Weather Forecasts, Advisory and Farmer Belief Formation

How to cite

Surendra, Vaishnavi. 2024. "Weather Forecasts, Advisory and Farmer Belief Formation." AEA RCT Registry. AEARCTR-0014522. https://doi.org/10.1257/rct.14522-1.1

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

Vaishnavi Surendra*

* Primary investigator

Registry ID
AEARCTR-0014522
DOI
https://doi.org/10.1257/rct.14522-1.1
First registered
Last updated
Status
On going
Trial dates
2024-04-01 – 2025-06-30
Countries
Private
JEL codes
D81, O12, O13, Q54
Registry record
http://www.socialscienceregistry.org/trials/14522

Abstract

Weather uncertainty amplifies agricultural production risk, and weather forecasts are a technology that can help farmers better cope with such risk with improvements in forecast-skill and increasing weather variability due to climate change. Accurate, granular weather forecasts at seasonal, medium-range and short-range timescales can help farmers form more accurate expectations about upcoming weather. Accurate expectations about upcoming weather could translate into optimal action for eventually-realized weather when farmers are equipped to act on their beliefs. We provide rainfall forecasts to farmers who are users of an existing mobile-phone-based agricultural advisory service, and focus on two aspects of forming beliefs based on weather forecasts and acting on them: (a) we assess farmer’s responses to inaccurate forecasts (false alarms), and whether a light-touch nudge reminding farmers of the probabilistic nature of forecasts impacts these responses and their engagement with the service; (b) we assess whether farmers’ engagement with/trust in the service depends on whether they receive forecasts alone or forecasts along with actionable advice.

Intervention

Farmers receive access to audio weather forecasts provided over voice-calls through a mobile-phone based advisory service

Experimental design

Farmers enrolled on a mobile-phone baed agricultural advisory service receive access to a an audio weather forecast service. In an initial phase, farmers assigned to receive the new forecast service are randomized into two experimental groups: one receiving probabilistic forecasts, and the other receiving the same forecasts along with a nudge at the village level. In the next phase (in a different region from the previous phase), villages are randomized into four experimental groups: probabilistic forecast only, probabilistic forecast with nudge, deterministic forecast only, and a control group. In these phases, we also track the impact of incorrect forecasts (cases where predicted weather does not occur) on outcomes.

In the final phase of the study (in a different region from the previous phases), villages are randomized into five experimental groups: probabilistic forecast only, deterministic forecast only, probabilistic forecast along with related agricultural advisory, deterministic forecasts along with related agricultural advisory, and a control group.

Randomization is stratified at the forecast-grid level. A forecast-grid cell is a sub-block geographic area, 0.2° ✕ 0.2° or 18 km ✕ 18 km (324 km2), which is the resolution at which forecasts are currently disseminated.

We focus on farmers' engagement with this service, and survey a small sub-sample of service-users over the phone in short surveys.

Design details

Randomization method
Randomization is conducted using a random number generator in Stata.
Randomization unit
Village level
Clusters
approximately 65 villages in the first phase, approximately 571 villages in the second phase, approximately 335 villages in the third phase.
Observations
~1,145 individuals in the first phase; ~14,644 individuals in the next phase; ~15,106 individuals in the final phase
Arms
1. 34 villages in the forecast only arm; 31 villages in the forecast + nudge arm 2. 160 villages in the probabilistic forecast only arm; 166 villages in the probabilistic forecasts + nudge arm; 167 villages in the deterministic forecasts only arm; 78 villages in the control group 3. 73 villages in the probabilistic forecasts only arm; 73 villages in the probabilistic forecasts with advisory arm; 80 villages in the deterministic forecasts only arm; 72 villages in the deterministc forecasts with advisory arm; 37 villages in the control group
Primary outcomes
Engagement with weather forecast service (pick-up rate, listening rate)

Cited by

No citation data available for this registration yet.

Keywords