Excerpt from an article in Mother Jones magazine, October 2017
Until we figure out how to fairly distribute the fruits of robot labor, it will be an era of mass joblessness and mass poverty. Working-class job losses played a big role in the 2016 election, and if we don’t want a long succession of demagogues blustering their way into office because machines are taking away people’s livelihoods, this needs to change, and fast. Along with global warming, the transition to a workless future is the biggest challenge by far that progressive politics—not to mention all of humanity—faces. And yet it’s barely on our radar.
Last year at Christmas, I was over at my mother’s house and mentioned that I had recently read an article about Google Translate. It turns out that a few weeks previously, without telling anyone, Google had switched over to a new machine-learning algorithm. Almost overnight, the quality of its translations skyrocketed. I had noticed some improvement myself but had chalked it up to the usual incremental progress these kinds of things go through. I hadn’t realized it was due to a quantum leap in software.
But if Google’s translation algorithm was better, did that mean its voice recognition was better too? And its ability to answer queries? Hmm. How could we test that? We decided to open presents instead of cogitating over this.
But after that was over, the subject of erasers somehow came up. Which ones are best? Clear? Black? Traditional pink? Come to think of it, why are erasers traditionally pink? “I’ll ask Google!” I told everyone. So I pulled out my phone and said, “Why are erasers pink?” Half a second later, Google told me.
Not impressed? You should be. We all know that phones can recognize voices tolerably well these days. And we know they can find the nearest café or the trendiest recipe for coq au vin. But what about something entirely random? And not a simple who, where, or when question. This was a why question, and it wasn’t about why the singer Pink uses erasers or why erasers are jinxed. Google has to be smart enough to figure out in context that I said pink and that I’m asking about the historical reason for the color of erasers, not their health or the way they’re shaped. And it did. In less than a second. With nothing more than a cheap little microprocessor and a slow link to the internet.
(In case you’re curious, Google got the answer from Design*Sponge: “The eraser was originally produced by the Eberhard Faber Company…The erasers featured pumice, a volcanic ash from Italy that gave them their abrasive quality, along with their distinctive color and smell.”)
Still not impressed? When Watson famously won a round of Jeopardy! against the two best human players of all time, it needed a computer the size of a bedroom to answer questions like this. That was only seven years ago.
What do pink erasers have to do with the fact that we’re all going to be out of a job in a few decades?
Consider: Last October, an Uber trucking subsidiary named Otto delivered 2,000 cases of Budweiser 120 miles from Fort Collins, Colorado, to Colorado Springs—without a driver at the wheel. Within a few years, this technology will go from prototype to full production, and that means millions of truck drivers will be out of a job.
Automated trucking doesn’t rely on newfangled machines, like the powered looms and steam shovels that drove the Industrial Revolution of the 19th century. Instead, like Google’s ability to recognize spoken words and answer questions, self-driving trucks—and cars and buses and ships—rely primarily on software that mimics human intelligence. By now everyone’s heard the predictions that self-driving cars could lead to 5 million jobs being lost, but few people understand that once artificial-intelligence software is good enough to drive a car, it will be good enough to do a lot of other things too. It won’t be millions of people out of work; it will be tens of millions.
This is what we mean when we talk about “robots.” We’re talking about cognitive abilities, not the fact that they’re made of metal instead of flesh and powered by electricity instead of chicken nuggets.
In other words, the advances to focus on aren’t those in robotic engineering—though they are happening, too—but the way we’re hurtling toward artificial intelligence, or AI. While we’re nowhere near human-level AI yet, the progress of the past couple of decades has been stunning. After many years of nothing much happening, suddenly robots can play chess better than the best grandmaster. They can play Jeopardy! better than the best humans. They can drive cars around San Francisco—and they’re getting better at it every year. They can recognize faces well enough that Welsh police recently made the first-ever arrest in the United Kingdomusing facial recognition software. After years of plodding progress in voice recognition, Google announced earlier this year that it had reduced its word error rate from 8.5 percent to 4.9 percent in 10 months.
All of this is a sign that AI is improving exponentially, a product of both better computer hardware and software. Hardware has historically followed a growth curve called Moore’s law, in which power and efficiency double every couple of years, and recent improvements in software algorithms have been even more explosive. For a long time, these advances didn’t seem very impressive: Going from the brainpower of a bacterium to the brainpower of a nematode might technically represent an enormous leap, but on a practical level it doesn’t get us that much closer to true artificial intelligence. However, if you keep up the doubling for a while, eventually one of those doubling cycles takes you from the brainpower of a lizard (who cares?) to the brainpower of a mouse and then a monkey (wow!). Once that happens, human-level AI is just a short step away.
This can be hard to imagine, so here’s a chart that shows what an exponential doubling curve looks like, measured in petaflops (quadrillions of calculations per second). During the first 70 years of the digital era, computing power doubled every couple of years—and that produced steadily improving accounting software, airplane reservation systems, weather forecasts, Spotify, and the like. But on the scale of the human brain—usually estimated at 10 to 50 petaflops—it produced computing power so minuscule that you can’t see any change at all. Around 2025 we’ll finally start to see visible progress toward artificial intelligence. A decade later we’ll be up to about one-tenth the power of a human brain, and a decade after that we’ll have full human-level AI. It will seem like it happened overnight, but it’s really the result of a century of steady—but mostly imperceptible—progress.
Are we really this close to true AI? Here’s a yardstick to think about. Even with all this doubling going on, until recently computer scientists thought we were still years away from machines being able to win at the ancient game of Go, usually regarded as the most complex human game in existence. But last year, a computer beat a Korean grandmaster considered one of the best of all time, and earlier this year it beat the highest-ranked Go player in the world. Far from slowing down, progress in artificial intelligence is now outstripping even the wildest hopes of the most dedicated AI cheerleaders. Unfortunately, for those of us worried about robots taking away our jobs, these advances mean that mass unemployment is a lot closer than we feared—so close, in fact, that it may be starting already. But you’d never know that from the virtual silence about solutions in policy and political circles.
I’m hardly alone in thinking we’re on the verge of an AI Revolution. Many who work in the software industry—people like Bill Gates and Elon Musk—have been sounding the alarm for years. But their concerns are largely ignored by policymakers and, until recently, often ridiculed by writers tasked with interpreting technology or economics. So let’s take a look at some of the most common doubts of the AI skeptics.
#1: We’ll never get true AI because computing power won’t keep doubling forever. We’re going to hit the limits of physics before long. There are several pretty good reasons to dismiss this claim as a roadblock. To start, hardware designers will invent faster, more specialized chips. Google, for example, announced last spring that it had created a microchip called a Tensor Processing Unit, which it claimed was up to 30 times faster and 80 times more power efficient than an Intel processor for machine learning tasks. A huge array of those chips are now available to researchers who use Google’s cloud services. Other chips specialized for specific aspects of AI (image recognition, neural networking, language processing, etc.) either exist already or are certain to follow.
What’s more, this raw power is increasingly being harnessed in a manner similar to the way the human brain works. Your brain is not a single, superpowerful computing device. It’s made up of about 100 billion neurons working in parallel—i.e., all at the same time—to create human-level intelligence and consciousness. At the lowest level, neurons operate in parallel to create small clusters that perform semi-independent actions like responding to a specific environmental cue. At the next level, dozens of these clusters work together in each of about 100 “sub-brains”—distinct organs within the brain that perform specialized jobs such as speech, visual processing, and balance. Finally, all these sub-brains operate in parallel, and the resulting overall state is monitored and managed by executive functions that make sense of the world and provide us with our feeling that we have conscious control of our actions.
Modern computers also yoke lots of microprocessors together. As of 2017, the fastest computer in the world uses roughly 40,000 processors with 260 cores each. That’s more than 10 million processing cores running in parallel. Each one of these cores has less power than the Intel processor on your desktop, but the entire machine delivers about the same power as the human brain.
This doesn’t mean AI is here already. Far from it. This “massively parallel” architecture still presents enormous programming challenges, but as we get better at exploiting it we’re certain to make frequent breakthroughs in software performance. In other words, even if Moore’s law slows down or stops, the total power of everything put together—more use of custom microchips, more parallelism, more sophisticated software, and even the possibility of entirely new ways of doing computing—will almost certainly keep growing for many more years.
#2: Even if computing power keeps doubling, it has already been doubling for decades. You guys keep predicting full-on AI, but it never happens. It’s true that during the early years of computing there was a lot of naive optimism about how quickly we’d be able to build intelligent machines. But those rosy predictions died in the ’70s, as computer scientists came to realize that even the fastest mainframes of the day produced only about a billionth of the processing power of the human brain. It was a humbling realization, and the entire field has been almost painfully realistic about its progress ever since.
We’ve finally built computers with roughly the raw processing power of the human brain—although only at a cost of more than $100 million and with an internal architecture that may or may not work well for emulating the human mind. But in another 10 years, this level of power will likely be available for less than $1 million, and thousands of teams will be testing AI software on a platform that’s actually capable of competing with humans.
#3: Okay, maybe we will get full AI. But it only means that robots will act intelligent, not that they’ll really be intelligent. This is just a tedious philosophical debating point. For the purposes of employment, we don’t really care if a smart computer has a soul—or if it can feel love and pain and loyalty. We only care if it can act like a human being well enough to do anything we can do. When that day comes, we’ll all be out of jobs even if the computers taking our places aren’t “really” intelligent.
#4: Fine. But waves of automation—steam engines, electricity, computers—always lead to predictions of mass unemployment. Instead they just make us more efficient. The AI Revolution will be no different. This is a popular argument. It’s also catastrophically wrong.
The Industrial Revolution was all about mechanical power: Trains were more powerful than horses, and mechanical looms were more efficient than human muscle. At first, this did put people out of work: Those loom-smashing weavers in Yorkshire—the original Luddites—really did lose their livelihoods. This caused massive social upheaval for decades until the entire economy adapted to the machine age. When that finally happened, there were as many jobs tending the new machines as there used to be doing manual labor. The eventual result was a huge increase in productivity: A single person could churn out a lot more cloth than she could before. In the end, not only were as many people still employed, but they were employed at jobs tending machines that produced vastly more wealth than anyone had thought possible 100 years before. Once labor unions began demanding a piece of this pie, everyone benefited.
In fact, it’s even worse. In addition to doing our jobs at least as well as we do them, intelligent robots will be cheaper, faster, and far more reliable than humans. And they can work 168 hours a week, not just 40. No capitalist in her right mind would continue to employ humans. They’re expensive, they show up late, they complain whenever something changes, and they spend half their time gossiping. Let’s face it: We humans make lousy laborers.
If you want to look at this through a utopian lens, the AI Revolution has the potential to free humanity forever from drudgery. In the best-case scenario, a combination of intelligent robots and green energy will provide everyone on Earth with everything they need. But just as the Industrial Revolution caused a lot of short-term pain, so will intelligent robots. While we’re on the road to our Star Trek future, but before we finally get there, the rich are going to get richer—because they own the robots—and the rest of us are going to get poorer because we’ll be out of jobs. Unless we figure out what we’re going to do about that, the misery of workers over the next few decades will be far worse than anything the Industrial Revolution produced.
Wait, wait, skeptics will say: If all this is happening as we speak, why aren’t people losing their jobs already? Several sharp observers have made this point, including James Surowiecki in a recent issue of Wired. “If automation were, in fact, transforming the US economy,” he wrote, “two things would be true: Aggregate productivity would be rising sharply, and jobs would be harder to come by than in the past.” But neither is happening. Productivity has actually stalled since 2000 and jobs have gotten steadily more plentiful ever since the Great Recession ended. Surowiecki also points out that job churn is low, average job tenure hasn’t changed much in decades, and wages are rising—though he admits that wage increases are “meager by historical standards.”
True enough. But as I wrote four years ago, since 2000 the share of the population that’s employed has decreased; middle-class wages have flattened; corporations have stockpiled more cash and invested less in new products and new factories; and as a result of all this, labor’s share of national income has declined. All those trends are consistent with job losses to old-school automation, and as automation evolves into AI, they are likely to accelerate.
That said, the evidence that AI is currently affecting jobs is hard to assess, for one big and obvious reason: We don’t have AI yet, so of course we’re not losing jobs to it. For now, we’re seeing only a few glimmers of smarter automation, but nothing even close to true AI.
Remember that artificial intelligence progresses in exponential time. This means that even as computer power doubles from a trillionth of a human brain’s power to a billionth and then a millionth, it has little effect on the level of employment. Then, in the relative blink of an eye, the final few doublings take place and robots go from having a thousandth of human brainpower to full human-level intelligence. Don’t get fooled by the fact that nothing much has happened yet. In another 10 years or so, it will.
So let’s talk about which jobs are in danger first. Economists generally break employment into cognitive versus physical jobs and routine versus nonroutine jobs. This gives us four basic categories of work:
Routine physical: digging ditches, driving trucks
Routine cognitive: accounts-payable clerk, telephone sales
Nonroutine physical: short-order cook, home health aide
Nonroutine cognitive: teacher, doctor, CEO
Routine tasks will be the first to go—and thanks to advances in robotics engineering, both physical and cognitive tasks will be affected. In a recent paper, a team from Oxford and Yale surveyed a large number of machine-learning researchers to produce a “wisdom of crowds” estimate of when computers would be able to take over various human jobs. Two-thirds said progress in machine learning had accelerated in recent years, with Asian researchers even more optimistic than North American researchers about the advent of full AI within 40 years.
But we don’t need full AI for everything. The machine-learning researchers estimate that speech transcribers, translators, commercial drivers, retail sales, and similar jobs could be fully automated during the 2020s. Within a decade after that, all routine jobs could be gone.
Nonroutine jobs will be next: surgeons, novelists, construction workers, police officers, and so forth. These jobs could all be fully automated during the 2040s. By 2060, AI will be capable of performing any task currently done by humans. This doesn’t mean that literally every human being on the planet will be jobless by then—in fact, the researchers suggest it could take another century before that happens—but that’s hardly any solace. By 2060 or thereabouts, we’ll have AI that can do anything a normal human can do, which means that nearly all normal jobs will be gone. And normal jobs are what almost all of us have.
2060 seems a long way off, but if the Oxford-Yale survey is right, we’ll face an employment apocalypse far sooner than that: the disappearance of routine work of all kinds by the mid-2030s. That represents nearly half the US labor force. The consulting firm PricewaterhouseCoopers recently released a study saying much the same. It predicts that 38 percent of all jobs in the United States are “at high risk of automation” by the early 2030s, most of them in routine occupations. In the even nearer term, the World Economic Forum predicts that the rich world will lose 5 million jobs to robots by 2020, while a group of AI experts, writing in Scientific American, figures that 40 percent of the 500 biggest companies will vanish within a decade.
Not scared yet? Kai-Fu Lee, a former Microsoft and Google executive who is now a prominent investor in Chinese AI startups, thinks artificial intelligence “will probably replace 50 percent of human jobs.” When? Within 10 years. Ten years! Maybe it’s time to really start thinking hard about AI.
And forget about putting the genie back in the bottle. AI is coming whether we like it or not. The rewards are just too great. Even if America did somehow stop AI research, it would only mean that the Chinese or the French or the Brazilians would get there first. Russian President Vladimir Putin agrees. “Artificial intelligence is the future, not only for Russia but for all humankind,” he announced in September. “Whoever becomes the leader in this sphere will become the ruler of the world.” There’s just no way around it: For the vast majority of jobs, work as we know it will come steadily to an end between about 2025 and 2060.
So who benefits? The answer is obvious: the owners of capital, who will control most of the robots. Who suffers? That’s obvious too: the rest of us, who currently trade work for money. No work means no money.
But things won’t actually be quite that grim. After all, fully automated farms and factories will produce much cheaper goods, and competition will then force down prices. Basic material comfort will be cheap as dirt.
Still not free, though. And capitalists can only make money if they have someone to sell their goods to. This means that even the business class will eventually realize that ubiquitous automation doesn’t really benefit them after all. They need customers with money if they want to be rich themselves.
One way or another, then, the answer to the mass unemployment of the AI Revolution has to involve some kind of sweeping redistribution of income that decouples it from work. Or a total rethinking of what “work” is. Or a total rethinking of what wealth is.
Let’s consider a few of the possibilities.
The welfare state writ large: This is the simplest to think about. It’s basically what we have now, but more extensive. Unemployment insurance will be more generous and come with no time limits. National health care will be free for all. Anyone without a job will qualify for some basic amount of food and housing. Higher taxes will pay for it, but we’ll still operate under the assumption that gainful employment is expected from anyone able to work.
This is essentially the “bury our heads in the sand” option. We refuse to accept that work is truly going away, so we continue to punish people who aren’t employed. Jobless benefits remain stingy so that people are motivated to find work—even though there aren’t enough jobs to go around. We continue to believe that eventually the economy will find a new equilibrium.
This can’t last for too long, and millions will suffer during the years we continue to delude ourselves. But it will protect the rich for a while.
Universal basic income #1: This is a step further down the road. Everyone would qualify for a certain level of income from the state, but the level of guaranteed income would be fairly modest because we would still want people to work. Unemployment wouldn’t be as stigmatized as it is in today’s welfare state, but neither would widespread joblessness be truly accepted as a permanent fact of life. Some European countries are moving toward a welfare state with cash assistance for everyone.
Universal basic income #2: This is UBI on steroids. It’s available to everyone, and the income level is substantial enough to provide a satisfying standard of living. This is what we’ll most likely get once we accept that mass unemployment isn’t a sign of lazy workers and social decay, but the inevitable result of improving technology. Since there’s no personal stigma attached to joblessness and no special reason that the rich should reap all the rewards of artificial intelligence, there’s also no reason to keep the universal income level low. After all, we aren’t trying to prod people back into the workforce. In fact, the time will probably come when we actively want to do just the opposite: provide an income large enough to motivate people to leave the workforce and let robots do the job better.
Silicon Valley—perhaps unsurprisingly—is fast becoming a hotbed of UBI enthusiasm. Tech executives understand what’s coming, and that their own businesses risk a backlash unless we take care of its victims. Uber has shown an interest in UBI. Facebook CEO Mark Zuckerberg supports it. Ditto for Tesla CEOElon Musk and Slack CEO Stewart Butterfield. A startup incubator called Y Combinator is running a pilot program to find out what happens if you give people a guaranteed income.
There are even some countries that are now trying it. Switzerland rejected a UBI proposal in 2016, but Finland is experimenting with a small-scale UBI that pays the unemployed about $700 per month even after they find work. UBI is also getting limited tryouts by cities in Italy and Canada. Right now these are all pilot projects aimed at learning more about how to best run a UBI program and how well it works. But as large-scale job losses from automation start to become real, we should expect the idea to spread rapidly.
A tax on robots: This is a notion raised by a draft report to the European Parliament and endorsed by Bill Gates, who suggests that robots should pay income tax and payroll tax just like human workers. That would keep humans more competitive. Unfortunately, there’s a flaw here: The end result would be to artificially increase the cost of employing robots, and thus the cost of the goods they produce. Unless every country creates a similar tax, it accomplishes nothing except to push robot labor overseas. We’d be worse off than if we simply let the robots take our jobs in the first place. Nonetheless, a robot tax could still have value as a way of modestly slowing down job losses. Economist Robert Shiller suggests that we should consider “at least modest robot taxes during the transition to a different world of work.” And where would the money go? “Revenue could be targeted toward wage insurance,” he says. In other words, a UBI.
Socialization of the robot workforce: In this scenario, which would require a radical change in the US political climate, private ownership of intelligent robots would be forbidden. The market economy we have today would continue to exist with one exception: The government would own all intelligent robots and would auction off their services to private industry. The proceeds would be divided among everybody.
Progressive taxation on a grand scale: Let the robots take all the jobs, but tax all income at a flat 90 percent. The rich would still have an incentive to run businesses and earn more money, but for the most part labor would be considered a societal good, like infrastructure, not the product of individual initiative.
Wealth tax: Intelligent robots will be able to manufacture material goods and services cheaply, but there will still be scarcity. No matter how many robots you have, there’s only so much beachfront property in Southern California. There are only so many original Rembrandts. There are only so many penthouse suites. These kinds of things will be the only real wealth left, and the rich will still want them. So if robots make the rich even richer, they’ll bid up the price of these luxuries commensurately, and all that’s left is to tax them at high rates. The rich still get their toys, while the rest of us get everything we want except for a view of the sun setting over the Pacific Ocean.
A hundred years from now, all of this will be moot. Society will adapt in ways we can’t foresee, and we’ll all be far wealthier, safer, and more comfortable than we are today—assuming, of course, that the robots don’t kill us all, Skynet fashion.
But someone needs to be thinking hard about how to prepare for what happens in the meantime. Not many are. Last year, for example, the Obama White House released a 48-page report called “Preparing for the Future of Artificial Intelligence.” That sounds promising. But it devoted less than one page to economic impacts and concluded only that “policy questions raised by AI-driven automation are important but they are best addressed by a separate White House working group.”
Regrettably, the coming jobocalypse has so far remained the prophecy of a few Cassandras: mostly futurists, academics, and tech executives. For example, Eric Schmidt, chairman of Google’s parent company, believes that AI is coming faster than we think, and that we should provide jobs to everyone during the transition. “The country’s goal should be full employment all the time, and do whatever it takes,” he says.
Another sharp thinker about our jobless future is Martin Ford, author of Rise of the Robots. Mass joblessness, he warns, isn’t limited to low-skill workers. Nor is it something we can fight by committing to better education. AI will decimate any job that’s “predictable”—which means nearly all of them. Many of us might not like to hear this, but Ford is unsentimental about the work we do. “Relatively few people,” he says, are paid “primarily to engage in truly creative work or ‘blue sky’ thinking.”
So how do we get these ideas into the political mainstream? One thing is certain: The monumental task of dealing with the AI Revolution will be almost entirely up to the political left. After all, when the automation of human labor begins in earnest, the big winners are initially going to be corporations and the rich. Because of this, conservatives will be motivated to see every labor displacement as a one-off event, just as they currently view every drought, every wildfire, and every hurricane as a one-off event. They refuse to see that global warming is behind changing weather patterns because dealing with climate change requires environmental regulations that are bad for business and bad for the rich. Likewise, dealing with an AI Revolution will require new ways of distributing wealth. In the long run this will be good even for the rich, but in the short term it’s a pretty scary prospect for those with money—and one they’ll fight zealously. Until they have no choice left, conservatives are simply not going to admit this is happening, let alone think about how to address it. It’s not in their DNA.
Other candidates are equally unlikely. The military thinks about automation all the time—but primarily as a means of killing people more efficiently, not as an economic threat. The business community is a slave to quarterly earnings and in any case will be too divided to be of much help. Labor unions have good reason to care, but by themselves they’re too weak nowadays to have the necessary clout with policymakers.
Nor are we likely to get much help from governments, which mostly don’t even understand what’s happening. Google’s Schmidt puts it bluntly. “The gap between the government, in terms of their understanding of software, let alone AI, is so large that it’s almost hopeless,” he said at a conference earlier this year. Certainly that’s true of the Trump administration. Asked about AI being a threat to jobs, Treasury Secretary Steven Mnuchin stunningly waved it off as a problem that’s still 50 or 100 years in the future. “I think we’re, like, so far away from that,” he said. “Not even on my radar screen.” This drew a sharp rebuke from former Treasury Secretary Larry Summers: “I do not understand how anyone could reach the conclusion that all the action with technology is half a century away,” he said. “Artificial intelligence is transforming everything from retailing to banking to the provision of medical care.”
So who’s left? Like it or not, the only real choice to sound the alarm outside the geek community is the Democratic Party, along with its associated constellation of labor unions, think tanks, and activists. Imperfect as it is—and its reliance on rich donors makes it conspicuously imperfect—it’s the only national organization that has both the principles and the size to do the job.
Unfortunately, political parties are inherently short-term thinkers. Democrats today are absorbed with fighting President Donald Trump, saving Obamacare, pushing for a $15 minimum wage—and arguing about all those things. They have no time to think hard about the end of work.
Nonetheless, somebody on the left with numbers, clout, power, and organizing energy—hopefully all the above—had better start. Conventional wisdom says Trump’s victory last year was tipped over the edge by a backlash among working-class voters in the Upper Midwest. When blue-collar workers start losing their jobs in large numbers, we’ll see a backlash that makes 2016 look like a gentle breeze. Either liberals start working on answers now, or we risk voters rallying around far more effective and dangerous demagogues than Trump.
Despite the amount of media attention that both robots and AI have gotten over the past few years, it’s difficult to get people to take them seriously. But start to pay attention and you see the signs: An Uber car can drive itself. A computer can write simple sports stories. SoftBank’s Pepper robot already works in more than 140 cellphone stores in Japan and is starting to get tryouts in America too. Alexa can order replacement Pop-Tarts before you know you need them. A Carnegie Mellon computer that seems to have figured out human bluffing beat four different online-poker pros earlier this year. California, suffering from a lack of Mexican workers, is ground zero for the development of robotic crop pickers. Sony is promising a robot that will form an emotional bond with its owner.
These are all harbingers, the way a dropping barometer signals a coming storm—not the possibility of a storm, but the inexorable reality. The two most important problems facing the human race right now are the need for widespread deployment of renewable energy and figuring out how to deal with the end of work. Everything else pales in comparison. Renewable energy already gets plenty of attention, even if half the country still denies that we really need it. It’s time for the end of work to start getting the same attention.
Kevin Drum is a political blogger for Mother Jones. Email Kevin firstname.lastname@example.org.
We Already Have a Solution for the Robot Apocalypse. It’s 200 Years Old.
From the window of his university office in Louvain-la-Neuve, Belgium, philosophy professor Philippe Van Parijs—considered by many to be Europe’s most prominent advocate for the idea that the state should provide a regular income to every citizen—can see the mailbox where he sent off invitations to the first “basic income” conference more than 30 years ago. “I’m quite amazed by the seed we threw on the ground now,” he says.
After decades of obscurity, the idea is suddenly in fashion. Politicians around the world are interested and a handful of governments, such as Finland and the Canadian province of Ontario, are planning or considering basic-income pilot projects.
But the idea of basic income has been around for more than 200 years, rising on waves of political and economic turmoil only to disappear in calmer times. Here are some of the highlights of its long, turbulent history:
1797: Thomas Paine proposes taxing landowners to provide a £10 annual stipend for people over 50 and a one-time £15 payment at 21 years of age.
1848: Belgian Joseph Charlier writes the first fully fledged proposal for basic income. It is virtually unknown until two academics stumble upon it 150 years later.
1918: After World War I, English Quakers call for a weekly “state bonus” for all citizens—a proposal ultimately rejected by the Labour Party.
1934: Amid the Great Depression, Louisiana Sen. Huey Long proposes confiscating wealth from the rich to provide guaranteed income for all families. The “Share Our Wealth” movement is cut short by Long’s assassination in 1935. That year, President Franklin D. Roosevelt creates Aid to Families With Dependent Children—the beginnings of “welfare.”
1940s: Conservative economists Milton Friedman and George Stigler propose a “negative income tax” (NIT): low-income households would receive government payments, and as earnings increased, so would the tax burden. Friedman believed an NIT would address poverty without adding to the state bureaucracy he reviled.
1962-63: Amid the Great Migration of blacks to the North, critic Dwight MacDonald argues for guaranteed income in an influential New Yorker article. Friedman presses for an NIT in Capitalism and Freedom, and liberal economist Robert Theobald floats a “Basic Economic Security Plan.”
1966: President Lyndon Johnson’s economic advisers say an NIT “would be the most direct approach to reducing poverty.” By 1968, a surprising cast of characters, including Martin Luther King Jr. and CEOs, support the idea. Some 1,000 economists sign a statement advocating a “national system of income guarantees and supplements.”
1974: A tiny Canadian town provides residents an annual income of $15,000 for a family of four. The data is forgotten until 1,800 dusty boxes are found in a Winnipeg warehouse 40 years later.
1976: As the Trans-Alaska Pipeline nears completion, Alaskans approve a measure to pay dividends to all residents. Commencing in 1982, and paying an average of $1,150 a year to eligible residents ever since, the state oil fund is the first basic-income system in the United States.
1978: US government NIT trials involving 4,800 families in two cities find a small reduction in “willingness” to work but a large increase of divorce in two cohorts. The divorce finding is later disputed, but the damage is done; Moynihan renounces guaranteed income. But by then, Nixon has pushed through Supplementary Security Income (for the old and disabled) and the Earned Income Tax Credit (an NIT for the poor).
1997: Mexico launches a program of conditional cash transfer to poor households. Brazil and Colombia follow suit. While CCTs come with requirements, they assume people will spend grants wisely. CCTs spread rapidly across Latin America and parts of Asia and Africa. Tens of millions of people worldwide now receive aid through CCTs funded by governments, aid organizations, and nonprofits.
2013: Basic-income experiments begin in rural India, involving more than 6,000 individuals. (The research team wanted to test the hypothesis that people are generally capable of making their own decisions and do so in the best interests of themselves, their children and their families, rather than spending it on private bads such as alcohol. While the team firmly believed that this hypothesis would hold true, that it in fact did was one the strongest findings of the study which resonated with top policymakers. the National Electronics Funds Transfer (NEFT) system that ensured real time transfer of money to individual accounts worked very well: while initial difficulties such as errors in account numbers resulted in 12% reversals in the first month, reversals went down to 0.5% by the fourth month and to 0% by the end of the project. The tribal pilot, in comparison, was more streamlined as payments were made in cash every month to all residents in the treatment village. Financial inclusion was rapid and near universal. Opening individual bank accounts was done intensively and within four months of the start of the general pilot, 95.6% of individuals had bank (or co-operative) accounts. For the remaining 4%, accounts were opened in the next three months. The basic living conditions in basic income villages improved starting with improvement in sanitation in villages covered by the general pilot. Some of the basic income was invested to get better access to drinking water, especially in tribal villages. Cooking and lighting energy sources also improved.
Many households in the general pilot used their basic income payments to change or improve their energy or lighting sources. According to the FES, 24.3% of basic income households covered under the general pilot had changed their main source of energy for cooking or lighting in some way in the previous 12 months, compared to just 10.6% in the control villages, with the difference being highly significant statistically. The tribal village too reported changes: 16% of households in the recipient village reported using a better cooking fuel and 14.5% reported improving their lighting, compared with practically no change in the control village.
The tribal villages, which were much poorer than the general villages, recorded significant increases in ownership of household assets. Some of the basic income money was spent by recipients on buying household assets in the general pilot, but it was not much. Households were more likely to buy productive assets to earn more income, rather than assets that would give them more comfort. However in the tribal villages families purchased all types of assets over the course of the project, but families receiving basic incomes were more likely to purchase them. For instance, transport is an important need for tribal families, given the remote location of both villages, particularly the basic income recipient village. So more families in the BI village purchased bicycles. In total, about 13 bicycles were purchased in the recipient village in comparison to only two in the control village. Further, in the basic income tribal village nearly 27% of households purchased a total of 32 scooters and motorcycles, whereas only two new two-wheel motor vehicles were purchased in the control village. In both the general and the tribal pilot, those who received basic income reported a statistically significant increase in their food sufficiency six months into the intervention. Receipt of basic income had a statistically significant impact on children’s nutrition, in both general and tribal villages, particularly on nutrition levels of female children. Disaggregating the weight-for-age scores by social groups, transfers under the general pilot were found to be progressively benefiting, in that children from ST households recorded the greatest improvements and the least improvements were recorded for children from general category households. In contrast, in the tribal pilot, while improvements in nutrition levels were recorded for basic income recipient households, the difference between them and the improvements recorded by control households was not statistically significant. This could be on account of the high levels of malnourishment in these villages, before the start of the project. So even though some improvements were observed, they were not enough to show up in a significant rise in numbers of ‘normal’ weight-for-age children.
Basic income improved capacity of households to buy from the market, resulting in a qualitative shift in their food basket; but more money did not result in more expenditure on alcohol. Households receiving the basic income reported a higher propensity to consume fresh vegetables and milk. Their ability to do so was more pronounced in the tribal pilot where basic income beneficiaries reported a substantial rise in consumption of more nutritious food like pulses, vegetables, eggs, fruits, fish and meat. No evidence was found of an increase in spending on alcohol, either in the general villages or the tribal pilot. If anything, when asked whether they were buying more or less of specific food items, a slightly higher proportion of households in basic income villages in both sets of pilots said they were buying less alcohol than before. Regular, basic income payments facilitated a more rational or considered response to illness, through more regular medication, and for some households, through more intake of food. While the period of the pilots was too short to expect any observable effects on health, interestingly enough households receiving the basic income reported a lower incidence of illness at the end of the intervention than those that had not been receiving them in both general and tribal villages. The difference was more striking in the tribal pilot: while households in the control village were more likely to report an incidence of illness (70% had at least one person ill in the three months before the end of the transfer), a lower proportion in the basic income village (about 58%) reported an illness in that period. A majority of basic income recipients in both pilots perceived an improvement in their health and attributed it to basic income payments. When asked how the transfers had helped, most in the general pilot agreed that the basic income had enabled them to buy medicines (66%); some spoke of having food more regularly (27%); while Majority of individuals receiving basic income in the general pilot perceived an improvement in their health and attributed it to receipt of the cash transfer. Of those who perceived an improvement, 66% said the improvement was because they could afford to take medicines more regularly. 12 some said that the payment had helped improve their health by reducing anxiety levels (16%). Interestingly, Scheduled Tribe respondents put more weight on regular food intake as a reason for a perceived improvement in their health, relative to other groups, emphasizing the importance of food sufficiency for this vulnerable group. Basic incomes also afforded families more choice in the type of health service to use and in the timing of seeking health care. Over the course of the pilots, the use of government hospitals as a first port of call when ill declined in the general basic income recipient villages slightly and the use of private doctors and hospitals increased.
Basic income payments reduced the burden of households to fund their health expenses through a vicious cycle of debt. Borrowing for hospitalization expenses was lower in basic income villages by the end of the general pilot (at 46%) compared to control villages (55%), with the difference being statistically significant. Instead, more cash recipient households said they had used their own income/savings to pay for hospitalization. What was even more encouraging was that SC and ST households in the general pilot tended to rely less on loans than their counterparts in control villages. So while around 64% of SC households and 68% of ST households with an incidence of illness in control villages had used loans and/or mortgaged their assets to fund hospitalization expenses, in basic income villages 52% of SC respondents and 46% of ST respondents did so. Consistent with the findings from the general pilot, BI recipients in the tribal pilot borrowed less on interest than households in the control village: some 50% borrowed to fund hospital treatment in the former, compared to 58% in the latter. Unconditional basic income payments had a salutary impact on enrollment levels, particularly that of girls, and more so girls in villages where SEWA was present. One of the strongest findings of the pilot was the ability of basic income to check the tendency of households pulling girls out from schools. While only 36% of girls of secondary school going age were enrolled in schools in the control villages in the general pilot, nearly 66% of girls of the same age cohort were going to school in basic income villages by the end of the intervention (Table 2). Interestingly, enrollment levels, more so for girls, were highest in basic income villages with a SEWA presence. In the tribal pilot, the basic income arrested the tendency of children dropping out from schools. So, while a 17-percentage point decline was observed in school enrolment in the control village, only a 3-percentage drop was seen in enrollment levels in the basic income village over the course of the tribal pilot. These correlations are encouraging in that they testify to a positive effect of the basic income on school enrollment, which importantly arises without imposing any conditionality.
Receipt of basic income also facilitated an increase in school spending – on items such as uniforms, shoes, and books in both pilots. Total expenditure by families on schooling as well as on different school objects was higher in basic income villages by the end of the general pilot. While no statistically significant differences were seen in villages where SEWA was not operative, households residing in villages with a SEWA presence and receiving the basic income spent nearly 82% more on sending their children to schools compared to households in control villages, with a SEWA presence. Further, and in what was a heartening trend, expenditure on schooling of girls was decidedly higher among households receiving the basic income in the general pilot, more so among households in SEWA villages. A similar development was observed in the tribal pilot.
Along with an increase in schooling, the basic income had a positive effect on waged child labour, especially in SEWA villages under the general pilot. There was a 20% reduction in child wage-labour in the general basic income villages compared to a 5% drop in control villages, with the difference being statistically significant. In the tribal pilot there was an interesting paradox as child labour for wages reduced and labour for 14 own-account work increased. Children in Ghoda Khurd (the basic income village) were more likely to work than those in Bhilami (control village). But their work was less likely to affect their schooling. So, 36% of children in Ghoda Khurd worked as opposed to 26% in Bhilami, but only 16% said their schooling was adversely affected, as opposed to 37% in the control village. The basic income did not reduce the level of migration but it did change the pattern, especially in the tribal villages.
One of the most important findings was the growth of productive work in both general and tribal villages, leading to a sustained increase in income. Nearly 21% of basic income recipient households in the general pilot reported an increase in incomeearning work or production, compared with just 9% of the control households (figure 5). The transfers also seemed to be progressive. More SC households receiving the basic income reported an increase in economic activity (19.4%), whereas only 7.2% of SC households in control villages said they had experienced an increase.
In the tribal villages, perhaps the biggest impact of the project was to enable small farmers to spend more time and also invest on their own farms as opposed to working as wage labourers. The monthly cash transfer ensured that daily expenses such as those on food could be met by tribal families, thereby allowing them with some extra funds to buy seeds and fertilizers. Figure 6 below shows the shift in how people reported what their primary occupation was in the tribal pilot baseline and then in the FES or the endline. Whereas in the baseline, less than 40% of households in the tribal cash transfer village said they were farmers, by the end of 12 months, this number had risen to over 62%. Conversely, only 35% of control village households said that they were farmers by the end of the project, the rest earning their living as labourers.
Multivariate analysis using data from the general pilot suggested that receipt of the basic income was strongly associated with diversification into a second activity combined with a primary one.
For the general villages, the multivariate analysis also revealed a positive and significant effect of basic income on the number of hours worked. Households receiving basic income under the general pilot had nearly 32% higher odds of working more hours than households not receiving the payment. Women too appeared to have had higher odds of putting in more hours in their main and secondary activity than men. Similar results were obtained in the tribal pilot: individuals in the village receiving the basic income significantly increased their days of work, whereas no change was seen in the control village. In fact, in the former village, by the end of the pilot, around 52% individuals reported getting 11-20 days of work in a month (up from 43.5% at the start of the pilot). In contrast, the percentage reporting getting that amount of work fell in the control village over the course of the pilot.
One of the reasons for increased income and productivity was the increase in productive assets, especially in the tribal village. In the general pilot, households that received the basic income used it to buy productive assets. There were increases especially in ownership of sewing machines and tube wells.
In the tribal pilot, however, there was a major increase in livestock in the cash transfer village, which had implications for economic activity and household income. In the said tribal village (i.e. Ghoda Khurd), small livestock increased from 424 to 633 in number and large livestock increased from 259 to 323 between the baseline and FES. During the same period, in Bhilami (the control tribal village), small livestock decreased in number from 466 to 355 and big livestock decreased from 207 to 190. Households in Ghoda Khurd also reported a statistically significant increase in wells and ploughs, by 34% and 48% respectively vis-à-vis a 13% and 9% increase in the control village. Women’s empowerment was one of the more important outcomes of this experiment; most women receiving the basic income thought they could participate in decisions on spending their basic income. In other words, the basic income appeared to have made household decision-making more equitable than before. In the general pilot, 54% of women in basic income villages reported that household income was shared equally, compared to 39% women in control villages. This was also true for decision-making dynamics in the tribal pilot. The change within the basic income households as compared to the control households was highly significant statistically. Individual accounts and individual transfers strengthened women’s control over finances.
Women and girls also benefited disproportionately from the basic income in terms of nutrition, health and education outcomes. As discussed earlier, the z-score index on nutrition suggested that girls experienced a greater drop in malnutrition than boys of the same age group in the general village pilot. There was also some evidence that girls gained parity in diets and as a result gained in relative terms. Into adulthood, there was evidence that women in general – and disabled women in particular – gained relatively more in terms of access to food and in their dietary balance. Female students benefitted more than boys with the secondary school enrolment going up among girls in families receiving the transfer under general pilot. In the tribal BI village, impacts were seen on women’s healthcare: more tribal women in the BI village (Ghoda Khurd) accessed health facilities and took medicines regularly than in the control village.
Women who received the basic income increased their labour and work relative to women who did not, particularly in the tribal village where women’s labour force participation increased by 16%, while it scarcely changed for men. One reason for this was the shift to own account work, which was particularly significant in the tribal village where the share of women doing it rose from 40% to 60%, while in the control village it actually shrank. Another reason for the shift was that small-scale and marginal farmers in the tribal village were able to farm their land. The share of women in the tribal BI village, whose primary activity was farming almost doubled, rising from 39% to 66%. There was a 6% increase in BI recipient households owning assets such as sewing machines, whereas the number fell among the control group. Similarly, assets such as livestock were also bought which had implications for household income and women’s work. Basic income had a direct impact on indebtedness of households. Households receiving the basic income in the general pilot villages were less likely to have increased their debt, six months into the intervention, and were in fact more likely to have reduced it, with the difference between them and households in control villages being statistically significant (figure 7). In the tribal pilot, while in the baseline both the control and transfer village had two-thirds of households in debt of some form or the other, in the latter, after six months, 18% of the households reported that their debt had reduced. After 12 months, 73% of basic income recipient households reported that their debt had reduced.
Basic income enabled households to shift away from harsher forms of borrowings to more benign forms. Figure 8 shows that during the most serious shocks, households in the general pilot villages usually depended on moneylenders, followed by relatives and then friends and neighbors. However, when we compare households in the BI villages with those in control villages, the latter were more dependent on moneylenders. In the BI villages, in contrast, reliance on relatives was much greater.
The basic income also enabled households living in the general pilot villages, access many government schemes. At the start of the general pilot, an assessment suggested that there were as many as 321 government schemes in the 20 villages covered by the general pilot that were aimed at addressing poverty and social protection. Most of them were targeted schemes with different types of conditions. The basic income helped households in these villages to obtain many of these schemes. For instance, the PFES examined access to 32 schemes in two cash transfer villages and found that having cash in hand allowed families in these villages to buy from the ration shops, take transport to government hospitals, open bank accounts etc.
Concluding Reflections The findings from the quantitative study combined with the qualitative case studies and focus group discussions led us to formulate some conclusions, which could show a way forward into policy directions.
Unconditional Cash Transfers are beneficial and the benefits build on one another. For one, our findings suggest that households use cash transfers wisely and do not dissipate it in wasteful ways like spending it on alcohol. This is even more important because the pilots did not impose any conditions. However, and crucially so, lack of conditions did not induce people to spend money in ways against their own interest. On the contrary they spent it on nutrition, health, education and productive assets among other things. This finding from the study removes one of the fears that is often voiced about cash transfers. Two, the unconditional nature of cash transfers meant that transfer became easy once a bank account was opened and recipients did not have to spend time and energy to get proof that they had fulfilled certain conditions, thereby increasing the take-up rate to over 98% of the households. Finally, the benefits of unconditional cash transfers usually built on one another, and therefore had a true emancipatory effect on households. For example, increased schooling increased schooling reduced child labour; productive assets increased income which increased access to nutrition; reduced debt freed up incomes for productive work etc. While this project was not intended as an attempt to compare conditional and unconditional cash transfers, data emerging from the pilots leave little doubt about the overall benefits of ‘unconditional’ cash transfers, including the ease of such transfer.
Universal financial inclusion is possible and desirable and cash transfers along with financial intermediation, further hasten the process. Coincidentally, while the pilots were taking place, there was an intense debate in the media on the need for financial inclusion. The pilots were able to demonstrate that a regular cash transfer, such as that in the project, led to rapid opening of bank accounts. They were also able to underscore the important role “financial intermediaries” can play. But underlying it all, was the empirical evidence that when given a reason, people do open accounts in financial institutions. Furthermore, they use these accounts not only to receive benefits, but also for savings and in some cases for accessing loans.
Deepening financial services requires doorstep banking and a better system of banking correspondents. Financial inclusion means more than just opening a bank account; it requires strengthening people’s capacity to actually operate that bank account, to save, borrow and undertake financial planning. Since the bank branches are far from villages and understaffed, doorstep banking is the only solution. Other than the banks, there are many financial institutions such as co-operatives, SHG federations, micro-finance agencies that do provide doorstep banking. This experiment demonstrates that using such institutions can facilitate more genuine financial inclusion. The Banks and the Reserve Bank of India have been promoting a system of “Banking Correspondents” (BCs) all over the country. Unfortunately, we found this system to be more or less non-operational, mainly because the BCs could not earn even a modest living from it, and because they did not get the full co-operation of banks.
Individual accounts and individual transfers lead to empowerment of the more vulnerable sections of people. During the course of the pilot there was internal debate in the project team as to whether the transfers should be individual or household based. Eventually, the project decided on paying money into while accounts, after extensive consultation, especially with villagers. The findings of the study further reaffirmed the need to do so: individual transfers in fact gave more control of money to vulnerable sections especially women, disabled and the elderly.
The involvement of a voice agency helps the basic income to work optimally. One of the unique features of this project has been that the pilots have been designed to try to identify the impact of both the basic income and independent voice, the underlying hypothesis being that the strength of positive effects of the basic income would be augmented by the existence of a collective body able to assist, advise and support vulnerable recipients. These expectations were borne out by the data in most respects, if not all. In fact, the data showed that basic income linked with SEWA activities produced better results vis-à-vis families using health and education services. Also being a member of SEWA tended to make households less averse to taking risks. Recommendation 5: While the project recognizes that SEWA is a particular type of collective organisation, which has stronger effects on some issues than on others, it is reasonably confident in recommending involvement of a body such as SEWA, so as to enhance the impact of cash transfers, as well as smoothen the process of financial inclusion. The main role of such an organization should be to help in the education of recipients on how to acquire and manage money and in how to protect their new economic and social right that an unconditional basic income provides. This educative function is vitally important in communities where cash in people’s hands has been relatively scarce. In other words, financial emancipation, not simple inclusion, should be the goal.
Tribal communities can be game changers. The tribal pilot conducted under the aegis of this project has shown that basic income can have particularly strong transformative potential in tribal villages. Basic income or other cash transfer systems should be phased in before existing subsidy schemes are replaced or phased out.10 This approach would pay social dividends later, since it would mean that low-income families would not face the initial risk and potentially severe cost of not obtaining the cash transfers while losing access to subsidised goods. 11 From http://unicef.in/Uploads/Publications/Resources/pub_doc83.pdf)
2016: Switzerland becomes the first country to vote on, and reject, a national basic income. Opponents argued that it would undermine the Swiss economy. But for advocates, the mere fact of the referendum is remarkable. Meanwhile, as Silicon Valley bigwigs like Elon Musk make ever scarier pronouncements about the threat of AI, startup incubator Y Combinator announces a pilot project to distribute about $1,500 a month to 100 families in Oakland, California. And US-based nonprofit GiveDirectly plans a 12-year trial involving thousands of Kenyans.
2017: Pilot projects are initiated in Finland, the Netherlands, and three cities in Ontario, Canada, where low-income residents will be paid $14,000 a year whether they work or not. It’s time to be “bold,” says Ontario’s premier. And in September, Hillary Clinton reveals she had considered an “Alaska for America” basic-income proposal, funded by carbon and financial transaction taxes, but “couldn’t make the numbers work.”
2017: Stockton, California, announces that in order to offset the economic inequalities brought on by the tech boom, it will begin a municipal pilot program known as SEED, or the Stockton Economic-Empowerment Demonstration. Some city residents may receive as much as $500 a month—not enough for anyone to subsist on exactly, but enough to ease some financial burden.
Moving Forward on Basic Income
We have a few updates we want to share on our Basic Income Project:
Our Research Director Elizabeth Rhodes is joining Basic Income Project as our Research Director. She recently completed a joint PhD in Social Work and Political Science at the University of Michigan, where her research focused on health and education provision in slum communities in Nairobi. We received over 1000 applications for this position (including tenured professors from Oxford, Columbia, and Harvard), and Elizabeth stood out as the right candidate based on her aptitude and her ambition. We’re very excited to work with her.
Pilot Study in Oakland We want to run a large, long-term study to answer a few key questions: how people’s happiness, well-being, and financial health are affected by basic income, as well as how people might spend their time. But before we do that, we’re going to start with a short-term pilot in Oakland. Our goal will be to prepare for the longer-term study by working on our methods–how to pay people, how to collect data, how to randomly choose a sample, etc. Oakland is a city of great social and economic diversity, and it has both concentrated wealth and considerable inequality. We think these traits make it a very good place to explore how basic income could work for our pilot.
It’s also close to where we live, which means we’ll be closer to the people involved. We think our local resources and relationships will help us design and run this study effectively, and we hope that will enable us to produce the best research possible.
In our pilot, the income will be unconditional; we’re going to give it to participants for the duration of the study, no matter what. People will be able to volunteer, work, not work, move to another country—anything. We hope basic income promotes freedom, and we want to see how people experience that freedom. If the pilot goes well, we plan to follow up with the main study. If the pilot doesn’t go well, we’ll consider different approaches.
And Some Thoughts on how We’re Thinking About Basic Income
We think everyone should have enough money to meet their basic needs—no matter what, especially if there are enough resources to make it possible. We don’t yet know how it should look or how to pay for it, but basic income seems a promising way to do this.
One reason we think it may work is that technological improvements should generate an abundance of resources. Although basic income seems fiscally challenging today, in a world where technology replaces existing jobs and basic income becomes necessary, technological improvements should generate an abundance of resources and the cost of living should fall dramatically.
And to be clear: we think of basic income as providing a floor, and we believe people should be able to work and earn as much as they want. We hope a minimum level of economic security will give people the freedom to pursue further education or training, find or create a better job, and plan for the future. We’ll be spending the next few months designing the pilot, and we welcome any input to help us do the best job possible—especially from the Oakland community. (1)
And again, we hope to follow-up with a long-term study on how people’s happiness, well-being, financial health, and time are affected. If you have thoughts on either, please get in touch at email@example.com. -Elizabeth Rhodes, Matt Krisiloff, and Sam Altman https://blog.ycombinator.com/moving-forward-on-basic-income/
1 – We’ve already been connecting with Oakland city officials and community groups for feedback, but we’re planning to host some public events in Oakland to get more voices involved. Details to come.
Why we should all have a basic income
Consider for a moment that from this day forward, on the first day of every month, around $1,000 is deposited into your bank account – because you are a citizen. This income is independent of every other source of income and guarantees you a monthly starting salary above the poverty line for the rest of your life.
What do you do? Possibly of more importance, what don’t you do? How does this firm foundation of economic security and positive freedom affect your present and future decisions, from the work you choose to the relationships you maintain, to the risks you take?
The idea is called unconditional or universal basic income, or UBI. It’s like social security for all, and it’s taking root within minds around the world and across the entire political spectrum, for a multitude of converging reasons. Rising inequality, decades of stagnant wages, the transformation of lifelong careers into sub-hourly tasks, exponentially advancing technology like robots and deep neural networks increasingly capable of replacing potentially half of all human labour, world-changing events like Brexit and the election of Donald Trump – all of these and more are pointing to the need to start permanently guaranteeing everyone at least some income.
A promise of equal opportunity
“Basic income” would be an amount sufficient to secure basic needs as a permanent earnings floor no one could fall beneath, and would replace many of today’s temporary benefits, which are given only in case of emergency, and/or only to those who successfully pass the applied qualification tests. UBI would be a promise of equal opportunity, not equal outcome, a new starting line set above the poverty line.
It may surprise you to learn that a partial UBI has already existed in Alaska since 1982, and that a version of basic income was experimentally tested in the United States in the 1970s. The same is true in Canada, where the town of Dauphinmanaged to eliminate poverty for five years. Full UBI experiments have been done more recently in places such as Namibia, India and Brazil. Other countries are following suit: Finland, the Netherlands and Canada are carrying out government-funded experiments to compare against existing programmes. Organizations like Y Combinator and GiveDirectly have launched privately funded experiments in the US and East Africa respectively.
I know what you’re thinking. It’s the same thing most people think when they’re new to the idea. Giving money to everyone for doing nothing? That sounds both incredibly expensive and a great way to encourage people to do nothing. Well, it may sound counter-intuitive, but the exact opposite is true on both accounts. What’s incredibly expensive is not having basic income, and what really motivates people to work is, on one hand, not taking money away from them for working, and on the other hand, not actually about money at all.
Basic income in numbers
What tends to go unrealized about the idea of basic income, and this is true even of many economists – but not all – is that it represents a net transfer. In the same way it does not cost $20 to give someone $20 in exchange for $10, it does not cost $3 trillion to give every adult citizen $12,000 and every child $4,000, when every household will be paying varying amounts of taxes in exchange for their UBI. Instead it will cost around 30% of that, or about $900 billion, and that’s before the full or partial consolidation of other programmes and tax credits immediately made redundant by the new transfer. In other words, for someone whose taxes go up $4,000 to pay for $12,000 in UBI, the cost to give that person UBI is $8,000, not $12,000, and it’s coming from someone else whose taxes went up $20,000 to pay for their own $12,000. However, even that’s not entirely accurate, because the consolidation of the safety net and tax code UBI allows could drive the total price even lower.
Now, this idea of replacing existing programmes can scare some just as it appeals to others, but the choice is not all or nothing: partial consolidation is possible. As an example of partial consolidation, because most seniors already effectively have a basic income through social security, they could either choose between the two, or a percentage of their social security could be converted into basic income. Either way, no senior would earn a penny less than now in total, and yet the UBI price tag could be reduced by about $220 billion. Meanwhile, just a few examples of existing revenue that could and arguably should be fully consolidated into UBI would likely be food and nutrition assistance ($108 billion), wage subsidies ($72 billion), child tax credits ($56 billion), temporary assistance for needy families ($17 billion), and the home mortgage interest deduction (which mostly benefits the wealthy anyway, at a cost of at least $70 billion per year). That’s $543 billion spent on UBI instead of all the above, which represents only a fraction of the full list, none of which need be healthcare or education.
So what’s the true cost?
The true net cost of UBI in the US is therefore closer to an additional tax revenue requirement of a few hundred billion dollars – or less – depending on the many design choices made, and there exists a variety of ideas out there for crossing such a funding gap in a way that many people might prefer, that would also treat citizens like the shareholders they are (virtually all basic research is taxpayer funded), and that could even reduce taxes on labour by focusing more on capital, consumption, and externalities instead of wages and salaries. Additionally, we could eliminate the $540 billion in tax expenditures currently being provided disproportionately to the wealthiest, and also some of the $850 billion spent on defence.
Universal basic income is thus entirely affordable and essentially Milton Friedman’s negative income tax in net outcome (and he himself knew this), where those earning below a certain point are given additional income, and those earning above a certain point are taxed additional income. UBI does not exist outside the tax system unless it’s provided through pure monetary expansion or extra-governmental means. In other words, yes, Bill Gates will get $12,000 too but as one of the world’s wealthiest billionaires he will pay far more than $12,000 in new taxes to pay for it. That however is not similarly true for the bottom 80% of all US households, who will pay the same or less in total taxes.
To some, this may sound wasteful. Why give someone money they don’t need, and then tax their other income? Think of it this way: is it wasteful to put seat belts in every car instead of only in the cars of those who have gotten into accidents thus demonstrating their need for seat belts? Good drivers never get into accidents, right? So it might seem wasteful. But it’s not because we recognize the absurd costs of determining who would and wouldn’t need seat belts, and the immeasurable costs of being wrong. We also recognize that accidents don’t only happen to “bad” drivers. They can happen to anyone, at any time, purely due to random chance. As a result, seat belts for everyone.
The truth is that the costs of people having insufficient incomes are many and collectively massive. It burdens the healthcare system. It burdens the criminal justice system. It burdens the education system. It burdens would-be entrepreneurs, it burdens both productivity and consumer buying power and therefore entire economies. The total cost of all of these burdens well exceeds $1 trillion annually, and so the few hundred billion net additional cost of UBI pays for itself many times over. That’s the big-picture maths.
The real effects on motivation
But what about people then choosing not to work? Isn’t that a huge burden too? Well that’s where things get really interesting. For one, conditional welfare assistance creates a disincentive to work through removal of benefits in response to paid work. If accepting any amount of paid work will leave someone on welfare barely better off, or even worse off, what’s the point? With basic income, all income from paid work (after taxes) is earned as additional income so that everyone is always better off in terms of total income through any amount of employment – whether full time, part time or gig. Thus basic income does not introduce a disincentive to work. It removes the existing disincentive to work that conditional welfare creates.
Fascinatingly, improved incentives are where basic income really shines. Studies of motivation reveal that rewarding activities with money is a good motivator for mechanistic work but a poor motivator for creative work. Combine that with the fact that creative work is to be what’s left after most mechanistic work is handed off to machines, and we’re looking at a future where increasingly the work that’s left for humans is not best motivated extrinsically with money, but intrinsically out of the pursuit of more important goals. It’s the difference between doing meaningless work for money, and using money to do meaningful work.
Basic income thus enables the future of work, and even recognizes all the unpaid intrinsically motivated work currently going on that could be amplified, for example in the form of the $700 billion in unpaid work performed by informal caregivers in the US every year, and all the work in the free/open source software movement (FOSSM) that’s absolutely integral to the internet.
There is also another way basic income could affect work incentives that is rarely mentioned and somewhat more theoretical. UBI has the potential to better match workers to jobs, dramatically increase engagement, and even transform jobs themselves through the power UBI provides to refuse them.
A truly free market for labour
How many people are unhappy with their jobs? According to Gallup, worldwide, only 13% of those with jobs feel engaged with them. In the US, 70% of workers are not engaged or actively disengaged, the cost of which is a productivity loss of around $500 billion per year. Poor engagement is even associated with a disinclination to donate money, volunteer or help others. It measurably erodes social cohesion.
At the same time, there are those among the unemployed who would like to be employed, but the jobs are taken by those who don’t really want to be there. This is an inevitable result of requiring jobs in order to live. With no real choice, people do work they don’t wish to do in exchange for money that may be insufficient – but that’s still better than nothing – and then cling to that paid work despite being the “working poor” and/or disengaged. It’s a mess.
Basic income – in 100 people
Take an economy without UBI. We’ll call it Nation A. For every 100 working-age adults there are 80 jobs. Half the work force is not engaged by their jobs, and half again as many are unemployed with half of them really wanting to be employed, but, as in a game of musical chairs, they’re left without a chair.
Basic income fundamentally alters this reality. By unconditionally providing income outside of employment, people can refuse to do the jobs that aren’t engaging them. This in turn opens up those jobs to the unemployed who would be engaged by them. It also creates the bargaining power for everyone to negotiate better terms. How many jobs would become more attractive if they paid more money or required fewer hours? How would this reorganizing of the labour supply affect productivity if the percentage of disengaged workers plummeted? How much more prosperity would that create?
Consider now an economy with basic income. Let’s call it Nation B. For every 100 working age adults there are still 80 jobs, at least to begin with. The disengaged workforce says “no thanks” to the labour market as is, enabling all 50 people who want to work to do the jobs they want. To attract those who demand more compensation or shorter work weeks, some employers raise their wages. Others reduce the required hours. The result is a transformed labour market of more engaged, more employed, better paid, more productive workers. Fewer people are excluded, and there’s perhaps more scope for all workers to become self-employed entrepreneurs.
Simply put, a basic income improves the market for labour by making it optional. The transformation from a coercive market to a free market means that employers must attract employees with better pay and more flexible hours. It also means a more productive work force that potentially obviates the need for market-distorting minimum wage laws. Friction might even be reduced, so that people can move more easily from job to job, or from job to education/retraining to job, or even from job to entrepreneur, all thanks to more individual liquidity and the elimination of counter-productive bureaucracy and conditions.
Perhaps best of all, the automation of low-demand jobs becomes further incentivized through the rising of wages. The work that people refuse to do for less than a machine would cost to do it becomes a job for machines. And thanks to those replaced workers having a basic income, they aren’t just left standing in the cold in the job market’s ongoing game of musical chairs. They are instead better enabled to find new work, paid or unpaid, full-time or part-time, that works best for them.
The tip of a big iceberg
The idea of basic income is deceivingly simple sounding, but in reality it’s like an iceberg with far more to be revealed as you dive deeper. Its big picture price tag in the form of investing in human capital for far greater returns, and its effects on what truly motivates us are but glimpses of these depths. There are many more. Some are already known, like the positive effects on social cohesion and physical and mental health as seen in the 42% drop in crime in Namibia and the 8.5% reduction in hospitalizations in Dauphin, Manitoba. Debts tend to fall. Entrepreneurship tends to grow. Other effects have yet to be discovered by further experiments. But the growing body of evidence behind cash transfers in general point to basic income as something far more transformative to the future of work than even its long history of consideration has imagined.
It’s like a game of Monopoly where the winning teams have rewritten the rules so players no longer collect money for passing Go. The rule change functions to exclude people from markets. Basic income corrects this. But it’s more than just a tool for improving markets by making them more inclusive; there’s something more fundamental going on.
Humans need security to thrive, and basic income is a secure economic base – the new foundation on which to transform the precarious present, and build a more solid future. That’s not to say it’s a silver bullet. It’s that our problems are not impossible to solve. Poverty is not a supernatural foe, nor is extreme inequality or the threat of mass income loss due to automation. They are all just choices. And at any point, we can choose to make new ones.
Based on the evidence we already have and will likely continue to build, I firmly believe one of those choices should be unconditional basic income as a new equal starting point for all. https://www.weforum.org/agenda/2017/01/why-we-should-all-have-a-basic-income/