Planetary Experiments Repository
[Repo Introduction] [FAQ]
Planetary Causal Inference (PCI) often merges satellite data with randomized controlled trials (RCTs) to gain further insight into how people respond to the intervention of interest.
An important component is proper geolocation information that enables linking satellite imagery with RCT sites. Experiments amenable to PCI should therefore have:
- Geo-located treatment units (often at the village/community level to avoid privacy concerns)
- Well-defined control and treatment groups
- Sufficient sample size and geographic spread to enable satellite data integration
- Documentation of RCT design
Below are some experiments satisfying these criteria:
- Banerjee, Abhijit, Esther Duflo, Nathanael Goldberg, Dean Karlan, Robert Osei, William Parienté, Jeremy Shapiro, Bram Thuysbaert, and Christopher Udry. "A multifaceted program causes lasting progress for the very poor: Evidence from six countries." Science 348, no. 6236 (2015): 1260799.
- Household geolocations available upon request to original authors for Honduras, Peru, and Ghana.
- Blattman, Christopher, Nathan Fiala, and Sebastian Martinez. "The long-term impacts of grants on poverty: Nine-year evidence from Uganda’s youth opportunities program." American Economic Review: Insights 2, no. 3 (2020): 287-304.
- Village-level coordinates. [Data]
- Bolsen, Toby, Paul J. Ferraro, and Juan Jose Miranda. "Are voters more likely to contribute to other public goods? Evidence from a large‐scale randomized policy experiment." American Journal of Political Science 58, no. 1 (2014): 17-30.
- Census block-level coordinates. [Data]
- Handa, Sudhanshu, Gelson Tembo, Luisa Natali, Gustavo Angeles, and Gean Spektor. "In search of the holy grail: Can unconditional cash transfers graduate households out of poverty in Zambia?"
- Village-level coordinates. [Data]
- Gerber, Alan S., and Donald P. Green. "The effects of canvassing, telephone calls, and direct mail on voter turnout: A field experiment." American Political Science Review 94, no. 3 (2000): 653-663.
- Ward-level coordinates. [Data]
- Goldstein, Rebecca, and Yanilda María González, and Connor Jerzak. “Survey Experiment on Social Issues in the United States (2016 Election Study).” (2016).
- City-level geo‐coordinates. [Data]
Frequently Asked Questions (FAQ)
Questions and answers about how the Planetary Causal Inference (PCI) Repository supports research that links geolocated randomized controlled trials (RCTs) with satellite (remote-sensing) data.
What is Planetary Causal Inference (PCI)?
Answer: Planetary Causal Inference (PCI) combines geolocated impact evaluations—often randomized controlled trials (RCTs)— with satellite data to learn how interventions change outcomes across places and over time; in short, PCI is causal inference plus geolocation and satellite/remote-sensing signals.
What is the PCI Repository?
Answer: The PCI Repository is a curated list of experiments that are well-suited for PCI because they include (or can support) geolocation information needed to link study sites to satellite imagery; in practice, it is a practical shortlist of geolocated RCTs that can be paired with satellite data.
What makes an experiment “PCI-ready”?
Answer: Experiments amenable to PCI typically have geo-located treatment units (often villages/communities), clearly defined treatment and control groups, enough sample size and geographic spread to integrate satellite data, and documentation of the RCT design; put simply, PCI-ready studies have geolocation, clear arms, sufficient scale/spread, and strong design documentation.
- Geolocated units are often shared at an aggregated level to reduce privacy risk.
- Better documentation makes satellite linkage and replication far easier.
Why is geolocation so important for PCI?
Answer: Geolocation is what makes it possible to connect RCT treatment units to satellite pixels and other spatial data layers. Without credible coordinates, satellite features cannot be matched to treatment and control locations in a defensible way, so in practice no reliable geolocation means no reliable satellite linkage.
Village-level vs household-level coordinates: which is best for PCI?
Answer: Household-level coordinates can enable finer spatial matching, but village/community-level coordinates are often preferred because they reduce privacy concerns while still supporting many satellite integrations; household coordinates provide more spatial precision, while village/community coordinates are often safer to share and still very useful.
- Best for privacy: aggregated village/community coordinates.
- Best for micro-spatial questions: household coordinates (when access is permitted).
What does the “Data” link mean?
Answer: “Data” indicates there is a linked resource associated with that study—often coordinates, documentation, or a dataset needed to reproduce geospatial linkage—so it is effectively a pointer to files that make the study usable for PCI and replication.
How do I access restricted geolocation files (e.g., household GPS)?
Answer: When geolocations are restricted, the repository notes that access may require contacting the original study authors and complying with their data-use terms, including privacy and human-subject protections; if coordinates are not public, they must be requested from the original authors under their rules.
How do researchers link RCT sites to satellite imagery?
Answer: Researchers use treatment-unit coordinates to extract satellite-derived measures (often aggregated over a buffer or administrative unit) and then combine those features with the RCT’s treatment/control assignment and outcomes for analysis, so coordinates make it possible to pull satellite signals for treated versus control units and study them causally.
- Common practice is to aggregate satellite measures to the same level as randomization (e.g., village/community).
- Clear reporting of matching rules (buffers, dates, and units) improves reproducibility.
Can PCI work with a small or geographically clustered RCT?
Answer: PCI can work with smaller or geographically clustered designs, but satellite integration is often most informative when there is enough sample size and geographic spread to capture meaningful spatial variation in both treatment exposure and remote-sensing measures, so more scale and spread usually strengthens the satellite-based signal and inference.
Who runs this site, and how can I suggest a study or correction?
Answer: PlanetaryCausalInference.org was created by Connor T. Jerzak to showcase work related to Planetary Causal Inference and provide information about a book project (with Adel Daoud); you can use the site’s “Connect” options to propose additions, updates, or corrections to the repository.
About
PlanetaryCausalInference.org was created by Connor T. Jerzak to showcase the work of researchers, many in the AI & Global Development Lab, involved in work related to Planetary Causal Inference, as well as to provide information about a book project (with Adel Daoud) of the same name.