Rural-urban differences in 
children's dietary diversity in 
Ethiopia 
A Poisson decomposition analysis 

Kalle Hirvonen  

May 2016 

ESSP WORKING PAPER 89 



 

 

TABLE OF CONTENTS 

Abstract ........................................................................................................................................................................................ 2 
1. Introduction ........................................................................................................................................................................ 2 

2. Data ................................................................................................................................................................................... 2 

3. Econometric approach ....................................................................................................................................................... 3 

4. Results ............................................................................................................................................................................... 4 

5. Concluding remarks ........................................................................................................................................................... 5 
References ................................................................................................................................................................................... 6 

 
LIST OF TABLES 

Table 2.1: Summary statistics ...................................................................................................................................................... 3 
Table 4.1: Decomposition of the rural-urban gap in children's dietary diversity .......................................................................... 5 

 
LIST OF FIGURES 

Figure 1.1: Children's dietary diversity by age in sub-Saharan Africa ......................................................................................... 2 

Figure 3.1: Fitting a Poisson distribution on dietary diversity score ............................................................................................. 4 

 



 

2 

ABSTRACT 

An emerging body of literature shows how low diversity in diets is associated with increased risk of chronic undernutrition 
and micro-nutrient deficiencies in young children. The latest available Demographic and Health Survey data for Ethiopia 
reveals unusually large differences in children's dietary diversity between rural and urban areas. Applying recently developed 
non-linear decomposition methods, this large rural-urban gap in dietary diversity can almost entirely be explained by 
differences in household wealth, parental education, and access to health services between rural and urban areas. 

Keywords: child dietary diversity, complementary feeding, count data, decomposition analysis 

JEL codes: Q18, C35 

1. INTRODUCTION 

An emerging body of literature shows how low diversity in diets is associated with increased risk of chronic undernutrition 
and micro-nutrient deficiencies in young children (Arimond and Ruel 2004; Kennedy et al. 2007; Moursi et al. 2008). A 
comparison of children's diets between Ethiopia and the rest of sub-Saharan Africa reveals two striking features (Figure 1.1). 
First, Ethiopian children consume a diet that is one of the most undiversified on the continent. Second, there exists an 
extraordinary large rural-urban gap in children's dietary diversity.  

Figure 1.1: Children's dietary diversity by age in sub-Saharan Africa 

 
Note: Local polynomial regression.  
Source: Own calculation from DHS data for 20 sub-Saharan African (SSA) countries. 

Recent research suggests that low child dietary diversity in Ethiopia is due to a combination of poor access to 
nutritious foods and limited knowledge about appropriate feeding practices (Hirvonen and Hoddinott 2014; Stifel and Minten 
2015; Hirvonen et al. 2016). This research paper examines the second striking feature observed in Figure 1.1: the large 
rural-urban gap in children's dietary diversity. Using the latest available Demographic and Health Survey (DHS) data for 
Ethiopia and applying recently developed non-linear decomposition techniques (Yun 2004; Bauer and Sinning 2008; Park 
and Lohr 2010), this gap is shown to be almost entirely due to differences in wealth and parental education levels, as well as 
unequal access to health services (antenatal care).  

2. DATA 

The analysis is based on the nationally representative 2010/11 Demographic Health Survey (DHS) data for Ethiopia. The 
2010/11 survey instrument contained a standard module to collect information about children's diets. Specifically, the 
questionnaire included a series of Yes/No questions about children's food consumption in the previous day. Following WHO 



 

3 

(2008) guidelines for assessing the feeding practices of children between 6 and 23 months of age, the responses were 
grouped into the following seven food group categories: grains, roots and tubers (e.g. barley, enset, maize, teff, and wheat); 
legumes and nuts; dairy products (milk, yogurt, cheese); flesh foods (meat, poultry and fish products); eggs; Vitamin A rich 
fruits and vegetables; and other fruits and vegetables. Totalling the number of food groups consumed by a child yields a 
dietary diversity score ranging in value from zero to seven. This simple indicator is considered in the literature as a good 
proxy of the quality of children's diets (Ruel 2003; Steyn et al. 2006; Kennedy et al. 2007; Moursi et al. 2008).  

The DHS surveys routinely collect information on household characteristics, including education levels, assets and 
access to health services. This information is used to construct the covariates used in the decomposition analysis. After data 
cleaning, the final sample used in this analysis includes 2,898 children (2,383 rural and 515 urban) aged 6 to 23 months. 
Table 2.1 provides the summary statistics for all variables used in this study.1 

Table 2.1: Summary statistics 
 rural urban difference 

Dietary diversity score 1.491 1.847 -0.36*** 
 (1.043) (1.371)  
Child’s age (months) 13.90 13.67 0.23 
 (5.175) (5.134)  
Female child (0/1) 0.495 0.505 -0.01 
 (0.500) (0.500)  
Maternal education (years) 0.863 2.519 -1.66*** 
 (1.739) (2.693)  
Paternal education (years) 1.753 3.400 -1.65*** 
 (2.398) (2.412)  
Mother's age (years) 28.24 28.06 0.18 
 (6.539) (5.969)  
Four or more antenatal visits (0/1) 0.125 0.487 -0.36*** 
 (0.331) (0.500)  
DHS asset index -53,900 87,757 -141,657*** 
 (30,346) (86672)  
Owns cows, bulls, bulls, goats or sheep (0/1) 0.881 0.292 0.59*** 
 (0.324) (0.455)  
Owns chickens (0/1) 0.617 0.202 0.42*** 
 (0.486) (0.402)  
observations 2,383 515  

Note: (0/1) indicates a binary (dummy) variable.  
Statistical significance based on a two-tailed t-test and denoted at *** p<0.01, ** p<0.05, * p<0.1. 

3. ECONOMETRIC APPROACH 

Following Blinder (1973) and Oaxaca (1973), the difference in the mean dietary diversity (𝐷𝐷�) between rural (subscript r) and 
urban (subscript u) areas is formally expressed as: 

𝐷𝐷�𝑟𝑟 − 𝐷𝐷�𝑢𝑢 = �𝑓𝑓�𝑋𝑋�𝑟𝑟�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟�� + �𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑢𝑢��, (1) 

where 𝑋𝑋� refers to a vector of covariates at mean values and �̂�𝛽 to estimated coefficients. The first part of the right-hand side 
of the equation �𝑓𝑓�𝑋𝑋�𝑟𝑟�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟�� captures the 'explained' component, which is due to differences in child or household 
characteristics between rural and urban areas (in coefficients estimated for the rural sample). The second part is the 
'unexplained' component �𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑢𝑢��, which is due to differences in the estimated coefficients. 

The functional form (𝑓𝑓) depends on the underlying data generating process (linear or non-linear). Our dependent 
variable of interest – the number of food groups consumed by the child (dietary diversity score) – takes only non-negative 
integer values. This warrants a count-data modelling approach (Winkelmann 2008). Fortunately, the Poisson distribution fits 
the unconditional distribution extremely well (Figure 3.1). Of note is that over-dispersion in the form of excess zeroes, does 
not seem to be a concern for the analysis. The Poisson model can be used to estimate the �̂�𝛽 coefficients in equation (1). 
                                                           
1 The choice of the covariates is motivated by Headey (2014) who offers a careful analysis of the long-run trends in child nutrition in Ethiopia. 



 

4 

Specifically, a maximum likelihood method is used to estimate the following Poisson model separately for the rural and urban 
samples, for child i residing in household h: 

𝐷𝐷�𝑖𝑖ℎ = exp(𝑐𝑐𝑖𝑖ℎ′ 𝛾𝛾 + 𝑥𝑥ℎ′ 𝛿𝛿 + 𝜀𝜀𝑖𝑖ℎ),  (2) 

where 𝑐𝑐𝑖𝑖ℎ′  is a vector of child level characteristics (sex and age in months) and 𝑥𝑥ℎ′  is a vector of household level 
characteristics that includes maternal and paternal education in years, a wealth index, mother's age, and livestock 
ownership. 

Figure 3.1: Fitting a Poisson distribution on dietary diversity scores 

 
Source: Own calculation from 2010/11 Demographic Health Survey for Ethiopia 

The contribution of each variable in Equation (1) to the overall difference in dietary diversity between rural and urban 
areas is also examined. In the case of a non-linear decomposition, the results of such detailed decomposition are sensitive 
to the order in which the variables enter the decomposition equation. The solution proposed by Yun (2004) is to apply 
weights that are proportional to the overall contribution of the characteristics or coefficient portion to the difference. The 
equation for the detailed decomposition for K explanatory variables can now be expressed as: 

𝐷𝐷�𝑟𝑟 − 𝐷𝐷�𝑢𝑢 = ∑ 𝑤𝑤∆𝑋𝑋
𝑖𝑖 �𝑓𝑓�𝑋𝑋�𝑟𝑟�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟��𝐾𝐾

𝑖𝑖=1 + ∑ 𝑤𝑤∆𝛽𝛽
𝑖𝑖 �𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑟𝑟� − 𝑓𝑓�𝑋𝑋�𝑢𝑢�̂�𝛽𝑢𝑢��𝐾𝐾

𝑖𝑖=1  (3) 

where  

𝑤𝑤∆𝑋𝑋𝑖𝑖 = �𝑋𝑋𝑟𝑟𝑖𝑖−𝑋𝑋𝑢𝑢𝑖𝑖 �𝛽𝛽𝑟𝑟𝑖𝑖

(𝑋𝑋𝑟𝑟−𝑋𝑋𝑢𝑢)𝛽𝛽𝑟𝑟
    and    𝑤𝑤∆𝛽𝛽

𝑖𝑖 = 𝑋𝑋𝑢𝑢𝑖𝑖 �𝛽𝛽𝑟𝑟𝑖𝑖−𝛽𝛽𝑢𝑢𝑖𝑖 �
𝑋𝑋𝑢𝑢(𝛽𝛽𝑟𝑟−𝛽𝛽𝑢𝑢) (4) 

The sum of each weight category (𝑤𝑤∆𝑋𝑋
𝑖𝑖  and 𝑤𝑤∆𝛽𝛽

𝑖𝑖 ) equals to one. The decomposition was implemented using Stata 14.1 
statistical software using the user-written mvdcmp command (Powers, Yoshioka, and Yun 2011).  

4. RESULTS 

Table 4.1 provides the results of the Poisson decomposition exercise. First, we see that differences in observed 
characteristics (or endowments) explain about 72 percent of the difference in children's dietary diversity between rural and 
urban areas. A detailed decomposition reveals that the rural-urban gap is mostly explained by wealth differences. Moreover, 
differences in parental (mostly maternal) education and access to antenatal care also have strong influences on this divide. 
Interestingly, the larger likelihood of owning livestock in the rural areas observed in Table 2.1 is found to narrow the 
difference in dietary diversity between the two groups. This is possibly due to the increased consumption of animal source 
foods in livestock owning households (Hoddinott, Headey, and Dereje 2015). 



 

5 

Table 4.1: Decomposition of the rural-urban gap in children's dietary diversity 

 Estimate 
Standard 

error Percent 
Explained: due to differences in characteristics -0.390*** 0.149 71.8 
Unexplained: due to differences in coefficients -0.153 0.164 28.2 
Raw difference -0.543*** 0.066 100.0 
Due to differences in characteristics:    

Age in months 0.004*** 0.000 -0.7 
Female child  0.000 0.002 0.1 
Maternal education (years) -0.128*** 0.030 23.5 
Paternal education (years) -0.058*** 0.020 10.7 
Mother's age 0.001 0.002 -0.1 
Four or more antenatal visits  -0.124*** 0.037 22.8 
DHS asset index -0.282* 0.147 52.0 
Owns cows, bulls, bulls, goats or sheep  0.099* 0.053 -18.2 
Owns chicken  0.099*** 0.025 -18.3 

Due to differences in coefficients:    
Age in months -0.413 0.252 76.2 
Female child  -0.024 0.078 4.4 
Maternal education (years) 0.103 0.081 -18.9 
Paternal education (years) 0.182 0.106 -33.5 
Mother's age 0.335 0.370 -61.7 
Four or more antenatal visits  -0.007 0.100 1.2 
DHS asset index 0.005 0.136 -0.9 
Owns cows, bulls, bulls, goats or sheep  -0.026 0.052 4.9 
Owns chicken  0.048 0.032 -8.8 

Note: A detailed non-linear (Poisson) decomposition. The 'Percent' column gives the contribution of each variable to the overall difference in children's 
dietary diversity between rural and urban areas. This is computed by dividing the estimate by the overall difference (0.543).  
Statistical significance denoted at *** p<0.01, ** p<0.05, * p<0.1. Sample: 2,898 children who are 6-23 months of age. 

5. CONCLUDING REMARKS 

The non-linear decomposition exercise reveals that the rural-urban gap in children's dietary diversity is mostly due to 
differences in household wealth and parental education, as well as unequal access to health care. Therefore, on the policy 
front, Ethiopia should continue its efforts to expand access in rural areas to education and health care services, e.g., nutrition 
counselling. Carefully designed livestock interventions may also generate significant positive outcomes in diets, as rural 
livestock ownership is found to narrow the rural-urban gap in children's dietary diversity. Finally, even though dietary diversity 
in the urban areas is higher than in the rural areas, the average child in both these areas is far from meeting the minimum 
dietary diversity (four or more food groups per day) for infant and young children as specified by the World Health 
Organization (WHO 2008).  

 



6 

REFERENCES 

Arimond, M., and M.T. Ruel. 2004. "Dietary diversity is associated with child nutritional status: evidence from 11 
demographic and health surveys." The Journal of Nutrition 134 (10): 2579-2585. 

Bauer, T.K., and M. Sinning. 2008. "An extension of the Blinder–Oaxaca decomposition to nonlinear models." Advances in 
Statistical Analysis 92 (2): 197-206. 

Blinder, A.S. 1973. "Wage discrimination: reduced form and structural estimates." Journal of Human Resources 8 (4): 436-
455. 

Headey, D. 2014. An Analysis of Trends and Determinants of Child Undernutrition in Ethiopia, 2000-2011. ESSP Working 
Paper 70, Washington, DC: International Food Policy Research Institute. 

Hirvonen, K., and J. Hoddinott. 2014. Agricultural production and children’s diets: Evidence from rural Ethiopia. ESSP 
Working Paper 69, Washington, DC: International Food Policy Research Institute. 

Hirvonen, K., J. Hoddinott, B. Minten, and D. Stifel. 2016. Children’s diets, nutrition knowledge and access to markets. ESSP 
Working Paper 84, Washington DC: International Food Policy Research Institute. 

Hoddinott, J., D. Headey, and M. Dereje. 2015. "Cows, missing milk markets and nutrition in rural Ethiopia." Journal of 
Development Studies 51 (8): 958–975. 

Kennedy, G.L., M.R. Pedro, C. Seghieri, G. Nantel, and I. Brouwer. 2007. "Dietary diversity score is a useful indicator of 
micronutrient intake in non-breast-feeding Filipino children." The Journal of Nutrition 137 (2): 472-477. 

Moursi, M.M., M. Arimond, K.G. Dewey, S. Trèche, M.T. Ruel, and F. Delpeuch. 2008. "Dietary diversity is a good predictor 
of the micronutrient density of the diet of 6-to 23-month-old children in Madagascar." The Journal of Nutrition 138 (12): 
2448-2453. 

Oaxaca, R. 1973. "Male-female wage differentials in urban labor markets." International Economic Review 14 (3): 693-709. 
Park, T.A., and L. Lohr. 2010. "A Oaxaca–Blinder decomposition for count data models." Applied Economics Letters 17 (5): 

451-455.
Powers, D.A., H. Yoshioka, and M.-S. Yun. 2011. "mvdcmp: Multivariate decomposition for nonlinear response models." 

Stata Journal 11 (4): 556-576. 
Ruel, M.T. 2003. "Operationalizing dietary diversity: a review of measurement issues and research priorities." The Journal of 

Nutrition 133 (11): 3911S-3926S. 
Steyn, N., J. Nel, G. Nantel, G. Kennedy, and D. Labadarios. 2006. "Food variety and dietary diversity scores in children: are 

they good indicators of dietary adequacy?" Public Health Nutrition 9 (5): 644-650. 
Stifel, D., and B. Minten. 2015. Market Access, Welfare, and Nutrition: Evidence from Ethiopia. ESSP Working paper 77, 

Washington, DC: International Food Policy Research Institute. 
WHO (World Health Organization). 2008. Indicators for assessing infant and young child feeding practices: Part 1: 

definitions. Conclusions of a consensus meeting held 6-8 November 2007 in Washington DC, USA. Geneva: WHO. 
Winkelmann, R. 2008. Econometric Analysis of Count Data. 5th ed. Berlin and Heidelberg: Springer-Verlag. 
Yun, M.-S. 2004. "Decomposing differences in the first moment." Economics Letters 82 (2): 275-280. 



INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE 
2033 K Street, NW | Washington, DC 20006-1002 USA 
T: +1.202.862.5600 | F: +1.202.457.4439 
Skype: ifprihomeoffice | ifpri@cgiar.org | www.ifpri.org  

ETHIOPIAN DEVELOPMENT RESEARCH INSTITUTE 
P.O. Box 2479, Addis Ababa, Ethiopia 
T: +251.11.550.6066; +251.11.553.8633 | F: +251.11.550.5588 
info@edri-eth.org | www.edri-eth.org 

IFPRI–ESSP ADDIS ABABA 
P.O. Box 5689, Addis Ababa, Ethiopia  
T: +251.11.617.2000 | F: +251.11.646.2318 
mahlet.mekuria@cgiar.org | http://essp.ifpri.info 

The Ethiopia Strategy Support Program (ESSP) is financially supported by the United States Agency for International Development (USAID) and the Department for International 
Development (DFID) of the government of the United Kingdom and is undertaken as part of the CGIAR Research Program on Policies, Institutions, and Markets (PIM) led by the 
International Food Policy Research Institute (IFPRI). This publication has been prepared as an output of ESSP and has not been independently peer reviewed. Any opinions 
expressed here belong to the author(s) and do not necessarily reflect those of IFPRI, the Ethiopian Development Research Institute, USAID, DFID, PIM, or CGIAR. 

Copyright © 2016 International Food Policy Research Institute. All rights reserved. To obtain permission to republish, contact ifpri-copyright@cgiar.org. 

About the Author 

Kalle Hirvonen is a Research Fellow in the Development Strategy and Governance Division of IFPRI, working under the 
Ethiopia Strategy Support Program (ESSP) jointly with the Ethiopian Development Research Institute (EDRI) in Addis Ababa. 

Acknowledgment 

Funding for this work was received through the Feed-the-Future initiative funded by the United States Agency for International 
Development (USAID). I thank the Demographic and Health Surveys (DHS) Program for making the data available.  

About ESSP 

The Ethiopia Strategy Support Program is an initiative to strengthen evidence-based policymaking in Ethiopia in the areas of rural 
and agricultural development. Facilitated by the International Food Policy Research Institute (IFPRI), ESSP works closely with 
the government of Ethiopia, the Ethiopian Development Research Institute (EDRI), and other development partners to provide 
information relevant for the design and implementation of Ethiopia’s agricultural and rural development strategies. For more 
information, see http://www.ifpri.org/book-757/ourwork/program/ethiopia-strategy-support-program; http://essp.ifpri.info/; or 
http://www.edri-eth.org/. 

About these working papers 

The ESSP Working Papers contain preliminary material and research results from IFPRI and/or its partners in Ethiopia. The 
papers are not subject to a formal peer review. They are circulated in order to stimulate discussion and critical comment.  

mailto:ifpri@cgiar.org
http://www.ifpri.org/
mailto:info@edri-eth.org
http://www.edri-eth.org/
mailto:mahlet.mekuria@cgiar.org
http://essp.ifpri.info/

	Table of Contents
	List of Tables
	Abstract
	1. Introduction
	Figure 1.1: Children's dietary diversity by age in sub-Saharan Africa

	2. Data
	Table 2.1: Summary statistics

	3. Econometric approach
	Figure 3.1: Fitting a Poisson distribution on dietary diversity scores

	4. Results
	Table 4.1: Decomposition of the rural-urban gap in children's dietary diversity

	5. Concluding remarks
	References