To what extent do beliefs that determine health-seeking for children vary with age and gender? A vignette experiment in Zambia.

How to cite

Grabowska, Marta. 2024. "To what extent do beliefs that determine health-seeking for children vary with age and gender? A vignette experiment in Zambia.." AEA RCT Registry. AEARCTR-0013085. https://doi.org/10.1257/rct.13085-2.0

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

Marta Grabowska*

* Primary investigator

Registry ID
AEARCTR-0013085
DOI
https://doi.org/10.1257/rct.13085-2.0
First registered
Last updated
Status
Completed
Trial dates
2024-02-26 – 2024-12-31
Countries
Zambia (Copperbelt), Zambia (Luapula)
JEL codes
I12, C93
Registry record
http://www.socialscienceregistry.org/trials/13085

Abstract

A child’s health is an important determinant of their educational attainment and their subsequent lifetime earnings (Dercon and Porter, 2014; Bhalotra et al., 2022). In LMICs, poor health is a major cause of absenteeism in schools (Banda, 2017; Kabanga and Mulauzi, 2020) and is known to lower cognitive abilities (Clarke et al., 2017). A key determinant of child health are parental decisions to seek appropriate care in a timely manner or to invest in preventative measures. However, these decisions may not be optimal: for example, 40% of severe malaria cases could be averted if treatment was sought within the first day of symptoms (Mousa et al., 2020). In addition, there is evidence of differential decision making between younger and older children, which matches policy prioritisation of children under 5. Children over 5 are less likely to sleep under a bed net (Olapeju et al., 2018; Walldorf et al., 2015), parents are less likely to seek treatment when older children are ill (Walldorf et al., 2015), and when they do, they are less likely to seek care from formal providers (Coalson et al., 2019). As a result, this age group faces a considerable burden of disease (Hill, Zimmerman and Jamison, 2017). In Zambia, the setting for the study, the prevalence of malaria is highest in children aged 5-17, with 40% of children testing positive in endemic areas (Pinchoff et al., 2016).

In this study, I set out to measure parental beliefs regarding the costs of and returns to health-seeking behaviour in children of different age and gender in order to capture whether inaccurate beliefs lead to the observed inefficiencies. I will elicit beliefs and preferences using hypothetical vignettes, which will provide a detailed and novel dataset on parental beliefs by child characteristics. The use of vignettes enables me to vary these characteristics (age and gender) which could not otherwise be randomised. Using three vignettes, about diarrhoea, malaria, and a common cold, I will estimate the following parameters of parental beliefs: (1) expectations regarding costs of treatment and prevention measures, (2) expected returns to prevention measures, captured by changes in the expected likelihood of an illness episode with and without the measure, and (3) expected returns to acute care, or the difference between expected illness duration for the child with and without any treatment.

Intervention

The intervention is a variation in the vignette presented to the parent, as part of a within-survey vignette experiment. There are 3 scenarios:

1. The opening for the malaria scenario is: "Please imagine a mother and father named Pauline and Felix. They live in a village similar to your village, in a house like yours. Pauline and Felix have a [son/daughter] named [male/female name] who is [3/4/5/6/7] years old."

2. The opening for the diarrhoea scenario is: "Now imagine a different mother and father named Mildred and Richard. They live in a village similar to your village, in a typical house for the area. Mildred and Richard have a [son/daughter] named [male/female name] who is [3/4/5/6/7] years old."

3. The opening for the cold scenario is: "Finally, please imagine a mother and father named Sara and Rodger. They live in a village similar to your village, in a house like yours. Sara and Rodger have a [son/daughter] named [male/female name] who is [3/4/5/6/7] years old. Now, imagine [male/female name] wakes up with a cough and looks tired."

Experimental design

The experiment is conducted within the survey tool, which randomly allocates child gender and child age for the parents for each scenario. Half of the parents are randomly allocated to see each scenario with a girl subject, the other half with a boy. Likewise, one of five ages is randomly allocated: 3, 4, 5, 6, and 7. This creates 10 age-gender groups for each scenario. The key comparisons of interest are: boys versus girls (controlling for age group) and above or below 5 years old (with 5 serving as a benchmark for the comparison).

Design details

Randomization method
Randomisation done within survey software (SurveyCTO) using a randomisation command
Randomization unit
Individual
Clusters
0
Observations
13,500 parents or guardians of primary school aged children
Arms
For each of the 3 experiments, the sample is divided euqally into 10 groups by: 1. Age: 3, 4, 5, 6 and 7 year old 2. Gender: female or male. Each group will have approximately 1,350 observations.
Primary outcomes
For each of malaria, diarrhoea, and cold, I analyse the: (1) expectations regarding costs of treatment and prevention measures (2) expected returns to prevention measures, captured by changes in the expected likelihood of an illness episode with and without the measures (3) expected returns to acute care, or the difference between expected illness duration for the child with and without any treatment.

Cited by

No citation data available for this registration yet.

Keywords