The Paper Boy by Mark Stone
January 2016. We tell ourselves two great fables when trying to take a positive view of the modern global economy. First we tell ourselves that it's inevitable that nations will evolve from agricultural to industrial to service economies. Second we tell ourselves that the Sharing Economy democratizes services in a way that benefits the producers in the Sharing Economy. Is either of these fables really true?
This question struck me while visiting my parents over the holidays, and watching the newspaper delivered. This was a job I used to have, and for my generation "paper boy" was a common entry point to the work force.
AirBnB and Uber are the flagship Sharing Economy companies, although arguably eBay was the first entrant. All of these companies have a common thread; they tap a produtivity inefficiency in the lives of ordinary people and turn it into a marketable good. eBay puts the unused stuff lying around your house up for sale; AirBnB finds the empty beds in your house or second home and rents them out; Uber finds the empty seats in your car and rents them out.
These forces work in conjunction with the evolution to, and democratization of the service economy. GDP statistics tell the story we have come to accept:
- The least developed countries in the world have economies that are still primarily reliant on agriculture. Sub-Saharan Africa is an oft-cited example; Chad's GDP is 52% agriculture; the Central African Republic's DGP is 58% agriculture.
- Developing countries move to an industrial / manufacturing base, often driven by a single natural resource, like oil, that is plentiful. GDP in Brunei is 67% industry;Iraq's GPD is 60% industry.
- Developed countries have made the transition to a primarily service-based economy. China has only recently made this transition. GDP in China is 43% industry and 48% services. The United States is much farther along this spectrum with 78% of GDP coming from services.
While we all recognize this pattern, we don't talk much about the economic drivers behind it. The two key dynamics are infrastructure cost and commoditization. Infrastructure cost refers to how much supporting infrastructure it takes to enable what an economy produces. Commoditization refers to how differentiable what an economy produces is. The forces tend to work in opposition: infrastructure is expensive to build, but has a relatively low maintenance cost. Differentiation is relatively easy to initiate, but more difficult to maintain.
Let's unpack this a bit. Agricultural economies tend to have a relatively low infrastructure barrier to entry, and if the local environment is suitable for agriculture at all, this tends to produce competitive differentiation among growers. Thus in a fertile agricultural area like California's central valley, produce is diverse and tends towards higher value crops (grapes, olives, fruits) while shying away from entry level crops like corn and wheat that can be grown in a wider range of environments.
With agriculture the limiting factor is transportation. As our transportation infrastructure has globalized via cheaply operated, high capacity container ships, growers that once enjoyed a differentiating advantage now find it harder to maintain differentiation. California avacados compete with goods from Mexico and Chile. Coffee grown in Africa competes with coffee grown in South America. So differentiation is a competitive advantage, but a fleeting one as the forces of commoditization undercut temporary differentiating advantage.
Which is why developing countries tend to shift from agriculture to industry. With some capital investment a more sophisticated infrastructure can be built, enabling the economy to shift from agricultural to manufactured goods. At the low end of industry are goods that require simple raw materials, basic assembly line processes, and are durable and easily transportable. Textiles is a common entry point to the manufacturing economy, and in my lifetime we've seen the primary source of shoes and clothes shift from Taiwan and Korea to China and Bangladesh, and more recently to Vietnam.
If simple industries like textiles offer a relatively easy way out of the agriculture economy, they also offer proportionately little protection from commoditization. Developing countries must do exactly that: develop. The capital from industry must be reinvested in infrastructure, enabling the manufacture of more sophisticated goods that offer greater differentiation.
However, even the most sophisticated manufacturing processes are vulnerable to commoditization. When I graduated college almost all the silicon chips in the world were manufactured in the United States or Europe. Film and optics companies, working with some of our most advanced technologies, were largely based in the United States. Today most silicon chips are manufactured in China, and American companies like Kodak are now bankrupt.
Which is why developed countries shift from an industrial economy to a service economy. While services includes many unskilled jobs without strong differentiation (janitor, receptionist), the engines of the service economy are educated, high skill workers offering highly differentiated services (health care providers, legal advisors, financial advisors).
Yet even a service economy is vulnerable to commoditization. We've seen the global telecommunication infrastructure impact services the way the global transportation infrastructure impacted agricultural and industrial economies. From call centers to software development, we're seeing service jobs move to developing countries. With a global telecommunications infrastructure in place, basing those jobs in the U.S. or Western Europe is no longer a strong differentiator.
None of this is particularly new. What is new, and what will lead to the post-service economy, is the alignment of global telecommunication with two new factors: data analytics and work force automation.
Data analytics is what makes possible the democratization of service back the sharing economy. Thousands of empty bedrooms and vacation homes have no value if we have no means to harvest, analyze, and productize the data about those vacancies. Data analytics is what makes AirBnB possible. Traffic flow and commute patterns are highly dynamic. Trying to harness vacant seats in cars and develop and adaptive pricing system to value those seats is a huge data analytics challenge, daunting enough that Uber would not have been possible in the original Dot Com boom because we simply lacked the computational resources and data tools to make Uber possible at scale.
What often goes unexamined in praising the sharing economy, however, is that the work enabled by sharing efficiencies is both fractional and independent. In other words, being a Uber driver provides no guarantee of 40 hours a week of work. A room or home rented out through AirBnB may not be full all the time. Data analytics doesn't care how much income you need as a sharer; analytics relentlessly finds need and pairs it with demand, but only for exactly as much need as there is at that moment. And these kinds of sharing jobs are often not highly differentiable. In a dense urban area one Uber driver is very much like another (or a Lyft driver). The work is uneven, and pressure on income from this kind of work is downward.
Because the work is independent -- freelance rather than under the auspices of an employer -- sharers do not enjoy any of the economies of scale in benefits that acrue to a large, organized work force. For sharers there is no company health plan, there are no 401Ks or pension plans, there is no vacation pay.
The sharing economy faces friction. Tradition and entrenched interests push back against sharers. Hotel operators rail against AirBnB as competition that doesn't have to play by the rules (no hotel tax, no Service Employees Union to deal with, etc.). Taxi companies level the same complaint against Uber and Lyft. Companies that worked hard and paid significantly for scarce taxi operator licenses are being undercut by sharers who simply go around the system.
In the long run tradition and entrenched interests cannot win. In fact the scarcity of licensed taxi drivers is entirely artificial. The right they enjoy -- the right to chaffeur a passenger from one place to another -- is a right conferred on all of us the moment we get a driver's license. The only additional right taxi drivers are granted is the right to charge for this service, and there is no differentiating skill taxi drivers possess that warrants their unique access to this right. In the ruthless world of data analytics, data will find a path by which to deliver this service without the inefficiency. Barriers to productivity like the taxi monopoly will fall before the power of data.
Which brings me back to newspaper delivery.
In my neighborhood delivery routes were handed down from one teenage boy to the next, and it seemed as if it had been that way forever. I "inherited" my route from Danny across the street when he got old enough to move on to a regular job at the local McDonald's. I had roughly 120 houses to deliver to every morning in a dense 3 block by 3 block territory.
In truth it was a pretty terrible job. Paper boys were not employees of the newspaper, though they could be fired by the paper for under performing. Essentially the paper boy bought the papers from the newspaper, and along with that purchase the paper boy acquired the right to collect the bill from subscribers. Once you collected enough from subscribers to pay off the papers you'd purchased, all the additional subscriber money was profit.
I got up at 4:30 ever morning. All the papers had to be delivered by 6:30. Weather conditions didn't matter; you still had to deliver on time. It was a 7 day a week job. Twice a week you had to do inserts, which meant you got stacks of advertising in addition to stacks of papers, and each paper had to have an ad section inserted into it. One evening a week I would make the rounds of my route trying to get people to pay their subscription. Most people did without complaint, but there were quite a few who regularly put me off, and I had little authority to force them to pay (the newspaper did not allow me to withhold their paper). I had to keep track of who was on vacation and wanted their paper held, who was on vacation and wanted their paper just cancelled for the short term. There were a couple of customers who left for work before 6:30 and had put in special requests with the newspaper to have theirs delivered early. I was expected to honor this request.
This was hard work, and the bookkeeping was complicated. On top of that I was a freelancer without benefits, working part time, earning less than minimum wage. And I hated the bill collecting part of it. After about a year I was so far behind in bill collecting that I wasn't making any money at it. My dad gently intervened at that point, and Brian one block over took over my route while I moved on to looking for regular work.
With hindsight the job looks like fairly exploitative child labor, taking advantage of a loophole in the labor laws.
Which made it all the more shocking this December, when, standing in the driveway of my parents' house, I saw a middle-aged woman drive up in a station wagon, get out, and hand me the newspaper. Times have changed.
For one thing routes are much less dense now. In my day 90% of the houses in our neighborhood were subscribers. Today, a 120 subscriber territory would cover miles, impossible for a kid on foot to cover.
More importantly, though, this middle-aged woman needs this job; our economy can no longer afford the luxury of letting a kid do work that adults are begging to have. This is the dark side of maximizing GDP. As regular, full time employment dwindles, people are forced to make do with an ensemble of lower paying fractional jobs.
Today's "paper boy", this station wagon driving woman, is still a freelancer. She still isn't guaranteed a minimum wage. She has no employer provided health coverage or other benefits. The burden of bill collecting has thankfully been automated away from her, but in all other respects she has no greater earning capacity than I had as a 15 year old kid.
What we don't talk about when we celebrate the democratizing power of the Sharing Economy is just how prevalent this "paper boy" type of work is in the Sharing Economy. The typical Uber driver isn't someone making a little spare change along a route they happen to be driving anyway. The typical Uber driver relies on Uber as a job, a job with no benefits, no minimum wage, and no job security because what they do is completely undifferentiated from what anyone else with a driver's license and a vehicle can do. In all likelihood this person drives for Uber during peak hours, and works one or two other jobs in between that are likely also free lance, part time, and without the guarantees of benefits, wage, or job security. They might be a gopher doing odd jobs on Task Rabbit. Or trolling for questions to research and answer for pay on Mechanical Turk. This is what the service economy looks like for all but a privileged, highly educated few.
We still have a few bastions or organized labor in teacher's unions, public employee unions (policemen, firemen, and the like). But even these are under siege. Between online courses, home schooling, and charter schools, even primary and secondary education teachers are being eroeded away. Technology and automation have made homes and buildings much safer, so we simply don't need very many fire fighters. Indeed, fighting fires represents a small minority of the calls to which they respond.
Regulatory measures still protect some jobs. In most cities Uber still can't send drivers to the airport. Cities like San Francisco are cracking down on Airbnb hosts that are effectively hotel services.
But unions and regulators are fighting a losing battle. In the end these barriers are inherently artificial, and the reach of data analytics will find its way into every nook and cranny of the economy and root out these inefficiencies. Big data will inevitably match every fractional piece of work with any available sliver of time someone has to give. Productivity will sky rocket, and service workers will be marginalized as a result, commoditized like their agrarian and industrial predecessors.
The commoditization of service isn't even the worst of it. We are now on the brink of the coming automation revolution. Automation will align with global transportation, global telecommunication, and data analytics like the fourth horseman of the apocalypse.
Those low skill service jobs? They will be gone entirely, swept away in a wave of automation. Already the manufacturing industry in developed countries has been transformed by automation. In the past 50 years the number of manufacturing jobs has declined moderately. Over the same time period manufacturing productivity has doubled, sustaining the United States as one of the strongest manufacturing economies in the world. Manufacturing employment is depressed, and where there is job growth in manufacturing, those jobs require a higher than ever level of education and technical training. The relentless force of automation has shaped American manufacturing into a highly efficient production engine. This outcome is great for the economy, but problematic for low skill workers who counted on dwindling manufacturing jobs to secure their economic future.
In manufacturing we've seen the automation transformation coming for a long time, and while it hasn't been easy at least it has been expected. What's less anticipated is the impact of automation in the service economy.
We're all familiar with call center automation as an attempt to automate service jobs. No one sees this as a clear success. The process may save some companies some money, but it doesn't deliver a better customer service experience. Some high profile companies have consciously backed away from this approach. If you call Nike's customer support line you won't encounter an automated system, you won't talk to someone at an outsourced call center in India. Instead you'll be talking to someone seated right at Nike HQ in Beaverton, Oregon. Does this deliver a better customer service experience? Yes. But that's not the only reason Nike takes this approach; they also believe it saves them money overall in the long run.
So it's easy to look at the prominent call center example and say that automating service jobs is just too hard without accepting a significant drop in service quality. And so it's easy to overlook some of the surprising places that automation is not just replacing service work, but improving service quality.
It takes a lot of training for a psychologist to be an effective counselor: to listen objectively, to reassure the client that they are listening without judgment, to observe not just what a person says, but what their body language conveys. Even with the best training, all of these skills are still influenced by the counselor's own mood, thoughts, and context. When we combine data analytics with automation, it's possible to come up with a proxy counselor who may be easier to talk to about some difficult subjects, and who may be more accurate at noticing patterns of response that signal important, and sometimes hard to identify conditions like depression or PTSD. This is exactly the approach the Defense Department is exploring with a "robot" named Ellie.
Does this kind of technology replace a clinical psychologist? Not exactly, and not yet. But it can greatly extend the reach of an understaffed counseling team working with veterans, and compliment needed human skills with the objectivity and analytic efficiency of technology.
Or consider one of the oldest service jobs in the world: the waiter / waitress. Could this job be automated? Perhaps not literally, not yet. But table turn time directly affects restaurant profitability, so any steps that can be sped up through automation benefit the restaurant. That's why chains like Applebee's are experimenting with digital tools that help make parts of the table waiting process self service. The key here is not just to affect a faster table turn rate; what's delivered is an overall better customer experience: fewer mistaken orders placed, less time waiting for the wait person to collect and process a bill once the customer is ready to go. Step by step even this venerable service job will be automated.
Which makes me think again of the paper boy. This job is on a shorter path to automation than many. My guess is this job will be gone entirely in ten years. In its place will be a self driving car, or even more likely, a drone. We may not mourn the loss of the paper boy. It's a niche job in a dying industry that was always exploitative. However, delivery automation will have broad impact. The most common job title in the United States is "truck driver". What happens when all of those jobs go away?
We are fast approaching an inflection point, not just for our economy, but for civilization. Will we see that inflection point as a crisis or an opportunity? Because we can ask one of two questions about this new world we are about to enter:
- What happens when so many people are unable to work?
- Or, what happens when so many people no longer need to work?
Two different worlds; which will we choose to live in?