Grain subsidies and junk food purchases among low-income individuals

by · Ideas for India (I4I) · Join

While governments rely on expensive food subsidy programmes to address malnutrition among low-income communities, their impact is unclear as only self-reported data on food purchase decisions are available. Based on an experiment in Mumbai using data from point-of-sale scanners, this article finds that low-income individuals – especially those living in households with children – who received a wheat and rice subsidy spent less on junk food and more on spices and accompaniments that complement grains in home cooking. 

Malnutrition – including undernutrition and obesity – remains a significant global issue, affecting more than 2.5 billion people worldwide as of 2022. As per the World Health Organization (WHO), India alone accounts for a third of global undernutrition. The second Sustainable Development Goal (SDG) calls for improved nutrition and an end to hunger by 2030 (United Nations Department of Economic and Social Affairs). Nutrition is also a key enabler for achieving other SDGs such as those related to education and economic growth. 

Do food subsidies solve the problem of malnutrition?

For decades, both government and non-government organisations have relied on food subsidies to address persistent malnutrition. These programmes typically involve the distribution of basic foodstuffs, such as cereals and grains, to low-income communities. In India, over 800 million people benefit from regular access to cereals and grains through the country’s Public Distribution System (PDS) programme. This aid is distributed via ‘fair price shops.’ 

The cost of India’s PDS is substantial. In 2021-22, the central government faced direct costs from the programme estimated at US$16.5 billion  (Puri and Pingali 2024) and likely incurred substantial indirect environmental and economic costs, which include fertiliser subsidies, scarce water use, and greenhouse gas emissions (Tata-Cornell Institute for Agriculture and Nutrition, 2022). 

Despite the significant financial burden food subsidies place on governments and non-government bodies globally, there remains a lack of consensus regarding their efficacy. Existing research on the impact of food subsidies has largely relied on self-reporting by recipients using mechanisms such as national nutrition surveys (Jensen and Miller 2008, Banerjee et al. 2018, Shrinivas et al. 2018). Given the low presence of digitised payment and record-keeping systems in low-income communities, these surveys have provided policymakers with insights that might otherwise be unavailable. 

However, self-reporting through surveys may not offer an objective record of food purchasing decisions and consumption habits. The potential for biases, misremembering, and misreporting limits their usefulness as a method for measuring the effectiveness of food subsidies. For example, the difficulty in identifying and recording consumer packaged goods within surveys – particularly if respondents are unable to recall and quantify the volume and nutritional value of goods consumed – reduces our ability to measure the consumption of junk foods, which are often sold as packaged goods. Moreover, the cross-sectional (rather than longitudinal) nature of such surveys means that the effects of a subsidy may be hard to estimate in an experimental study, due to systematic variation across individuals. 

Assessing the impact of food subsidies 

Our study introduces a novel methodology for collecting comprehensive and objective data on the impact of food subsidies, specifically examining how they alter the food basket and nutritional purchases of low-income individuals (Aouad, Ramdas and Sungu 2024). 

We equipped 39 of the 52 food sellers in a low-income settlement in Mankhurd area of Mumbai, with point-of-sale (POS) digital scanners to record all transactions before, during and after a ‘randomised controlled trial’ (RCT).1 The scanners recorded the individual shopper via a study ID, as well as the products purchased, their prices and quantities, and the time of the purchase. Each transaction was also assigned a unique ID, and the nutritional value of items purchased was calculated.2 Full-time research assistants and ‘mystery shoppers’ were used to verify the accuracy of the data collection process. 

In total, 768,074 product transactions by 23,717 customers were recorded from March to October 2022. Of these transactions, 84.05% were tagged with a unique customer identifier, allowing separate transactions by the same individual to be recorded. 

A ‘subsidy store’ similar to the fair price shops that distribute PDS subsidies was also established in Mankhurd. This design removed supply-side uncertainty and allowed the effect of demand-side factors on purchasing decisions to be isolated. We invited a random subset of the shoppers, identified via prior transactions at our partner stores, to participate in our RCT, which took place between 1 July and 4 September 2022. Our inclusion criterion was that the first shopping instance was at least five weeks prior to the start of the experiment, and the last shopping instance was at most two weeks prior. 

A total of 1,255 Mankhurd residents were randomly enrolled into a ‘treatment arm’ and a ‘control arm’. Participants in the ‘treatment group’ received a rice and wheat subsidy on a weekly basis for six weeks; individuals in the ‘control group’ received a delayed subsidy distributed after the experiment’s completion. The purchasing habits of participants in the treatment and control groups were recorded using the installed POS scanners at participating vendors. On average, individuals in our treatment and control groups made 26.135 shopping trips at our partner vendors and visited 3.16 of these stores during the six-week experiment period. 

This methodology provides a new data source that addresses some of the limitations of self-reporting and allows for individual-level changes in purchasing decisions to be measured over an extended period. To the best of our knowledge, this experiment provides the first objective record of food purchases in a low-income community and the impact of subsidies in emerging markets. 

Findings

We found that, in the absence of food subsidies (that is, for the control group), shoppers at the 39 stores where we collected data spent as much on unhealthy, ultra-processed foods (snacks, soft drinks) as they did on wheat and rice combined. Individuals in households with children spent 60% more on snacks than those in households without children, pre-treatment. 

Participants who received the rice and wheat subsidy (the treatment group) significantly decreased their expenditure on soft drinks (by 31%), as well as snacks and sugar (by 16%) during the experiment period, relative to the control group. This effect was most pronounced among treatment group participants in households with children. Spending on spices and accompaniments (complements to the subsidised grains) increased by 27% and 21% respectively, compared to participants who did not receive the food subsidy. The total amount spent across both the control and treatment groups remained stable. The observed increased spending on accompaniments and spices, coupled with decreased expenditure on junk foods, suggests that access to subsidised grains encourages home cooking and reduces reliance on high-glycemic-index carbohydrates. This suggests that the subsidies that the Indian government is distributing are improving the nutrition of their beneficiaries. 

We found no evidence of any negative substitution patterns from access to food subsidies, such as reducing total purchased nutrients or food expenditure. Both the control and treatment groups purchased similar amounts of calories, fat, protein, carbohydrates, and several micronutrients throughout the six-week experiment. That said, the change in the distribution of spending across food categories suggests that treatment participants shifted to healthier, lower-glycemic index sources of nutrition due to the subsidy. These findings continue to hold up in the sub-sample that is already receiving government subsidies. This suggests that the government subsidies already being disbursed in India might have an even larger positive effect on nutrition. 

Policy implications

This study provides policy insights that support ongoing efforts to address malnutrition globally. Our novel methodology offers high-quality longitudinal data on the purchasing habits of individuals in low-income communities and the impact of basic food subsidies on their buying decisions. 

The evidence collected makes a strong case for the continuation and expansion of food subsidy programmes in low-income communities. It confirms that reliable access to wheat and flour subsidies decreases expenditure on ‘empty calorie’ foods and drives greater spending on ingredients that encourage home cooking. Crucially, these subsidies do not negatively impact recipients’ total nutritional purchases. Given the importance of good nutrition in supporting early development, it is also significant that the observed substitution effect is most pronounced in households with children. 

We also present evidence that purchasing decisions are influenced by the type of food subsidy offered. Policymakers aiming to improve nutritional outcomes must consider how certain cereals and grains are likely to be adopted by low-income populations. For instance, introducing a less popular but more nutritious cereal, such as millet, into subsidy programmes may require further investigation to understand its impact on purchasing decisions.

Notes:

  1. An RCT involves randomly assigning individuals into groups – the 'treatment group' which is exposed to the intervention, and the 'control group' which is not. Both groups are statistically similar, and therefore any differences between the treatment group and control group detected after the intervention can be attributed to it.
  2. We constructed a database that contains information on all products sold in our partner stores, including each item's net weight, nutritional components, and category. The nutrition mapping is built using different strategies for loose, packaged, and ‘self-packaged’ food items. More details are provided in Aouad, Ramdas and Sungu (2024).

 Further Reading