Does Market Access Mitigate the Impact of Seasonality on Child Growth? Panel Data Evidence from Northern Ethiopia

Abstract

Seasonality in agricultural production continues to shape intra-annual food availability in low-income countries. Using high-frequency panel data from northern Ethiopia, this study attempts to quantify seasonal fluctuations in children’s weights. Consistent with earlier studies, we document considerable seasonality in children’s age and height adjusted weights. While children located closer to local food markets are better nourished compared to their counterparts residing farther away, their weights are also subject to considerable seasonality. Further analysis shows that children located closer to food markets consume more diverse diets than those located farther away but the content of the diet varies across seasons. This leads us to conclude that households located near these food markets are not able to insulate their children from seasonal weight fluctuations. We discuss some policy options with potential to address this threat to child wellbeing.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Horton and Steckel (2013) use a database of mean adult heights to estimate the economic losses from malnutrition. The central assumption is that adult height is a relatively good indicator of nutritional status in early childhood. The authors first collect estimates from a large body of literature that studies the effect of increased adult height on wages. These estimates are then used, together with the database on mean adult heights, to derive a height loss function as a per cent of Gross National Product.

2. The seasonality literature in the development studies field is too large to be fully covered here. A good starting point is the seminal book by Sahn (1989). In the Ethiopian context, Dercon and Krishnan (2000b) use a panel survey with six-month intervals to show how household consumption and poverty rates fluctuate across agricultural seasons.

3. A recent study by Darrouzet-Nardi and Masters (2015) focused on the impact of season of birth on HAZ scores measured later in life. Using two rounds of DHS data for the Democratic Republic of Congo, the authors find that children born in the lean season have systematically lower HAZ scores than other children, and that being located farther away from towns and cities exacerbates this adverse seasonality effect. The focus in the current paper is different. Using high-frequency panel data over a two-year period, we study how seasonality affects children’s growth patterns (after their birth) and whether closer proximity to markets mitigates these effects.

4. The Z-scores measure the distance to the median weight of a healthy and well-nourished reference population of same sex and age/height. This distance is measured in terms of standard deviations of that same reference population. We used the WHO (2006) as the reference population and computed the z-scores using zanthro06 command (Leroy, 2011) in Stata 14. Z-score observations that were below −5 or above +5 standard deviations were considered outliers and dropped from the analysis.

5. For Ethiopia, Gilbert el al. (2015) use monthly price data for 2003–2012 from 11 wholesale markets.

6. For example, the seasonal price gap in Ethiopia for oranges was 21 per cent and for tomatoes more than 35 per cent. In the case of cereals, the seasonal price gap for maize was estimated to be 20 per cent, for teff 10 per cent, and for sorghum nearly 14 per cent. For more information, see Gilbert et al. (2015).

7. A child is categorised stunted if his or her height-for-age is below −2 standard deviations, underweight if his or her weight-for-age is below −2 standard deviations and wasted if his or her weight-for-height is below −2 standard deviations.

8. This refers to households where the head is less than 18 years of age.

9. For more information about the survey and the impact evaluation, see Berhane et al. (2015).

10. 1280 households were interviewed in Abi Adi woreda.

11. As a result, their production is less affected by seasonality in rainfall patterns.

12. The Euclidian distances are based on GPS coordinates that were recorded for each household and each food market. We calculated the distance to the nearest market and then took the median distance for each kushet to minimise the role of measurement error. However, the results are robust to using distances defined at the household level as well (results available upon request).

13. These distances are not atypical in the Ethiopian context. For example, our calculations using the nationally and regionally representative Ethiopia Socioeconomic Survey (ESS) 2013/2014 survey data (CSA & the World Bank, 2015), the mean (median) distances to the nearest weekly market in rural areas in the main regions are the following: Tigray: 15.4 km (15.0 km); Amhara: 11.1 km (8.0 km); Oromia 9.9 km (6.0 km); SNNPR 9.5 (6.0).

14. Most areas of the country have similar bi-modal distribution of rainfall (Tadesse et al., 2006). At the national level, meher (kiremt) rains are more important with 90 per cent of the total crop production taking place during this season (Taffesse, Dorosh, & Gemessa, 2012). There is little regional variation in this regard. For example, in 2010/2011 season, 99 per cent of the total annual grain output in the Tigray region was produced during the meher season while the corresponding figure for SNNP is 85 per cent (CSA, 2011a, 2011b).

15. The rains during the short rainy season (belg) are small in magnitude and not reliable and therefore only little cultivation takes place during this period.

16. Of note is that Hintalo Wajirat is part of Ethiopia’s Productive Safety Net Programme (PSNP), a large government and donor managed programme aiming to improve food security in the country. About 52 per cent of all households in our sample are PSNP beneficiaries. All regression results are robust to the inclusion of a dummy that captures these PSNP beneficiary households. Results available upon request.

17. We use a spline function with knots at 6, 12 and 24 months to model the WHZ/WAZ – age relationships reported in Figures 4 and 5. Suits, Mason, and Chan (1978) offer an accessible introduction to the use of spline functions in econometrics.

18. Earlier literature documents considerable month-of-birth effects on children’s anthropometric outcomes (Lokshin & Radyakin, 2012), implying that changes in food availability affects intra-uterine growth trajectories. The results reported here are not driven by month-of-birth effects. Including month-of-birth dummies to the specification does not affect the main coefficients of interests. Interestingly, we find that the month-of-birth dummies are jointly statistically significant in the case of WAZ but not WHZ. This suggests that month-of-birth affects children’s WAZ scores, over and above the contemporaneous seasonal effects.

19. Our results are robust to using different cut-off points (for example 3, 4.5, 5 and 5.5 km). These results are available upon request.

20. These seasonal weight fluctuations are of similar magnitude that was found earlier in south-central parts of Ethiopia by Ferro-Luzzi et al. (2001). However, it is worth noting that the z-score values are not directly comparable to the ones reported in the current study as the authors of the earlier study used a different reference population to calculate the weight-for-height and weight-for-age z-scores.

21. For example, the difference in mean WHZ between children whose mother has no education and children whose mother has completed primary school is 0.2 units of standard deviation (Central Statistical Agency, & ICF International, 2012). The same difference holds for WAZ.

22. These food items are: Injera; Other foods made with grains; Roots and tubers; Orange coloured vegetables; Leafy dark green vegetables; Other vegetables; Fruit; Meat; Eggs; Fresh, canned or dried fish or other seafood; Legumes; Dairy product; Fats and oils; Sugar, honey, sweets; and Coffee, tea, soft drinks.

23. This type of simple dietary diversity score has been shown to be correlated with children’s anthropometric measures in different contexts (Arimond & Ruel, 2004; Jones et al., 2014), including Ethiopia (Disha, Rawat, Subandoro, & Menon, 2012).

24. The specification here follows Equation (2). Using a specification that is based on Equation (1) yields similar conclusions.

25. In the pooled sample, 26 per cent of the children consumed animal source foods and 11.9 per cent consumed dairy products in the past seven days.

26. Note that there is no (seasonal) malaria risk in this area due to the high altitude.

Additional information

Funding

This work was supported by the United States Agency for International Development (Feed-the-Future Initiative [Ethiopia]).