition
JEL classiication: I12, I15, F63, F68, Q13, Q18
INTRODUCTION
Economic growth, urbanisation, and foreign direct investment have all contributed to the rapid rise of domestic and international supermarket chains in developing and emerging economies, including Indonesia (Faiguenbaum, Berdegue,
and Reardon 2002; Reardon et al. 2003). The transformation of the food-retail
sector, or the so-called supermarket revolution, is having a profound effect on
the market conditions faced by both producers and consumers in many countries (Faiguenbaum, Berdegué, and Reardon 2002). Previous research on the
* We thank Wahida from the Indonesian Center for Agricultural Socio Economic and Policy Studies (ICASEPS) and Randy Stringer from Global Food Studies, at the University of
Adelaide, for their invaluable insight and support during the research-design and datacollection phases of this research. We also thank the ICASEPS research team for their assistance with all phases of the research project. The project was funded by the Australian
Centre for International Agricultural Research (grant ADP/2005/006). All views, interpretations, and conclusions are those of the authors and not necessarily those of the supporting or cooperating institutions.
ISSN 0007-4918 print/ISSN 1472-7234 online/15/000389-15
http://dx.doi.org/10.1080/00074918.2015.1111997
© 2015 Indonesia Project ANU
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
390
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
emergence of supermarkets in developing countries has focused on their impact
on smallholder farmers, rural development, small traders, and consumer purchasing habits,1 but only a handful of studies have considered their impact on foodconsumption patterns and the related health implications (such as Asfaw 2008;
Tessier et al. 2008; Banwell et al. 2013; Zhang, Van der Lans, and Dagevos 2012;
and Kelly et al. 2014). There is, however, increasing speculation that supermarket
penetration is one cause of the dramatic shift in Asian diets towards more westernised diets, typiied by the increased consumption of reined carbohydrates,
fats, and oils at the expense of grains, fresh fruits, and vegetables (Popkin 1999,
2006; Mendez and Popkin 2004; D’Haese and Van Huylenbroeck 2005; Asfaw
2008).
The World Health Organization is concerned about the impact of dietary
changes on the health status of consumers in many developing Asian countries, because the consumption of highly processed foods has been linked to an
increased risk of non-communicable chronic diseases such as cardiovascular disease and Type 2 diabetes (Asfaw 2008). The diffusion of supermarkets can have
important dietary implications, both positive (such as more diverse diets, lower
food prices, and increased food accessibility) and negative (such as greater inequalities in food accessibility and increased consumption of nutrient-poor, highly
processed foods) (Hawkes 2008; Banwell et al. 2013).
Previous empirical studies, using data from Tunisia, Guatemala, and Thailand,
have examined the relation between supermarket use and diet quality, but
their results have been mixed (Tessier et al. 2008; Asfaw 2008; Kelly et al. 2014).
Additional research is needed to understand the implications of supermarket-sector growth and diet transition in Indonesia. We use data from a survey of urban
households in Java to examine whether a higher share of food expenditures in
modern food-retail outlets (supermarkets) is associated with greater consumption
of speciic ‘unhealthy’ food products. We apply methods similar to those of Volpe,
Okrent, and Leibtag (2013) to develop household-level measures for, irst, shares
of expenditure on ‘healthy’ and ‘unhealthy’ foods, and, second, shares of food
expenditure in supermarkets, which include all types of modern food retailers—
hypermarkets, supermarkets, and convenience stores (mini-marts)—in Indonesia.
Indonesia, particularly Java, is an appropriate setting in which to explore the
relation between diet quality and supermarket use. Indonesia has been undergoing signiicant economic growth and urbanisation. Furthermore, its food-retail
sector has been transforming since 1998, when the Indonesian government began
to allow foreign direct investment in food retailing (World Bank 2007; Reardon
1. For the impact of supermarkets on smallholder farmers, see, for example, the work of
Bignebat, Koç, and Lemeilleur (2009); Hernández, Reardon, and Berdegué (2007); and
Michelson, Reardon, and Perez (2012). On rural development, see that of Dries, Reardon,
and Swinnen (2004); Hu et al. (2004); Neven et al. (2009); Rao and Qaim (2011); Reardon,
Stamoulis, and Pingali (2007); Reardon et al. (2003); and Weatherspoon and Reardon
2003). On small traders, see studies by Cadilhon et al. (2006); Faiguenbaum, Berdegué,
and Reardon (2002); Gorton, Sauer, and Supatpongkul (2011); Schipmann and Qaim (2011);
Suryadarma et al. (2010); and Zhang and Pan (2013). On consumer purchasing habits, see
the work of Bai, Wahl, and McCluskey (2008); D’Haese and Van Huylenbroeck (2005);
Neven et al. (2006); and Rodriguez et al. (2002).
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
391
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
and Timmer 2014). Between 1999 and 2009, the number of supermarkets increased
by 67%, with around three-quarters of these owned by multinational companies
(Dyck, Woolverton, and Rangkuti 2012). The growth was much more intense in
Java, particularly around the greater Jakarta metropolitan area, Jabotabek, which
has a population of around 28 million and includes the cities of Bogor, Depok,
Bekasi, and Tangerang.
DATA AND METHODS
Data
The analysis in this article uses data from our survey of urban consumers, which
includes a sample of 1,180 urban households in Surabaya, Bogor, and Surakarta,
representing large, medium, and small cities, respectively (Umberger et al. 2015).
These three cities also represent different stages of urbanisation and supermarket
penetration, with higher modern food-retail penetration in Bogor and Surabaya
than in Surakarta. In addition, it is likely that there are ethnic and cultural differences in the populations of these cities. Within each city, we used stratiied
random sampling to select households, oversampling higher-income neighbourhoods and areas close to supermarkets. We calculated sampling weights on the
basis of the inverse of the probability of selection, and used them to calculate all
results presented here.
The survey was implemented from November 2010 to February 2011 by trained
and experienced enumerators, who conducted the interviews at the homes of the
respondents. The questionnaire covered household composition, housing and
asset ownership, shopping behaviour at different types of outlets, food-expenditure patterns, and perceptions of each type of food retailer. The food-expenditure
module collected information on the amount that households spent on 67 food
items and the most important type of retailer for each item. We identiied and considered shopping behaviour at seven types of food-retail outlets: hypermarkets,
supermarkets, convenience stores, small shops (warung), semi-permanent stands,
traditional wet markets, and peddlers. In this study, we use the terms ‘modern
food retailers’ and ‘supermarkets’ to mean hypermarkets, supermarkets, and convenience stores.2
Econometric Approach
To explore the possible link between diet quality and supermarket use in
Indonesia, we estimate the household-level share of food expenditures spent on
healthy food products as a function of the household-level share of food expenditures at modern retail outlets. Similar to Volpe, Okrent and Leibtag (2013), we
use the United States Department of Agriculture’s Guidelines for Healthy Eating
(USDA 2010) to categorise food items as either healthy or unhealthy. According to
these guidelines, healthy items include grains, roots and tubers, fruit, vegetables,
2. Hypermarkets are very large modern stores that occupy more than 8,000 square metres,
have at least 10 or more cash registers, and sell both food and other groceries. Supermarkets
are medium to large modern stores that occupy between 300 and 8,000 square metres, have
3–9 cash registers, and sell both food and other groceries. Convenience stores, or minimarts, are small modern stores with 1–2 cash registers (Suryadarma et al. 2010).
392
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
unprocessed meat, milk, and eggs. Unhealthy items include sugar, sweetened
food and beverages, fats, oils, and processed foods such as snack foods, processed
meat, and ready-to-eat meals. The general form of the econometric model used in
this study is as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
HealthyShare = f (Pct_modern, Female, Age, Age2, Education, Hourjob, Label,
Income, Household_size, Children5, Domestic, Activity, FAFH, Surabaya, Bogor) (1)
Table 1 provides the deinitions, means, and standard errors of all variables.
Following the approach of Asfaw (2008) and Volpe, Okrent, and Leibtag (2013),
we use as the dependent variable a continuous variable, HealthyShare, that represents the share of household food expenditures on healthy food products.
Pct_modern is the explanatory variable of interest. It is a continuous variable
representing the household share of food expenditures in supermarkets. Excluded
are food expenditures made at traditional retail outlets, including small shops,
semi-permanent stands, traditional wet markets, and peddlers. As discussed earlier, consumers who buy a larger share of their food at modern retail outlets may
buy fewer healthy food products for their household.
The survey enumerators interviewed the primary person responsible for making food-purchasing decisions for each household. We include several variables
representing respondents’ socio-demographic characteristics, because these characteristics may inluence household food expenditures on certain foods (Huston
and Finke 2003; Schroeter, Anders, and Carlson 2013; Umberger et al. 2015).
Female is a dummy variable equal to one if the respondent was female. Age and
Age2 represent the age of the respondent and the squared age of the respondent.
The partial effect of age on the share of healthy food expenditures is expected
to be positive, because previous studies have shown that individuals become
more concerned about the healthfulness of their diet as they get older (Frazao and
Allshouse 2003).
Education is the number of years of education completed by the respondent.
Respondents with higher levels of education may be more aware of the relation
between diet, nutrition and health, and thus decide to buy healthier food for their
households (Turrell and Kavanagh 2006). More educated consumers may also
have a lower future discount rate, making them more likely to buy healthy food as
an investment in their family’s future (Huston and Finke 2003; Schroeter, Anders,
and Carlson 2013).
Hourjob is the number of hours per week that the respondent spends at work.
Respondents who work longer hours may have less time to spend on household
food shopping and food-purchasing decisions (Mancino and Kinsey 2004). Label
is a factor score, based on a Likert-scale response, that represents respondents’
attitudes to a series of questions about their use of nutrition labels and their nutrition information preferences. Respondents with a larger positive factor show a
preference for nutrition information and use nutrition labels more frequently.
Respondents who are more concerned about nutrition are expected to spend more
on healthy food (Huston and Finke 2003).
We include several household characteristics that may inluence the share of
total food expenditures on healthy food. We use Income as a proxy for per capita household expenditures. Households with higher incomes may face fewer
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
393
TABLE 1 Deinitions and Descriptive Statistics for Variables
Variable
Deinition
% of food expenditures classiied
as healthy
Pct_modern
% of food expenditures made at
modern retail outlets
Female
Gender (0 = male; 1 = female)
Age
Respondent’s age (years)
Respondent’s age (years)2
Age2
Education
Years of education completed
by respondent
Hourjob
Number of hours per week
spent at respondent's job
Label
Factor scores representing
respondents’ nutrition concerns,
food label use & perceived
nutrition knowledge
Income
Household income ($ ’000/year)
Household_size Number of family members
living in household
Children5
Dummy for household with
children under 5 years old
(1= yes; 0 = otherwise)
Domestic
Dummy for households with a
domestic employee
(1 = yes; 0 = otherwise)
Activity
Average hours per week adults
in the household exercise
FAFH
% of food expenditures spent
on meals and beverages eaten
outside of the home
Surabaya
Dummy for Surabaya
(1 = Surabaya; 0 = otherwise)
Bogor
Dummy for Bogor
(1 = Bogor; 0 = otherwise)
Time_modern Time needed to get to the
nearest modern food retailer
(mins)
Food_safety
Dummy variable (1 = respondent
indicated that a retail outlet
offering food-safety assurances
is important or extremely
important when deciding
where to purchase food;
0 = otherwise)
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
HealthyShare
Mean
SD
Min
Max
68.71
15.05
2.93
98.52
13.39
14.97
0
83.71
0.89
43.02
2,004
9.35
0.32
12.40
1,152
4.52
0
15
225
0
1
83
6,889
22
21.11
25.01
0
130
–0.07
1.06
–4.94
2.42
4.93
4.41
4.91
1.76
0.32
1
77.73
12
0.58
0.49
0
1
0.16
0.37
0
1
2.11
3.13
0
35
7.41
10.75
0
95.39
0.61
0.49
0
1
0.21
0.41
0
1
18.98
10.45
1.5
120.0
0.96
0.19
0
1
Note: SD = standard deviation
economic barriers to buying healthy food, particularly if healthy food is more
expensive than unhealthy food (Smith, Bogin, and Bishai 2005). We include
Household_size because households with more family members to support may
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
394
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
have less disposable income (Schroeter, Anders and Carlson 2013). Children5 is a
dummy variable equal to one if there are children under ive years of age in the
household. Households with young children may make different food-purchasing
decisions, owing to the speciic dietary needs of young children (Umberger et al.
2015). Domestic is a dummy variable equal to one if the household has a domestic
employee who helps with household duties, including shopping for food.
Activity is a continuous variable representing the average number of hours
per week that adults in the household spend exercising. Households with adults
who exercise more may place more value on foods that are healthy, and therefore may have a higher share of food expenditures on healthy food (Huston and
Finke 2003; Komlos, Smith, and Bogin 2004; Mancino and Kinsey 2004). FAFH is
an additional measure of the impact of the food-market environment on foodconsumption behaviour. It is the household’s share of total food expenditures
on food-away-from-home (FAFH), including meals and beverages prepared and
eaten outside the home. Previous research has found that individuals who consume more FAFH tend to have poorer diets as a result of other factors in the
food-market environment (Bezerra and Sichieri 2009; Drichoutis, Nayga, and
Lazaridis 2012).
Surabaya and Bogor are dummy variables representing city-level effects.
Compared with Surakarta, Surabaya and Bogor are more metropolitan and have
more mature modern food-retail sectors. There may also be unobservable citylevel factors (such as social norms, cultural traditions, and climatic conditions)
that have an effect on food-expenditure shares (Umberger et al. 2015).
Estimation and Instrumental Variables
We irst estimate equation (1) using ordinary least squares (OLS), controlling for
the set of observable factors that may inluence food-purchasing decisions. Under
the OLS approach, however, the coeficient on the modern-share variable (Pct_
modern) may not accurately measure the effect of shopping at modern retailers on
healthfulness of food-consumption patterns, for three reasons. First, there may
be reverse causation: a household that prefers to eat unhealthy food may decide
to shop at supermarkets, potentially resulting in negative bias of the Pct_modern
coeficient in equation (1). Second, an unobserved variable may be inluencing
both the modern share and the unhealthy share of food spending. For example,
if watching television exposes viewers to advertisements for supermarkets as
well as advertisements for unhealthy foods, the OLS coeficient estimate of the
Pct_modern variable will be negatively biased. Third, the estimates of food expenditures may be subject to measurement error (Gibson and Kim 2007) and therefore
the OLS estimation of the impact on food-consumption patterns of using modern
retailers will also be biased. In all three cases, the explanatory variable is correlated with the error term, thus violating one of the assumptions behind OLS and
giving biased and inconsistent estimates of the coeficient.
We can address these problems by using an instrumental variable (IV) technique, in which we replace the potentially endogenous explanatory variable, Pct_
modern, with an estimated value of the variable. The estimated value is based on a
supplementary regression of the potentially endogenous explanatory variable as
a function of one or more IVs (or instruments). Under the IV approach, the main
equation is as follows:
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
yi = α 0 + α 1 M i + βi Xi + ε i
395
(2)
where yi is the share of food expenditures of household i spent on healthy food; Mi
is the estimated share of food expenditures of household i purchased at modern
food-retail outlets; Xi is a vector of the independent variables; α0, α1, and βi are
parameters to estimate; and εi is the error term. We estimate Mi as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
M i = β i Xi + γ Z i + υ i
(3)
where Zi is a vector of instrumental variables, γ is an estimated parameter, and υi
is the error term.
One dificulty in implementing IV estimations is inding valid instruments.
Instruments must meet the relevance condition, meaning that they must be
strongly correlated with the potentially endogenous explanatory variable (Pct_
modern). They must also meet the exclusion restriction, meaning they cannot be
correlated with any omitted variable that may help to explain the dependent variable (that is, they must not be correlated with the error terms) (Stock and Watson
2012).
A standard IV used in studying the impact of supermarket use is the distance
from the household to the nearest retail outlet (Volpe, Okrent, and Leibtag 2013).
In Indonesia, however, distance is not a good measure of access to shopping outlets and food markets, owing to the absence of good public transportation and
related trafic congestion in urban areas. As suggested by Narayan, Rao, and
Sudhir (2012), the cost of buying a product from a particular retail outlet includes
the cost of travelling to the store. This suggests that travel time to a particular
type of outlet will be negatively related to the likelihood of using that outlet. In
this study, we use as an IV a variable representing the time it usually takes the
respondent to get to the nearest modern retail outlet, Time_modern, even though
using this instrument may raise other concerns. Some consumers, for example,
may use different transportation modes to access retail outlets. To address this
issue, we apply different identiication strategies to check the robustness of the
instruments.
We include, as an additional IV, the dummy variable Food_safety, which equals
one if the respondent indicated that food-safety assurances are an important or
extremely important attribute when deciding where to purchase food. We consider this to be a valid IV because previous studies suggest that consumer concerns
about food safety have increased the use of modern retail outlets in developing
countries (Gorton, Sauer, and Supatpongkul 2011; Minten and Reardon 2008).
Inappropriate IVs could potentially give estimations inferior to OLS results. To
mitigate this risk, we also use Lewbel’s (2012) two-stage, heteroskedasticity-based
estimator, which involves irst creating ‘higher moment’ internal instruments that
are the product of a set of selected demeaned regressors and the residuals obtained
from the irst-stage regression. According to Lewbel, these instruments should,
irst, satisfy the exclusion restriction if they are uncorrelated with the other regressors and, second, meet the relevance condition if hetroskedasticty exists in the
irst stage. Lewbel cautions that ‘higher moments can lead to noisier, less reliable
estimates than exclusion based on standard exclusion restrictions’ (78). They may
be more appropriate, however, if the usual instruments are weak.
396
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
We follow methods used by Emran and Shilpi (2012); Schroeter, Anders, and
Carlson (2013); Volpe, Okrent, and Leibtag (2013); and Umberger et al. (2015). We
estimate equation (4) as follows:
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
M i = π ′Z i + ξi
(4)
where Zi is a vector of internal instruments that, according to Lewbel, can be either
a subset of or identical to Xi (Zi = Xi). In this analysis, we consider three exogenous
variables (Education, Income, and Label) for Zi. We estimate equation (4) by regressing (Mi) on this subset of exogenous variables, obtaining the residual ξ i. We then
use this residual to generate the higher-moment instruments
(Z
i
− Zi ) ξ̂i
where Zi is the mean of Zi and ξ̂i is the residual from equation (4). We use Z! i to
denote the higher-moment instruments. We then use two-stage least squares and
re-estimate equation (3), substituting Z! i for Zi in equation (3), in the irst stage,
and then re-estimating equation (2). Finally, to test the robustness of the estimates,
we conduct a fourth estimation, using two-stage OLS: we estimate equation (3)
by using both the original IVs (Time_modern and Food_safety) and the highermoment instruments ( Z! i ) in the irst stage. We use the Breusch-Pagan test to test
for heteroskedasticity:3
(Cov ( Z ξ ) ≠ 0 )
2
i, i
RESULTS
Table 2 reports the results of the estimations of the healthy expenditure share
models. As discussed earlier, the Pct_modern variable may be endogenous, and
the OLS results provided in the irst column of table 2 may be biased. We therefore provide in the other columns the coeficients estimated using the three IV
methods. These results give an insight into potential biases that may result from
endogeneity.
The results are similar across all four models, particularly for our OLS, Lewbel,
and IV-Lewbel estimations. The coeficient for the main explanatory variable of
interest, Pct_modern, is consistently statistically signiicant and negative, which
suggests that an increase in the share of food expenditures in modern retail outlets
is associated with a decrease in the share of food expenditures on healthy food.
There are some notable differences between the results of the standard IV estimation and the others. The estimated coeficient for Pct_modern from the standard
IV regression is much larger than the OLS, Lewbel, and IV-Lewbel coeficients.
Additionally, four variables (Hourjob, Domestic, Surabaya, and Bogor) are signiicant only in the standard IV estimation. Our results are consistent with those of
other studies (such as Gao and Smyth 2015 and Sabia 2007) in that the Lewbeland OLS-estimated coeficients are similar and both much higher than those estimated using the standard IV approach.
Before discussing additional signiicant covariates, we consider the validity
of the results of the alternative IV approaches using several test statistics. The
3. We use the ivreg2 command developed by Baum, Schaffer, and Stillman (2010).
TABLE 2 Regression Results: Estimation of Shares of Expenditure on Healthy Food
Pct_modern
Female
Age
Age2
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Education
Hourjob
Activity
Surabaya
Bogor
FAFH
Label
Income
Household_size
Children5
Domestic
Constant
Observations
R2
First-stage regression
Time_modern
Food_safety
First stage (Stock & Yogo) F-test
Endogeneity test
Breusch-Pagan chi2
Hansen J-statistic
OLS
IV
Lewbel
IV-Lewbel
–0.20***
(0.03)
1.83
(1.35)
0.62***
(0.19)
–0.01***
(0.00)
0.04
(0.01)
–0.02
(0.01)
0.18*
(0.10)
0.74
(0.95)
1.05
(1.03)
–0.19***
(0.04)
0.12
(0.42)
0.00
(0.00)
–0.39*
(0.220)
1.19
(0.73)
1.69
(1.05)
60.77***
(4.66)
–0.70**
(0.31)
3.69*
(1.96)
0.59***
(0.22)
–0.01**
(0.00)
0.76
(0.47)
–0.04*
(0.02)
0.24*
(0.13)
3.08*
(1.85)
2.91*
(1.66)
–0.22***
(0.05)
0.70
(0.60)
0.00
(0.00)
–0.40*
(0.24)
1.25
(0.82)
3.19**
(1.52)
57.92***
(5.76)
–0.15**
(0.07)
1.64
(1.36)
0.62***
(0.19)
–0.01***
(0.00)
–0.03
(0.13)
–0.02
(0.01)
0.18*
(0.10)
0.50
(1.00)
0.86
(1.06)
–0.19***
(0.04)
0.06
(0.42)
–0.00
(0.00)
–0.39*
(0.22)
1.19
(0.73)
1.54
(1.07)
61.07***
(4.67)
–0.17**
(0.07)
1.72
(1.36)
0.62***
(0.19)
–0.01***
(0.00)
–0.00
(0.13)
–0.02
(0.01)
0.18*
(0.10)
0.59
(0.99)
0.94
(1.05)
–0.19***
(0.04)
0.08
(0.42)
0.00
(0.00)
–0.39*
(0.22)
1.19*
(0.72)
1.60
(1.07)
60.95***
(4.67)
1,180
0.13
1,180
–0.16
1,180
0.12
1,180
0.125
60.21***
–0.11
(0.06)*
5.03
(1.46)***
41.36***
–0.10
(0.05)*
4.71
(1.50)***
7.22
2.78
(0.09)
189.06
(0.00)
0.63
(0.73)
5.33
(0.26)
Note: The dependent variable is HealthyShare. Standard errors are in parentheses. IV uses two-stage
least squares with Time_modern and Food_safety as external instruments. Lewbel uses the heteroskedasticity-based estimator, which takes ( Zi − Zi ) ξi as instruments, where Zi includes the variables Education,
Income, and Label. IV-Lewbel uses both the external instruments and the Lewbel approach.
* p < 0.10; ** p < 0.05; *** p < 0.01.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
398
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
endogeneity test result (2.78) is signiicant, suggesting that the Pct_modern variable is endogenous. The two instruments, Time_Modern and Food_Safety, are both
statistically signiicant in explaining Pct_modern. As we expected, their signs are
based on previous literature and suggest that they satisfy the relevance condition.
However, the irst-stage Stock and Yogo (2005) F-test from the standard IV estimation (table 2) indicates that these variables are relatively weak instruments. In this
case, the endogeneity test is unreliable.
To support our use of the Lewbel and IV-Lewbel models, we consider the results
of two additional tests. First, the Breusch-Pagan chi-square statistic (189.06) is signiicant (table 2) and we can reject the null hypothesis of homoskedasticity of the
error term (ξ i). Second, the Hansen J statistics, used to test for overidentiication of
all instruments for the Lewbel and the IV-Lewbel models, are not signiicant, which
conirms that the sets of instruments included in these models are valid. These
diagnostic test results support the use of the Lewbel and IV-Lewbel approaches
as an alternative to the standard IV method.4 Thus, after controlling for various
respondent and household characteristics, we ind that a one-percentage-point
increase in the share of food purchased at a modern outlets (Pct_modern) results in
a decrease of 0.15 to 0.17 percentage points in the share of healthy food purchases.
This inding supports the notion that increased use of supermarkets for food shopping may lead to less healthy food-purchasing behaviour among households.
In both models, the coeficient on Age is positive and signiicant, while the coeficient on Age2 is negative and signiicant. This implies that the share of expenditures on healthy food rises and then falls with the age of the person responsible
for shopping for food for the household. More speciically, according to the coeficients for the Lewbel and IV-Lewbel models, the healthy share reaches its maximum at the age of approximately 57.5 Contrary to expectations, we found that
older people (above 57) spend less on healthy food for their household, other
factors being equal.
The coeficients for Activity are positive and signiicant, which is consistent
with previous studies and suggests that households with active adults are also
more likely to buy healthier food. As expected, FAFH was negative and highly
signiicant (at 1%) in both models, implying that households spending more on
food-away-from-home tend to buy less healthy food.
The coeficient on Income was not signiicant in explaining healthy expenditure
shares in either estimation. This result is surprising, but there are two possible
explanations. First, many low-income households receive food subsidies from the
government, especially for rice, allowing them to spend more of their income on
other healthy foods (such as unprocessed meats, fresh milk, and eggs).6 Second,
4. The coeficients estimated using the standard IV approach (table 2) are not reliable, owing to weak instruments, and we provide them only for comparison. As such, we focus
the remaining discussion on the results from the estimations of the Lewbel and IV-Lewbel
models.
5. Taking the derivative of the regression equation with respect to age, the maximum occurs when b + 2(a)(Age) = 0, where b is the linear coeficient and a is the quadratic coeficient. Rearranging terms, we get Age = –b/2a = –0.622/0.011 = 56.5.
6. Since 1998, the Indonesian government has provided food subsidies for low-income
households as part of its Raskin rice-subsidy program (Sumarto and Bazzi 2011)
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
399
high-income households are better able to afford processed foods and convenience foods, which are considered to be less healthy.
A related variable, Household_size, was negative and statistically signiicant in
both models, perhaps because, for a given income, larger households have less
disposable income to spend on more expensive, healthier foods, so household
members may have poorer diets. The coeficient on the Children5 variable in the
IV-Lewbel model (table 2) is positive and statistically signiicant, suggesting some
evidence of a positive relation between having young children in the household
and having healthier household food-expenditure shares.
There was no statistically signiicant difference in the share of healthy food
expenditures in Surabaya, Bogor, and Surakarta. Any differences across cities are
due to differences in other explanatory variables.
CONCLUSIONS AND IMPLICATIONS
There is increasing speculation that consumers’ dietary practices and health outcomes can be affected by the strategic decisions of modern retailers. It is unclear
from the literature, however, whether there is a connection between the use of
modern retail outlets and unhealthy dietary patterns. Tessier et al.’s (2008) study,
for example, which used data from Tunisia, determined that the diet quality of
frequent users of supermarkets was slightly higher than that of less frequent
supermarket users. On the other hand, studies using data from Guatemala (Asfaw
2008), Thailand (Kelly et al. 2014), and Kenya (Rischke et al. 2015) found higher
supermarket use to be associated with an increased consumption of highly processed foods, which are generally considered to be unhealthy.
We have attempted here to build on previous empirical studies and shed light
on the relation between supermarket shopping behaviour and diet transition in
Indonesia, a topic that has not previously been researched. Using cross-sectional
data from a survey of Indonesian urban households, we found a negative and signiicant relation between the share of food expenditures in modern food retailers
and the healthfulness of consumer food expenditures. Given, however, that the
coeficients on the supermarket variables (Pct_modern) estimated using the OLS,
Lewbel, and IV-Lewbel models are relatively small, so, too, is the actual magnitude or impact of changing food expenditures on diet quality. Speciically, after
controlling for other variables, we found that an increase of one percentage point
in the share of food expenditures in supermarkets is associated with a relatively
small decrease (of 0.15 to 0.17 percentage points) in the household’s share of food
expenditures on healthy food.
Nevertheless, our results support those of Asfaw (2008), Kelly et al. (2014),
and Rischke et al. (2015). As in Thailand, Guatemala, and Kenya, supermarket
penetration in urban areas of Indonesia is relatively high, and supermarkets are
generally accessible to both low-income and high-income households. As Kelly et
al. (2014) explain, modern food retailers, including hypermarkets, supermarkets,
and convenience stores, may be able to compete with traditional retailers by providing less expensive processed foods, which tend to be higher in fat, salt, and
sugar content, and fewer fresh foods, particularly fruits and vegetables.
We also found that diet quality is negatively associated with greater
food-expenditure shares on meals and beverages bought outside the home.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
400
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
Policymakers in Indonesia who are concerned about diet transition may want
to consider aspects of the food environment other than just supermarket penetration. The development of one-stop shopping centres (such as shopping malls
where supermarkets are typically co-located with food courts), for example, may
exacerbate the negative effects of supermarket penetration on diet quality.
This study has several limitations. We used cross-sectional data, so we were
unable to capture dynamic patterns of consumption; panel data would have
yielded more robust estimates of causal relations. We did not measure the quantity of food consumed by individuals in households, which made it dificult to
isolate the effect of food expenditure shares on the intake of calories and nutrients.
And we focused on urban households in Java, where modern food-retail penetration is much greater than in rural areas and in other provinces. The relation
between modern food-retail penetration and health outcomes, in Indonesia and
elsewhere, requires more research.
REFERENCES
Asfaw, Abay. 2008. ‘Does Supermarket Purchase Affect the Dietary Practices of Households? Some Empirical Evidence from Guatemala’. Development Policy Review 26 (2):
227–43.
Bai, Junfei, Thomas I. Wahl, and Jill J. McCluskey. 2008. ‘Consumer Choice of Retail Food
Store Formats in Qingdao, China’. Journal of International Food and Agribusiness Marketing 20 (2): 89–109.
Banwell, Cathy, Jane Dixon, Sam-ang Seubsman, Suttinan Pangsap, Matthew Kelly, and
Adrian Sleigh. 2013 ‘Evolving Food Retail Environments in Thailand and Implications
for the Health and Nutrition Transition’. Public Health Nutrition 16 (4): 608–15.
Baum, Christopher F., Mark E. Schaffer, and Steven Stillman. 2010. ‘IVREG2: Stata module
for extended instrumental variables/2SLS and GMM estimation’. http://ideas.repec.
org/c/boc/bocode/s425401.html.
Bezerra, Ilana Nogueira, and Rosely Sichieri. 2009. ‘Eating Out of Home and Obesity: A
Brazilian Nationwide Survey’. Public Health Nutrition 12 (11): 2037–43.
Bignebat, Céline, Ahmet Ali Koç, and Sylvaine Lemeilleur. 2009. ‘Small Producers, Supermarkets, and the Role of Intermediaries in Turkey’s Fresh Fruit and Vegetable Market’.
Agricultural Economics 40 (s1): 807–16.
Cadilhon, Jean-Joseph, Paule Moustier, Nigel D. Poole, Phan Thi Giac Tam, and Andrew P.
Fearne. 2006. ‘Traditional vs. Modern Food Systems? Insights from Vegetable Supply
Chains to Ho Chi Minh City (Vietnam)’. Development Policy Review 24 (1): 31–49.
D’Haese, Marijke, and Guido Van Huylenbroeck. 2005. ‘The Rise of Supermarkets and
Changing Expenditure Patterns of Poor Rural Households Case Study in the Transkei
Area, South Africa’. Food Policy 30 (1): 97–113.
Drichoutis, Andreas C., Rodolfo M. Nayga, Jr, and Panagiotis Lazaridis. 2012. ‘Food Away
from Home Expenditures and Obesity among Older Europeans: Are There Gender Differences?’. Empirical Economics 42 (3): 1051–78.
Dries, Liesbeth, Thomas Reardon, and Johan F. M. Swinnen. 2004. ‘The Rapid Rise of
Supermarkets in Central and Eastern Europe: Implications for the Agrifood Sector and
Rural Development’. Development Policy Review 22 (5): 525–56.
Dyck, John, Andrea E. Woolverton, and Fahwani Yuliati Rangkuti. 2012. Indonesia’s Modern Retail Food Sector: Interaction with Changing Food Consumption and Trade Patterns.
Economic Information Bulletin 97. Washington, DC: United States Department of
Agriculture.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
401
Emran, M. Shahe, and Forhad Shilpi. 2012. ‘The Extent of the Market and Stages of Agricultural Specialization’. Canadian Journal of Economics 45 (3): 1125–53.
Faiguenbaum, Sergio, Julio A. Berdegué, and Thomas Reardon. 2002. ‘The Rapid Rise of
Supermarkets in Chile: Effects on Dairy, Vegetable, and Beef Chains’. Development Policy Review 20 (4): 459–71.
Frazao, Elizabeth, and Jane Allshouse. 2003. ‘Strategies for Intervention: Commentary and
Debate’. Journal of Nutrition 133 (3): 844S–847S.
Gao, Wenshu, and Russell Smyth. 2015. ‘Education Expansion and Returns to Schooling in
Urban China, 2001–2010: Evidence from Three Waves of the China Urban Labor Survey’. Journal of the Asia Paciic Economy 20 (2): 178–201.
Gibson, John, and Bonggeun Kim. 2007. ‘Measurement Error in Recall Surveys and the
Relationship between Household Size and Food Demand’. American Journal of Agricultural Economics 89 (2): 473–89.
Gorton, Matthew, Johannes Sauer, and Pajaree Supatpongkul. 2011. ‘Wet Markets, Supermarkets and the “Big Middle” for Food Retailing in Developing Countries: Evidence
from Thailand’. World Development 39 (9): 1624–37.
Hawkes, Corinna. 2008. ‘Dietary Implications of Supermarket Development: A Global Perspective’. Development Policy Review 26 (6): 657–92.
Hernández, Ricardo, Thomas Reardon, and Julio Berdegué. 2007. ‘Supermarkets, Wholesalers, and Tomato Growers in Guatemala’. Agricultural Economics 36 (3): 281–90.
Hu, Dinghuan, Thomas Reardon, Scott Rozelle, Peter Timmer, and Honglin Wang. 2004.
‘The Emergence of Supermarkets with Chinese Characteristics: Challenges and
Opportunities for China’s Agricultural Development’. Development Policy Review 22
(5): 557–86.
Huston, Sandra J., and Michael S. Finke. 2003. ‘Diet Choice and the Role of Time Preference’. Journal of Consumer Affairs 37 (1): 143–60.
Kelly, Matthew, Sam-ang Seubsman, Cathy Banwell, Jane Dixon, and Adrian Sleigh. 2014,
‘Thailand’s Food Retail Transition: Supermarket and Fresh Market Effects on Diet
Quality and Health’. British Food Journal 116 (7): 1180–93.
Komlos, John, Patricia K. Smith, and Barry Bogin. 2004. ‘Obesity and the Rate of Time Preference: Is There a Connection?’. Journal of Biosocial Science 36 (2): 209–19.
Lewbel, Arthur. 2012. ‘Using Heteroskedasticity to Identify and Estimate Mismeasured
and Endogenous Regressor Models’. Journal of Business & Economic Statistics 30 (1):
67–80.
Mancino, Lisa, and Jean Kinsey. 2004. ‘Diet Quality and Calories Consumed: the Impact of
Being Hungrier, Busier and Eating Out’. Working paper 04-02, Food Industry Center,
University of Minnesota, St Paul, MN.
Mendez, Michelle A., and Barry M. Popkin. 2004. ‘Globalization, Urbanization and Nutritional Change in the Developing World’. Journal of Agricultural and Development Economics 1 (2): 55–80.
Michelson, Hope, Thomas Reardon, and Francisco Perez. 2012. ‘Small Farmers and Big
Retail: Trade-Offs of Supplying Supermarkets in Nicaragua’. World Development 40 (2):
342–54.
Minten, Bart, and Thomas Reardon. 2008. ‘Food Prices, Quality, and Quality’s Pricing in
Supermarkets versus Traditional Markets in Developing Countries’. Applied Economic
Perspectives and Policy 30 (3): 480–90.
Narayan, Vishal, Vithala R. Rao, and K. Sudhir. 2012. ‘Early Adoption of Modern Grocery
Retail in an Emerging Market: Evidence from India’. Johnson School Research Paper
27-2012, Samuel Curtis Johnson Graduate School of Management, Ithaca, NY.
Neven, David, Michael Makokha Odera, Thomas Reardon, and Honglin Wang. 2009. ‘Kenyan Supermarkets, Emerging Middle-Class Horticultural Farmers, and Employment
Impacts on the Rural Poor’. World Development 37 (11): 1802–11.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
402
Hery Toiba, Wendy J. Umberger, and Nicholas Minot
Neven, David, Thomas Reardon, Jonathan Chege, and Honglin Wang. 2006. ‘Supermarkets
and Consumers in Africa: The Case of Nairobi, Kenya’. Journal of International Food &
Agribusiness Marketing 18 (1–2): 103–23.
Popkin, Barry M. 1999. ‘Urbanization, Lifestyle Changes and the Nutrition Transition’.
World Development 27 (11): 1905–16.
———. 2006. ‘Global Nutrition Dynamics: The World is Shifting Rapidly toward a Diet
Linked with Noncommunicable Diseases’. American Journal of Clinical Nutrition 84 (2):
289–98.
Rao, Elizaphan J. O., and Matin Qaim. 2011. ‘Supermarkets, Farm Household Income, and
Poverty: Insights from Kenya’. World Development 39 (5): 784–96.
Reardon, Thomas, Kostas Stamoulis, and Prabhu Pingali. 2007. ‘Rural Nonfarm Employment in Developing Countries in an Era of Globalization’. Agricultural Economics 37
(s1): 173–83.
Reardon, Thomas, and C. Peter Timmer. 2014. ‘Five Inter-Linked Transformations in
the Asian Agrifood Economy: Food Security Implications’. Global Food Security 3 (2):
108–17.
Reardon, Thomas, C. Peter Timmer, Christopher B. Barrett, and Julio Berdegué. 2003. ‘The
Rise of Supermarkets in Africa, Asia, and Latin America’. American Journal of Agricultural Economics 85 (5): 1140–46.
Rischke, Ramona, Simon C. Kimenju, Stephan Klasen, and Matin Qaim. 2015. ‘Supermarkets and Food Consumption Patterns: The Case of Small Towns in Kenya’. Food Policy
52: 9–21.
Rodriguez, Elsa, Miriam Berges, Karina Casellas, Rosángela Di Paola, Beatriz Lupín, Laura
Garrido, and Natacha Gentile. 2002. ‘Consumer Behaviour and Supermarkets in Argentina’. Development Policy Review 20 (4): 429–39.
Sabia, Joseph J. 2007. ‘The Effect of Body Weight on Adolescent Academic Performance’.
Southern Economic Journal 73 (4): 871–900.
Schipmann, Christin, and Matin Qaim. 2011. ‘Modern Food Retailers and Traditional Markets in Developing Countries: Comparing Quality, Prices, and Competition Strategies
in Thailand’. Applied Economic Perspectives and Policy 3 (3): 345–62.
Schroeter, Christiane, Sven Anders, and Andrea Carlson. 2013. ‘The Economics of Health
and Vitamin Consumption’. Applied Economic Perspectives and Policy 35 (1): 125–49.
Smith, Patricia K., Barry Bogin, and David Bishai. 2005. ‘Are Time Preference and Body
Mass Index Associated? Evidence from the National Longitudinal Survey of Youth’.
Economics & Human Biology 3 (2): 259–70.
Stock, James H., and Mark W. Watson. 2012. Introduction to Econometrics: International Edition, 3rd ed. Boston, MA: Pearson Education.
Stock, James H., and Motohiro Yogo. 2005. ‘Testing for Weak Instruments in Linear IV
Regression’. In Identiication and Inference for Econometric Models: Essays in Honor of
Thomas Rothenberg, edited by Donald W. K. Andrews and James H. Stock, 80–108. New
York: Cambridge University Press.
Sumarto, Sudarno, and Samuel Bazzi. 2011 ‘Social Protection in Indonesia: Past Experiences and Lessons for the Future’. Paper presented at the 2011 Annual Bank Conference
on Development Economics, Paris, 30 May–1 June.
Suryadarma, Daniel, Adri Poesoro, Sri Budiyati, Meuthia Rosfadhila, and Asep Suryahadi.
2010. ‘Traditional Food Traders in Developing Countries and Competition from Supermarkets: Evidence from Indonesia’. Food Policy 35 (1): 79–86.
Tessier, Sophie, Pierre Traissac, Bernard Maire, Nicolas Bricas, Sabrina Eymard-Duvernay,
Jalila El Ati, and Francis Delpeuch. 2008. ‘Regular Users of Supermarkets in Greater
Tunis Have a Slightly Improved Diet Quality’. Journal of Nutrition 138 (4): 768–74.
Turrell, Gavin, and Anne M. Kavanagh. 2006. ‘Socio-Economic Pathways to Diet: Modelling the Association between Socio-Economic Position and Food Purchasing Behaviour’. Public Health Nutrition 9 (3): 375–83.
Downloaded by [Universitas Maritim Raja Ali Haji] at 23:17 17 January 2016
Diet Transition and Supermarket Shopping Behaviour: Is There a Link?
403
Umberger, Wendy J., Xiaobo He, Nicholas Minot, and Hery Toiba. 2015. ‘Examining the
Relationship between the Use of Supermarkets and Over-nutrition in Indonesia’. American Journal of Agricultural Economics 97 (2): 510–25.
USDA (United States Department of Agriculture). 2010. ‘Healthy Eating Index-2005
Development and Evaluation Technical Report Support Files’. Center for Nutrition
Policy and Promotion, USDA. http://origin.www.cnpp.usda.gov/HealthyEatingIndex2005report.htm.
Volpe, Richard, Abigail Okrent, and Ephraim Leibtag. 2013. ‘The Effect of Supercenter-Format Stores on the Healthfulness of Consumers’ Grocery Purchases’. American Journal of
Agricultural Economics 95 (3): 568–89.
Weatherspoon, Dave D., and Thomas Reardon. 2003. ‘The Rise of Supermarkets in Africa:
Implications for Agrifood Systems and the Rural Poor’. Development Policy Review 21
(3): 333–55.
World Bank, 2007. Horticultural Producers and Supermarket Development in Indonesia. Report
38543-ID. Washington, DC: World Bank.
Zhang, Qian Forrest, and Zi Pan. 2013. ‘The Transformation of Urban Vegetable Retail in
China: Wet Markets, Supermarkets and Informal Markets in Shanghai’. Journal of Contemporary Asia 43 (3): 497–518.
Zhang, Xiaoyong, Ivo van der Lans, and Hans Dagevos. 2012. ‘Impacts of Fast Food and
the Food Retail Environment on Overweight and Obesity in China: A Multilevel Latent
Class Cluster Approach’. Public Health Nutrition 15 (1): 88–96.