Published on Development Impact

Fruit Salad, Chocolate Cake, Cognitive Control, and Poverty

This page in:

In a psychology experiment from 15 years ago, participants were asked to remember a number – the number was randomly selected to either be a short two digit number or a seven digit number – and then to walk down a hallway to another room for an interview. As a seeming afterthought, they were told there is a snack cart in the hallway and to help themselves to one of the snacks. The snack choice was either fruit salad or chocolate cake.  The subjects asked to remember the two-digit number selected the fruit salad in equal proportions to the chocolate cake. The subjects tasked with remembering the longer seven digit number overwhelmingly chose the chocolate cake.

“How bizarre” I thought when I first read this study, “but what does this have to do with our understanding of poverty?” The answer, perhaps, is that the cognitive control or willpower necessary to resist the affective impulse to consume the cake is a depletable resource. When attention is focused elsewhere, such as on retaining a long number, there is less of this resource available to guide the decision over snack choice.

An emerging strand of thought in behavioral economics suggests that, akin to remembering the long number, the conditions of poverty exact a heavy toll on cognitive resources through the everyday challenges of scarcity. The repeated trade-offs confronting the poor in daily decision making – i.e. “should I purchase a bit more food or a bit more fertilizer?” – occupy cognitive resources that would instead lay fallow for the wealthy when confronted with the same decision. The rich can afford both a bit more food and a bit more fertilizer, no decision is necessary.

In this framework there is a clear interaction between poverty and decision-making that may lead to decisions against the long-term interest of the decision maker (for a theoretical overview from an economic perspective, see this working paper from Abhijit Banerjee and Sendhil Mullainathan). My impulsive desire may prefer the consumption good in front of me, but my cognitive control can resist that impulse and select the alternative investment good if it hasn’t already been depleted through recent repeated usage. If my control resource has been depleted through earlier use, then the conditions of poverty can induce behavior that in turn prolongs poverty.

I read about the aforementioned chocolate cake experiment in a new working paper by Dean Spears who attempts to measure a causal effect of decisions taken in an environment of scarce resources on subsequent behavioral control.

In Spears’ experiment, conducted in rural villages in southern Rajasthan, selected respondents were tasked with two activities. Study subjects were offered to purchase a well-known brand of soap at a highly discounted price and they were also asked to squeeze a mildly resistant handgrip. The order in which the two activities were given was determined on a random basis. Regarding the handgrip, respondents were asked to squeeze it for as long as possible (up to a maximum 180 seconds). Apparently handgrips are a common way to measure cognitive control in psychological experiments. What determines the duration of the handgrip squeeze, for the majority of people, is not physical endurance but something like mental willpower.

What did Spears find? If the hand grip came before the offer of discounted soap, both poor and rich respondents squeezed the grip for an average of two minutes. If the decision to purchase soap was taken before the hand grip exercise, the rich respondents still held the handgrip for an average of two minutes. The poor, however, gripped for a full 40 seconds less. The economic decision of the soap purchase had a negative effect on subsequent performance, but only when the respondent was poor.

In this sense, poverty is said to “cause” the under-performance. The conclusion is not that only the poor face difficult economic decisions, the rich certainly do as well. But rather because cognitive control is a depletable resource, the higher frequency of difficult economic decisions confronting the poor takes a toll on subsequent decisions.

This study presents corollary evidence from laboratory experiments in India and an analysis of time-use data collected from United States households. Spears’ finds that poorer individuals in the US are much more likely to eat while they shop than individuals with high incomes, but not more likely to eat during other common activities such as leisure, TV watching, work, and household chores. (Although this suggestive result highlights the pitfalls of observational analysis – Spears cannot control for the location of shopping. Perhaps the poor tend to shop in areas where there are more abundant affordable food options and thus face higher levels of temptation.)

As with any work in a burgeoning and promising field, there are caveats. This is one suggestive yet small experiment in one specific population. This is a nascent literature where terms need to be further hashed out, theory further developed, experiments further refined and then replicated. The dynamic interactions between cognitive control and wealth obviously need to be better understood – there is surely heterogeneity in cognitive control endowments across individuals and these endowments are developed starting in early childhood, or even in utero, due to conditions that may in turn be related to wealth. Presumably cognitive control can also be strengthened in adults, but how? And so on.

The to do list is long and the need for further understanding and evidence is great. However if this framework captures a significant force operating in the world, the ramifications for poverty policy strike me as … possibly … profound.


P.S. I was first made aware of Dean Spears' study from this nice TNR article which is worth reading in its entirety.


Jed Friedman

Lead Economist, Development Research Group, World Bank

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000