Almost every weekday, some arm of the US government issues some sort of economic statistic. News media and financial analysts review and report it. Then 99.9% of the adult population, and probably 90% of the financial industry, forget all about it. And they’re probably right to do so.
The monthly jobs report isn’t like that. Yes, any single month doesn’t tell us much. Yes, the Labor Department’s methodology has some flaws, both major and minor. But imperfect as it is, the jobs report is our best look at the economy’s pulse. Jobs matter in a visceral way to almost all of us, as you know well if you’ve ever lost one. Almost any survey that asked questions around employment would reveal the angst that many Americans feel about the possibility of losing their jobs.
Right now, automation tops the list of things that might threaten our jobs. Artificial intelligence and robotics technology are rapidly learning to do what human workers do, but better, faster, and cheaper.
I’ve use the following chart before, but it’s a compelling illustration of how technology is reducing employment. It shows the rising rig count in the oil patch since mid-2016 – and yet the number of workers on those rigs is actually still falling. This is the impact of a new robot called an iron roughneck: Tasks that used to require 20 people now need only five. And the iron roughneck is not even that widely deployed in the oil and gas industry – the trend will hit hard in the coming decade. Roughneck jobs are relatively high-paying; it takes a great deal of training and skill to be able to do them.
A Decelerating Job Picture
Also, we must smooth out all the post storm disruptions. This give us a 3-month average monthly job gain of 170k, a 6-month average of 178k, and a year-to-date average of 174k. These numbers compare with average job growth of 187k in 2016, 226k in 2015, and 250k in 2014. Again, the slowdown in job creation is a natural outgrowth of the stage of the economic cycle we are in where it gets more and more difficult finding the right supply of labor.
The growth in wages is also decelerating. I was talking with Lacy Hunt this morning about the jobs report. He noted that real wage growth for the year ending November 2015 was 2.8%, while for the year ending November 2016 it was just 1%. The savings rate is now the lowest in 10 years. The velocity of money is still slowing, which means that businesses have to do everything they can to hold down costs, and one of those things is to rein in wages.
And yet the Federal Reserve has a fetish for this thing called the Phillips curve, a theory that was thoroughly debunked by Milton Friedman early on and later by numerous other economists as having no empirical link to reality. But since the Fed has no other model, they cling desperately to it, like a drowning man to a bit of driftwood. Basically, the theory says that when employment is close to being as full, as it is right now, wage inflation is right around the corner. According to the Phillips curve, then, the FOMC needs to be tightening monetary policy. Later we’ll see how the FOMC’s faulty tool is likely to lead to a major monetary policy error.
Basically, the Federal Reserve looks at history and tries to conjure models of future economic performance based on it – even as everyone in the financial industry goes on intoning that past performance is not indicative of future results. But all the Fed has is history, and they cling to it. My contention is that the near future is not going to look like the near or the distant past, and so we had better throw out our historical analogies and start thinking outside the box. Now let’s look at some real problems that will impact the future of employment.
Every year, reports like this reflect a process that’s occurred many times in human history. People discover or invent something useful: fire, the wheel, iron, gunpowder, coal, oil, the steam engine, electricity, the automobile, the airplane, the computer, etc. Life changes as the new knowledge spreads. People either adapt or they don’t. Those who don’t adapt fade into the background. In the last few decades of their working lives, they end up taking the very lowliest of jobs in order to get some food, clothing, and shelter; but it’s not a comfortable life. There was no government safety net for most of our history. But most people tried hard to adapt their skills to the new changes. And as we adapted to radically disruptive inventions like the steam engine, automobile, and computer, hardly anyone had the necessary skills, and so everyone had to learn.
Today, things are different. Fifteen percent of men between the ages of 25 and 54 – who should be in their most productive years of contributing to their families and society – don’t even want a job. That’s up from 5% in the mid-’60s, and the number has been steadily rising. Fifty-six percent of these people receive federal disability payments, averaging about $13,000, which is roughly equivalent to the pay for a minimum-wage job, after taxes – except that disability comes with free Medicare. Unless these people find ways to develop needed skills, there is not much financial incentive for them to look for jobs.
The rest of the people who don’t want jobs are mostly early retirees, homemakers, caregivers, or students. And roughly 1/3 of the 10 million+ men who have dropped out of the workforce have criminal records, which is often a barrier to work. Only about 3–4% are actually discouraged workers who might take a job if a job is available. That picture should be worrying. It is one reason why GDP has not increased all that much. Remember that GDP is proportional to the number of workers available times their productivity. Taking 10 million workers out of the workforce reduces GDP.
The problem for most of us now is that we don’t want to simply fade into the background like so many people have done with each major shift in technology; yet new knowledge spreads around the globe now in seconds instead of centuries. It’s easy to feel that the walls are closing in, because for many of us they are. The McKinsey report makes that crystal clear. They project that technology will replace as many as 800 million workers worldwide by 2030. Displacement is not just a US or developed-world phenomenon; it will show up in the emerging and developing markets as well.
McKinsey draws a distinction that we should all remember. The problem is less about jobs disappearing than about the automation of particular tasks that are part of our jobs. In most cases, employers can’t simply fire a human, plug in a robot, and accomplish all the same things at the same or better performance level but lower cost. You have to zoom in closer and look at the tasks that each job entails, and ask which of them can be automated. The roughneck jobs in the oilfield are a good example: The Iron Roughneck doesn’t replace all workers on the rig, just some of them.
So when McKinsey says that 23% of US “current work activity hours” will be automated by 2030, that’s not the same as saying 23% of jobs. The shift will affect almost all jobs to some degree. That 23% figure is their “midpoint” scenario, too. In the “rapid” scenario it’s 44% of US current work activity hours that will be handed over to machines.
In other words, whatever your job is, some part of it will likely get automated in the next decade or so. That might be good news if the machines can take on the repetitive drudgery that you don’t enjoy. Automation could free you to do things that are more interesting to you and more valuable to your employer. But outcomes are going to vary widely. Here’s a chart on sector and occupation employment shifts from McKinsey. (This one is for the US; their report has sections for other countries as well.)
he circles on the right are the translation of those task-hours into numbers of workers. As you can see, in their rapid automation scenario, by 2030 – just 12 years from now – 73 million people out of a workforce of 166 million will have been displaced, with 48–54 million of them needing to change occupations completely.
In other words, a full third of the workforce may have to change career fields. That’s going to be a problem. Yes, Americans change jobs more frequently now than they used to, but the changes tend to be evolutionary: We gain new skills, find a better place to apply them, acquire new contacts, seek out new opportunities, and so on. The personal transformation happens slowly enough to be manageable. That’s going to change.
My friend Danielle DiMartino highlights another of the amazing charts in the McKinsey study, one that analyzes US job-market susceptibility to automation scale: (See source)
This chart demonstrates that it’s not just the low-skilled workers who are at risk. It’s also mid-level and even some high-level people. There is more job risk than many of us imagine. That is why I break the world up into the Unprotected, the Protected, and the Vulnerable Protected classes. The latter group doesn’t even realize their vulnerability.
Worse, I think the shift to automation may come even faster than McKinsey’s rapid scenario suggests. Recently I ran across an artificial intelligence story that’s almost terrifying. You might have heard about AlphaGo, the AI system created by Google subsidiary DeepMind. It plays the very complex board game called Go.
In 2015, DeepMind became the first computer to beat a human professional Go player. It learned how to do this by analyzing many thousands of games played by humans. Impressive, but only the beginning.
This year, DeepMind introduced AlphaGo Zero, a new system that quickly acquired the same skills with no human help at all. The programmers simply gave it a blank board and the rules of the game. It then played millions of games against itself. Here’s the chilling quote from DeepMind CEO Demis Hassabis:
The most striking thing is that we don’t need any human data anymore.
It gets more unnerving. On December 5 (yes, last week), DeepMind published a scientific paper that sounds straight out of science fiction. I added the bold print.
The AlphaGo Zero program recently achieved superhuman performance in the game of Go, by tabula rasa reinforcement learning from games of self-play. In this paper, we generalise this approach into a single AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in many challenging domains.
Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case.
That’s startling, so let me repeat it slowly. In one day, starting from nothing at all (“tabula rasa”), AlphaGo Zero learned to play chess, shogi, and Go at a superhuman level, beating the same systems that had beaten the best humans in the world.
That’s how fast the technology is evolving. I suspect some of the rapid acceleration came from faster processor chips – Moore’s law says they should double in power every two years. But this was far more than a doubling; this was exponential.
Systems like that are coming for your job. So if you think you’re safe because you aren’t an assembly-line worker or a retail cashier and don’t work at the level of rote repetition, you could be wrong. These systems will only get better and take on ever more complex jobs.
Could DeepMind build a system that reads my archives, monitors my email, and then writes Thoughts from the Frontline at a level where you couldn’t tell the difference between it and me?
How do you know it hasn’t?
Those who control the tech are intent on bringing the era of superautomation forward as fast as possible. I talk a lot about incentives and the way people and businesses respond to them. Identifying incentives is a key tool in analyzing trends and forecasting what different players will do next. Well, between dicey Federal Reserve policies and possible tax reforms, businesses are getting new incentives to automate sooner rather than later.
Diminished earning power has, in turn, robbed businesses of pricing power and forced them to cut costs ruthlessly. One way you slash costs is by automating. In this week’s Outside the Box I shared a story about how Amazon is now hiring robots faster than it is human employees. Amazon is in the lead, but other companies aren’t far behind. This trend limits wage gains even more, and the situation is getting worse as the technology gets better and cheaper. (The fact that San Francisco has limited the number of robots per company and limited the speed of robotic delivery simply ensures that San Francisco will be behind the rest of the country in terms of growth and productivity within a few years.)
Of course, there’s absolutely nothing wrong with making your business more efficient. You have to survive against the competition. But in this case the competition is not happening naturally or according to market forces. The Fed has kept market forces from working and has created an environment that would never have occurred otherwise. You can argue whether a laissez faire market would have worked better or worse, but it’s pretty clear we haven’t had one.
Now add in tax policy. I explained early this year in my open letters to the new US president that we would all be better off with a consumption tax like a VAT rather than we are currently with the income tax. Alas, I did not get my wish. Congress is right now “improving” the tax code in ways that may actually accelerate the automation trend.
For instance, one proposal is to allow equipment purchases to be expensed immediately instead of amortized over time. That’s not a bad idea on its own. However, it effectively subsidizes companies to upgrade their equipment and technology to the latest state of the art. And, as we saw above, the state of the art is automated devices that need little human help.
The accelerated shift to automation may help explain a Business Roundtable survey that showed some odd results. As reported by the Wall Street Journal last week, CEOs say their plans for capital investment have risen to the highest level since the second quarter of 2011. That’s good news: Businesses see growth opportunities and want to add production capacity to meet them. But the same survey shows CEO hiring expectations going in the opposite direction. Hiring is not plummeting by any means, and many do plan to increase hiring over the next six months; but the majority say they will keep their headcount where it is or lower it. General Electric will cut 12,000 jobs from its power business, roughly 18% of that division’s total employment, in order to cut costs and reduce overcapacity.
How do we explain a situation in which capital spending rises but employment stay the same or falls? Automation is one answer. It lets you increase capacity without increasing headcount and expenses – you may even reduce them.
Not coincidentally, the new tax bill may remove the Obamacare individual mandate, but the employer mandate is staying in place – and healthcare costs are still rising. That too incentivizes businesses to use machines instead of people wherever possible.
So where do all these factors leave human workers? The McKinsey forecasts fall more or less at the midpoint of those in other reports I am reading. We’re facing a perfect storm: Technological, monetary, and political entities are joining forces to stir up a maelstrom of change that is going to bombard all of us. I’m not an exception, and neither are you.
We can’t control these giant forces, but we can control our responses. Whatever your job is now, you need to think about how vulnerable it may be and what else you might do. If you need to acquire new skills, start doing it now. If you have young adult or teen children, help them with their education and career choices. That art history degree may not be much in demand in 2030. Or even in 2020.
Full article: Automatic Job Storm Coming (Mauldin Economics)