In the past few years, we’ve all seen the impact of technology on jobs: It kills many (average-paying) jobs to create fewer highly-skilled and high-paying new jobs. That is, technology is a net job killer. A few days ago, in a great post titled “The jobless future”, Jeff Jarvis claimed that
Our new economy is shrinking because technology leads to efficiency over growth.
In other words, technology is killing jobs and don’t expect those jobs to come back. I couldn’t agree more. Then Paul Graham of Y Combinator started a great discussion about this on Hacker News by asking the following question:
“why now? Technology has been killing (but not net killing) jobs for centuries. It’s possible that technology could start to net kill jobs. But why now, when it hasn’t in the past?”
And that’s what I’d like to answer here. But first, here’s my take on technology as a job killer.
Create a job, destroy two.
There are two ways that new technology transformations impact incumbents and jobs: 1- By fundamentally destroying value in the sector (think of the disappearing newspaper industry), 2- By offering more labor-efficient ways to provide the same value. I would argue that #1 is actually less relevant to net job loss in the long term because if you don’t have to buy newspapers to get your news fix, that’s more dollars left for you to spend on some other goods or services that create jobs. That is, in the long term, these kinds of disruptions shift jobs around (unless of course that new sector of the economy where you’ll shift your spending to is less labor intensive).
#2, however, is the heart of the issue here. Amazon can provide the same value to me as other brick and mortar retailers (buy my favorite items) in a much more labor efficient way. There isn’t much value destruction here but fewer jobs are needed to provide the same value. If you find this obvious, please skip to the next section to see why I think these job losses are permanent. Otherwise, read the rest of this section for more explanation.
Labor productivity (worker’s real output per man-hour) has been monotonically increasing for many many years. Here’s the labor productivity of the nonfarm sector from the Bureau of Labor Statistics — indexed to 100 for year 2005. You can download the raw data here:
The common wisdom is that this is happening due to technology and outsourcing. Sounds great, we are becoming more effective workers. But that can cut both ways. What if the total output doesn’t grow as rapidly or stays constant? Then fewer workers are needed to create the same output, hence more structural unemployment.
Here’s another way to think about it. As a proxy for output per employee, think of sales or revenue per employee per year. Interestingly, if you look at that metric for large successful companies in the consumer internet sector, it turns out that most of them fall somewhere between $0.5M-$1.5M per head per year. This is amazingly high! Think Zynga ($0.25B quarterly revenue divided by 2000 employees(?) times 4), Amazon ($5.4B quarterly domestic sales divided by 34,000 employees times 4), Google ($9B quarterly revenue divided by 29000 times 4). I’ve been guestimating this metric for various tech startups of different sizes for a while now and I can tell you with certainty that it is trending towards even more efficiency.
So what happens when you have a stagnating total output with a growing and increasingly efficient tech sector? For every dollar that the consumer spends to get something on Amazon.com or buy a tractor on Farmville, a dollar is not spent somewhere else in the economy.
Don’t fool yourself… Stagnating total output in the past few years has essentially made this a zero sum game!
What is the alternative to buying something online? Get in your car or get a taxi and go to the mall (car maintenance, parking garage fee, taxi fare), and then pay your local retailer. All those players are losing your business and they are a lot more labor intensive than, let’s say, Amazon. Take Macy’s, for example.
Depending on which ratio you believe is more relevant, at $5.8B quarterly net sales with 180,000 retail employees, Macy’s is at least 3-5 times more labor intensive than Amazon. So as long as Amazon’s top line is growing at the expense of Macy’s top line, we are losing 3-5 retail jobs for every job that Amazon creates. This is a massive net loss!
And it’s not just ecommerce. That $10 that you spent on Farmville means you have $10 less to spend at your local bowling alley (which is more labor intensive than Zynga). This is happening across many verticals, every day.
Why now? What is different this time?
In a great post, Tom Murphy calculates that economic growth cannot grow at a sustainable exponential rate (say, a fixed percentage every year) because the “physical” sector of the economy (i.e. the part that requires energy to do things, such as transportation, manufacturing etc.) depends on gains in energy efficiency to continue growing. If we stop gaining energy efficiency, “the energy growth rate in any form of technology leads to a thermal reckoning in just a few hundred years (not the tepid global warming, but boiling skin!).” The bad news is that energy efficiency is bound by laws of physics and cannot drop below certain theoretical thresholds. And even worse, in many areas we are already close to these theoretical limits. (If you are a science geek like me, please read the article it’s a fascinating read.)
So the only way for the total economic activities to grow is to grow in the “unphysical” sectors, i.e. those activities that require little energy input. But then he argues:
The important result is that trying to maintain a growth economy in a world of tapering raw energy growth (perhaps accompanied by leveling population) and diminishing gains from efficiency improvements would require the “other” category of activity to eventually dominate the economy. This would mean that an increasingly small fraction of economic activity would depend heavily on energy, so that food production, manufacturing, transportation, etc. would be relegated to economic insignificance. Activities like selling and buying existing houses, financial transactions, innovations (including new ways to move money around), fashion, and psychotherapy will be effectively all that’s left. Consequently, the price of food, energy, and manufacturing would drop to negligible levels relative to the fluffy stuff. And is this realistic—that a vital resource at its physical limit gets arbitrarily cheap? Bizarre.
So how does this relate to permanent job losses? Let me create an imaginary world where economic activities are cleanly divided it into two pieces: physical and unphysical. And let’s assume that there are a fixed number of people in this world who can work in either sector and can easily transition between the two. Yes, in real world there is friction in the labor force. But if the timescale is long enough, shifts will eventually happen in the real world just the way I’ll describe for this imaginary model. The size of each circle shows the size of the total economic activity in that sector.
Now let innovative minds invent mass production, robotic manufacturing, better transportation and energy grids etc. What happens?
Stage 1 – There’s an explosion in the physical sector. As the red circle expands, so does the productivity of each worker in this sector. That is, each worker in the physical sector is generating more value and therefore more wealth. As those workers become richer, they consume more unphysical services. Therefore the unphysical sector grows accordingly. Think of this as the beginning of the industrial revolution. As labor efficiency in the physical sector continues to grow, the unphysical sector grows in response. Some workers migrate from physical sector (which is becoming more labor efficient) to the unphysical sector. Total economic activity is growing really rapidly in both physical and unphysical sectors. This is what we end up with:
Stage 2 – Physical sector is limited by its most important input, energy. As the demand for energy goes up, physical sector relies more heavily on energy efficiency gains to grow. So it grows somewhat slower. In the meantime, it becomes more labor efficient due to technology (and outsourcing in real world.) Still there is huge demand for unphysical services so that sector grows faster than the physical sector. Even though the productivity in the unphysical sector starts to increase thanks to the invention of computing devices, the growth in the demand for those services outpace labor efficiency gains. In summary:
- Physical sector is growing somewhat slower but its continued albeit sluggish growth is allowing the unphysical sector to continue its rapid growth.
- In the physical sector: Labor efficiency gains starts to outpace the growth of the sector. So there is net job loss in this sector.
- In the unphysical sector: Labor efficiency gain is still less than the growth in the sector. So there is net job gain in this sector. More workers migrate from physical to unphysical sector.
And that’s what happened in the last two decades of the 20th century (plus outsourcing which you can lump together with technology because they both cause labor efficiency.)
Stage 3- Growth in the physical sector really slows because energy efficiency gains slow. If you buy the argument that the unphysical sector of the economy is hinged to the physical sector (i.e. you cannot have an economy where only 1% of the economy is physical stuff and the remaining 99% is people selling things to each other or babysitting each other’s kids), that causes the growth in the unphysical sector (which has ballooned to become a big chunk of the economy) to slow down as well. In the meantime, labor efficiency gains continue to increase in this sector of the economy thanks to the Internet. At some point, labor efficiency gain outpaces the growth in the unphysical sector and people become unemployed. This is structural unemployment. This is what has been happening in the past decade.
And that’s what’s different this time. In the past, every time there was a sector where labor productivity outpaced growth to create structural unemployment, we were dealing with a physical sector. That is, we were losing farmers, manufacturing workers, construction workers etc. Those kind of job losses were being offset by a faster growing unphysical sector. In the past decade or so, however, the slowdown in the unphysical sector of the economy (because the unphysical sector is tied to a physical sector which cannot grow forever due to real world limitations and laws of physics) along with the continued growth in labor productivity in this sector has created a situation where we are net losing workers in the unphysical sector of the economy as well. We are talking bank clerks, retail jobs, journalists, paralegals, etc. And you cannot really expect this trend to reverse anytime soon unless there is a huge paradigm shift in the physical sector, e.g. we start populating other planets, which will in turn kickstart another period of fast growth in the unphysical sector and create new jobs.
The physical sector of our total economic activity is bound by laws of physics. Even if you find new sources of inexhaustible energy and invent new technologies, you are still bound by theoretical limits of efficiency (again, please refer to Tom Murphy’s article for further explanation.) As labor productivity increased in this sector in the past, we offset that by creating a fast growing unphysical economy. But now, since unphysical economy is hinged to the physical economy, the growth in that sector has also slowed down and cannot be expected to pick up again (because it is tied to the physical economy and that sector won’t grow at the same pace it was before… ever!). In the meantime, the explosion of computing and the internet has significantly increased the labor productivity in the unphysical sector. And the resulting unemployment is permanent.
Invention of new technologies, new services and industries, or discovery of new sources of energy or other resources will not reverse this trend. In another post, I’ll discuss a few thoughts on how to deal with this.