Survey will open in a new window or redirect momentarily, please keep this tab open! If the survey does not open, please enable popups and refresh the page.

Status:

Dismiss
A link to the study pre-analysis plan allows forecasters to read further if they want additional information.
Study: Direct and Spillover Impacts of a Community-Level HIV/AIDS Program: Evidence from a Randomized Controlled Trial in Mozambique

It is important to inform respondents about the intervention context. Here we wanted to highlight high baseline prevalence of HIV, and the fact that the intervention was delivered by a local implementing partner.
Introduction: Mozambique has high levels of HIV (7.1% of the population).  This study examines the impacts of Força à Comunidade e Crianças (FCC, “Strengthening Communities and Children”), a U.S. government-funded program targeting households with orphaned and vulnerable children which is designed to combat HIV/AIDS. The focus of this study is to understand how home visits by local community workers (hired by a local implementing partner of the FCC program) impact HIV-related outcomes in these households.

Here we introduce home visits visits, which are essential to understanding the treatment.
Home visits: Local community workers conduct home visits to identify households with orphans and vulnerable children, which are  then linked to appropriate programs and services in communities, schools, and health facilities. Local community workers are 80% female and usually between 25 and 40 years old.

The FCC program has many facets. Here we highlight the parts of the program most relevant to the experiment.
Home visits and HIV testing: A key component of the home visits is referrals for HIV testing at the nearest affiliated health clinics. All FCC beneficiaries (both adults and children of all ages) who do not know their HIV status (or were negative and have not been tested in the last 12 months) are supposed to be referred by the community workers to HIV testing services. Those testing positive for HIV are referred to receive antiretroviral therapy (ART) through a nearby affiliated clinic. Local community workers follow up with individuals initiating ART to promote ART adherence on an ongoing basis. During these home visits, the local community workers try to increase HIV testing rates through:
  • Since these interventions are fairly complex, we try to summarize only the key features.
  • Note that since the intervention fidelity is unknown, we only say what community workers are expected to do.
  • We provide a link where interested respondents can read more about the intervention.
  • Information related to HIV/AIDS: FCC beneficiaries receive information on HIV/AIDS, such as on methods of disease transmission, progression of the disease, treatment, HIV testing, and locations of health clinics providing testing.
  • Discussions to reduce stigma concerns: FCC beneficiaries also engage in discussions to reduce stigmatizing attitudes among program beneficiaries. Community workers are expected to provide psychosocial support, gradually gaining program beneficiaries’ trust over repeated interactions.
  • Education: In home visits, community workers are also expected to give caregivers advice and encouragement regarding children’s education.
  • Other components: Households are connected to other relevant services after the home visits, based on needs assessments conducted by the local community workers. These other services are expected to reach only a relatively small fraction of those reached by home visits. More information on these subcomponents can be found in this document (pages 5-6).
Experimental design: Randomization took place at multiple levels. In this survey, we focus on the two types of households depicted in the figure below:
This study contained a large number of treatments. In order to reduce survey burden, we elected to elicit forecasts of the simplest of these interventions/ randomizations.

Here we highlight that not all targeted households may receive a visit. This is relevant since elicit predictions of the effect of being assigned to treatment.

At baseline, the average household contained 5.9 members. The experiment contains several other levels of randomization, including an individual level of randomization that took place in all villages. For simplicity, this survey focuses on those households that were not assigned to receive any intervention beyond the FCC program. For more information, see this document (pages 7-9)

Study sites: Communities were selected on the basis of being close to health clinics offering HIV testing and treatment, having sufficient populations of orphans and vulnerable children, and having no other active donor-funded HIV/AIDS programs.

Respondents should have detailed information on the treatment timing relative to outcome measurement. Here we highlight variation in treatment exposure.
Timing: The FCC program began activities in early 2017. Over the calendar year they gradually enrolled beneficiaries and scaled up program activities. The follow-up survey began in May 2019, and was scheduled to be completed near the end of 2019. Households in treated communities can therefore have had up to two years of exposure to the FCC program at the time of the follow-up survey, but some households may have had a few months' less program exposure, if they happened to have been enrolled in the program towards the end of 2017.

Respondents should be provided with a comprehensive description of the predicted outcome.
We are interested to hear your predictions about the effects of this intervention on one outcome: self-reported HIV testing. Even if you do not have strong beliefs about the effects of the intervention on this outcome, we are still interested in your best guess.

Respondents should understand how the predicted outcome is coded/measured.
Self-reported HIV testing was measured in the endline survey. Respondents were asked if anyone in the household had been tested for HIV in the last 12 months. The outcome is a household-level variable equal to 1 if at least one household member is reported to have had an HIV test in the last 12 months, and 0 otherwise.

Please predict the difference in self-reported HIV testing (in percentage points) between households assigned to be visited by local community workers and the control group.
Notes:
  • This link takes participants to the two page description.




  • We provide the control groups mean and standard deviation as a reference.
  • An example can be useful for helping ensure respondents understand their predictions.
Notes:
  • Click here for a reminder of the interventions and study background, which will open in a new window.
  • Timing: Households in treated communities can have had up to two years of exposure to the FCC program at the time of the follow-up survey, but some households may have had a few months' less program exposure, depending on enrollment date.
  • Reference: As a reference, at baseline about 41.9% (with a standard deviation of 49.4 percentage points) of households self-reported having any household member receive HIV testing in the last 12 months.
  • As an example, if you enter 14.7 it means that you think self-reported HIV testing will be 14.7 percentage points higher in the group assigned to be visited by local community workers. If you enter -14.7 it means that you think self-reported HIV testing will be -14.7 percentage points lower in the group assigned to be visited by local community workers. If you enter 0 it means you think the treatment had no impact.

There are many ways to elicit predictions. In DellaVigna et al. (2020) we outline four variations in elicitation strategy (two examples are provided below): (1) small versus large reference values (see last bullet above); (2) whether predictions are in raw units or standard deviations; (3) text-entry versus slider responses; and (4) small versus large slider bounds. Our results suggest that reference values and units seem to have little effect on responses, though wider slider bounds are associated with higher forecasts.

We bound numeric responses at +-1 SD to avoid confusion among survey respondents.

Here we bounded our slider scale at +-2 SD, but elicited predictions in raw units. To avoid acquiescence, the slider must be moved to continue to the next question (the default is not 0).