Syndicate content

Scenes from a Data Dive: The Final Presentations

Itir Sonuparlak's picture

Read our interview with the fraud and corruption data ambassadors, the poverty data ambassadors and a recap of Data Dive 2013.

Following last month's DataDive 2013, we continue our commitment to the work that started during the event. It’s astonishing to reflect on the accomplishments of a mere 24 hours of work. These events are particularly exciting and important because they bring many ideas and new possibilities to problems that seemed intractable for so long.

In case you weren’t able to attend, or if you did and would like to relive the weekend, here are videos from each team’s final presentation. Take a look for yourself and let us know which of these findings best fit your work and goals.

Scraping Websites to Collect Consumption and Price Data
Using a software technique for extracting data from websites, the presentation by Team Ndizi, which means banana in Swahili, unearthed data sources on food prices and consumption of bananas and rice in Kenya, as well as Indonesia. The team hopes that this analysis will serve as a faster and easier estimate of inflation.

Predicting Small-Scale Poverty Measures from Night Illumination
This team worked to understand the relationship between poverty and night illumination. The team worked to find a relationship between light illumination patterns by overlaying poverty metrics from district-based census measures.

Listening to Latin America and Caribbean (L2): High Frequency Poverty Data using Mobile Phone Surveys
Looking for new ways to track welfare in Latin American countries, this team looked at the collection and response rate of poverty data collected using mobile devices. The goal of their work was to determine whether it is possible to draw inferences at a national level on welfare changes.

Measuring Socioeconomic Indicators in Arabic Tweets
This team explored approaches to inferring socioeconomic status from tweets in Arabic. The team also looked at whether socioeconomic indicators can be identified by observing conversations in Arabic on Twitter. Examples include listening for poverty terms or human development phrases such as “no medicine”, “bankrupt”, or “bad education.”

Analyzing the World Bank’s Project Data for ‘Signals’
Using "Wayback Machine" to scrape archival data, this team used search APIs to identify patterns to determine the likelihood of fraud and corruption.

Analyzing World Bank Supplier Profiles
This team analyzed detailed profiles of World Bank suppliers to understand their relationships and potentially identify suppliers who might be at high risk of fraud.

UNDP Resource Allocation
The United Nations Development Programme (UNPD) examined the relationship between the composition of its workforce and program performance, using expenditure data to estimate correlations between workforce characteristics and performance.

Heuristics Auditing Tool for All
This presentation displays a tool developed to function as a reverse Google to identify fraud and corruption.

Evaluating Project Performance and Outcome
This project uses data to identify and evaluate performance, and analyzes whether development project objectives are being met. The findings reflect that while roughly 75 percent of projects meet the baseline standards, about $10-12 billion a year of projects do not achieve objectives and/or are not meeting expectations.

Comments

Add new comment