Does farm level diversification improve household dietary diversity? Evidence from Rural India
CHATTERJEE, TIRTHA (2016). 'Does farm level diversification improve household dietary diversity? Evidence from Rural India' Paper presented at the annual conference of the HDCA, Tokyo 2016.
Malnutrition is recognized as a major issue among low-income households in developing countries (Black et al., 2008; FAO, 2010 and 2012; WFP, 2012; and IFAD, 2012). An emerging line of research has been to tackle this problem through the channel of agriculture. Among the different pathways (identified by Gillespie and Kadiyala; 2012, Hawkes and Ruel, 2008 among many others) through which agriculture and nutrition are interlinked, one of the most direct ones is as source of food and diversity in food basket. To empirically understand the link between agriculture and nutrition, in this paper we exploit the relationship between farm production diversity and dietary diversity at the household level in rural India. India makes for an interesting case study to explore this relationship because of (1) a significant rural population (approximately 68%) (2) very poor nutrition status and (3) agriculture still remains as one of the most dominating sectors in terms of livelihood generation in the economy. The objective of this paper is to use a propensity score matching technique to establish a causal relationship between the two and identify if households which are exclusively cereal growers have a different dietary pattern compared to households with more diverse production basket. Findings from this empirical exercise will substantiate evidence in favour of positive impacts of agricultural farm level diversification not only on income, reduction of poverty but improvement of nutritional status as well.
We use data from a nationally representative survey of farm households conducted by the National Sample Survey Organization (NSSO) of the Government of India in 2003 (GoI, 2005). To estimate the impact of farm level diversification on dietary diversity, we use a propensity score matching (PSM) methodology wherein our outcome variable is the log of count of number of food groups consumed by the household in the last 30 days. The food groups used for calculating the count measure are: cereal, pulses, gram, milk and milk related products, oil and oil related products eggs, fish, meat, vegetables, fruits, dry fruits, sugar, salt, spices, beverages, cooked meal and processed food. Our treatment variable is a binary indicator variable and equals ‘1’ if households grow only cereals (treatment group) and ‘0’ if households grow cereals and other crop groups (control group). PSM framework will help us identify if given all the observable differences among treatment and control group if dietary basket of a household would be different had he not been given the treatment (i.e. counterfactual). Our hypothesis is that on an average household who are exclusively cereal producers will consumer a less diversified diet compared to those who produce both cereals and other high value crops.
First, the propensity score is estimated using a logistic regression model, in which treatment status is regressed on observed characteristics of the farmer household. The estimated propensity score is therefore the predicted probability of a household producing only cereals given its characteristics. Second, a matching method is selected to be used to match treatment and control group. A number of matching methods have been suggested in the literature. Three matching methods have been used in this study for robustness: (1) nearest neighbour or one to one matching; (2) caliper matching; and (3) the kernel-based (KM) matching. The propensity score distribution for the treated and untreated groups and region of common support indicate that there is overlap in the distributions of the propensity score. We find that after matching difference in means for each of the covariates are not statistically different from each other between the two groups. Further, the relatively low pseudo- R square and the p-values of the likelihood ratio test of joint significance of regressors confirm that, after matching, there are no systematic differences in the distribution of covariates between the two groups. Our results show that indeed after controlling for the differences in observed characteristics of treatment and control group, farmer households who are exclusively cereal producers have a statistically significant less diversified diet compared to farmer households who produce both cereals and high value crops. Robustness checks corroborate our findings. With the evidence presented in this paper, farm diversification can be seen to be a potential strategy to improve the dietary behaviour and therefore the nutrition status of households.
 Latest Situation Assessment Survey was done in 2013-14, 70th round. However, it does not provide detailed breakup of food consumption expenditure of the household.
 Source: http://data.worldbank.org/indicator/SP.RUR.TOTL.ZS