Poverty Prediction Challenge

Estimate individual and aggregate household consumption from limited survey data. #development

$10,000 in prizes
8 weeks left
40 joined

Overview

Accurate poverty measurement is essential for directing global development efforts and informing evidence-based policies for poverty reduction and equity enhancement, yet many countries lack recent data due to the high costs and complexity of collecting comparable comprehensive household expenditure surveys.

This challenge simulates a common real-world scenario faced by World Bank economists, who are tasked with producing up-to-date poverty measurements and additional welfare indicators, even in cases where fully detailed recent information on household expenditure is unavailable. Your goal is to develop survey-to-survey imputation models that predict both poverty rates and per capita household consumption from anonymized historical survey data.

Competition Notes

Research focus. This competition is designed for research purposes to explore imputation methods, and has a limited test set reflecting the real-world challenges faced by researchers. As a result, the submission rate is limited and the public leaderboard for this competition may not be indicative of final rankings.


Competition End Date:

Feb. 4, 2026, 11:59 p.m. UTC

Place Prize Amount
1st $4,000
2nd $2,500
3rd $1,500
Bonus $2,000

Bonus prize: best poverty rate predictions

The team that submits poverty rate predictions with the lowest error on the private test set and whose solution is verified will receive a $2,000 bonus prize.


How to compete

  1. Click the "Compete!" button in the sidebar on the competition homepage to enroll in the competition.
  2. Get familiar with the problem through the overview and problem description. You might also want to reference additional resources available on the about page.
  3. Download the data from the data tab.
  4. Create and train your own model.
  5. Use your model to generate predictions that match the submission format.
  6. Click "Submit" in the sidebar, and then "Make new submission". You’re in! You can make up to three submissions per week.
  7. Before the submission deadline, select the solution you wish to use for final evaluation. You may select only one solution for final evaluation.

Competition rules

The challenge rules are in place to promote fair competition and useful solutions. If you are ever unsure whether your solution meets the competition rules, ask the challenge organizers in the competition forum or send an email to info@drivendata.org. A few key rules are highlighted below. For more details, see the official rules page.

Eligibility

Individuals currently affiliated with the World Bank Group, including staff and consultants, are encouraged to participate but are not eligible to receive any prizes or awards.

Use of test data

A unique feature of this challenge is that the prediction of poverty rates relates to the prediction of household consumption, and in this sense your predictions on the test set are not "independent". For the purposes of this competition, you should treat each survey (i.e., collection of household responses) as an independent sample. The imputed values for an individual household's consumption and the overall poverty rate may leverage multiple households’ responses but the predicted poverty rates must be a function of each survey alone. Your models should be capable of accepting an arbitrary number of surveys and survey responses within each survey, and the predictions for households within each survey and the poverty rates predicted from a survey should be made independently of the other surveys.

Use of external data

The use of external data is not allowed in this competition. You may only use the data provided in the training and test sets, as well as any additional information provided in the problem description and about pages.

Tool use

The use of pre-trained models is allowed, as long as the models used were 1) available freely and openly in that form at the start of the competition and 2) not trained on any data associated with the ground truth data for this challenge.

Participants may not upload competition data to tools like ChatGPT or Gemini, but may load open-source model weights into an environment that they can wipe the data from afterwards, such as their local machine or a cloud compute environment.

PRIVACY NOTICE

The World Bank Group, (“we” or “us”), respects your concerns about your privacy and is committed to protecting it by establishing and complying with this Privacy Notice which adheres to the core privacy principles set forth in the Personal Data Privacy Policy of World Bank Group Institutions.

The WBG has engaged a vendor, DrivenData, to host the competition on the DrivenData Platform

Participation in the competition is entirely voluntary. By registering for the competition, you acknowledge that you will be sharing personal data directly with DrivenData. They will process your personal data according to their privacy policy and notice and in compliance with applicable laws.


This challenge is sponsored by the World Bank Group.


Images courtesy of the World Bank Group