Whenever the subject comes up of how automation is eliminating blue collar jobs, and has eliminated a huge number of pink and white collar ones as well, inevitably someone claims that “the market” will produce new jobs.
The problem is, this belief has no basis in fact and is unsupported by any hard evidence. The market, the capital behind the market, and the men and women controlling the capital all have no particular interest in producing new jobs. If new jobs are going to be created, they’ll be created where labor is cheapest - in China or other low wage countries.
I’ve come to realize that this is not a recent trend. It started a while ago. Pretty much at the birth of the human race.
Farming food was a labor-saving measure. A handful people collecting seed and planting rows of crops was far more efficient than the huge number of people needed to gather food in the wild.
The domestication of animals was another labor-saving measure. The amount of time one person would spend caring for a pen full of domesticated pigs is far less than the number of hours required for hunters with spears to take down the same number of wild boars. The plow and fence freed everyone else who would have hunted or farmed to enter into other trades, and invent new ones.
But as these new trades entered the machine age, innovations help reduce the number of people required for the trade. Consistently, the labor force produces the same amount of product with fewer people, or greater amounts of product with the same number of people.
My rule of thumb is this: If it can be automated, it will be. If it can’t be automated, but can be exported to a low wage country, it will be. Even good paying jobs that have formed the basis of the American middle class.
Take banking. When I was a kid, every bank had a large number of tellers. It was a decent paying job employing a large number of people. My ex-sister-in-law was a teller at a bank. I visited that bank recently to deposit some checks in my brother’s account. The area that used to have a dozen teller windows is now executive offices. To perform a walk-up transaction, I was directed to a small booth outside where I called one of the two remaining tellers via video monitor and sent in the checks via pneumatic tube. Two tellers handled the drive-through and walk-up. The bank has more customers than before, but the majority of the transactions are handled via automation. Automatic Teller Machines. Web. Text message. E-mail. Any aspect of banking that required human beings has either been automated as much as possible, or transitioned to a different method that requires less labor.
Retail in general has been moving in this direction for years. I’m not saying this hasn’t been good for consumers. Anyone who can use a computer regularly searches on-line for the best price and buys through a retailer’s web site - a method that employs a much smaller number of people per unit. Amazon employs a large number of people, but it is tiny when compared to the number of retail employees who would have been employed to sell the same amount via “brick and mortar” stores.
Some retail categories have been decimated. The city of Chicago has no large record stores anymore. Tower Records went bankrupt. The Virgin Megastore closed. There are a handful of independent record stores still managing to hang on, but what has happened in this retail category is a clear precursor to the rest of retail. Again, this has been an improvement for most folks. Web retailers have offered many more choices than even the biggest record store - with a dramatically smaller number of man-hours per sales unit. iTunes and other digital retailers can sell the same product over and over again with little additional labor beyond ripping the original CD and entering the data, and little cost beyond paying the artist.
Any physical store of sufficient size is either experimenting with self-checkout systems or has already converted a percentage of their checkout lanes. The same retailer pressure that caused manufacturers to place UPC bar codes on every product will result in universal adoption of RFID tags. We’re promised in the very near future the customer won’t have to manually scan every item, or even to take it out of their cart. A passive radio-frequency ID tag will also enable automated systems to stock shelves by providing location information by RF triangulation and precise dimensions and weight of each item.
From what I’ve seen, when a new product category is created it is manufactured using the latest techniques to minimize labor costs. Take the GPS category. Garmin’s headquarters are in Olathe, KS. But the official corporate offices are a mail drop in the the Cayman Islands to skirt US tax laws. I believe the engineering is done in India. The manufacturing is in China.
Not even the worst, lowest-paid jobs are immune to being automated. McDonald’s has been testing a touch-screen ordering system that will help minimize data entry errors - and jobs (The Burger King at Atlanta’s airport has this system already.) A recently built McDonald’s near me has a system that dispenses drinks, the cups on a conveyor line ready to go as soon as the car pulls up to the window of the drive through. Reportedly, they are even trying to develop a burger-flipping machines. In some markets, they have even out-sourced drive-through order taking to operators in locations with lower labor costs.
Even the handful of jobs that legally pay less than minimum wage, wait staff and agricultural workers, are automating. The California citrus growers associations are paying for the development of an automated system to harvest oranges - one machine that performs a 3D scan of the whole tree identifying the location and ripeness of each orange, the second to follow this 3D map and harvest the fruit with a vacuum system. Not even the fact that fruit pickers can be paid only $2.20 an hour will prevent this from happening. Once you have a machines to do a particular job, that job will continue to be done by machine. Few farmers are going to abandon their combine harvester in favor of manually picking corn.
There are few jobs outside of managerial, creative and personal services that can’t be automated.
Take truck driving. Long distance trucking has been a reasonably decent paying job. But search for the “DARPA Grand Challenge”. The US military has decided that 33% of all their vehicles need to be capable of driving themselves to a particular location by 2015. In order to accomplish this, the Defense Advanced Research Projects Administration has had teams from companies and universities competing for large cash prizes to develop self-driving vehicles. This has pushed the field of real-world computer vision systems further in a few years than it had in decades before. Earlier competitions have been staged in the desert, with competing vehicles having to maneuver around parked cars, drive under bridges, read road signs, tell the difference between a tumbleweed and a boulder. In the most recent competition the vehicles have had to pass all tests while following all California driving laws. I believe an upcoming competition will involve a race, in live traffic, between Las Vegas and Los Angeles.
Some elements of this technology are in current production cars. Automatic parallel parking. Adaptive Cruise Control that maintains a fixed distance. If a Humvee will be able to autonomously drive to a destination through an urban environment by 2015, how long will it take for Kenworth and Mac to bring this technology to US highways?
How many jobs will be eliminated when UPS and FedEx adopt automated driving systems? They don’t need a Asimo-like robot to be able to carry packages into a business. The automated delivery vehicle will just phone the business to let them know when it will be outside for the pick-up or delivery, and a door will open to an isolated compartment to accept or deliver the packages. The businesses will accept this new model in the same way that US drivers have accepted pumping their own gas, booking their own travel, etc. UPS and FedEx will be able to charge less or at least avoid raising prices when they eliminate a big chunk of their labor costs.
And human labor is one of the few constant and (usually) increasing costs to businesses. Most businesses have implemented voice mail systems, or even more sophisticated systems. I’m writing this on an Amtrak train. I purchased my tickets via their “Julie” voice-response system. It asked and understood what city I was in, what city I was travelling to and the dates I wanted to travel. The “Julie” system probably cost quite a bit of money to develop, but I’m confident that development cost has been paid back in wages not paid to human operators.
I recently saw the film “Changeling”. It provides a good example of how evolves to eliminate jobs. Angelina Jolie plays a supervisor of telephone operators in the late 1920s to early 1930s. When the telephone system first appeared, every call required at least one operator to manually connect it, and there were thousands of people employed to do it. Direct dialing eliminated the majority of these jobs, and operators were only needed for long distance and directory assistance. Even that has been automated as much as possible, with the actual greeting and recitation of the number being done via computer voice. The current state of the art seems to be Google’s 1-800-GOOG-411 that provides free, completely automated business directory assistance. It works well, and employs a small number of programmers. Eventually, they’ll have residential number service, and the job of telephone operator will go the way of whalebone corset maker.
Construction trades have been another source of good-paying jobs. But materials and construction techniques have changed, principally to minimize labor. At the time of the turn of the previous century, walls were constructed by nailing thin strips of wood to wooden studs, then applying layers of cement and plaster. This was replaced with metal studs and plaster wallboard, and can be installed in a fraction of the time. Most industrial buildings used to be made of bricks, hand-laid in a process that took weeks. Now, most are “tilt-ups”, huge pre-cast slabs of concrete delivered to the site, set on end, tipped up into place and held together by the roof trusses and the other slabs in a process that can be accomplished in a few days.
Assuming the advances in computer vision generated by the DARPA Grand Challenge, how long before automated equipment is capable of completely building a typical industrial park building? I’ve got a good imagination and an inventive turn of mind (references on request) and can easily conceptualize machines that can accomplish each construction task - from the concrete work to the steel fabrication to the plumbing and electrical to wallboard installation, plastering and painting. Given how automated the job of producing the considerably more complex automobile has become, why should something as simple as a typical industrial park building be any trouble?
Given this evidence, why do people insist that the market will produce new jobs to replace all the old ones?
