The year is 2014, this is Siaya county in rural Kenya, it is about to become part of one of the biggest free money experiments in history, distributing millions of Dollars to one of the poorest areas on earth. This program is about to fundamentally change the lives of people that are often struggling for food. It is also the first thorough experimental study ever to test the theories from two of the greatest economists who ever lived, Milton Friedman’s theory that rapid increases in the supply of money cause inflation, and John Maynard Keynes’s theory that big spending programs often benefit the economy by a multiple of what was spent.
So, what happened? Did giving out so much free money actually lift people out of poverty in the long term, as Keynes would have predicted? Did it work better than traditional development aid? Or, did handing out so much free money cause inflation, as Milton Friedman would have predicted?
To answer these questions, I made this video with GiveDirectly —the charity involved in giving out the money—, and interviewed Oxford professor Dr. Dennis Egger, who was part of the team that scientifically studied the program, and summarized their research into key insights about exactly: (1) what their findings mean for Friedman’s teachings about inflation, (2) what it means for Keynes’s theory about big spending programs, and finally, how big cash transfer programs like these will change development aid forever.
But, to answer these questions, we first need to get into
How experiments are revolutionizing macroeconomics
Macroeconomics focusses on the performance of entire economies. So, it can be about what happens to the income of all people in a certain area, it’s GDP. Or what happens to all prices at the same time, inflation.
Like other scientists, macroeconomists follow the scientific method, which involves using theory to formulate a hypothesis, testing that hypothesis through experiments, then analyzing the data from these experiments, and then confirming or rejecting the hypothesis. However, when it comes to experiments, macroeconomists face a much greater challenge than other smaller scale social sciences.
For example, in medicine, the gold standard for experiments are so-called randomized controlled trials, RCTs for shorts. In such a trial, there are typically two groups. The first group is the intervention group, which receives the treatment or intervention being tested. The second group is the control group, which does not receive the treatment. To ensure that the two groups are as similar to each other as possible, subjects are randomly assigned to either groups. If an RCT is done correctly, and we compare the start and end results of both groups, we can be pretty sure that, if there is a change in the intervention group, but not in the control group, this change was CAUSED by the intervention.
There are of course ethical considerations when “testing” anything with human subjects However, RCTs are considered a relatively fair way of distributing aid, especially in cash. After all, there is never enough aid for everyone in need, so if we’re going to distribute aid to a limited number of people anyways, doing it randomly is both fair and allows us to study the impact.
Randomized controlled trials have already revolutionized how individual development aid programs are evaluated. For example, a recent evaluation of over hundred randomized controlled trials has shown that unconditional cash transfer
“generate strong and positive average treatment effects”
such that those that received cash had significantly higher incomes, food security, psychological, well-being. They owned more and could afford more. Their children were going to school more and were less likely to die before they reached 5 years old. These studies also find that there was no evidence that people who received cash became lazy, worked less, or spent the money on unproductive things like booze. In fact, there is now so much experimental evidence for the effectiveness of giving out cash, that it is quickly becoming the default benchmark for doing development aid. For example, in a recent randomized controlled trial, Craig McIntosh & Andrew Zeitlin found that a nutrition program in Rwanda did not improve core child outcomes, while just giving the aid recipients $500 did improve child nutrition and growth.
However, up to now, these studies could not tell us anything about macroeconomics. After all, if we want to study problems on an economy-wide scale, we’d need thousands of participants, and millions if not billions of Dollars. Money that scruffy university researchers simply do not have.
This is where GiveDirectly came in, a charity that, thanks to generous donations, had just over 10 million Dollars that it could distribute. In a rich country, you cannot perform a proper macroeconomic experiment with 10 million Dollars.
But, in this poor area, with average household income of just above 1000 US Dollars per year, this program represented a proper macroeconomic experiment, given that the 10M US dollars represented roughly 15% of the areas GDP. On top of that, the study area is a relatively closed-off economy in which most people travel to weekly markets by foot and mainly consume locally produced goods.
So, how did Dr. Egger and his colleagues Johannes Haushofer, Ted Miguel, Paul Niehaus, and Michael Walker, do a proper randomized controlled trial on macroeconomic scale?
Well, in this case, the treatment was a so-called unconditional cash transfer worth roughly 1000 USD, paid in three installments by GiveDirectly, in the form of a local currency payment directly to the phone of the head of households, that were sufficiently poor to be eligible.
But, how were eligible households randomly assigned to either the intervention or control group? In my recent interview with him, Dr. Egger told me that randomization took place on two levels, where
we start randomizing at the Village level. We had 700 Villages some Villages get money all of them all of the eligible people in a treated Village get money and in the control Village they don’t. And what we did on top of that is we had a second layer of randomization so in some regions 2 third of villages were randomly chosen to receive uh money in other regions only one third of villages was treated.
This is how the researchers ended up with this map, which shows untreated and control group villages. It also shows that in regions like this, a lot of villages were treated, while in regions like this only a few villages were treated.
This first layer of randomization is basic for any randomized controlled trial. These groups should on average be the same thanks to the random selection. These villages all feel the effects of Kenya’s broad economic trends like its economic growth and inflation or local trends like weather events that impact farming. However, due to the nature of macroeconomics, some money given to treatment villages will automatically spill-over to non-treated villages. Therefore, unlike in a medicine experiment, non-treated subjects will also benefit from the treatment. To calculate how big the effect is, the researchers applied the second layer of randomization, where some areas have more treated villages than others. In high treatment areas, the researchers expected to see bigger spill-over effects, such as more economic growth & inflation than in low treatment areas.
But, was that actually the case?
Thanks to a host of surveys asked to individuals, firms, and market participants before, during and after the experiment, the economists were in the unique position of having extremely detailed information about all prices and transactions in the area over the course of two-and-a-half years.
So, what happened to this economy? Did GiveDirectly’s cash injection cause inflation as Milton Friedman’s theory would suggest, and or did the economy expand by a multiple of the money introduced, as Keynes’s theory suggests?
Let’s start with the most obvious one.
What happened to inflation
Amongst many other things, Friedman theorized that inflation is typically caused by a rapid increase in the money supply. His reasoning is informed by the famous quantity theory of money, which states that, if the rate at which people spend money is roughly stable, and economic production does not increase, then blindly pumping a bunch of money into an economy will only raise prices.
For example, assume, just for the sake of simplicity, that in Kenya’s economy people go to the barber a lot. Barber shops are always full. But, now a charity gives the poorest in society a lot of money. Now, these people can go to the barber as well. It improves their lives for sure. But, because the barber shops were already full, now, not every customer can be helped. More demand than supply means that barbers will be able to raise their prices. Because getting a haircut dominates economic activity in this example, raising barbershop prices equals inflation.
So, because the barbershops were already full this money injection caused inflation. But, crucially, it did not cause people to be actually better off, on average. After all, the amount of haircuts given remains the same. Of course, in the long run, we might expect some unemployed people to become barbers, but this will take time.
Using a more advanced version of this logic, Friedman famously said about inflation that
it’s always and everywhere a result of too much money, of a more rapid increase in the quantity of money than in output.
In that same speech he also said that,
“There has never in history been an extremely rapid increase in the quantity of money without an inflation.”
Based on this logic, you’d expect that the result of the 10 million Dollars distributed by GiveDirectly was either that people worked more hours or that it would cause inflation, or at least some combination of the two.
But, this is not what happened. Instead, Dr. Egger and his colleagues observed that the intervention caused hardly any inflation at all. Treated villages did not have higher inflation than untreated villages, and high intensity treatment areas had about the same inflation as low intensity areas.
So, then based on Friedman’s framework, the researchers expected that they’d see a big increase in hours worked. But, digging deeper into their data, Egger and his colleagues were surprised to discover that while working hours of both recipients and non-recipients only increased slightly, the effect that was so small that it was not statistically significant.
So, what happened, was Friedman wrong?
Not entirely. To find out what was going on, Egger and his colleagues dove deep into their data and observed that in Kenya, by far most firms are operated by a single person. In these types of firms people officially work the whole day. But, in practice, they don’t have customers during much of the day.
So, we need to change our barber example. In this case, the barbers did not have enough customers.
In econ-speak, I would say, this economy was constrained by demand. Or, there was ‘slack’ in the economy.
So, if now again a charity comes in to drop some cash, again demand will increase. But, prices will not increase since these barbers had so few customers that they could easily absorb the extra customers without raising their prices.
So, as in the barber shop example, GiveDirectly’s big money injection did not cause inflation because the extra demand it generated was immediately matched by extra supply by small entrepreneurs who did not have enough customers.
Therefore, I would argue that his statement about inflation that
“There has never in history been an extremely rapid increase in the quantity of money without an inflation.”
has now been proven wrong by this experiment. There was clearly a rapid increase in the supply of money. That being said, if we go back to Friedman’s statement that inflation
it’s always and everywhere a result of too much money, of a more rapid increase in the quantity of money than in output.
then this could still be true. After all, you could say in this case that there was no inflation because while the money supply increased rapidly, this was matched by an equally rapid increase in economic output, because there was so much slack in this economy.
So, now the question becomes: will that increase in economic output be equal to, or more than the increase in money? That is the essence of the second famous theory we this experiment put to the test, which is
The Keynesian multiplier
The multiplier concept was introduced by British economist John Maynard Keynes in his influential book “The General Theory of Employment, Interest, and Money” in 1936.
In his book Keynes theorizes that an initial injection of money into an economy by the government can lead to a larger overall increase in employment, income and spending as the money circulates. This is because when one person spends money, it becomes income for another person, who then spends a portion of it, and so on. However, because most people save a bit of their income every time, and because some people spend some of it in neighboring economies, some new money will continuously leak out of the economy, until all of it is spent.
Notably, because poor people tend to spend more of the money they get than rich people, Keynes predicted that
the multiplier is larger in a poor community
Therefore, according to this theory, the cash introduction by GiveDirectly should increase the local economy by more than the money introduced. In other words, the multiplier should be quite a bit bigger than 1.
By following up all transactions over two and a half years, and by comparing low to high treatment areas, Egger and his team calculated that overall GDP was boosted by just over $2.5 for each Dollar distributed by GiveDirectly. This multiplier is higher than most estimates for multipliers in richer economies such as Mexico, Brazil and the U.S.
So, I’d say that this experiment provides evidence to support Keynes’s theory that spending in poor economies gives you a lot of bang for your buck.
That means that cash transfers were much more effective than you would have guessed if you had just looked at the direct impacts on recipients.
This finding could have massive
Implications for development aid
which in previous decades has come under intense scrutiny for how of every dollar donated only a fraction ever reached the poor people for which it was intended.
In reaction to this scrutiny, a new, much simpler, more direct way of doing development aid was pioneered by groups like GiveDirectly. What if you just send money directly to the poorest people on the planet, on their phone. And, if they don’t have a phone, by giving them a phone.
Of course, as you can see here: there are still a few costs involved here, such as financial transaction costs, the costs of following up, and buying people a SIM card, and potentially a phone. But, using this approach people actually receive most of the money donated, and it is clear to the donor exactly where costs are being made.
This is great in itself. But, what’s even better is that the study we discussed today shows that Dollars that do reach their recipients, benefit the economy by roughly 2.5 times. This ads to the ever increasing pile of evidence that every Dollar you donate via an unconditional cash transfer program ends up benefiting people by much more than a Dollar donated to traditional aid programs.
This study and Givedirecly’s work in general got me really excited for two reasons. First, their program represents the best way to actually help really poor people that I’ve seen so far. Second, their work with researchers like Dr. Egger is helping us finally scientifically test some of macroeconomics biggest unsolved mysteries.
This is why, for this video, I’ve decided to work with GiveDirectly, directly. You see, I started this channel back in 2020 with the aim of providing more accessible information to people all over the world about economics, with the hope that they would then use this knowledge to actually improve their economies.
However, if an economy like that of Kenya is demand constrained, I think study has shown that you have to wonder if training sessions will even help, if people don’t have money to start a business and customers don’t have money to buy their products. This, could explain why so many randomized controlled trials have found that giving free stuff or training is likely not the most effective way to help. But giving cash has proven benefits for the recipient WITH spillovers for their neighbors.
Therefore, as we are nearing the holiday season in much of the rich world, I would like to urge you, on top of buying some gifts for your loved ones who possibly already have a lot, please consider supporting a new village in rural Kenya through GiveDirectly by using the YouTube fundraiser feature on the right of the screen or by going to GiveDirectly.org/macro
If you support GiveDirectly, your money will be delivered to [all 164 families in this village early next year. The recipients can themselves decide how they spend the money. Isiah says he’ll spend on school fees for his kids. Jenniffer plans to restart her fishing business she lost to a fire. Esther’s going to build a house and buy goats to rear.
After that, as can be seen in Egger’s paper or in my in-depth interview with him, Isiash, Jenniffer, Esther and their neighbors will see long term reductions in poverty and an improved quality of life. And, so will those who live outside their village thanks to spillovers and Keynes’s multiplier.
So, happy holidays everyone, and please donate to a village in Kenya via Givedirectly using the links below.