Looking Past the Indian Calorie Debate: What is Happening to Nutrition Transition in India

Abstract

We utilise large national household datasets for 1993–1994, 2004–2005 and 2011–2012 to analyse factors influencing changing patterns in per capita calorie consumption in India. Our study findings demonstrate the significance of the disease environment in which people live, with those living in healthy areas having lower calorie consumption than those living in less healthy ones. Calorie intake has been falling in India, but the study findings reveal that fat calorie intake has been rising successively over time among the rural and poorer urban sub-populations raising concerns for policy-makers that non-communicable diseases are expected to rise for these vulnerable population groups.

Acknowledgements

The authors are able to make data and code available to bona fide researchers upon request. Please email Dr Md Zakaria Siddiqui at [email protected]

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The national household survey data reports use total household monthly expenditure as a proxy measure for household income. We use the same proxy measure when referring to household income and quintile income groups throughout the study.

2. The FDI is calculated as the sum of squares of the shares of the various food items represented in the consumption basket (Gaiha et al., 2013). We use nine food groups to construct our FDI measure: cereals and cereal substitutes; roots and tubers; sugar and honey; pulses, nuts and oilseeds; vegetables and fruits; meat, eggs and fish; milk and milk products; oils and fats; miscellaneous food products and beverages.

3. Although we also ran models at the district level, our analysis is at the state level. Much of health policy decision-making is made at the state level and given that geographical variations are more readily discernible between 17 major states than between 540 districts; state level analysis provides a more meaningful approach to present this. We discuss district issues again later (see also Note five below).

4. Although State MPCE is derived from household-level MPCE, we aggregate 47,249/32,523 rural/urban households (for 2011–2012) into 17 major states for the purpose of analysing state level differences. As expected, with this level of aggregation, model results revealed no specification issues. Significantly, our results reported later reveal that State MPCE and household MPCE exert the opposite effect on calorie intake. Whilst it is acknowledged that State MPCE is not an ideal proxy measure to capture state-level human development, we do test alternative approaches to measure the same derived from other data sources (for example, state GDP/capita) – which we also discuss later (see Note six below).

5. We also ran model(s) incorporating the disease environment covariate at the district level (rather than state-level). Whilst district level model results revealed a strong negative relationship with respect to calorie consumption the overall explanatory power (R²) was lower than the state level model. We present the district level results in the Supplementary Materials (Table S2) where we compare this with the state-level model.

6. We also used alternative measures derived from other data sources for human development including state GDP/capita, government health expenditure/capita, and healthcare workers/population. Whilst all yielded negative state-level coefficients with varying degrees of statistical significance, the explanatory power (R²) and the intra class correlation (ICC) performance for all these alternative models were less than the model reported in Table 1 using state MPCE. For reference, we present model results for state GDP/capita and government health expenditure/capita in the Supplementary Materials (Table S3).

7. For presentation purposes, we use the same model specifications of household covariates and contextual factors, but use per capita calorie consumption as the dependent variable.