We’re building on the success of our pilot to test the impact of giving lump sum cash grants to refugees at scale.
Through this experiment, we hope to inform some of the biggest challenges facing humanitarian aid today, help us explore how we can go beyond helping refugees to survive, and instead empower them to prosper.
GiveDirectly originated as a giving circle started by Paul Niehaus, Michael Faye, Rohit Wanchoo, and Jeremy Shapiro, students at MIT and Harvard, based on their research into philanthropy. In 2012 they formalized their operation into GiveDirectly.
In December 2012, GiveDirectly received a $2.4M Global Impact Award from Google. In June 2014, the founders of GiveDirectly announced plans to create a for-profit technology company, Segovia, aimed at improving the efficiency of cash transfer distributions in the developing world. In August 2015, GiveDirectly received a $25M grant from Good Ventures.
In April 2016, GiveDirectly announced a $30M initiative to test universal basic income in order to "try to permanently end extreme poverty across dozens of villages and thousands of people in Kenya by guaranteeing them an ongoing income high enough to meet their basic needs" and, if it works, pave the way for implementation in other regions. The initiative launched in November 2017 and is set to run for 12 years.
GiveDirectly set up two emergency response program to the COVID-19 pandemic: one in the US, for which it has raised US$118 million, and one in African countries, for which it has raised US$76 million. The organization has sent cash relief to 116,000 families in the US and 342,000 families in Kenya, Liberia, Malawi, Rwanda and Togo.
In Togo GiveDirectly used satellite and cellphone data to target needy people, delivering cash completely contactless. It built on an existing cash transfer program called Novissi introduced by the government of Togo. Money is paid via mobile money technology, with beneficiaries withdrawing money at local shops. GiveDirectly helped expand the program to certain rural areas where the government found it particularly difficult to identify the poorest beneficiaries. The machine learning algorithm uses two steps. First, it finds the poorest villages by analyzing, among other things, roof material, sizes of farm plots and the presence of paved or unpaved roads through satellite images. Second, it finds the poorest individuals within a village by analyzing their mobile phone data like lengths and frequency of phone calls, number of inbound versus outbound calls, and amount of mobile data used. After the poorest individuals are identified, they are asked to enroll via mobile phone.
Basic income experiment
In April 2016 GiveDirectly announced that they would be conducting a 12-year experiment to test the impact of a universal basic income on a region in Western Kenya. More than 26,000 people will receive some type of cash transfer, with more than 6,000 receiving a long-term basic income.
Working in rural Kenya, it plans to conduct a randomized control trial comparing four groups of villages:
Long-term basic income: 40 villages with recipients receiving roughly $0.75 (nominal) per adult per day, delivered monthly for 12 years
Short-term basic income: 80 villages with recipients receiving the same monthly amount, but only for 2 years
Lump sum payments: 80 villages with recipients receiving a lump sum payment equivalent to the total value of payments of the short-term stream
Control group: 100 villages not receiving cash transfers
In November 2019, an economics paper on the GiveDirectly experiment was published. It found the local fiscal multiplier to be around 2.6x (in other words, each dollar provided by cash transfers increased local economic activity by $2.60).
After the release of GiveDirectly's impact self-evaluation in October 2013,World Bank economist David McKenzie praised the robustness of the study's design and the clear disclosure of the study lead's conflict of interest, but raised two concerns:
The use of self-reporting made the results hard to interpret and rely on (this being a feature of any study that attempted to measure consumption).
The subdivision of the sample into so many different groups meant that there was less statistical power that could be used to clearly decide which group had better outcomes.
Nonprofit Tax Code Designation: 501(c)(3) Defined as: Organizations for any of the following purposes: religious, educational, charitable, scientific, literary, testing for public safety, fostering national or international amateur sports competition (as long as it doesn’t provide athletic facilities or equipment), or the prevention of cruelty to children or animals.
Donations to this organization are tax deductible.