Rethinking efficiency to better reflect our diverse cash programs

Occasionally, to “unlock” institutional gifts, we contribute some of your individual donations in a new program. Here’s how that has worked in one case:

🇲🇿 Example: your donations unlocked $17.5M in USAID funds for Mozambicans in poverty. 

• The bespoke program that allowed us to start working in Mozambique in 2022 was a two year, $40/month basic income for rural households, funded with $5M from USAID and a $3M match from your donations, ran at a below-target efficiency (~60%), in part due to fixed costs of launching a new country office. 
• This year, we’ve expanded our work with USAID Mozambique with a higher (~75%) efficiency project, funded by $5M from them and $1.3M from your donations. 
• Across Mozambique, $6M in match dollars from individual GiveDirectly donors has helped unlock $17.5M in U.S. government funds –– new money that would not have otherwise reached recipients as cash transfers. 
• In the first two years of this new country, we’ve reached 18k recipients, and will reach ~5k more by year end, doing more good for people in poverty.

Our flagship poverty relief programs are the simplest, most cost-effective way to deliver cash to Africans in extreme poverty. Today, we run versions of our flagship program in five countries: Kenya, Malawi, Mozambique, Rwanda, and Uganda. 

These programs are uniquely low-cost because they…

• 💰deliver a relatively large amount per recipient, offsetting enrollment costs
• 🛖enroll entire villages, simplifying operations at scale
• ☮️run in stable areas, limiting security costs and delays
• 🗓️run for a long-time in the same areas, delivering $10M+ and recouping start-up costs

In our first (small & scrappy) years, we kept efficiency at ~90%. However, over time our work changed in important ways that increased those costs:

• 🥾reaching even poorer communities, who are often more remote, raising travel costs
• 👥adding more safeguarding/fraud prevention and community/government relations staff, vital to keep our operations running safely and smoothly
• 📋adding more monitoring and evaluation to better understand recipients’ experience and outcomes, helping us continuously improve
• 💵paying our in-country staff fairer, higher salaries⁴
• 📈rise in costs for all goods and services over the past decade while the amount we give recipients has remained mostly fixed

Keeping our flagship program efficiency at ~85% is an ambitious but achievable target. Here’s why:

• A USAID paper on African cash programs of various sizes found 20% was the average share of budgets that went to direct costs, not including indirect organizational costs as we do in our efficiency calculations –– more here →
• The European Union emergency relief (ECHO) agency target “for cash programmes equal to or above €10M… partners are expected to reach a minimum cost-efficiency ratio of 85:15,” though, in practice, this is not often achieved.

If you want your donation to only go to this flagship program, give through this page →

Our newer, bespoke programs use innovative ways to find, target, and support vulnerable African populations in poverty left behind by traditional aid. Typically, we co-design these programs with specific donors. 

These programs may have a lower target than our flagship as some of them cost more to deliver because they may… 

 🥾→🏍️Reach underserved, hard-to-reach communities 

• Design: The areas with the highest extreme poverty rates are often the most remote, both spread out and with limited infrastructure like paved roads or functioning bridges. And this can get much worse, e.g. during rainy seasons
• Efficiency: Reaching high-need, remote areas requires more staff time, field offices, and transit costs. Damage to infrastructure can result in costly delays. In some cases, we’ve invested in telco infrastructure to bring mobile money access where it was not previously. 

💰→💵Give cash in smaller installments: 

• Design: Delivering small recurring payments ($20-50 per month for a year or two) is proven to improve consumption and promote wellbeing for more vulnerable populations including parents of young children. 
• Efficiency: Delivering $1,000 over twenty monthly installments costs more than delivering it all at once, because each payment requires follow-up from our team to confirm there were no issues  

 🗓️→⏱️Smaller scale, delivering less cash in the short-term

• Design: New programs often start as a limited pilot, delivering a relatively small amount (<$5M) over a year, as the funders reassess continuing. 
• Efficiency: New programs have upfront costs for bespoke planning and design, and often mean launching field offices and staff in a new location. If this program closes a year later, delivering <$5M, initial investments in planning, design, and field offices may not be fully recouped.

Some bespoke programs will easily clear 75% efficiency and others, by nature of their design, will be below. Based on the above factors, we believe keeping the average efficiency of our newer, bespoke program at ~75% is an ambitious but achievable target. This is in line with outside analysis that scale, duration, accuracy of targeting, and timeliness of delivering cash are the key drivers of program efficiency. 

The benefits of these programs outweigh their higher cost. This is a trade-off we’re comfortable with, as these higher cost programs mean we’re reaching more vulnerable populations – refugees or people living with HIV/AIDS or cancer to name two – with life-changing cash, often from funders that were not giving cash before (see above). We leverage such unique programs to unlock new funds for recipients.

If you want your donation to only go to our flagship poverty relief programs, and not these bespoke ones, give through this page →

Our global emergency relief programs help people in moments of crisis very quickly, averting suffering and preventing them from falling into extreme poverty. 

This humanitarian work has an even lower target as it costs more to deliver because the programs… 

☮️→⛑️Reach insecure regions: 

• Design: We’re reaching recipients in more insecure regions, which is vital as half of the 700M people in extreme poverty live in so-called ‘fragile states.’ 
• Efficiency: Less secure areas like conflict or disaster zones require additional security expenses and can face costly delays.

💰→💵Give smaller amounts of cash: 

• Design: Delivering a smaller total amount ($100-400) is shown to help survivors of disaster not fall deeper into poverty and allows us to reach more people with the same budget.
• Efficiency: It costs about the same amount to enroll and pay someone $100 as it would $1,000, so many programs with smaller transfer sizes are less efficient. 

🛖→🛰️Select individual recipients rather than entire villages:

• Design: Reaching specific groups –– disasters survivors, refugees, urban poor – requires more novel targeting approaches than simply picking an entire village, ranging from using lists from other groups to developing machine-learning technology to identify recipients.   
• Efficiency: These targeting approaches are often more labor intensive than picking an entire village, and the tech-based ones require set-up costs that are only recouped at scale. 
  • Investing in partnerships and infrastructure to target and deliver cash remotely (which we call “MobileAid”) will drive up efficiency over time, especially when done at a large scale. For instance, we used MobileAid to help the Government of Togo deliver COVID-19 relief to 138k people at an 87% efficiency.

🗓️→⏱️Smaller scale, delivering less cash in the short-term:

• Design: Disaster relief programs often run for just a short period after the crisis, delivering a relatively small amount (<$5M). They prioritize speed (reaching people quickly), as emergency cash has more impact the sooner it reaches survivors. And they are often in more dangerous places. 
• Efficiency: Quickly responding where emergencies occur often means launching new field offices and staffing in new locations, increasing fixed upfront costs. If this program closes a year later, delivering <$5M, this setup is a higher share of the program cost than in longer term, larger programs. Additional security requirements also bring more costs.

Keeping our emergency relief program efficiency ~60% is an ambitious but achievable target. Here’s why:

• A review of 47 EU-funded humanitarian cash programs found an average efficiency of 44%. 
• An earlier review of 37 EU-funded rural humanitarian cash programs found an average efficiency of 56%

The benefits of these programs outweigh their higher cost. This is a trade-off we’re comfortable with, as these higher-cost programs mean we’re reaching communities at their most vulnerable moments after disasters or during conflicts.

If you want to give to these programs, give through this page →

If you want your donation to only go to our flagship poverty relief program, and not global emergency relief programs like this, give through this page →

Many of our U.S. programs provide low-income Americans with a guaranteed minimum income so they can build a foundation for economic success. Other programs are targeted at families in vulnerable periods, during emergencies, after hurricanes, or in their first year of parenthood.  

The efficiencies of U.S. programs, whose transfers and direct costs are paid for by U.S.-only donations, range widely because the programs may… 

💰→⚖️Vary how we deliver cash to study what works best: 

• Design: We know much less about how to make cash work best in high-income countries like the U.S. In order to learn more, our programs have ranged from monthly payments for two years to a single upfront lump sum to a blend where recipients choose their own transfer size and timing.  
• Efficiency: Designing new, novel programs requires upfront costs, but as we innovate with giving a single large lump sum, we see higher efficiency – it costs less to deliver a one-time $10k than spreading that same amount out over 1-3 years. 

💰→💰💰Give much larger total amounts: 

• Design: Recipients in our U.S. programs receive 10-30x as much cash as those in our African programs, as it costs more to create comparable outcomes in countries with higher costs of living.
• Efficiency: It costs about the same amount to enroll and pay someone $10,000 as it would $1,000, so programs with such large transfer sizes are more efficient. Our largest U.S. programs regularly exceed 85% efficiency.

🛖→📲Deliver large emergency relief payments remotely:

• Design: In response to COVID-19, we targeted and delivered $1,000 to nearly 200k families fully remotely through a food-stamp benefits app. We’ve used the same approach in response to Hurricanes Ian, Helene, and Milton.
• Efficiency: Delivering large sums of money fully remotely at scale is very efficient –– up to 98% for our COVID-19 program.

➕🏛️Come with unique regulatory compliance requirements:

• Funding: Most of our U.S. guaranteed income programs are paid for by federal funding from the American Rescue Plan. 
• Efficiency: Government funded programs have higher, distinct compliance costs than many of our international programs.

Because of this variance, we don’t currently set a target efficiency for U.S. work. As we build up our portfolio, we’ll know more about which types of programs we pursue more often: larger, higher efficiency emergency relief or novel, lower efficiency poverty relief pilots and studies. We want to prove direct cash can be as efficient as national scale U.S. safety net programs like SNAP (food stamps), which delivers $118B a year at ~95% efficiency. 

Read more on why we work in the U.S. If you want your donation to go to our U.S. programs, give through this page →

📋→📲 Our numerator used to be dollars “committed” to recipients; now it’s dollars “delivered” 

• Once we enroll a recipient to receive funds, we’ve “committed” to pay them a certain amount, even though some time may pass between enrollment and the last of the cash is “delivered.”  
• Until recently, we used “committed” as the numerator for calculating efficiency, rather than “delivered.”
• When we mostly ran our one-time lump sum flagship program, the time between committing funds and delivering all of them was a matter of months, so it was largely the same measure. 
• Now that we run more long-term basic income programs, using dollars “committed” would skew our annual efficiency calculations, as a large amount of money “committed” one year would still be getting delivered years later. 
• We’ve switched to using dollars “delivered” in the numerator to give you a more accurate picture of our efficiency year over year.

💱🛑 We stopped treating accounting adjustments to translate local currencies to USD$ as an operational cost

• When we launch a project, our budget is based on our best guess of what the exchange rate will be between USD$ (which we hold donations as) and the local currency that will be delivered to recipients via mobile money.
• Historically, there’d be some minor changes in the value of those local currencies against the U.S. dollar, which we’d represent as a positive cost (if the money became less valuable) or a negative cost (if the money became more valuable). 
• These adjustments almost always netted out across all our programs as a negative cost or “gain,” and was so small as to have a negligible impact on efficiency calculations. 
• However, recently local currencies (e.g. Malawi) have seen sudden, large devaluations, which would look like very large negative costs if we kept calculating in the way we had before. That would artificially make those programs look more efficient than they really are.
• So now we’re changing to treat currency fluctuation gains as revenue, not as a negative operating cost. This means it is no longer impacting how we calculate and present efficiency.

We raised 5.6x more in 2020-21 than the two years prior ($97M→$544M), so hired more staff to deliver more cash. Then in 2022-23 we raised half as much as the two years prior ($544M→$243M), so needed to downsize. The lag between raising less money than we expected and contracting our operations was a main driver in our decreased efficiency, for example staff contracts and operational leases take time to terminate.

How we’re improving

We’ve developed more predictable funding flows to all our countries of operation, which enables better planning to manage key costs such as staffing. This includes building multi-year pipelines with institutional donors, and earlier internal alignment on how our flexible individual donations will be allocated.

COVID-19 was an emergency situation, so we prioritized moving fast to deliver hundreds of millions in new donations. We prioritized reaching people in need quickly rather than scaling more slowly to maximize efficiency. Only once the dust settled from this rapid expansion, did we look to reduce our staff size and other operational costs.

How we’re improving

We’ve reduced our staff size by 20% since 2022, in order to make sure our team is right-sized for our funding levels. We’ve also refined our operations, consolidating technology costs and moving some outsourced accounting and hiring functions in-house now that we’re a larger organization. We’re pursuing other cost-savings ideas like integrating directly with select telcos (reducing fees) and creating modularized program design technology (reducing program startup costs).

Specifically, we opened offices in Ethiopia, Nigeria, and Yemen to reach more of the extreme poor on the heels of 2020-21’s fundraising. When we raised less in 2022-23, we decided to close these newer offices to focus our work in the countries where we were better established. Closing these offices meant  not offsetting upfront costs as much as if they’d remained open delivering cash. These “back-office” costs are shared across all our programs.

How we’re improving

We’re holding a higher bar for opening new country offices, not growing our footprint. Also, we developed a rapid response model for emergency relief in new countries that does not require opening permanent offices.

In the past two years, we’ve given large lump sums to twice as many people in Malawi, Rwanda, and Uganda than in the two years prior. Working at a larger scale and giving out much more cash requires coordinating with higher levels of government, in some cases integrating with their social registries. While this can raise costs or cause delays, not coordinating well has its own cost, as we saw in Uganda in 2020 when our operations were paused due to a misunderstanding with upper levels of government.

How we’re improving

Completing these social registry integrations with governments comes with upfront costs, but we expect these to be offset over time as we continue scaling.

In the wake of a large fraud in the DRC in 2022, we paused all operations in the country. This loss and the costs of the resulting long pause as we rehired a new team and improved systems lowered efficiency across all our programs.

How we’re improving

We tightened controls, built new detection systems, and instituted back-end machine learning flags to prevent future major fraud loss.