In venture capital we come across startups all the time that are building something that has been tried in the past and failed. It would be very easy to dismiss these opportunities based on a naive “pattern matching” approach to investing. Instead at USV we always ask ourselves “what is different now?” to understand whether the lessons from the past do in fact apply or something profound has changed.
This is important because things that failed in the past tend to eventually succeed due to changes in technology, the market or society / behavior. It is useful to keep some big examples in mind to remind oneself of this. Here are a few: Apple had an epic failure with the Newton but eventually a huge success with the iPhone (changes: further turns of Moore’s law, cellular network buildout). Car companies had many failed stabs at electric cars until Tesla came along (changes: more expensive gasoline, cheaper batteries, critical mass of affluent environmentally conscious buyers). There were attempts at digital currencies (remember Flooz anyone?) which failed where bitcoin is appearing to succeed (changes: technical breakthrough, growing distrust of governments).
Or how about this one? Social networks in the past failed or didn’t make much money (friendster, MySpace) until Facebook came along (and recently inspired a whole series of tweets by Marc about failed FB predictions). Nassim Taleb in Antifragile and elsewhere has referred to the “it hasn’t happened so it won’t happen” argument as the Turkey fallacy (the Turkey is happy about the farmer feeding it every day until Thanksgiving).
All of this is to say that I strongly disagree with Marc’s assessment that people who worry about technology’s impact on the labor market today are either Luddites or socialists simply because this argument has been wrong in the past. There are plenty of things that have changed fundamentally since 1964. Most importantly, the cost of computing has plummeted and the power of computing has skyrocketed (see my comparison btw NASA’s compute power in 1960s and Raspberry Pi).
During the first industrial revolution people worried about machines replacing human workers because machines provided mechanical power. Well, it turned out that humans were still needed because we supplied brain power. This time round though, at the dawning of the “Second Machine Age" we are worrying because machines are providing brain power. That’s a new and different set of circumstances and so we should rightly re-examine this question and not just take a no answer for granted.
Marc has written that he is “way long human creativity” to say that even if computers replace humans for some thinking tasks we will be the ones who are creative (and come up with interesting things for ourselves to do). Even if you believe that — and I happen to share this belief — that still doesn’t mean that we shouldn’t worry about the transition period.
In fact if we want to look at history for a lesson we would do well to keep in mind that industrialization was incredibly ugly and had us go through lots of revolutions and two world wars. Agricultural jobs were lost at a far faster pace (due to mechanization) than industrial jobs appeared. And the early industrial jobs were awful involving horrid working hours and conditions.
In summary: something has changed (computing) and it is entirely appropriate to worry about existing jobs disappearing. At a minimum the worry is that they will disappear far faster than we can figure out how to be creative and at a maximum that we won’t in fact figure out what to substitute for them (I am personally less worried about the latter but not dismissive).
There was a fun Twitter convo about strong AI between Patrick Collison and Marc Andreessen. I also love speculating about this topic but before I engage in that I want to point out that from an employment perspective this is a red herring. A car is not a horse. And yet cars replaced horses in transportation. A tractor is not a horse. And yet tractors replaced horses in agriculture. A tank is not a horse. And yet tanks replaced horses in war.
It is irrelevant from an employment perspective as to how a robot or machine intelligence solves a problem. What matters is whether or not it solves that problem at a lower unit cost than a human. In fact, even the problem statement itself is up for change, meaning that a machine maybe solves a slightly different problem or a sub-problem that can then be used (together with cheaper — read less trained — humans) to accomplish the same overall outcome.
Let me illustrate that last statement with two examples. In order to get a car to its destination you could either use a trained driver who knows where everything is in the city or an untrained one who simply knows how to drive together with routing software (see Semil’s blog post). Similarly you could use skilled drivers of forklift trucks in a traditionally laid out warehouse or you can dramatically change the layout so robots can move the shelves.
That’s why I think the “it will be a longtime before computers/robots can do x” argument *underestimates* the labor market impact. As always, it will take a bit longer than we assume for these impacts to actually arrive but when they do it will be more profound than we anticipate.
So now for the speculation on strong AI. Personally, I am in the same camp as Patrick. Just because it hasn’t happened in the first 70 years of having computers barely moves my prior on whether it can happen. After all, human intelligence took millions of years to emerge. Also, I don’t really subscribe to the idea that there is anything more to the functioning of our intelligence than, well, our brains.
Here too I think we need to be careful about what exactly we expect to see. Airplanes fly but they don’t fly exactly like birds. Yet, I think we can all agree that the similarity in effect is more important than the differences in technique. The same goes for intelligence. Machines may exhibit human like intelligence and yet use a somewhat different way of getting there.
On Twitter this morning I received a link to a Guardian piece titled “In the digital economy, we’ll soon all be working for free – and I refuse” by Suzanne Moore. What caught my attention was not so much the piece itself, which has many arguments that I have been making on Continuations for some time about technology being the driving force behind income inequality and the disappearance of work. Instead I was intrigued by the “I refuse” part. There is, unfortunately, not much elaboration on this other than a link to a Facebook post (irony?) exhorting creatives to “stop working for free.”
This is of course wishful thinking. Most people do not have the ability to stop working because they are in need of an income. We are facing a problem of dramatic oversupply of labor that we will ultimately only be able to address with redistribution of wealth most likely in the form of some kind of guaranteed base income and/or negative income tax rates at low income levels. That is not an individual choice but rather something we have to do as a society.
The topic has been on my mind because when I just went to Germany, Der Spiegel had a big feature on the impact of the further legalization of prostitution in Germany. Instead of the intended improvement in the safety of prostitutes the primary result has been a dramatic increase in supply (partially through human trafficking) and an attendant erosion in prices (and safety). Why is this important? Because the combination of technology and a huge imbalance in the labor market acts to drive down incomes very broadly.
For instance, much as I am a fan of the idea of letting people share rides in their cars (because moving only 1 person with a vehicle designed for 4 is incredibly inefficient for the environment) we have to also be cognizant that this has the potential to destroy driving a taxi as a job that someone can earn a living with. That is not to say let’s ban ridesharing. Just like I personally don’t believe in banning prostitution. But naive deregulation of labor markets at a time of huge income inequality can have dramatic negative consequences and we cannot be naive here.
So let me repeat my earlier contention. If we want the benefits of the Internet (which I for one very much do), we need to work on the social systems that will make that new society sustainable. I will write a lot more about what I believe those are.
Yesterday I attended MIT's Roundtable on the Digital Economy. It was a gathering convened by Erik Brynjolfsson and Andrew McAfee, the authors of Race Against the Machine, to discuss “Work and Value in the Digital Economy.” The gathering brought together an interesting mix of academics, startups, investors (yours truly) and policy experts.
I don’t have time this morning for more detailed thoughts, but I came away with a clearer grasp of some of the central questions:
1. Is work actually diminishing more permanently or do we just have a disappearance of “jobs” as we know them? The latter might be easier to deal with than the former.
2. Do people need jobs or can we deliver what jobs provide some other way and in a potentially unbundled fashion? The “jobs of a job” include income, structure, social connections, meaning, and at least in the US, access to healthcare.
3. Are emerging marketplaces such as Etsy, Airbnb, ODesk, Relayrides (these were all present) a potential solution? There are important differences between markets that potentially drive down the payout to (commodity) labor (a la Amazon Mturk) and those that may create sustainable activities with the potential for a stable income. More on that in a future post as I had a big Aha moment.
4. How meaningful is the shift from producer surplus (profits) to consumer surplus (value creation) in the digital economy (eg Khan Academy) and how should it be measured? This is an important question as we think about how to regulate some of these emerging providers and marketplaces.
Unfortunately I could not stick around for the last session on regulation, so I have to wait to see some of the transcribed notes from that. All in all it was a very thought provoking day on a topic that I care a great deal about, ie how do we get the transition that’s before us right instead of winding up with massive social unrest.
In part 3 of this series I argued that the changes in employment are the main driver behind the massive rise in inequality in the US. That inevitably brings out a number of responses for how to address inequality, none of which I believe will actually work. At the end of that post I already showed a graph that debunks the notion that this can be solved through better education (at least of the traditional variety). Similarly we now have calls for better labor organizing and higher minimum wages. Unions played an important role during industrialization so why not now? As even Robert Reich concedes, if there are too many people without a job it becomes very hard to organize those who have one. And even though I am sympathetic to an increased minimum wage as a short term measure, in the long run it will only hasten the demise of jobs as machines do not qualify (ditto for healthcare).
What all of these old solutions have in common is that they are premised on the continued growth of traditional jobs. If we want to make progress we first have to abandon that notion and have to think about a world in which there are fewer and fewer traditional jobs. Here then are three proposed solutions aimed at such a world:
1. Encourage and support the creation of new marketplaces that support micro-entrepreneurship, such as Etsy, Airbnb, Taskrabbit, Kitchensurfing, Sidecar, Skillshare and Shapeways to name just a few (including several USV investments).
2. Stimulate demand in areas of the economy that are mispriced or underpriced and hold significant employment potential for instance by taxing carbon and problematic food ingredients and providing prizes for space exploration and medical innovation.
3. Take measures aimed directly at inequality and its consequences, such as unbundling healthcare from employment and changing the tax code to affect some level of redistribution.
Each of these will eventually get its own detailed treatment in a separate blog post. But for now I want to provide some more justification for each and more importantly point out how they work together given the changes in employment.
Micro-entrepreneurship is exciting because it allows people to earn an income often with minimal need for capital. Many (but not all) of these marketplaces are aimed at providing unique products or experiences or otherwise ad hoc tasks that will not easily be substituted by machines any time soon. So from a regulatory perspective we should be encouraging the growth of these marketplaces and the creation of new ones. Instead, we are seeing a fair bit of hasty action, such as the cease and desist orders against Lyft and Sidecar, that could harm the development of these marketplaces.
The second part of the approach is likely to be more controversial. But I firmly believe that it is the central role of government to deal with externalities. And we have a bunch that are not properly addressed today that could create significant net new employment. When people protest a carbon tax by pointing to jobs that would go away they fail to point to the new jobs that would get created. In fact, a carbon tax would mostly hit industries that are heavily substituting machines for humans whereas it would benefit efforts such as reforestation that are labor intensive. The same goes for processed foods compared to local farming and farm-to-table. I strongly favor approaches such as taxes and prizes where the government just sets the framework and lets private activity unfold over having this activity be part of the government itself. In fact, this is a strong complement to the micro-entrepreneurship. For instance, taxing highly processed foods while simultaneously fostering farm-to-table marketplaces (e.g. Farmigo and Good Eggs) will push economic activity from capital intensive to labor intensive which is what we need. Similarly, posting prizes for medical innovation and space exploration will help foster collaboration in emerging research networks such as Research Gate.
Some of the likely criticism of both of these recommendations is that people cannot earn enough money via micro-entrepreneurship and that taxing carbon and processed foods will make staples more expensive thus compounding the inequality problem. Which is exactly why we need to attack inequality and its consequences more directly. In fact all indirect approaches which make labor more expensive are the exact opposite of what we need in a world which is substituting away from labor. Direct approaches aimed at making healthcare accessible and affordable (one of the major issues with inequality) and redistribution through the tax code don’t have that problem.
The goal for today’s post was to get these three pillars of what I think of as the “new new deal” out there. I will provide more meat on each of these in subsequent posts. But I want to address one other immediate question as to how we can afford all of this. Leaving aside for a moment that we cannot afford the alternative, which ultimately heads towards massive social unrest, I am proposing new revenue sources for government. I also believe that by shifting more activity into new and emerging marketplaces we can actually reduce some costly government programs. In fact, re-inventing how we tax and allocate budgets to avoid the silliness that is the Washington’s fiscal cliff could and should be a fourth component.
What I am most hoping for with this post is to move the debate past the old solutions and past the old divisions towards a different set of approaches that do not fall along existing party lines. All of the proposals above really are aimed at moving us away from the hierarchical organization of production which is dominated by a substitution of capital for labor towards a peer economy and society.
This is the third post in my mini series on employment. Part 1 illustrated how agriculture and manufacturing, the two historically large areas of employment, have collapsed over time. Part 2 drilled into the services sector which today accounts for the bulk of employment to show that many categories there have stopped to grow and are under pressure from the twin forces of automation and globalization. Why does all of this matter? My overriding argument is that we have not found a new source of jobs. The result is an unprecedented pressure on wages that is the primary driver of the astounding growth in inequality in the United States.
Much has been written recently about inequality but still the numbers are so extraordinary that it is worth calling some of them out. Maybe the most radical summary came in the form of a tweet by Senator Bernie Sanders in which he claimed that the Walton family owns more wealth than the bottom 30% of Americans. Subsequent fact checking showed that he was off with the real number being above 40%. For some more similar examples see this post. How is such an extreme situation possible? Of course on one end it is the exceptional fortunes at the top of the pyramid. But these wouldn’t be so much a problem in isolation.
The real issue, as fas as I am concerned, is just how tough things have gotten for a large and growing part of the population. A recent report by the Census Bureau showed a further decline in median household income (image taken from the full PDF report which is worth studying)
When looking at this chart it is critical to keep in mind that these are median household incomes. That means half of all households make *less* than that and in some cases by a lot, with almost 10 million households living in poverty.
To see how inequality is driven by what is happening in the labor market consider first the following chart from a terrific blog post titled “Is Decoupling Real?" that investigates the relationship between GDP and income growth:
The chart shows average household income tracking GDP growth but median and Q3 (middle quintile) income detaching around 1980 and slowing down substantially. The only way for the average to grow faster than the median (for any measure) is for the high end of the distribution to grow faster. Translation: incomes at the top have been rising while incomes at and below the media have been stagnant or declining.
But what is behind that decoupling? The next chart shows what has happened to labor’s share of national income. It is from a great post from two researchers at the Federal Reserve in Cleveland titled “Labor’s Declining Share of Income and Rising Inequality”
The two obvious questions are: why is labor’s share declining and what is growing in its stead? Let’s start with the second question as it is covered by the same set of statistics: income from capital has been rising rapidly. The answer to the first question are the forces I have described in part 1 and part 2 of this series: globalization and automation. Both of these will make capital more important than labor. And that is exactly what the statistics show!
The rising importance of capital is a double whammy for inequality. First, wealth is already highly unevenly distributed which means that the returns to capital are highly concentrated among a small part of the population. Second, the shift is responsible for the decoupling between average and media wages shown above. For those with jobs that are complementary with capital (eg, investors, CEOs) incomes have been rising with the increased importance of capital. The latter effect is made even more pronounced as automation is driving increasing returns to scale with winner-take-all (or nearly all) effects in many markets.
And to tie all of this back to the basic premise of my series of posts: traditional remedies simply won’t do. We are experiencing a fundamental restructuring of economic activity on par with the shift from agrarian to industrial society. Pressure on labor will only rise from here on out. We need to think beyond fiscal stimulus and even beyond a simplistic and unreflected mantra of “more education” as the following chart shows (taken from a post title “It’s the distribution, stupid”):
The red and blue lines indicate that a growing percentage of the population has at least completed high school (blue) and even more gains have been achieved in the percentage of the population with a college degree (red). Over that time period productivity (yellow, measured here in change in GDP per capita) has also grown substantially due to automation. All the while median wages (green) have stagnated as we have already seen.
What is the answer then? I don’t know either but in the next post I will take some stabs in what I think is the right direction. It should come as no surprise to regular readers of this blog that much of it will involve the peer progressive agenda.
Last weekend I wrote a post about what is happening with employment. The gist was that employment in the US in agriculture and manufacturing has collapsed as a percentage of total employment and declined even in absolute terms mostly as a result of increased productivity and globalization. All employment growth has been concentrated in services. Today I want to take a deeper dive on services, but before doing so I should point out that I am optimistic about technology in the long run. My point is not at all that we can’t get to a better place eventually — it is that we are being naive about the degree of the dislocation that’s ahead of us and what it will take to get to a better place.
In the previous post my primary goal was to show that entire sectors of the economy can decline in employment. There I treated all of services as one big bucket labeled “other.” Now here is a breakdown of service employment over the last 10 years based on data from the Bureau of Labor Statistics:
The chart shows that the total number of employees has been flat or has even declined somewhat in most categories other than Education & Health, Professional & Business Services and Leisure & Hospitality. Again, keep in mind that over the same time period the US population grew by almost 10%!
One way to gauge how robust the increased demand for labor is in these three areas is to look at what has happened to wages in these three sub sectors (data also from the BLS):
Leisure and hospitality has been relatively flat and isn’t much above minimum wage to begin with. Wage growth in professional and business services started to level of in 2009 and even in education and health services which had the strongest employment growth wages started to level off last year. The pressure on wages even in these areas of growing employment comes from the high supply of labor (as people cannot find jobs elsewhere) but also because even here technology is becoming an even strong substitute for labor.
So what exactly are the mechanics by which increased productivity operates in these services sectors? Here are five different scenarios
Elimination of jobs: these are not being done by a machine, instead the task has disappeared entirely, such as filing papers (when there are no more papers to be filed).
- Complete replacement: examples here are wide ranging from the bank teller being replaced by an ATM machine to the dental technician replaced by a 3D printer.
- Increased human effectiveness: a car mechanic using computer diagnostics can determine in minutes what’s wrong with a car or an executive assistant with email and and Internet can schedule meetings faster.
- Distribution of work to the consumer: for instance with online travel booking the task of picking a flight is shifted to individual end users, reducing the number of travel agents that are required.
- Shifting overseas: many information based jobs from low end data validation to high end engineering can now be carried out anywhere in the world.
- Insufficient seller rents: this last one is a bit tricky but the idea here is that especially with digitally delivered goods the price is dropping to zero so that fewer people can earn a living (at least for now, more on how that might change in the future in a subsequent post) — eg regional newspapers would each employ a sports writer but on the Internet one sports writer can reach the entire nation.
For many of these mechanisms we are just at the beginning of what technology can do. For instance, in legal services computer programs are rapidly replacing humans during the discovery phase in scanning and summarizing documents. We now have driverless cars and in some big open pit mines the trucks are self driving already. Each of these are currently available only at the high end but as the cost of the technology declines will become more pervasive. Education is another example where we have just started a potentially massive transformation. Previously a single teacher could at best teach a few hundred students in a large lecture course. Now someone like Sal Khan has millions of students and large MOOCs have hundreds of thousands of students.
The bottom line here is that it would be foolish to count on continued growth in the services economy to fill the widening employment gap. Already today many of the components of services have either stopped to grow or even started to shrink and the remaining areas of growth are under pressure. And we are only in the early stages of what technology can do in these areas.
In the next post in this little mini series I will talk about how these changes are the leading cause of the massive growth in income and wealth inequality in the United States. I will also come back to the point that we have been using government and consumer debt as a way to try to counteract these changes. That is unsustainable and will set the stage for a subsequent post about the emerging “peer economy” (think Etsy, Uber, airbnb and more) together with thoughts on a more radical view of the future.
One of the truly depressing things about listening to the presidential debates was that both sides seem to be disregarding a fundamental factor in the economy: the displacement of humans by “machines” (in quotes because these days the machines in question are computers). Because politicians are largely ignoring the importance of this trend, the proposed policy solutions all belong to the same tired arsenal of either supply side or demand side economics. It’s a bit as if you were on the Titanic *after* hitting the iceberg and arguing about turning to port versus starboard instead of getting everybody into the life boats.
For context, here is an interesting little chart that I pulled together mostly from US Census data. It shows the composition of employment over the last 200 years.
As you can see the percentage of people employed in the US in agriculture collapsed from a high of over 80% of all people employed in 1810 to below 3% in 2010. Manufacturing (tightly defined) peaked in the 1950s and 1960s, accounting for about 1 in 4 jobs, but has been declining since 1970 and by 2010 was down to only 10% of employment. The rest of employment has gradually shifted into “other” which includes services, transportation, government and more and now accounts for over 85% of all employment in the US.
Now of course over the same time period total employment has grown in absolute numbers and as the following chart shows even as a percentage of the population (again mostly based on Census data)
This growth over time from about one third to almost one half of the population being employed has been driven in no small part by the increased participation of women in the workforce. As far as total job creation goes this is an even bigger accomplishment given the tremendous population growth over this time period. In 1810 the US had barely over 7 million people in it and in 2010 there were over 300 million Americans for a growth of more than 40x!
Against this backdrop of a growing population and growing workforce participation, the collapse of agriculture and manufacturing is even more dramatic. Both of these sectors are in fact down in *absolute* numbers of people employed off their highs. Agriculture peaked in the early 20th century at around 12 million people and is now down to 3 million. Manufacturing peaked in the 1980s at around 22 million and is now down to below 15 million employees.
What is behind the shrinking of agriculture and manufacturing? Innovation and globalization. Innovation has allowed us to be vastly more productive in agriculture and manufacturing (meaning fewer people can achieve the same output) and globalization has allowed us to import food and products from other parts of the world. Innovation is at work everywhere though and so if you assembled similar charts for say China or India you would find the percentage of the population working in agriculture there declining as well. I am not sure whether manufacturing is already declining there, but the announcement that Foxconn is planning to deploy 1 million robots suggests the eventually inevitable direction.
Now the obvious argument here is that surely there must be parts of the “other” category for which employment is growing. That’s something I will dig into more in a second post. But at a high level there is no reason why the growing parts of “other” *have* to add employees faster than the contracting parts are shedding them. Put differently: we are already using fewer people to feed ourselves and fewer people to make stuff so it is entirely conceivable that we will need fewer people for everything else! Essentially all it takes for employment to drop is for productivity growth in “other” to outstrip the growth in demand.
To be continued!
I have been struggling to articulate why I feel quite so concerned about the economy but in a dinner conversation last night I had a kind of Eureka moment. It is now well understood that for at least two decades or so we had economic growth that was fueled (at least in the US) by consumer debt, which in turn was made possible by a housing bubble.
Much of the political discussion today seems premised on the idea that once we are done with de-leveraging, households will return to spending and eventually growth and employment will follow. But let’s examine how the economy grew when it was not fueled by consumer debt. For many decades (arguably since World War II and possibly even further back) we had a successful loop working: consumers wanted to buy new things, companies made those new things and in the process paid out wages that allowed consumers to buy those things. This is a virtuous cycle. Make more stuff, have more income, buy more things which you or someone else makes, who then has income, and so on.
What happened though is that technology has gotten in the way. In particular, we have gotten so good at making things (i.e., productivity has gone up so much) that how much we need to pay to labor has started to decline. This breaks the virtuous cycle. Yes we make more things but it no longer results in higher wages and so we can no longer afford those things. We were able to hide this breakdown using debt. Now that the debt is disappearing, the breakdown is becoming obvious. We have had the least jobs created in any recovery from a recession. And I believe that this is just the beginning. My former thesis advisor at MIT, Erik Brynjolfsson, has really been digging into this. It is worth reading “Race Against the Machine" co-authored by Erik with Andrew McAfee, which lays out the case that technology is rapidly displacing labor.
So just to restate my epiphany: the technology-induced break in the virtuous economic growth cycle didn’t just happen but goes back maybe as far as two decades. We simply used debt to hide it. If that is indeed the correct explanation then the employment situation is going to continue deteriorating.
PS I have made a fair number of assertions in this post that I should really link to data for but I am on the road in London and a bit pressed for time. It would be great if anyone reading this wants to provide some links as part of the comments.
PPS Globalization also plays a role here. Apple yesterday announced the iPhone 5, which will generate billions in US sales and even profits to Apple. But very little of this will go towards paying US domestic wages.
Last year when there was a big debate about carry taxation, I came out in support of taxing carry as income (even though that will clearly mean a much higher tax bill for me personally). With the budget battle in Washington reaching ever new heights this issue is coming back in full force. I missed Nic Kristof’s column on this last week. I am completely in agreement until he gets to the point about taxing founders:
One important proposal has to do with founder’s stock, the shares people own in companies they found. Professor Fleischer has written an interesting paper persuasively arguing that founder’s stock is hugely undertaxed. It, too, is essentially a return on labor, not capital, and shouldn’t benefit from the low capital gains rate.
The Professor Fleischer quoted here is a law professor at the University of Colorado who has apparently not spent a lot of time with entrepreneurs. To characterize their gains as a “return on labor” is completely misguided. In reality this is a “return on passion, ingenuity and creativity” and it should be taxed at *less* than capital gains from trading!
Why? Because we desperately need more of it. Employment is created by new companies in new industries. Taxes are all about influencing incentives and if anything given the terrible employment numbers we want to tilt incentives dramatically towards starting new companies.