Sunday, June 4, 2017

Cash transfers in scale

Sceptics of cash transfers argue that the beneficiaries would fritter away the cash on temptation goods like alcohol and entertainment. But there is fairly strong evidence refuting this hypothesis. People, after all, behave as rational individuals and use the cash for their felt needs. 

But these are findings from small sample field experiments. In contrast, the findings of a real world scale cash transfer program is not encouraging. The Times points to a report by the Department of Agriculture which finds that the largest purchases by the beneficiaries of Supplemental Nutrition Assistance Program (SNAP), the $ 74 bn food stamp program that covers 43 million poorest Americans, are soft drinks, which accounted for about 10% of the dollars spent on food. The report concludes that a disproportionate amount of money was going to unhealthy foods,
Over all, the report found, SNAP households spent about 40 cents of every dollar at the grocery store on “basic items” like meat, fruits, vegetables, milk, eggs and bread. Another 40 cents of every dollar was spent on “cereal, prepared foods, dairy products, rice and beans.” Lastly, 20 cents of each dollar was spent on a broad category of junk foods that included “sweetened beverages, desserts, salty snacks, candy and sugar.” SNAP households spent 9.3 percent of their grocery budgets on soft drinks alone. That was slightly higher than the 7.1 percent figure for households that do not receive food stamps.
Now, it is possible that SNAP households would have, even without cash transfers, consumed more junk food. But that does not take away from the fact that SNAP households actually spend more on junk food than their non-SNAP counterparts. And it does pose worrying questions for those advocating the replacement of an in-kind food/nutrition security program with cash transfers. 

This is another reminder about the need for caution in interpreting the findings of field experiments.  It is likely that the scale dynamics associated with such transformations work in an unpredictable manner. In particular, the general equilibrium effects from a scale implementation can be very different from the partial equilibrium of a pilot.


Anonymous said...

Could it be that welfare recipients are exercising their individual autonomy by purchasing junk food, even if they know it is against the rules and not the best use of benefits they receive? It could just be that cash transfers may do good only if policymakers viewed as a means of transferring benefits and recognized that ultimately it is up to recipients. How those benefits should be used should be up to the individual. Micromanagement is only going to dehumanize welfare recipients and in a just society, lead to electoral blow back.

Karthik Dinne said...


1. The temptation goods studied in field experiments are only "tobacco" and "alcohol". Not junk food.

2. Several field experiments document consumption/spending on "sugar" or "tasty" food. This is sort of well established.

Poor Economics book has a separate section on "Why do the poorest people in the Indian state of Maharashtra spend 7 percent of their food budget on sugar?"

3. Percentage comparison is misleading here because of the huge difference in the absolute budgets of the rich and poor. In absolute terms, the per capita junk food consumption turns out to be much higher for richer households.

Even considering %, the difference mentioned in Times report is miniscule (2.2%). Since the food stamps are $256 per household per month, this turns out to be an extra spending of $5 (2.2% of $256) per month per household. Per capita, this is an extra spending of $1 per person per month.

Extra spending of $1 per person per month shouldn't matter much I believe.

Gulzar Natarajan said...

Anon, your questions go to the heart of debates on libertarian paternalism. My personal view is neither the micro-management nor the full free-choice, but something in the middle, though more towards the latter side.

Karthik, yes I know about the defn of temptation goods there, but that is becos there is no junk food comparable. More than sugar, tea can be the closer comparator (actually most sugar is used with tea!).

But your observations on the scale of the effect is interesting. Because that is a point I have made in multiple posts, but in the opposite direction - to downplay the importance of several innovations that have been RCT tested. If you take a very large share of innovations, from incomes to savings to weight gains to DALYs to job hiring to learning outcomes, you will find that the statistically significant improvement is on a very small baseline. In other words the 35% increase on a baseline of 8%! Papers get published in peer reviewed journals without even a cautionary line somewhere about the base effect... So it is interesting that you raise it now:)

On a serious note, it will be useful for someone to do a metastudy on such base effect distortions.