Syndicate content

Two-stage offers and deception: clarification from Beaman and Magruder

David McKenzie's picture

Last week I posted about a nice experiment that Lori Beaman and Jeremy Magruder had done to understand the role networks play in job-referrals. I used this discussion to bring up the issue that several recent experiments have used two-stage offers, in which individuals are offered one thing (a price, interest rate, or in their case payment schedule) to see how this determines selection into something (buying a good, taking a loan, or in their case, the type of worker they bring along), and then once they have taken this first action, the variation in the first stage is effectively negated in a second step (in their case by then paying everyone the same flat wage). Berk followed up with more discussion on the ethics of deception in experiments.

Lori and Jeremy wrote to clarify the procedure used in their study, which includes some important steps they took to ensure that deception did not take place in their case – as well as setting the record straight, these seem like useful guidelines for others considering this sort of two-stage offer to consider:

“As a result, even though multiple IRBs across the country have determined that not revealing that you intend to give subjects extra payments constitutes incomplete disclosure rather than deception (in both our study and the others cited by David), we took additional lengths to ensure that no deception took place.  In particular, in the consent process, we informed subjects that we might pay them more than we had indicated - that is, we had all of our subjects explicitly consent to the possibility that we would give them more income.  Then, upon announcing contracts, all payments were phrased in terms of "at least," so that all payments made were consistent with both verbal consent and with the specific language of the contract.  There was no deception here; our subjects had made informed consent.  Unsurprisingly, the possibility of additional unexpected pay did not cause any subjects to withdraw from the study.  It's understandable that you were unaware of this: given that the profession (and many of the country's IRBs) have determined that this sort of payment scheme is not deception, we did not use up valuable text space with a more lengthy discussion of the consent process.  We also note that Berk's suggestion that actually providing performance pay would solve these concerns was incorrect - if respondents are splitting their finders fee with referrals than referrals in performance pay treatments would still have have resulted in increased incentives for referrals in performance pay treatments regardless of the referral's own contracting terms.”


Submitted by Berk on
Hi Jeremy and Lori, Thanks for the clarification, which is very useful. I actually think that the wording here was important not just for ethics purposes but also for the readers to be able to exactly interpret the experimental findings. On my suggestion, you're right that it does not solve the problem, it just makes it less of a concern - which is what I said in my post. It was suggested as the kind of compromise that researchers and IRBs may come to in order to avoid deception. Since you had already covered this nicely with full informed consent (very similar to the mystery client/placebo discussion I had), you did not need such a compromise and could get identification more clearly. Thanks again for the clarification. Berk.