We ran a lab-in-the-field experiment in Santa Rosa de Copán (Honduras) from May 1-14, 2019, and randomly selected 360 households from four different districts (Osorio, El Carmen, Prado Alto, and Santa Teresa) to interview.
Participants were asked two separated questions: i) their knowledge about the age of menarche (self-report SR), and ii) to predict the modal response of the other participants regarding the same question (modal guess MG). Appendix A shows the original instructions (in Spanish) and B the translation in English. It is important to highlight that SR and MG are not necessarily correlated. While SR captures an individual’s knowledge about the age of menarche, MG measures society’s knowledge about the same subject.
Our design considers both incentives and possible order effects. We used a monetary incentive in the MG task (a monetary award was given if the mode was hit and 0 otherwise) to reduce social desirability bias. Given that the order of the questions may also contribute to bias (see Brañas-Garza et al., 2021), we randomized the question order using p=0.5 to SR à MG and 1-p to MG à SR. As a result, half of the participants (n=186) answered SR à MG, and the other half (n=174) MG à SR (see Appendix A and B).
We also collected participant sociodemographic characteristics to assess possible biases, primarily: sex, education, ethnic group, and socioeconomic status, as well as the composition of the household in terms of girls and boys.
The field experiment was conducted by a Honduran organization, PILARH. Enumerators were trained on the objectives of the study, how to conduct the survey, confidentiality, and informed consent. The method was pre-tested in the field with 24 participants and reviewed before implementation.
Enumerators used paper-based questionnaires and received a list of households they had to visit, including the type of questionnaire (treatment) they had to implement. Face-to-face interviews were conducted in households and only one experimental subject was interviewed per household (father, mother, or guardian). The random allocation of participants into (order) treatments was made prior to the visit, therefore the enumerators had no influence on the selection.
The field study was pre-registered in AsPredicted before execution. The documentation can be consulted here: https://aspredicted.org/ps766.pdf.
Sample and outcome variables
To ensure the sample selection included households from different socioeconomic groups, we considered 11 schools with populations reflecting different socioeconomic levels. The respondents from the socioeconomic groups were divided as follows: 31% were from low-income, 34% from middle income, and 35% from high income households.
Of the total 360 participants, 50 were men and 310 women. The age of respondents varied from 22 to 78 with the following frequencies: 22-25 (15%), 26-30 (28%), 31-35 (21%), 36-40 (16%), 41-45 (9%), 46-50 (4%), and over 50 (7%). Most respondents over 50 were grandparents.
The respondents cover the entire spectrum of level of education, although most are concentrated in the lower levels: 49% primary education (6 years of schooling) or less, while only 3% held a university degree or higher (see Appendix C).
As regards ethnicity: 11% were Chorti, 7% Lenca, almost 8% Maya Chorti, 70% Mestizo, and 4% were from other groups.
In order to assess poverty levels, respondents were asked about access to food in the week previous to the survey: 23% responded that they did not have enough money to feed their children.
Household composition was also determined to assess whether parents with at least one daughter were more informed than those with only sons: 35% of responders had only male children, whilst 65% had at least one female child, but only 21% had at least one daughter who was at least 12 or older (experience).
The main objective of this study was to determine whether parents had knowledge of the age of menarche and what they believe regarding other parents’ knowledge. The age of menarche varies across countries and time, yet is considered healthy when it happens starts between the ages of 9 and 16. In Honduras, a recent study found that 93.3% of respondents had their first menstruation at 12 (Vides Torres et al., 2017).
As can be seen below, our sample average is 12.13 with a mode exactly equal to 12. Using the data on the mean age of menarche in Honduras and the modal value of 12 from the sample, we defined the following outcome variables:
- Self-report: SRHit (takes the value of 1 if respondents answer 12 and 0 otherwise), SRUnder (=1 if reported age is lower than 12 and 0 otherwise) and SROver (=1 if reported age is higher than 12 and 0 otherwise).
- Modal guess: MGHit (takes the value of 1 if respondents guess 12 and 0 otherwise), MGUnder (=1 if respondents guess lower than 12 and 0 otherwise) and MGOver (=1 if respondents guess higher than 12 and 0 otherwise).
Therefore, the first set of items – SRHit, SRUnder, and SROver – determine whether parents have accurate knowledge while the second set – MGHit, MGUnder, and MGOver – explores whether they think other parents are also well informed.
1The ex-ante answer might be any positive (integer) number, although we expected numbers between 8 and 16.
2 Although the study was conducted in the municipality of Choluteca, we consider it to be a good proxy for the situation of Santa Rosa de Copán as a whole.