AEA RCT Registry

Fields Changed

Registration

Field	Before	After
Abstract	"Nudges" have increasingly shown to be cost-effective tools for promoting a wide range of behaviors, from medication adherence to saving to energy efficiency. But most research evaluates one intervention in isolation on target outcomes. As such, we have little understanding of how campaigns might interact with one another, or whether they generate spillovers in unanticipated domains. This paper explores the hypothesis that such campaigns might interfere with one another due to limited attention. I propose a simple framework, motivated by a taxonomy of attention from the psychology literature that distinguishes between “internal” and “external” attention. I test the predictions of the framework using an experiment in which individuals receive combinations of messages and incentives for two healthy behaviors. I then estimate the model in order to run two important sets of counterfactuals. First, I vary characteristics of the behaviors—the returns they generate and their difficulty (which can potentially be measured with simple survey questions)—to explore interference between interventions that target different behaviors from the ones in the experiment. Second, I look at how interference changes when we move from interventions that involve lots of external stimuli to interventions that involve none.	"Nudges" have increasingly shown to be cost-effective tools for promoting a wide range of behaviors, from medication adherence to saving to energy efficiency. But most research evaluates one intervention in isolation on target outcomes. As such, we have little understanding of how campaigns might interact with one another, or whether they generate spillovers in unanticipated domains. This paper explores the hypothesis that such campaigns might interfere with one another due to limited attention. I propose a simple framework, motivated by a taxonomy of attention from the psychology literature that distinguishes between “internal” and “external” attention. I test the predictions of the framework using an experiment in which individuals receive combinations of messages and incentives for two healthy behaviors.
Trial Start Date	September 05, 2018	January 01, 2018
Trial End Date	March 14, 2019	May 01, 2019
Last Published	December 27, 2018 05:57 AM	January 31, 2019 05:40 AM
Intervention End Date	February 28, 2019	February 23, 2019
Primary Outcomes (Explanation)		See pre-analysis plan.
Experimental Design (Public)	Some participants will receive message or incentive programs for only one behavior, and some will receive messaging programs for both behaviors. The key outcome of interest will be whether or not participants engage in each action at the individual-day level. By looking at spillovers and interactions between interventions, we will be able to distinguish between two types of limited attention -- internal and external -- with important policy implications.	Some participants will receive message or incentive programs for only one behavior, and some will receive messaging programs for both behaviors. The key outcome of interest will be whether or not participants engage in each action at the individual-day level. By looking at spillovers and interactions between interventions, we will be able to distinguish between two types of limited attention, internal and external.
Randomization Method	Randomization is done in office by a computer, re-randomized on six variables (see Pre Analysis Plan)	Randomization is done in office by a computer with re-randomization (see Pre Analysis Plan)
Planned Number of Observations	3000	3780
Intervention (Hidden)		See pre-analysis plan.
Secondary Outcomes (End Points)	Opt-outs, score on quiz about information sent via messages	Expectations, opt-outs, response to a surprise raffle via SMS, score on quiz about information sent via messages, health
Secondary Outcomes (Explanation)	Opt-outs are whether or not participants opted out of the SMS programs at any point. The score on the quiz will be the fraction of questions answered correctly in the endline survey.	See pre-analysis plan.