More about Dupas and Robinson on the Psychology of Saving Behavior

My last post (#42) provided a layman’s summary for laymen of the paper by Pascaline Dupas and Jonathan Robinson (D & R) on “Why Don’t the Poor Save More? Evidence from Health Savings Experiments” in Kenya.  My Freedom from Hunger colleagues found this very helpful for their own work; they had some comments I share more broadly here.

First, they found it hard to keep the four saving devices straight in their heads regarding which was what and how these related to the results, so I cooked up this table to summarize more compactly.  The top six rows show the features of the four devices and the control.  The bottom three rows show the results in very general terms.

Second, D & R use the terms “mental accounting,” “earmarking” and “commitment” in confusing, overlapping ways.  I have tried to sort out their meanings in the labels I provide for the device features in the table above.  In particular, they use “earmarking” to label what many of us think of as “commitment,” as in a “commitment savings account” — that is, the earmarking seems to lock up the saved money for a specified purpose, and the unlocking is controlled by others to enforce the intended use of the saved money (note that this earmarking results in “low liquidity”).

D & R seem to use the term “commitment” to refer to a mutual agreement among savers in the same ROSCA to save a certain amount for a specific purpose (like purchase of a mosquito net) at the time and place of the ROSCA meetings.  Actually, D & R use “social commitment” to label this closely monitored mutual obligation among ROSCA members.

Each of these features could be called “mental accounting,” but I think D & R use this term to label a personal decision to set aside money for a certain purpose when there is no obligation to anyone else to actually honor this decision.

Clearly, standardized use of these terms is a thing of the future.  Meanwhile, we have to read carefully!

Third, one colleague observed that earmarking required people to be specific about their future health needs and to decide if these are important enough to commit to save for them. If so, it would be quite rational to avoid commitment to save for particular health products when health needs can be quite varied and unpredictable — not like saving for school fees or even a bag of fertilizer, which are much more predictable with respect to amount and timing.  Such commitment to save for unknown future health needs would be particularly challenging for the “present-biased” person.

Fourth, another colleague noted the study raises important operational questions for practitioners. For example:

  • Will we find similar positive results among populations that are not particpating in ROSCAs?
  • How can the Health Pot and HSAs be added to other types of groups besides ROSCAs?
  • How much training do groups require to understand and self-manage the technologies? What problems do they face?
  • How can the technologies be adapted for low-literacy groups?
  • What is the cost-benefit of each technology?
  • What are the longer-term health impacts of these technologies? Better health? Better coping with health shocks?  Improved consumption smoothing?

Fifth, the first colleague wonders whether greater focus on health savings is likely to increase all savings.  D & R suggest this may be the case; the health saving devices dependent on ROSCA management (Health Pot and HSA) increase total saving in ROSCAs, but this may displace some saving in other mechanisms.  Their data also suggest that the autonomy made possible by having an individual saving device, like the Safe Box, may undermine use of ROSCAs.  So far this is a muddy picture at best.

Clearly, the D & R research has revealed some very promising design options and more general implications for practitioners promoting saving by the poor.  If you have other observations or questions you would like to add to these above, please send me a comment.  I will add it to this post.