Measurement and Survey Response

How to cite

Duffy, John. 2019. "Measurement and Survey Response." AEA RCT Registry. AEARCTR-0004122. https://doi.org/10.1257/rct.4122-1.3

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

John Duffy*

* Primary investigator

Registry ID
AEARCTR-0004122
DOI
https://doi.org/10.1257/rct.4122-1.3
First registered
Last updated
Status
Completed
Trial dates
2019-04-03 – 2020-06-30
Countries
United States of America ()
JEL codes
C80, C90
Registry record
http://www.socialscienceregistry.org/trials/4122

Abstract

This project explores the impact of measurement on survey responses. We hypothesize that the very act of measuring survey responses can affect the distribution of those responses. We design and implement an experiment that mimics the famous double slit experiment of Thomas Young (1801) demonstrating wave-particle duality for light depending on whether its passage through double slits is subject to detection of which slit the light particle went through. In our experiment, subjects are asked to respond to a survey question by providing a numerical response on a given scale. In the control treatment, their numerical answer is recorded and reported back to participants. This is analogous to the physics experiment where a detector identifies the slit through which a particle went through. Alternatively, in a second treatment, we measure whether subjects' response "passes," through one of the slits but without eliciting information to identify which slit. In either treatment, subjects who "pass" through are then asked to consider the same survey question again and to provide a second numerical answer which is again recorded and reported back to subjects. More precisely, the main treatment modifies the control treatment in one important respect. Subjects are asked to “think” about their numerical response to the same survey question posed in the control treatment and to hold that number in memory. Their response to the survey question is NOT directly measured/observed. Instead, they are then asked whether their numerical answer lies within the two subsets of answers used in the control treatment (the two slits). If it is, then, as in the control treatment, they will be asked to evaluate the survey question once again, and in this second instance, their answers to the question will be recorded and reported back to them. We hypothesize that, like light (and other subatomic particles), survey responses by subjects in our main treatment, where survey questions are not directly measured tin the first round measurement, will display a wave-like interference pattern in the second round measurement. We also hypothesize that such an interference pattern will not be observed in our control treatment, where survey answers are directly measured to determine which slit the subjects responses pass through.

Intervention

We will randomly assign subject participants to the control or to the treatment group.

Experimental design

In a control treatment, subjects will be asked to respond to a survey question using a numerical scale. Their answer will be measured/observed. Depending on the response that they give (whether they pass through the slit) they will or will not be asked to answer the same question again. If they are asked to respond to the same survey question a second time, their answer to that question will also be measured/observed. if not, there is no further action for them to take. After answering the second question, if asked to do so, they proceed to answer several survey questions regarding their age, gender, race and postal code. Then, the survey is over.

In the main treatment, subjects will be asked to think about their response to the same survey question posed in the control treatment, but their response to this question will NOT be directly measured/observed. Instead, they will be asked if their response was in a certain range of values (the two slit ranges) without indicating which slit or the precise numerical value of their answer. If they indicate that their response was in one of the two specified slit ranges, then they will next be asked to again consider the same survey question and this time to provide a numerical answer that is measured/observed. If their initial response to the survey question was outside the values of the two slits, the question is not asked a second time. After subjects have considered their answer to the survey question and addressed the survey question a second time, depending on their initial response, they then proceed to answer several demographic questions.

Design details

Randomization method
Subjects are randomly assigned to either treatment by the computer program. Subjects only participate in a single treatment
Randomization unit
Individual
Clusters
1, Mturk North America population
Observations
1000 or more depending on budget constraints
Arms
We will aim to have approximately equal numbers of observations in both our control and main treatment sessions.
Primary outcomes
The key outcome variable is the distribution of survey responses that pass through the two slits.

Cited by

No citation data available for this registration yet.

Keywords