Policy Research Working Paper 8823 Food Prices, Access to Markets and Child Undernutrition in Ethiopia Paul Brenton Mike Nyawo Macroeconomics, Trade and Investment Global Practice April 2019 Policy Research Working Paper 8823 Abstract This paper looks at how changing food prices affect child studies, rising crop prices are positively associated with undernutrition in Ethiopia. It derives height for age (stunt- improved child stunting rates for children between ages 6 ing) and weight for height (wasting) as indicators of child months and 5 years, while the results for wasting are not undernutrition from the two most recent years of the Liv- conclusive. These results suggest that across the board policy ings Standards Measurement Survey and utilizes market interventions that seek to suppress cereal price increases prices for key cereals, teff, wheat, and maize at the zone level may have adverse effects on poverty reduction in the long across all regions of the country. Using a panel data fixed term by undermining potentially positive impacts on child effects model, the analysis finds that, contrary to previous nutrition. This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at pbrenton@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Food Prices, Access to Markets and Child Undernutrition in Ethiopia Paul Brenton1 and Mike Nyawo2 Keywords Food prices; Child undernutrition; Stunting; Wasting; Ethiopia, Livings Standards Measurement Survey JEL I15, O13, Q11 1 Global Trade and Regional Integration, World Bank Group 2 Global Trade and Regional Integration, World Bank Group We are grateful to Carlo Azzarri, Tom Bundervoet, Chris Garbers, Mombert Hoppe, Neil Rankin, Asmelash Tsegay and Manex Bule Yonis for comments and suggestions. This research benefited from the Umbrella Facility for Trade trust fund that is supported by the Governments of the Netherlands, Norway, Sweden, Switzerland and the United Kingdom. 1. Introduction Do higher food prices help or undermine efforts to reduce poverty in developing countries? This remains a critical issue in guiding policy, since most the world’s poor live in rural areas and work in agriculture. At the same time, expenditures on food account for the largest proportion of the budget of poor households. Global integration and international trade play a key role in determining food price developments that face poor farmers. The recent increases in food prices across the globe have led a number of governments to limit trade in agricultural products and intervene in agricultural markets, more often in the name of protecting the poor (Jacoby, 2013). Most studies that have sought to assess the impact of higher food prices on the welfare of poor households have focused on whether the household is a net consumer or net producer of food. Since most poor households have been found to be net consumers of food, analyses using this approach typically conclude that higher food prices are likely to increase poverty in the short run (see, for example, Ivanic and Martin ((2008), (2014)). However, there are some concerns with this approach. First, there is the issue of the reliability of the data on the net food position of households from household surveys. Substantial errors can arise from recall biases that lead to the underestimation of food production and from annualization of short recall responses that overstate consumption (see Headey (2016)). Second, there may be second-round effects beyond those on household income that can lead to households adjusting to higher food prices in ways that lead to lower poverty in the medium term. For example, food producing households could increase their use of fertilizers and/or invest in higher yielding seeds, that contribute to higher productivity and greater output. An increase in the demand for unskilled labor in rural areas, where poverty is concentrated, could result in higher wage income for the household. Finally, intra-household distributional decisions will influence the outcome of rising food prices in ways which are not captured in the net food position of the household. This paper explores an important channel from food prices to poverty which has received less attention in the literature; the effects of rising food prices on child undernutrition. Child undernutrition is closely related to household welfare but can reflect the impact of household distributional decisions to rising food prices. This is important given the recent finding of 1 Brown et al., (2017) that a high proportion of undernourished children in Africa reside in non- poor households. Poor developmental outcomes for children have a lifelong impact on their well-being. Childhood undernutrition increases the risk of death from common childhood illnesses such as diarrhea, measles, pneumonia and malaria. Black et al. (2013) argue that undernutrition is a cause of approximately 3.1 million child deaths annually, most of which occur in developing countries. Child undrnutrition increases the costs of health care and social safety nets and lowers the efficacy of investments in education. In addition, childhood undernutrition is associated with delays in cognitive development which can lead to poorer learning outcomes in school and decreased productivity and earning potential in adulthood.3 As a result, the costs of child undernutrition in Africa and Asia have been estimated to range from 4 percent to 11 percent of GDP. This paper uses detailed data on crop prices and nutritional outcomes for children under 5 from the Living Standards Measurement Survey (LSMS) to examine how food prices and accessibility to markets affect child undernutrition in Ethiopia. Approximately 80 percent of the population live in rural areas in Ethiopia and agriculture generates 40 percent of gross domestic product (World Bank 2016).4 The empirical analysis seeks to contribute to the debate on the role the government can play in preventing childhood deaths in a context where child undernutrition may contribute to 51 percent of childhood deaths annually (Government of Ethiopia (2013)).5 Despite substantial progress in reducing child undernutrition over the last decade, a large proportion of children under 5 years are still stunted and underweight (Cintron, Seff, & Baird, 2016). The remainder of the paper is structured as follows: in section 2 we briefly summarize the main features of the literature on the links between food prices and household welfare in the context of a simple conceptual framework by which to trace through the linkages between global price changes and farm-level prices and how these impact households. We then look at the 3 Estimates suggest that this can reduce a person’s potential lifetime earnings by between 10 and 50 percent. See http://www.worldbank.org/en/topic/earlychildhooddevelopment and http://1000days.unicef.ph/. 4 See http://www.worldbank.org/en/country/ethiopia/overview 5 https://www.undp.org/content/dam/ethiopia/docs/GTP%20APR%202004%20English%20Version_Sept%207.p df 2 framework in which to situate the limited work that looks at the impact of food prices on child undernutrition. Section 3 discusses the available data that we are able to utilize and Section 4 presents our empirical approach to the impact of changing food prices on child nutritional status. Section 5 presents the key descriptive statistics and then discusses the main results from the empirical analysis. Section 6 reviews the main findings in relation to previous studies, while Section 7 provides conclusions. 2. Food Prices and Child Undernutrition: A Framework and the Evidence There is a substantial literature on the impact of food prices on household welfare. For example, Headey (2016) examines the cross-country evidence from over 50 developing countries and finds that rising food prices are associated with reduced national poverty rates. The paper suggests this stems from: (i) the large number of poor people who still live in rural areas and that their welfare is largely determined by farm and non-farm agricultural incomes, (ii) that agricultural supply in many countries has increased quickly and substantially due to higher prices, and (iii) some limited evidence that wages rise quickly for rural but not urban workers. The linkages between changing prices and household welfare in an open economy in which shocks from the global economy are transmitted to local markets are shown in Figure 1. A range of studies suggest that the transmission of changes in global food prices to domestic food prices is far from perfect and that the extent of pass-through varies across countries and is influenced by factors including the degree of competition in domestic transport and distribution sectors. Minot (2011) for example, found a long-term relationship with world prices in only 13 of 62 African food prices examined. For Ethiopia specifically, a number of studies suggest little impact from the global market on local food prices (see for example Rashid (2010)) while others find significant price transmission (Leoning et al,. (2009)). Durevall et al., (2013) conclude that the external sector largely determines inflation in Ethiopia in the long run with domestic food prices adjusting to changes in world food prices. In this study, rather than looking at overall food price inflation, we look at the effects of changing prices for three different staples, teff, wheat and maize, for which interaction with the global market is quite different. These are all crops that are both produced and consumed in Ethiopia. Teff is a cereal produced almost exclusively in Ethiopia. Until recently exports were banned, although in 2015 the government started allowing exports of teff flour and injera (the bread made from teff). However, it is possible that there are and have been significant exports 3 of teff grains to neighboring countries such as Somalia and Djibouti that are not recorded. Wheat, on the other hand, is heavily imported, and the market has been subject to considerable intervention by the government and the World Food Programme (WFP). The elements of the value chain at the wholesale and retail levels that affect the translation of prices from world markets to the final consumer or producer in Ethiopia shown in Figure 1 may well differ for the different food products, for example, if there are different levels of competition among millers of the different crops. Figure 1: The transmission of trade shock Source: Winters (2002) There are however, a number of mechanisms which impact child nutritional outcomes beyond the level of poverty of the household, as suggested by the finding of Brown et al., (2017). A commonly applied framework developed by UNICEF identifies that child nutrition depends not only on food intake but also access to health, parental care, medical facilities, elementary education, drinking water, and sanitary facilities (UNICEF (2003)). Within this broad approach the price of food represents a key determinant of real income, access to food, dietary intake 4 and the nutritional status of the child.6 In regions in which agriculture is the dominant economic activity, food prices, through their effects on incomes and revenues for the state and local authorities, can also affect both the demand and provision of health, education and sanitation. Studies examining the association between changing food prices and child nutritional outcomes are limited. Arndt et al., (2016) use household survey data and food price inflation rates to study the relationship between shifts in food prices and child nutrition status in Mozambique. Using propensity score matching, the study shows that nutrition measures (weight for height and weight for age) improve significantly when the inflation rate for food products is low. However, their analysis focuses on general price inflation and covers changes between quarters within one year and so does not allow for longer term impacts. Torlesse, Kiess, and Bloem (2003) find that in the 1990s the price of rice was correlated with the percentage of underweight preschool children in rural Bangladesh. This seems to reflect a lack of substitution away from rice as the amount of rice consumed per capita did not change in response to price changes. As a result, household expenditure on rice varied with the fluctuations in price and less was spent on non-rice foods when rice expenditure increased. Expenditures on non-rice products were found to be negatively correlated with the percentage of underweight children. Vellakkal et al., (2015) investigate the association between food price hikes in the 2000s and child undernutrition in India. They first look at the association between changes in food prices and food consumption, and then the relation between these consumption changes and child nutrition. The study suggests that rising food prices were associated with an increased risk of undernutrition among children. Miller and Urdinola (2011) investigate how child survival in Colombia responds to fluctuations in world coffee prices. They argue that the health of children depends on time consuming tasks such as breastfeeding, bringing clean water from far away or taking a child to a health clinic for primary care services or vaccinations. When coffee prices increase, workers spend more 6 Azzarri et al (2014), for example, find a high elasticity of stunting to income in Mozambique. In addition to income, the most important factors correlated to chronic child undernutrition are mothers’ characteristics and fertility history, and parents’ human capital endowments. They conclude that improved income-generating opportunities are crucial for improving child nutritional status. 5 time tending the crop and less time on things that are good for children's health, and so child health outcomes can worsen even though incomes from coffee production increase. Finally, Woldemichael et al., (2017) look at the impact of price inflation of teff, maize and wheat on a child’s health at different periods of development in the months both before and after birth. Using data from different rounds of the Demographic and Health Surveys and separate data on prices at the local level, they find a u-shaped relationship between food price inflation and a child’s height for age (stunting). The impact is more pronounced for teff than for wheat and maize. For wasting (weight for height), there is no significant impact of food price inflation on a child’s development. Overall, a common theme that emerges from this literature is that there are significant risks to children’s nutritional development from rising food prices. 3. Data on Prices, Households and Child Nutritional Outcomes in Ethiopia In this study we use information on individual (child) outcomes and individual, household and community characteristics from the two most recent rounds of the Living Standards Measurement Survey (LSMS) undertaken in Ethiopia.7 The data set contains detailed information on health statistics for children under 5 years of age which allows the construction of measures of child undernutrition. The community data come with detailed geo-spatial household information including accessibility to markets. Table 1 summarizes the year, survey period and the number of children under 5 years. Table 1: Summary of Ethiopia LSMS data for children aged between 6 to 59 months Year Survey period No. of children (6-59 months) 2013/14 September 2013-March 2014 2,880 2015/16 September 2015-April 2016 2,993 We study individual nutritional outcomes of children and follow previous studies and compute indicators of undernutrition for children between 6 and 59 months.8 To achieve this we compute z-scores for height for age (stunting).9 A child is stunted if the height-for-age z-score is below -2 (that is two standard deviations below the mean of the reference group). According to WHO 7 There was also a first wave of the LSMS in Ethiopia in 2011/12. 8 Infants aged under 6 months are often excluded from nutrition surveys. 9 The z-scores are calculated using the 2006 World Health Organization (WHO) child growth guidelines. The analysis is done using zscore06 Stata module. 6 definitions stunting “reflects a process of failure to reach linear growth potential as a result of suboptimal health and/or nutritional conditions.” Stunting is an indicator of persistent, longer term chronic undernutrition and has long-term negative effects on the child’s development (Brown, 2017). We also calculate weight-for-height (wasting). A z-score of less than -2 for wasting suggests recent rapid and severe weight loss which is often associated with acute starvation and/or severe disease. We use market prices for three crops – teff, wheat and maize, collected through the community questionnaire in the LSMS.10 These are key crops in Ethiopia and account for the largest share of total cereal area cultivated (Taffesse, Dorosh, and Asrat, 2011). The data are collected at item level from two different vendors in each enumeration area across Ethiopia. An average price is computed from the two different vendors, which forms the price in that enumeration area. Table 2 presents summary statistics of the three crop prices in wave 1, wave 2 and wave 3.11 Table 2: Summary statistics of market prices in ETB/kg Teff Observations Mean Std Dev Min Max 2011/12 258 9.64 2.3 3.5 15.67 2013/14 347 13.1 2.68 4 18 2015/16 365 16.42 3.52 4 22 Wheat Observations Mean Std Dev Min Max 2011/12 248 8.2 2.58 4.46 15.67 2013/14 339 8.96 2.66 4.55 18 2015/16 365 10.48 3.43 4 22 Maize Observations Mean Std Dev Min Max 2011/12 268 5.33 2.11 3.5 15.69 2013/14 381 6.39 2.4 4 18 2015/16 389 6.4 3.18 4 22 Source: LSMS Survey dataset The mean price for all three crops increased considerably between the three waves. For example, the mean price for teff increased from 9.64/kg in wave 1 to 16.42/kg in wave 3; the 10 We explored using monthly data collected by the Ethiopian Central Statistical Agency on crop prices at the district level, as used by Woldemichael et al., (2017). However, in our case we faced the challenge of many missing observations and that results were sensitive to the approach adopted to dealing with these. 11 Table A.2 in the appendix presents the average price for each region. 7 mean price of wheat increased from 8.20/kg in wave 1 to 10.48/kg in wave 3; the mean price for maize increased from 5.33/kg in wave 1 to 6.40/kg in wave 3. For purposes of our analysis, we compute a yearly average price for each of the three crops at the zone level.12 13 4. Empirical Approach To assess how food prices affect child undernutrition, we estimate the following equation: (1) Where is the outcome indicator for child i at time t in terms of the z score for height for age and weight for height;14 is a vector of individual characteristics; is a vector of household characteristics; is a vector for community characteristics such as road access and access to markets. Our variable of particular interest, , is the price for each crop at the district/zone level within each region in Ethiopia while are region fixed effects15 and is the error term. For stunting, which measures persistent long-term child undernutrition, the impact of changes in prices is likely to take time to be realized. We therefore lag the crop price by one period before merging again with wave 2 and wave 3 of the LSMS data set and include in the model. For wasting, which captures more rapid weight loss, we use the contemporaneous price. The vector of control variables is drawn from previous studies that seek to explain child undernutrition rates (see Cintron, Seff, and Baird (2016), Edris (2007), Alemayehu et al. (2015), Liben, Abuhay and Haile (2016)). Specifically, we control for individual, household, and community factors that are associated with individual nutritional outcomes. Individual variables include a continuous variable for child’s age in months, child’s age in months squared since there might be a non-linear relationship and a binary variable for child’s sex. Household 12 Administratively, Ethiopia is divided into regions, zones, woredas and kebeles. Zones are a second-level sub- division of Ethiopia, below regions. They are subdivided into woreda’s or districts, which are further divided into kebele’s (wards). 13 Less than 1 percent of the observations are missing crop prices. In this case, we use the region average as the crop price for that zone. 14 We use continuous dependent variables rather than dichotomous (stunted/non-stunted, underweight/non- underweight) as categorization can weaken the power of statistical tests and may lead to residual confounding. A data determined 'optimal' cut-point can lead to serious bias (Royston et al., 2006; Weinberg, 1995). Using dichotomous variables instead of continuous would mean discarding an important dimension of the intensity of the phenomenon. 15 These regions include Tigray, Afar, Amhara, Oromia, Somalie, Benshagul Gumuz, SNNP, Gambelia, Harari, and Dire dawa. 8 characteristics include literacy of the father and the mother.16 Due to the depth of the LSMS survey, there are numerous household variables that capture the income status of the household which include floor type, roof type, and access to improved toilet facilities.17 Because child undernutrition is highly influenced by sanitation, we control for improved sanitation facilities.18 For community-level characteristics, we use distance to main road and distance to market as control variables. Since there is a non-linear relationship between the community variables, we control for distance to main road, distance to main road squared, distance to market and distance to market squared. We apply a panel data fixed effects approach to equation (1). We measure changes for children who were included both in wave 2 and wave 3 of the LSMS survey data set – those aged 6-41 months old in wave 2 and 24-59 months old in wave 3. A panel data fixed effects model controls for unobservable time-invariant factors, thereby focusing on what drives changes between periods and eliminating the effects of time invariant factors such as gender. We note that other characteristics, for example, the literacy rates of mothers and fathers, are very unlikely to change in a relatively short panel.19 5. Results and Analysis 5.1. Descriptive Statistics: Child Undernutrition in Ethiopia Table 3 summarizes the information on nutritional outcomes for children between 6 and 59 months from the LSMS data set. Stunting increased from 40 percent in wave 2 (2013/14) to 41.7 percent in wave 3 (2015/16). The LSMS data also show similar small increases in the proportion of children who were wasted. It is noteworthy that in Ethiopia there is little difference in the prevalence of stunting between boys and girls. For wasting the prevalence is 16 While some studies use mother’s education, we use mother’s literacy since the majority of our sample comes from rural areas and most of the women surveyed have no education. Cintron, Seff, and Baird (Cintron et al., 2016) use the same approach. 17 Since there is a possibility of correlation between household variables, we use the Pearson Correlation Coefficient to select variables that highly influence child undernutrition. The Pearson Correlation Coefficient is a number between -1 and 1 that indicates the extent to which two variables are linearly related. For this analysis, we choose 0.6 as the rule of thumb. If two variables have a correlation coefficient of greater than 0.6, we select one of the variables and drop the other. The variable selected is the one that highly influences child undernutrition. In this instance, we choose improved sanitation. 18 For a detailed analysis on how each household variable is associated with child nutritional outcomes, see Cintron, Seff, and Baird (2016). 19 Contrary to Cintron et al (2016) who apply a fixed-effects model including only independent variables for which 10% or more of the sample experienced a change between periods 1 and 2, we do not exclude any of our control variables from the analysis. 9 slightly lower among girls than boys. Child undernutrition is much lower in households in which the father is literate and similarly for the mother. Finally, we note that in both 2013/14 and 2015/16 the prevalence of child undernutrition appears to be somewhat lower in female headed households for both indicators. Table 3: Distribution of z-scores in Ethiopia, 2013/14 and 2015/16 (% below -2 S.D) for children between 6 to 59 months of age.20 2013/14 2015/16 Stunting Wasting Stunting Wasting Overall 40.0 10.9 41.7 10.3 Male 40.6 11.7 42.2 12.2 Female 39.4 10.0 41.2 8.3 Mother illiterate 43.3 12.1 45.5 11.8 Mother literate 31.3 7.6 35.9 9.5 Father Illiterate 47.7 14.3 45.3 9.9 Father literate 33.5 8.1 39.4 11.5 Female headed household 35.7  8.3  40.2 9.8 Table 4 shows the prevalence of child undernutrition for the two indicators according to the age group of the child. Children aged 12 – 24 months have higher rates of stunting compared to other age groups whereas for wasting the youngest group has the highest rate. For stunting, rates declined between the two survey years for children aged 24 to 36 months but increased for all other age groups. While for wasting, children between 24 and 36 and 36 to 48 months showed improvement. Table 4: Child nutritional status by age, 2013/14 and 2015/16 (% below -2 S.D) 2013/14 2015/16 Description Stunting Wasting Stunting Wasting Age       6 months - 12 months 28.3 11.6 36.1 16.7 12 months - 24 months 46.7 11.0 47.9 13.0 24 months - 36 months 43.3 11.3 41.7 10.4 36 months - 48 months 38.7 10.9 42.3 8.6 48 months - 59 months 34.5 9.4 41.0 10.5 20 We ran a t-test to check if the means of nutritional outcomes are statistically different over the two surveys. Our results indicate that the means of nutritional outcomes are not statistically different from each other although there were increases in stunting rates between the two surveys. Similar results were found for child nutritional status by age and by area and region. 10 Table 5 shows considerable variation in the prevalence of child undernutrition by region of Ethiopia. In 2013/14 stunting ranged from 32% in Oromia to 55% in Amhara, while wasting was lowest in SNNP (8%) and highest in other regions (20%). There are also substantial differences across regions in the change in undernutrition rates between 2013/14 and 2015/16. For example, stunting declined in Tigray and Amhara but increased in Oromia and SNNP.21 While stunting declined in Tigray, wasting increased. Table 5: Nutritional status by area and region, 2013/14 and 2015/16 (% below -2 S.D) for children aged 6 to 59 months old   Stunting Wasting Description 2013/14 2015/16 2013/14 2015/16 Area Rural 41.5 43.8 11.3 11.1 Small town 24.6 37.2 7.8 9.9 Region Tigray 49.9 46.7 11.4 18.6 SNNP 40.7 44.1 7.8 9.7 Amhara 55.1 45.2 10 9.8 Oromia 31.5 40.1 11.8 9.2 Other regions 34.8 37.0 19.4 24.4 5.2 Results from the Panel Data Fixed Effects Analysis We now turn to the panel data fixed effects22 model to examine if changes in crop prices are associated with changes in child nutritional outcomes. Table 6 presents the results of changes in crop prices on height for age. Table 6: Fixed effects regression results of crop prices on height for age (stunting) (1) (2) (3) (4) Other Teff Prices Wheat Prices Maize Prices Covariates (Prob (Prob > F (Prob > F = (Prob > F = VARIABLES > F = 0.0000) 0.0000) 0.0000) 0.0000) Age -0.0318** -0.0578*** -0.0383** -0.0390** (0.0150) (0.0163) (0.0157) (0.0152) Age squared 0.000372* 0.000332 0.000432* 0.000398* 21 There was a drought in some regions when the 2015/16 survey was undertaken. This could have contributed to the increase in child undernutrition rates in most of the regions between the two waves. 22 A Hausman test was used to determine whether a fixed effects or a random effects model was relevant. Our p- value was less than 0.05 for all the specifications, rejecting the null hypothesis that our fixed effects are not consistent and efficient. 11 (0.000221) (0.000232) (0.000230) (0.000221) Father literacy -0.0311 -0.135 -0.0418 -0.0655 (0.183) (0.198) (0.193) (0.183) Mother literacy 0.0419 0.0227 0.214 0.0281 (0.187) (0.208) (0.196) (0.187) Improved sanitation 0.450 0.657 0.519 0.526 (0.641) (0.629) (0.636) (0.638) Distance to main road -0.125* -0.194** -0.113 -0.115 (0.0702) (0.0778) (0.0724) (0.0699) Distance to market 0.0626 -0.0617 0.0544 0.0654 (0.0734) (0.129) (0.0800) (0.0731) Distance to main road squared 0.00361*** 0.00348** 0.00322** 0.00313** (0.00133) (0.00140) (0.00156) (0.00133) Distance to market squared -0.000120 0.000255 -8.74e-05 -0.000130 (0.000266) (0.000422) (0.000281) (0.000264) Teff pricet-1 0.231*** (0.0378) Wheat pricet-1 0.119*** (0.0383) Maize pricet-1 0.126*** (0.0430) Constant -5.753* 1.148 -6.271* -6.120* (3.331) (6.376) (3.503) (3.311) R-squared 0.022 0.063 0.034 0.031 Number of individuals 1,021 962 1,019 1,021 Observations 1,923 1,730 1,838 1,913 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 The coefficients on each of the crop prices are positive and strongly statistically significant. This suggests that changes in crop prices are associated with improved scores for height for age. For individual characteristics, age is negative and statistically significant. However, the impact of age is non-linear. Proximity of the community to a main road is associated with positive outcomes for height for age. This suggests that measures that increase the connectivity of rural communities will contribute to improving health and nutritional outcomes for children. Consistent with the results of Cintron et al (2016), many of the control variables are not significant since these time-varying dimensions are unlikely to change substantially in such a short panel. We now turn, in Table 7, to our other indicator of child undernutrition, weight for height (wasting). Table 7: Fixed effects regression of crop prices on weight for height (1) (2) (3) (4) Other Covariates Teff Prices Wheat Maize (Prob > F (Prob > F Prices(Prob Prices(Prob > F VARIABLES 0.0000) 0.0000) > F 0.0000) 0.0000) 12 Age -0.0286** -0.0289** -0.0237* -0.0317*** (0.0122) (0.0132) (0.0125) (0.0121) Age squared 0.000380** 0.000340* 0.000332* 0.000403** (0.000181) (0.000187) (0.000183) (0.000179) Father literacy -0.0817 -0.151 -0.0958 -0.0657 (0.149) (0.160) (0.155) (0.147) Mother literacy -0.260* -0.296* -0.290* -0.214 (0.152) (0.161) (0.159) (0.152) Improved sanitation -0.690 -0.692 -0.673 -0.708 (0.528) (0.528) (0.522) (0.520) Distance to main road 0.0159 0.0421 0.0590 0.0154 (0.0576) (0.0644) (0.0645) (0.0569) Distance to market 0.110* 0.149 0.200** 0.112* (0.0607) (0.102) (0.0837) (0.0599) Distance to main road squared -0.00183 -0.00195 -0.00274** -0.00181 (0.00112) (0.00119) (0.00133) (0.00110) Distance to market squared -0.000279 -0.000403 -0.000553** -0.000285 (0.000219) (0.000336) (0.000280) (0.000216) Teff prices 0.0247 (0.0266) Wheat prices -0.0239 (0.0297) Maize prices -0.0716** (0.0322) Constant -3.835 -6.964 -8.304** -3.481 (2.727) (4.949) (4.038) (2.701) Observations 1,949 1,833 1,870 1,934 R-squared 0.020 0.022 0.025 0.027 Number of individuals 1,022 986 1,004 1,022 N 1949 1833 1870 1934 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Contemporaneous crop prices are statistically insignificant when we use wasting as the dependent variable except for maize prices where the sign of the coefficient is negative. 6. Discussion of the Results This study investigated the association between food prices and child undernutrition in Ethiopia. Our results suggest that rising crop prices are associated with improved scores for child undernutrition of children aged 6-59 months. The largest effects on child undernutrition are observed for teff prices compared to maize and wheat prices. For example, a 10 percent increase in the price of teff leads to an improvement in height-for-age scores of children by 2.8 percent. Similarly, a 10 percent increase in the prices of maize and wheat leads to an 13 improvement in scores for height for age of children by 1.3 percent and 1.2 percent respectively. Results for weight for height are not conclusive. Our results differ from much of the previous literature that finds that rising food prices pose risks to children’s health (see Woldemichael, Kidane, & Shimeles (2017); Arndt et al., (2016); Vellakkal et al., (2015); Miller and Urdinola (2011); Torlesse et al., (2003)). There are a range of linkages from rising crop prices to child nutrition that go beyond the direct impacts on household incomes to include access to health. There are two important differences in our approach compared to the previous literature on food prices and child undernutrition. First, with regard to the price data used, the approach implemented here uses cereal price data collected from market stalls when the survey of child nutrition and household outcomes was conducted. Thus, our cereal prices may reflect more closely the prices households actually face compared to the price data collected by statistical agencies used in previous studies. The latter may be a less direct measure of the prices faced and typically suffer from the challenges of missing values and averaging. Second, our approach utilizes a panel data fixed effects model compared to the frequently used pooled regression analysis. Our results may be different since we track the same child between waves of the Living Standards Measurement Survey to examine the how changes in cereal prices are associated with changes in child undernutrition. Lastly, note that the majority of households in the LSMS survey come from rural areas and their welfare is likely to be closely related to agricultural development in terms of both farm and non-farm incomes.23 7. Conclusions The study investigated the association between food prices and child undernutrition in Ethiopia. The increase in crop prices between wave 2 (2013/14) and wave 3 (2015/16) in the LSMS survey created a unique opportunity to identify how food price changes affect child undernutrition. We use height for age (stunting) and weight for height (wasting) as indicators of child undernutrition and zone-level market prices of three cereals - teff, maize and wheat. Using a panel data fixed effects model, we show that increases in crop prices are associated with improved child stunting rates for children between 6 months and five years, while the 23 We could not run separate regressions to differentiate net buyers and net sellers since our survey does not provide sufficient information to make that distinction. 14 results for wasting are not conclusive. These results are consistent with the aggregate country- level results reported by Headey (2016), who finds that increases in food prices are associated with reductions in poverty in a large sample of developing countries. Our results question the notion that rising food prices necessarily have a detrimental impact on child nutritional outcomes in an economy in which the majority of the poor live in rural areas and are involved in the agricultural sector. A policy approach that seeks to suppress crop price increases may have long-term consequences for poverty reduction by constraining the potentially positive impact of rising crop prices on child undernutrition. An approach that avoids interventions to suppress prices and targets financial and other support to those among the poor who have been adversely affected by higher prices, as has happened in Ethiopia through safety net programs, will tend to be more effective in reducing poverty in the long term. Finally, teff is a crop that is only produced in the Horn of Africa, with the vast majority of production in Ethiopia. The analysis here suggests the need for further research to examine if the opening-up of export opportunities that in turn raise producer prices for teff farmers is likely to bring positive development outcomes in terms of additional reductions in child undernutrition rates. 15 References Alemayehu, M., Fitiwi, T., Kiday, H., Oumer, S., Gebremedhin, G., & Henock, Y. (2015). Undernutrition status and associated factors in under 5 children in Tigray, Northern Ethiopia. Nutrition, 31(7–8), 964–970. https://doi.org/10.1016/j.nut.2015.01.013 Arndt, C., Hussain, M. A., Salvucci, V., & Peter, L. (2016). Economics and Human Biology Effects of food price shocks on child malnutrition : The Mozambican experience 2008 / 2009. Economics and Human Biology, 22(September 2008), 1–13. https://doi.org/10.1016/j.ehb.2016.03.003 Azzarri, C., G. Carletto, B. Davis and A. Nucifora (2014) 'Child Undernutrition in Mozambique', Mimeo, World Bank Black, R. E., Victora, C. G., Walker, S. P., Bhutta, Z. A., Christian, P., Onis, M. De, & Ezzati, M. (2013). Maternal and Child Nutrition 1 Maternal and child undernutrition and overweight in low-income and middle-income countries, 6736(13). https://doi.org/10.1016/S0140-6736(13)60937-X Brown, C., M. Ravallion and D. van de Walle (2017). Are Poor Individuals Mainly Found in Poor Households ? Evidence Using Nutrition Data for Africa, NBER Working Paper No. 24047 Cintron, C., Seff, I., & Baird, S. (2016). Dynamics of Wasting and Underweight in Ethiopian Children. Ethiopian Journal of Economics, XXV(2). Edris, M. (2007). Assessment of nutritional status of preschool children of. Ethiopian Journal of Health Dev, 21(2), 125–129. Headey, D. D. (2016). Food Prices and Poverty. World Bank Policy Research Working Papers, (November). Ivanic, M., & Martin, W. (2008). Implications of higher global food prices for poverty in low- income countries. World Bank Policy Research Working Papers, 39(SUPPL. 1), 405–416. https://doi.org/10.1111/j.1574-0862.2008.00347.x Ivanic, M., & Martin, W. (2014). Short and Long-run Impacts of Food Price Changes on Poverty. World Bank Policy Research Working Paper, (December). Retrieved from http://iatrc.software.umn.edu/activities/annualmeetings/themedays/pdfs2011/2011Dec- TD-IvanicMartin_paper.pdf Jacoby, H. G. (2013). Food Prices, Wages, and Welfare in Rural India. World Bank Policy Research Working Paper, (April). Liben, M. L., Abuhay, T., & Haile, Y. (2016). iMedPub Journals Determinants of Child Malnutrition among Agro Pastorals in Northeastern Ethiopia : A Cross-Sectional Study, 1–10. Loening, J. L., Durevall, D., & Birru, Y. A. (2009). Inflation Dynamics and Food Prices in an Agricultural Economy : The Case of Ethiopia Inf. Gothenburg Centre for Globalization and Developmen, 2473(347). 16 Miller, G., & Urdinola, P. B. (2011). Cyclicality, Mortality, and the Value of Time: The Case of Coffee Price Fluctuations and Child Survival in Colombia. Journal of Political Economy, 118(1), 113–155. Rashid, S. (2010). Staple Food Prices in Ethiopia. Michigan State University, Department of Agricultural, Food, and Resource Economics, Food Security Collaborative Working Papers.. Taffesse, A. S., Dorosh, P., & Asrat, S. (2011). Crop Production in Ethiopia : Regional Patterns and Trends. Ethiopia Strategy Support Programme II Working Papers, 0016. Torlesse, H., Kiess, L., & Bloem, M. W. (2003). Community and International Nutrition Association of Household Rice Expenditure with Child Nutritional Status Indicates a Role for Macroeconomic Food Policy in Combating Malnutrition, (June 2002), 1320–1325. Vellakkal, S., Fledderjohann, J., Basu, S., Agrawal, S., Ebrahim, S., Campbell, O., … Stuckler, D. (2015). Food Price Spikes Are Associated with Increased Malnutrition among Children in, (C). https://doi.org/10.3945/jn.115.211250 Winters, L. A. (2002). Trade Liberalisation and Poverty : What are the Links ? Blackwell Publishers Ltd. Woldemichael, A., Kidane, D., & Shimeles, A. (2017). A Tax on Children? The Effects of Food Price Inflation on Child Health. African Development Bank Working Paper Series. Retrieved from http://pubdocs.worldbank.org/en/766351495654703618/B1- InflationChildHealthABCA.pdf 17 Appendix Table A.1 Description of variables used in the empirical analysis VARIABLES Description Gender Binary variable for sex of the individual Age Continuous variable for age of the individual Father literacy Binary variable for whether the Father attended formal education or otherwise Mother literacy Binary variable for whether the Mother attended formal education or otherwise Binary variable for whether the household has better sanitation facilities (flush Improved sanitation toilet) or not. Distance to main Distance to main road from the household where the individual lives road Distance to market Distance to market from the household where the individual lives Distance to main Quadratic distance to main road from the household where the individual lives road squared Distance to market Quadratic distance to market from the household where the individual lives squared Teff pricet-1 Average teff price at the district/zone level in the previous survey period Wheat pricet-1 Average wheat price at the district/zone level in the previous survey period Barley pricet-1 Average barley price at the district/zone level in the previous survey period 18 Table A.2: Average crop prices by year and region (ETB/kg) Teff Tigray Amhara Oromia SNNP Addis Ababa Other regions 2011/12 9.46 9.13 9.56 9.29 - 11.37 2013/14 13.38 12.72 12.36 12.37 16.33 14.75 2015/16 16.36 14.65 15.64 16.51 19.15 19.04 Wheat Tigray Amhara Oromia SNNP Addis Ababa Other regions 2011/12 8.21 8.41 7.87 7.45 - 9.24 2013/14 8.26 8.8 7.85 8.85 13.88 9.72 2015/16 10.51 9.82 9.55 10.64 12.7 11.42 Maize Tigray Amhara Oromia SNNP Addis Ababa Other regions 2011/12 4.99 5.67 5.25 4.51 - 6.18 2013/14 6.44 6.3 5.66 5.94 8.79 7.16 2015/16 8.47 5.88 5.58 5.44 8.14 7.23 19