Privacy and measurement error in phone surveys: the case of women’s agency.

How to cite

Romero, Javier; Agüero, Jorge; Perego, Viviana. 2021. "Privacy and measurement error in phone surveys: the case of women’s agency.." AEA RCT Registry. AEARCTR-0008551. https://doi.org/10.1257/rct.8551-2.1

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

Javier Romero* , Jorge Agüero , Viviana Perego

* Primary investigator

Registry ID
AEARCTR-0008551
DOI
https://doi.org/10.1257/rct.8551-2.1
First registered
Last updated
Status
On going
Trial dates
2021-01-01 – 2023-12-30
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/8551

Abstract

We propose a survey experiment (SE) that takes advantage of an ongoing impact evaluation (IE) in rural Guatemala. The IE aims to test the effect on female entrepreneurship and agency of an intervention providing women farmers with information on new market opportunities for their agricultural products.

Data collection will be carried out through phone interviews due to the pandemic. In particular, we will evaluate if adding privacy over the phone increases truthful reporting of women’s agency. We will randomly assign 1000 women to two treatment groups and one control group. The SE will be done during the enlinde data collection of the ongoing IE.

Specifically, both the treatment groups and the control group will receive a set of questions intended to measure women's empowerment. The control group will receive the questions directly from the enumerator during the phone interview. The first treatment group will receive the exact same set of questions, but respondents will answer using code words. For instance, the enumerator will ask “Who usually decides whether you can go to the market?”, to which women in the control group could reply “myself” or “my husband”, while women in the treatment group will use code words, such as, “one” or “two” for these alternatives. This allows for enhanced privacy as the respondent does not risk disclosing potentially sensitive information to individuals in her surroundings. The second treatment group will receive the exact same set of questions but will respond using the phone keypad. For instance, when the enumerators asks “Who usually decides whether you can go to the market?”, women in the second treatment group will press buttons in the keypad, such as “one” to answer "myself", "two" to answer "my husband", etc. The answers with the keypad produce a unique tone that will then be coded as numerical data by the research team. Thus, women in the second treatment arm do not risk disclosing potentially sensitive information to individuals in their surroundings or to the enumerator.

The null hypothesis is that there is no difference in the responses on agency elicited through the different methods. The random assignment to treatment or control in the SE will allow us to characterize the profiles of women that report differently under increased privacy. In addition, we will conduct 100 semi-structured interviews to shed light on the mechanisms behind the results. The semi-structured interviews will be coded into quantitative data to further study the characteristics of women that need more privacy as suggested by the interviews.

Intervention

Two treatment groups and a control group will receive a set of questions intended to measure women's empowerment through a phone interview. The control group will receive the questions directly from the enumerator during the phone interview. The first treatment group will receive the exact same set of questions, but respondents will answer using code words. For instance, the enumerator will ask “Who usually decides whether you can go to the market?”, to which women in the control group could reply “my-self” or “my husband”, while women in the treatment group will use code words, such as, “one” or “two” for these alternatives. The second treatment group will receive the exact same set of questions but will respond using the phone keypad.

There are a total of six questions that measure empowerment, which was adapted to the context of Guatemala from the shortlist provided by Jayachandran et al (2021) "Using machine learning and qualitative interviews to design a fi ve-question women's agency index".

Experimental design

We propose a survey experiment (SE) that takes advantage of an ongoing impact evaluation (IE) in rural Guatemala, which aims to test the effect on female entrepreneurship and agency of an intervention providing women farmers with information on new market opportunities for their agricultural products.

Data collection will be carried out through phone interviews due to the pandemic. In particular, we will evaluate if adding privacy over the phone increases truthful reporting of women’s agency. We will randomly assign 1000 women into two treatment arms and one control group.

Design details

Randomization method
Randomization at the individual level at the time of the survey (computer-based, using SurveyCTO)
Randomization unit
Individuals
Clusters
1000 women
Observations
1000 women
Arms
333 or 334 each
Primary outcomes
Six markers of empowerment: Are you allow to go alone to meet up with your girlfriends and friends for any reason (chat, have lunch, etc.)? Do you have to ask permission to buy vegetables or fruits; clothes for yourself; medications or personal supplies? Who usually decides if you can go to visit the house of a friend or friend / neighbor or neighbor? Who contributes the most to the decision to buy new assets (hoe, fumigation pump)? Who decides when to give support (exemplify support) to parents, in-laws, siblings, etc.? When making decisions regarding children's health care, who usually makes the decision?

Cited by

No citation data available for this registration yet.

Keywords