What if the government just gave everyone money?
In the 2016 presidential election campaign, Donald Trump campaigned heavily on the talking point that American industrial workers have lost their jobs because of bad trade deals. And it’s true that the U.S. lost 5.6 million manufacturing jobs between 2000 and 2010, according to a study by Ball State University’s Center for Business and Economic Research.
Yet there is a paradox when looking at the numbers. At the same time as factory workers were being laid off, the manufacturing sector was more productive than ever. The Brookings Institute, a Washington think tank, found that manufacturing output has gone up 250 percent since 1980 even as employment has fallen by 75 percent in the same time period.
This discrepancy is made possible by automation: the factory workers were not being replaced by foreign workers, for the most part, but by ever more sophisticated machines and robots that could take the place of human hands.
The data are clear: factory workers who have lost their assembly line jobs in recent years should blame automation, not foreigners or immigrants, for the loss of those jobs. And no trade deal is going to stop the inevitable advance of technology.
There is renewed interest in a simple idea that will give humans an economic place in a world where robots and computers do more of the work: just give everyone money, at least enough to live at a minimum standard. As crazy as it sounds, it has actually been tested extensively by Mathematica Policy Research, the Alexander Road-based research company that did pioneering studies on the idea in the 1960s and ‘70s.
Mathematica’s old studies on the topic are more relevant than ever, so much so that just this spring the firm announced that it has digitized an archive of the studies, which ran from 1967 through 1978, and posted them to its website, www.mathematica-mpr.com. And, if the government decides to seriously consider the idea again, it may well turn to the company for a new study on the issue.
Mathematica’s old studies on guaranteed minimum income are relevant again because the automation-driven fall in factory labor has rocked the social and economic landscape of the United States and arguably has contributed to the rise of right-wing politicians like Trump.
Even bigger changes loom on the horizon. Auto manufacturers are perfecting self-driving cars and trucks at a time when some 3.5 million people are professional truckers — and that figure does not include cab drivers, rideshare service drivers, and others who could be easily replaced by robot vehicles. The fast food industry is rapidly introducing self-service kiosks, putting cashiers out of work.
Robotic arms are taking the place of humans on food processing assembly lines, and the cost of industrial robots is falling so rapidly that even Chinese companies are starting to replace workers as labor costs rise.
Robots are getting into the service industry, too, with robots for food preparation, health care, cleaning, and elder care.
Automation of physical labor is only the beginning of the threat to U.S. workers. Desk jobs are already being taken away by artificial intelligence. The financial services industry is now offering “robo-advisors” instead of financial advisors to help clients build investment portfolios, and this shift to circuitry could put squishy-brained human wealth managers out of work (U.S. 1, April 26, 2017).
In 2016 Princeton computer science professor Ed Felten helped prepare a report for the Obama administration about artificial intelligence. The report noted that AI could drive astonishing economic growth, but that it also threatened employment (U.S. 1, March 8, 2017). A study by Carl Frey and Michael Osborne of Oxford University estimated that an incredible 47 percent of all U.S. jobs were at risk of being eliminated by computerization. A website, www.Willrobotstakemyjob.com, is based on their work and lets you plug in your job and see the odds that it will be automated based on Frey and Osborne’s calculations.
In the past technological advances have eliminated jobs but created new ones in their place. Human labor has been able to keep up with technological innovation by becoming educated and gaining new skills. But the coming AI onslaught is different since it threatens educated jobs as well as blue collar ones. Sophisticated algorithms are already being used to perform tasks in the medical, legal, and financial realms that were once thought to be strictly the domain of humans. For example, IBM’s Watson supercomputer is being used at Memorial Sloan-Kettering Cancer Center in New York to assist oncologists in diagnosing cancer.
It’s possible that the future could merely continue current trends. The computerization wave of the 1980s and 1990s eliminated many middle-class white collar jobs like bookkeeping, but it actually increased the demand for accountants and other highly educated professionals and also added service industry jobs at the low end. The result has been a more polarized job market with more low-end and high-end jobs, fewer middle class ones, and a worsened gap between the rich and the poor.
The Oxford University report predicts that cheap surveillance cameras will reduce law enforcement jobs, sensors on public utilities will reduce the need for maintenance workers, speech recognition software will replace call center jobs, automatic interactive tutors will replace teachers, and algorithms will replace some software programmers.
AI researchers have even made forays into the humanities, creating programs that draw, compose music and poetry, and write jokes.
No one really knows for sure what effect AI and robots will have on the job market. Nevertheless, the possibility of a robot-driven future raises the question: what will we all do when our services are no longer needed?
The possibility of super-advanced automation raises the possibility of a radically changed society. In his “Culture” series the science fiction author Iain M. Banks wrote of a world in which money was abandoned and everyone lived in luxury and worked only on things they enjoyed, leaving all unpleasant tasks and administration to superintelligent A.I.
But if the highly automated economy of the future is to resemble anything like the current money-based system, consumers will need a source of income to survive, and that will be very difficult if unemployment or underemployment becomes the norm as human labor becomes less and less valuable to businesses.
What about the simplest, most obvious solution to this quandary? What if the government just gave everyone money?
In 1967 the U.S. government was seriously considering doing this, not as a response to unemployment specifically, but as a general cure for poverty. To study the issue President Lyndon Johnson’s Office of Economic Opportunity hired the newly formed Urban Opinion Surveys Division of Mathematica, which had been founded in 1958 and had made a name for itself studying economics and defense issues for the government.
Researcher David Kershaw took the lead on the study, which lasted until 1974. It was Mathematica’s first public policy study and the first large scale social experiment conducted anywhere.
In a paper that Mathematica published in 1969, Kershaw laid out criticisms of the welfare system at the time, some of which are still valid today: The system does not reach all the poor, those it helps still remain in poverty, and complex bureaucracy makes everything costly to administer and open to abuse. The welfare systems of the 1960s also inadvertently discouraged applicants from working and caused families to break up because some programs were only for single mothers.
The Negative Income Tax (NIT), envisioned by conservative economist Milton Friedman in 1962, was a proposed solution to these problems. The NIT was an extension of the existing progressive tax system in which low income earners pay little or no taxes, while high income earners pay a higher tax rate. Under the NIT low income earners would be given a check by the government. This ensures that no matter how little money you make, your income can never fall below a certain level.
In one Mathematica experiment 1,300 low-income families from Trenton, Paterson, Passaic, Jersey City, and Scranton, Pennsylvania, were enrolled in a negative tax plan, under which they were given a maximum of $4,125 a year in reverse taxes, an amount equal to 125 percent of the poverty level at the time. Families were paid every two weeks and could do whatever they wanted with the money.
The study was not without controversy. Mercer County prosecutors even hauled Mathematica employees to court, claiming that they were encouraging welfare fraud among study participants.
The court case was dismissed, and after four years Mathematica had its results and was able to crunch the numbers. It turned out there was a modest reduction in hours worked — about 12 percent — but most of this came from the spouses of primary income earners reducing work hours to take care of children, and some from workers taking longer during unemployment to look for a better job.
Throughout the 1970s and 1980s, Mathematica and other groups conducted similar studies in Pennsylvania, Indiana, North Carolina, Iowa, Oregon, and Colorado.
The subsequent studies had findings similar to the initial New Jersey results, with a 13 percent reduction in work hours, for similar reasons.
Paul Decker, current president of Mathematica, says the results were widely misinterpreted in the public debate about the NIT. “What they found was sure enough, there was a negative impact on people’s hours worked and work intensity,” he said. “Researchers tended to view them as modest effects, but in political discussion they got played up as large effects.”
Overall, the experiments showed that receiving extra income did not cause people to stop working. Despite these positive results, and perhaps because NIT opponents exaggerated the work hour reduction effects, the policy idea died and was never brought back. (The current Earned Income Tax Credit program does bear some resemblance to the NIT.)
Lately there has been renewed interest in a Guaranteed Minimum Income, or a similar proposal, the Universal Basic Income. The UBI differs from the GMI in that everyone, regardless of income, would receive a UBI payment, much like everyone receives Social Security benefits regardless of income level. Proponents believe that making the income universal would guarantee that once implemented, it would be politically impossible to remove. When designing the Social Security program, Franklin Roosevelt famously said that the people who contributed taxes to it would feel a “legal, moral, and political right” to collect, and that “no damn politician can ever scrap my social security program,” and he has so far been proven correct.
The Universal Basic Income idea is being touted not only by left-leaning politicians, but by Silicon Valley businessmen. Tesla CEO Elon Musk is one of many tech executives who have floated the idea recently, foreseeing widespread unemployment from automation. “I think we’ll end up doing universal basic income,” he said in a speech at the World Government Summit in Dubai last November.
Currently, there are several basic income experiments running in Brazil, Canada, Finland, France, Kenya, India, Liberia, the Netherlands, and Uganda, and one in the U.S. However, Decker says that it’s possible that a larger one could be forthcoming, and that he has also seen renewed interest in looking at Mathematica’s old NIT studies from the 1960s through 1970s.
Even if the idea of universal basic income were accepted, many questions would remain: How much would it be? How would it be paid for? And would the basic income replace existing social safety net programs or supplement them? Those are all questions that lawmakers would want to answer before creating a program, and it’s likely that they would turn to Mathematica once again to help research it.
Mathematica was founded in 1958 by Oskar Morgenstern, a Princeton professor of economics who wanted to put the new ideas of “operations research” to use in business and government. Operations research, developed during World War II, was a scientific way of managing large organizations, making use of mathematics and rigorous analysis to make processes run efficiently. These techniques, developed in Britain and the U.S., are what allowed the Allies to produce a titanic arsenal of weapons, train vast armies, and build the logistics chain that could deliver them swiftly to battlefields around the world to overwhelm Germany and Japan.
Rapidly advancing electronic computer technology promised to supercharge the young field of operations research, and Princeton was the perfect place to do it, since the university had one of the world’s fastest computers.
Morgenstern, together with a group of other Princeton professors of math and economics, founded Mathematica with the idea of “linking business problems and the mathematical sciences.” The group of founders included Harlan D. Mills, a pioneer of computer science; George Dantzig, a mathematician who created linear computer programming; Albert Tucker, an early researcher of game theory; Ralph Gomry, credited as “the father of integer programming;” Michael Balinski, an operations research expert; William Baumol, an economic theorist; Harold Kuhn, another game theory and programming pioneer; and Richard Quandt, a specialist in mathematical economics.
With this brain trust, Mathematica established an office at 92A Nassau Street, where Hamilton Jewelers is now, and began taking federal contracts. In 1959, a year after its founding, Mathematica became a subsidiary of Market Research Corporation of America (MRCA), where Morgenstern was a board member.
The company studied nuclear deterrence, satellite tracking, naval logistics, commodity pricing, and other government problems. One of its most influential early studies was a report on the economics of the performing arts, in which William Baumol and William Bowen (who would later become president of Princeton University) studied the economics of dance, theater, and music. The study, published in 1968 and still widely cited, concluded that the performing arts would always need subsidies to survive.
In 1964 Mathematica moved to 1 Palmer Square and Tibor Fabian, an economist, became president. The company expanded rapidly, partly due to the negative income tax studies, and became independent of MRCA in 1969. By 1971 Mathematica went public with a stock offering.
In 1975 the Urban Opinion Surveys division, the group that conducted the NIT study, was spun off into a subsidiary led by Kershaw, who had also led the NIT experiment. According to a company history, the split happened because the UOS employees wanted more separation from Mathtech, another Mathematica division that had started doing work for the CIA, and they believed that any association with the intelligence agency would harm their research efforts. The employees threatened to resign en masse, and Tibor compromised by allowing Mathematica Policy Research to become a subsidiary company.
By 1972 Mathematica had more than 500 employees and consolidated into a building at Princeton Station Office Park. Over the next decade, Mathematica Policy Research continued its negative income tax experiments and also did studies of Canada’s national healthcare system and various welfare reform ideas. Mathematica also expanded, building offices in Denver, Colorado; Seattle, Washington; Madison, Wisconsin; and Washington D.C.
The Reagan era was not kind to Mathematica, as the conservative government lost enthusiasm for new welfare programs for Mathematica to study. Mathematica’s contracts dropped sharply in the early 1980s, and in 1983 it was sold to Martin Marietta Data Systems, an aerospace company that built the Disney World monorail system as well as nuclear missiles for the military. According to a company history, the two companies’ cultures were not a good fit. In 1986 Mathematica Policy Research employees reached an agreement with their parent company to buy out MPR and make it an independent, employee-owned company led by Chuck Metcalf.
At that time, Mathematica Policy Research only had 81 employees. After becoming an independent company, MPR delved into new areas of research, including early childhood development and education.
The company started expanding again and today has 1,200 employees spread out in offices at 600 Alexander Park and 707 Alexander Road in Princeton; Cambridge, Massachusetts; Ann Arbor, Michigan; Washington D.C.; Oakland, California; Chicago; and Baltimore. It is still employee-owned via an employee stock ownership program, in which every worker holds shares of the company as part of a retirement plan. Revenue in 2016 was $254 million.
Mathematica’s continued ability to get contracts depended on its reputation for objectivity and for sometimes giving clients bad news about their programs. For example, Decker says, in the 1970s Mathematica studied a job training program for people who had lost their jobs due to international trade. The study found that 72 percent of the workers in the program eventually returned to their former jobs, so the job training went to waste. The program was altered so that it targeted people who were actively seeking training.
The company took on an even more controversial topic in the late 1990s when it undertook a nine-year study of abstinence-only education programs that were receiving federal funding. The researchers found that the courses had substantially the same outcomes as standard sex education: students enrolled in them were no more likely to be abstinent, but they were also no more or less likely to have unprotected sex or become pregnant. “Congress got stingier with the dollars for abstinence-only sex education programs after that,” Decker says.
In the 1990s and 2000s Mathematica undertook another study that made headlines. Teach For America, a nonprofit group founded by Wendy Kopp, a 1989 Princeton graduate, had begun a program that put top college graduate teachers in disadvantaged school districts. While the program was intended as a way to give underprivileged children access to talented teachers, some teachers argued the program was harming students because they were being taught by inexperienced educators. The program was also unpopular with teachers’ unions, who did not like teachers with only five weeks’ training being put into classrooms.
The study was typical of the rigor that Mathematica applies to its research. The three-year, $2 million study involved 17 schools, 100 classrooms, and nearly 2,000 students, who were randomly selected to be in TFA classrooms or get traditional teachers.
Mathematica found that students actually did better in math with TFA teachers than regular teachers, though there was no difference in reading test scores.
The main purpose of all these studies is to take ideas that sound good on paper and see how well they work in the real world. Sometimes the results are disappointing.
In the 1980s and 1990s Mathematica carried out a series of experiments with unemployed workers to see whether several different incentives, including a cash “employment bonus,” would encourage them to find jobs faster. The “employment bonus” cash proved to be ineffective. Job-seekers did find work slightly faster if offered a cash bonus, but the program came nowhere close to paying for itself, Decker says. As a result, many states tailored their unemployment programs to match Mathematica’s findings, emphasizing job training.
Occasionally, Mathematica’s studies have shown unequivocally that a program works as intended. Studies of the Children’s Health Insurance Program (CHIP) showed that it was very effective in improving the health of children. Another recent study supported Medicaid reimbursing family members for caring for their sick relatives. Partly as a result of Mathematica’s studies, most states now allow the “cash and counseling” option for home healthcare.
All this objective evaluation might seem a bit threatening to an advocate of a well-meaning program being studied. From a certain perspective Mathematica might seem like cold-blooded number crunchers whose evaluation could give politicians reason to cut funding from a program that is helping people.
But Decker doesn’t mind the term “technocrat” being applied to his organization. To him, studying evidence objectively is not unmerciful at all. “Yes, it might lead to a program being cut. But don’t we want to know the truth if a program is in fact harmful despite its positive intent? Wouldn’t you want to know that and be able to adjust? It’s really tough to get past the original intent of programs.”
This scientific attitude towards policy took hold in Washington during the Obama years, with more grants and programs making funding dependent on evidence of effectiveness.
MPR continues to grow, and has expanded into the international arena, doing studies on behalf of USAid, the Millennium Challenge Corporation, and foundations. One current project is a study of an electrical grid in Tanzania.
Decker says Mathematica also does more problem-solving type work for organizations in the public and private sectors, using data and analysis to help managers run programs better. The company has also begun to focus on making sure its study results are understandable. Looking at operations of a program, processing of IRS forms — how do you make sure that’s being done in a way you’re producing valid results and doing it efficiently?
Decker’s background is typical of Mathematica staff. He grew up in Jacksonville, Illinois, where his father taught English and theater at MacMurray College and his mother taught government and economics at a high school. He went to college at William and Mary and earned a doctorate in economics at Johns Hopkins and joined Mathematica in 1988.
“When I started at Mathematica in 1988, the typical way that we worked was to write a 300-page report full of equations, not formatted very well, and put it on a shelf and hope that eventually that study would have an impact on policy,” he says. “We saw ourselves as kind of these detached scientists.”
Nowadays, Mathematica is much more focused on communicating with decision makers. Instead of just a 300-page report, Mathematica now releases its results in written documents and on websites, and directly engages with lawmakers, program administrators, and the researchers who are interested in the research results. “That way we know our research is being entered into the dialogue and that it’s being interpreted correctly,” he says. “It used to drive us crazy to see somebody else interpret the findings of a study and get it wrong and get notoriety for it.”
The staff makeup of Mathematica has changed over the years. What started off as a group of Princeton professors has become a much more diverse company. Mathematica now employs more data scientists and computer analysis experts.
Mathematica has always relied on computers, but technology is becoming even more integral to the way it conducts research. The company is currently working with data sets, including physician performance records from the Medicare and Medicaid programs, which were too massive to collect and manage 10 years ago. Mathematica’s Maryland location is physically close to the Centers for Medicare & Medicaid Services office and occasionally has to physically transport data that is too voluminous to send over the Internet.
In a 1961 brochure for Mathematica, Morgenstern was quoted as saying, “Wherever mathematics has entered, it has never again been pushed out by other developments. The mathematization of an area of human endeavor is not a passing fad. It is the prime mover of scientific and technological progress.”
Morgenstern has yet to be disproven. And while Mathematica relies more and more on computers, it’s unlikely that their jobs will be automated out of existence, even as that same technology has Mathematica pondering a post-work society. Willrobotstakemyjob.com gives “Operations Research Analysts” jobs only a 3.5 percent chance of being automated.
Mathematica Policy Research Inc., 600 Alexander Park, Suite 100, Princeton 08540. 609-799-3535. Paul Decker, president. www.mathematica-mpr.com