Agricultural Production, Dietary Diversity and Climate Variability

Abstract

Nonseparable household modelsoutline the interlinkage between agricultural production and household consumption, yet empirical extensions to investigate the effect of production on dietary diversity and diet composition are limited. While a significant literature has investigated the calorie-income elasticity abstracting from production, this paper provides an empirical application of the nonseparable household model linking the effect of exogenous variation in planting season production decisions via climate variability on household dietary diversity. Using degree days, rainfall and agricultural capital stocks as instruments, the effect of production on household dietary diversity at harvest is estimated. The empirical specifications estimate production effects on dietary diversity using both agricultural revenue and crop production diversity. Significant effects of both agricultural revenue and crop production diversity on dietary diversity are estimated. The dietary diversity-production elasticities imply that a 10 per cent increase in agricultural revenue or crop diversity result in a 1.8 per cent or 2.4 per cent increase in dietary diversity respectively. These results illustrate that agricultural income growth or increased crop diversity may not be sufficient to ensure improved dietary diversity. Increases in agricultural revenue do change diet composition. Estimates of the effect of agricultural income on share of calories by food groups indicate relatively large changes in diet composition. On average, a 10 per cent increase in agricultural revenue makes households 7.2 per cent more likely to consume vegetables, 3.5 per cent more likely to consume fish, and increases the share of tubers consumed by 5.2 per cent.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Acknowledgements

We appreciate the helpful comments of Gero Carletto, Alan de Brauw, John Maluccio, Paul Winters, an anonymous referee, and the participants of the Farm Household Production and Nutrition conference at the World Bank. The management of the Nigeria LSMS-ISA data collection process in Nigeria benefited from the collaboration of Gero Carletto, Kinnon Scott, the Nigerian Bureau of Statistics staff, and Colin Williams. Siobhan Murray provided invaluable support to organise the GIS and climate data. The data and dofiles for this paper are available upon request to the authors. All errors are our own.

Notes

1. Available at http://ndb.nal.usda.gov/index.html

2. There have been a few studies that have focused on effect of climate variability on agriculture in Nigeria, although most are state or region specific rather than nationally representative. Adamgbe and Ujoh (2012) examine the patterns and trends of the variations in the climatic parameters and the implications of such variations on efficient yield rates of some food crops in Benue state using data on climatic variables (rainfall, temperature, sunshine). Among the seven climatic parameters used in their study, sunshine and rain days have the highest influence on the yield of all the seven crops while dates of onset and duration have the least influence. Adejuwon (2005) examines the impact of climate variability on the yield of the major crops (cowpeas, groundnut, millet, maize, sorghum and rice) cultivated in the Nigerian Arid Zone, using Bornu and Yobe states as case studies. The author found that among the more powerful determinants of crop yield were rainfall at the onset and at the cessation months of the growing season and during the long periods with normal and above normal rainfall, crop yield sensitivity tends to be weak. However, Adejuwon (2005) found that during the years with unusually low precipitation, crop yield sensitivity becomes more pronounced. Ayinde and Muchie (2011) examine the effect of variability in rainfall and temperature on agricultural productivity in Nigeria and find strong effects of variability in rainfall while temperature appears not be as important for agricultural production in Nigeria. Temperature change was revealed to exert negative effect while rainfall change exerts positive effect on agricultural productivity but found that previous year rainfall was negatively significant in affecting current year agricultural productivity.

3. The data can be found at: http://power.larc.nasa.gov.

4. We include agricultural revenue as opposed to agricultural profit due to limitations in estimating the shadow value of household labour allocated to agricultural production. Revenue is directly observable in our data set while profit would have to be imputed.

5. In our data across season, we do not see large changes in agricultural capital stocks over time and this stylised fact is commonly indicated as a major determinant of yield gaps and low productivity in African agriculture.

6. Note that farmers could have farmed more than one crop within each dietary diversity group.

7. The specification in Table 7 was also estimated with several other sets of instruments, including a set of instruments that omitted the agricultural capital variable. This specification produced similar elasticities in sign and magnitude, but did not pass the IV tests. For this reason, the set of instruments including agricultural capital was chosen as the preferred specification. Agricultural capital does not vary within season in our sample and is unlikely to be correlated with consumption as low levels of capital are reported in most agricultural households.

8. Two alternative specifications were estimated for production diversity. The first uses the number of distinct crops grown by the household to measure production diversity instead of the number of crop groups. This yielded a very similar crop count-dietary diversity elasticity of 2.1 per cent, significant at the 10 per cent level. In the second alternative specification, the logs are dropped from dietary diversity and production diversity. The IV estimate of effect of the number of crop groups grown on dietary diversity was 1.05 (significant at the 10% level). This suggests producing an additional crop (food) group results in consumption of an additional food group. The results from both alternative specifications are available upon request.

9. Robustness checks of our results are presented in Tables A1 and A2 in the online appendix. In the first, tree crops are excluded since they are less subject to seasonal variation. Table 10 shows the results of the specifications in Tables 6 and 7 with the exclusion of tree crops. In the agricultural revenue-dietary diversity relationship, we find a slightly stronger effect on dietary diversity compared to our main results in Table 6. A 10 per cent increase in agricultural revenue increases dietary diversity by 2.1 per cent. In the production diversity-dietary diversity relationship, we found a similar 2.2 per cent increase in dietary diversity associated with a 10 per cent increase in production diversity. However this effect was not precisely estimated. The value of household durable assets is included as additional variable in the agricultural revenue-dietary diversity specification and the results are presented in Table A2. We find the revenue-nutrition diversity elasticity to be of the same magnitude (1.7%) as the result in Table 6.