bizmentor

View Original

Marshmallow Tests: Promise and Limits

by Gautam Brahma

In the ‘60’s Walter Mischel of Stanford University conducted a series of interesting experiments to study impulse control among children. He offered sweets to groups of 4 to 5 year olds with the condition that they could either have one sweet right away or get two sweets if they waited for a while. He found that some kids could not wait at all while some others could wait for as long as 20 minutes. The mean waiting time was 7-8 minutes. Mischel followed up these children regularly for four decades and found that those who exhibited greater impulse control scored better at school and college, stayed out of trouble with the law and got better paying jobs. These experiments are usually referred to as the Marshmallow tests after the eponymous sweet, though Mischel and later experimenters used a variety of sweets including chocolates.

 

The Marshmallow tests have been discussed threadbare, especially because other studies suggested that the exhibited impulse control was largely learnt and not entirely genetic in origin. This meant that children could be taught to control impulses and thereby assure a better life for themselves in the future. People have also speculated on policy implications of these tests assuming that the inferences hold for adults. For instance, one interesting experiment reported in 2013 actually extends the findings of the Marshmallow tests by suggesting that waiting by choice makes the reward even more desirable. Researchers from the Booth business school in Chicago University and from the Chinese University Hong Kong, asked volunteers to choose between small rewards that they could have early and large rewards which they could get if they get later if they waited. They found to their surprise that the percentage of people prepared to wait for a larger reward actually went up as the wait duration was increased. Not only that, the percentage choosing the larger and later option went up even further if the options were presented today but the respondents were given some days to mull over their final choice! The managerial implications for designing deferred payment schemes and financial savings instruments are obvious.

 

Managerial interest in the Marshmallow tests also stems from the finding that current performance (on the job or in the classroom) can be predicted by seemingly unrelated behaviour in the past (choosing to eat one sweet now or wait a while and eat two). One cannot go back in time and create tests for any personnel one is evaluating but one can dig into past records for other evidence of peripherally related but desirable behaviour. Moreover, with the availability of convenient data analysis tools one could look at multiple surrogate variables to improve the reliability of these evaluations. A recent piece in the New York Times lists many examples of this approach to predicting future individual behaviour . Financial services startups Upstart and Zest Finance are awarding loans for people without decent credit history by  processing variables like nature of school attended, marks obtained, and surrender of pre-paid phone connections. Workday offers HRM software for enterprises that processes 45 variables to predict likelihood of an employee quitting the company. Jure Leskovec of Stanford has used a similar data-driven approach to evaluate bail applications and found that this approach is 30% better than the one currently in place where a judge takes a look at the applicant and decides whether or not he or she is a risk.

 

Managers can not only leverage evidence of past, seemingly unrelated, behaviour to evaluate staff and customers, they can also design processes and incentives to improve impulse control in these people. This is another way of improving chances of future success. Rewarding people for long courses of study and well-crafted deferred payment schemes are just two examples. The larger field of behavioural economics subsumes such studies and one oft-quoted reference for examples of how even national policies can be driven using this approach is ‘Nudge’ by Richard H. Thaler of the University of Chicago and Cass R. Sunstein of the Harvard Law School. This vast field deserves a separate discussion.


Like every other prescription in Management, the desirability of impulse control is also context-specific. Detailed studies focussing on differences across socioeconomic groups by Gene H. Brody of the University of Georgia have shown that while the better-off benefit unconditionally by adopting impulse control the results are more nuanced in the case of the less privileged. Poorer people display the same long-term benefits at school and in jobs but also display worrying symptoms like high blood pressure, obesity and higher levels of stress hormones. Impulse-control is apparently a mixed blessing for the poor. What is even more amazing -and disturbing- is that higher impulse control in the poor was linked to faster cell ageing suggesting some sort of permanent damage. Why this happens and how how this can be offset will undoubtedly be studied by researchers, but till then this remains a cautionary input for managers interested in the Marshmallow tests and related insights.

-------

References:

1.     “When Waiting to Choose Increases Patience”; Xianchi Dai & Ayelet Fishbach; OB and Human Decision Processes; Vol 121 issue 2; July 2013

2.     “Using Algorithms to Judge Character”; Quentin Hardy; NYT;  July 26, 2015

3.     “Self Control Forecasts Better Psychosocial Outcomes but Faster Epigenetic Ageing in Low-SES Youth”; Gene H. Brody et al; Proceedings of the National Academy of Sciences of the USA; June 2015