Infant malnutrition, clean-water access and government interventions in India: a machine learning approach towards causal inference

ABSTRACT

We use a debiased Machine Learning technique to explore causes behind infant malnutrition for households below-poverty-line in India and examine effectiveness of various government interventions along with other factors. Our analysis reveals that access to clean water is one of the most crucial issues to focus on.

KEYWORDS:

• Infant malnutrition
• clean water
• debiased machine learning
• government policies

JEL CLASSIFICATION:

• C29
• O10

Acknowledgements

The authors are thankful to seminar participants at Western Michigan University, Reserve Bank of India, National Council of Applied Economic Research, and Brookings Institution India Center for their useful comments.

Notes

¹ They are states of Bihar, Jharkhand, Uttar Pradesh, Uttarakhand, Madhya Pradesh, Chhattisgarh, Rajasthan, Orissa and Assam.

² We also observe this using OLS in our data, as expected, results not reported for brevity.

³ Since inference (not predictive modelling) is the goal, cross-validation is not used for choosing ${\lambda }_0$ following their paper.

⁴ Their method is applicable when p is large, including when it is larger than n. In case of joint significance tests, one needs to be careful about the dimension of p (low dimension), especially when p is larger than n, none of which is our concern.

⁵ Very low values of WHZ in our data can possibly indicate ‘acute’ as opposed to ‘chronic’ malnutrition. Chronic malnutrition is indicated by other measures and is not under the purview of the paper. We are thankful to an anonymous referee for bringing our attention to this issue.

⁶ We are extremely thankful to an anonymous referee for bringing our attention to this important aspect for discussion.

⁷ See https://www.who.int/childgrowth/training/module_c_interpreting_indicators.pdf?ua=1

⁸ See http://rchiips.org/nfhs/NFHS-4Reports/India.pdf

⁹ Note the other two EAG samples have very small percentage of families in the data that further purifies water, possibly causing insignificance of this dummy.