Longing for good jobs: the deliveroo strike and the future of work

Photo by Russell Davies on Flickr

In this post, I start from the recent strike performed by riders of the UK delivery startup Deliveroo to look at alternative scenarios concerning the creation of “good jobs”. I claim that “good jobs” are not about the relationship between jobs and technology but rather a product of the total value created by it. Attempts to enforce the creation of good jobs through redistribution are bound to fail in the absence of sufficient value to be shared. On the other hand, protecting “poor jobs” by out-of-market subsidies like tax credits and basic income will result in a lose-lose situation where neither innovation nor fairness are achieved. Understanding the unintended consequences of alternative actions should lead us to a different approach: less interventionist and more evolutionary.

What is a good job? And, how can we protect the one we have while creating more of them?

The recent successful strike conducted by riders of the UK delivery startup Deliveroo offers a good chance to look at these questions in a practical case, outside generic and buzzword-laden discussions about the future of work.

Machine jobs, human jobs and “good” jobs

A typical mistake we make when talking about the future of work is to think of jobs created by new technology as involving, by default, sophisticated skills, new knowledge, or both. In short, an advancement on what we used to do before. While this is true for some people, it is not a general rule.

A more realistic approach is to look at jobs that humans can do because we are better at them than machines, where better doesn’t stand for “more intelligent”, but rather “diversely intelligent”. Historically, technology has relentlessly taken over tasks that were a good fit for it, humans have kept everything else (which, to be fair, includes a whole amount of new jobs being created). A cynical way of putting it would be to say that humans occupy the space left over by the machines [1].

The gig economy, and in particular food delivery, is no different in this sense. It is a product of information technology. On the buyer side, twenty years of e-commerce and now almost ten of smartphones have created the very basis for the on-demand movement, liberating latent desires to get everything we want, when we want it. On the the supply side, the same tech allows people to find work (and to be found by it) anywhere, anytime. All glued together by beautifully designed digital platforms. Discovery, recommending, searching, matching: all machine tasks. Delivering, especially if done by bike in a busy city: human task [2].

Seen in this light, there is nothing intrinsically bad about gig economy jobs. They are an updated manifestation of a machine-enabled division of labour started in the industrial age [3]. Pushing the argument a bit further, the emergence of the gig economy demonstrates what technology optimists keep saying: new tech creates new jobs, possibly more than those it deletes.

What makes then people so worried, and workers so upset? What makes these jobs worse, in the public eye, than “good old” industrial economy manufacturing (or clerical) jobs? Or, looking from the other side, what made these better? They didn’t need special skills, they didn’t need high education, and they could count on plenty of supply. The answer, as a survey of recent researches performed by the magazine fivethirtyeight shows, might come as a surprise:

“…this much is clear: For all of the glow that surrounds manufacturing jobs in political rhetoric, there is nothing inherently special about them. Some pay well; others don’t. They are not immune from the forces that have led to slow wage growth in other sectors of the economy. When politicians pledge to protect manufacturing jobs, they really mean a certain kind of job: well-paid, long-lasting, with opportunities for advancement. Those aren’t qualities associated with working on a factory floor; they’re qualities associated with being a member of a union.”

Which takes us back from where we started: the Deliveroo strike.

Bringing unions back to the gig economy

DonkeyHotey on Flickr

It would seem, therefore, that what gig economy jobs have lacked so far in order to improve their status is collective bargaining power. The same that led to the painfully achieved social contract of the post war years and turned industrial jobs into “good jobs” even when, as it proved later, this was a distortion of free market dynamics. I will get back to this point shortly.

To confirm this impression, it is sufficient to look at the reactions the Deliveroo strike generated among sympathetic observers. Like the New Economic Foundation (NEF), a “think tank promoting social, economic and environmental justice”, which in a recent comment to the events wrote:

“Insecure, badly paid work is bad news for the economy overall. A weak voice for employees’ rights leads to a lopsided share of national income going to company profits and sitting in banks, rather than into wages and spending in the real economy.”

“The divide and conquer approach of splitting up this workforce into thousands of ‘self-employed’ contractors, managed via an app, may seem appealingly efficient to an employer. But the power imbalance created is counterproductive if we look at the bigger economic picture.”

“This victory, however small, is symbolically very significant. The strike is an example of collective bargaining navigating a new frontier of app-enabled work — and the new power imbalances that come with it.”

According to the people at NEF, we are facing essentially a question of redistribution, where “collective bargaining” will prove key to navigate the “power imbalances” that come with “app-enabled work”.

What these commentaries miss is the reason, along with the possibilities created by technology, why gig economy jobs exist in the first place. They have not only suffered from power imbalances, they own their very existence to these imbalances. To put it more clearly: these jobs have emerged in the last years thanks to a regulatory loophole (the independent worker vs employee status) which has allowed the achievement of a lower-than-normally-permitted equilibrium between demand and supply (a “clearing price” if we want). But while this low clearing price has provided the fuel to kick start innovation in the on-demand segment, its days may be numbered. At least in the view of union leader Mags Dewhurst:

“..this kind of tech has been designed to isolate individuals and atomise work, deskilling the industry and driving down wages…”

“…the difficulty we faced had been accessing them. How do you reach thousands of people when they are spread around London? I was literally chasing them down the street if I saw a blue jacket.”


“The biggest problem people face is getting in contact with each other. Once they are in contact and they have decided to work with one voice, they have effectively unionised and the company is screwed.”

“This [the Deliveroo strike] is just a taste of what could happen when low-paid workers campaign together”

The recent strike may well put an end to the low clearing price of the gig economy (at least part of it). But is the overall economics of the model sufficient now to redistribute the pie? Or..

Can gig economy jobs be “good jobs”?

The answer to this question is likely to be negative. Looking at why this is the case can help us understand some of the characteristics that make a good job sustainable.

Let’s start by looking at the redistribution-through-unionisation option.

Gig economy jobs (delivery in particular) carrie an intrinsic weakness that practically makes it impossible for them to become good jobs just based on a redistribution of power between the parties. The short version is that they create too little additional value between the restaurant and the customer for both the platform and the worker to benefit from it [4]. Ultimately, delivery companies simply cannot afford to offer better conditions to their riders.

Workers are faced therefore by a real “squeeze”. In the short term, the acceptance of their requests (and thereby the removal of the low clearing price) has good chances to put their employers out of business. In the long term, increasing the cost of delivery might create a sufficient incentive to push for automation [5]. No matter how we see it, a better bargaining power from the workers side will likely destroy the long term prospect of delivery jobs to become sustainable [6].

An alternative option is what liberal thinkers in my twitter feed have been advocating: redistribution after market equilibrium.

This position is clear. Improved bargaining power (which is what will achieve minimum wage for Deliveroo riders) might be good in the short term, but it has an overall negative effect for the economy as it stiffens innovation and limits employment opportunities [7]. This is basically an argument to maintain the low clearing price mentioned above, without putting the burden entirely on the workers’ shoulders. The alternative is for someone else to take fill the gap, and that someone is the “public”. An example of this redistribution outside of labor market equilibrium is tax credits, advocated in the US by prominent business leaders like Warren Buffet as a measure to be preferred over the distorting effects of minimum wages.

But it is not so simple. On one hand, maintaining an artificially low cost of labour puts a (publicly subsidised) break on innovation, as the incentive for these companies to automate will be lower. On the other, it risks satisfying only partially the request of the riders, in particular their intangible, but equally important, sense of fairness. Their economic needs might be taken care of, but they will probably still feel “exploited” as competition for work coupled with a strong system of tax incentives pushes the salary per hour towards zero. Profit will be accumulated at the corporate level (in few hands) while society will foot the bill [8].

To recap:

Unionisation satisfies better our sense of fairness but ultimately has low chances of achieving the objective of turning bad jobs into sustainably good ones. Outside-of-market corrections seem to have a better chance of saving the jobs (assuming this is really what we want) but to do so require a privatisation of benefit and a socialisation of costs (both in terms of innovation and financially).

An interesting dilemma:

Fair but bad for jobs vs. Good for jobs but unfair

Other possibilities

Basic income:

A discussion about the future of work would not be complete these days without talking about basic income. Leaving for another time the different nuances of the concept, in this specific instance I see its effects in many ways similar to those of tax credits. The main difference between the two is that tax credits require the beneficiary to have a job, basic income doesn’t. This will likely generate less competition for delivery jobs and less downward pressure on wages, but I don’t expect basic come to simply make supply disappear (and therefore push for more automation). People on basic income will keep looking for jobs driven by the desire to increase their disposable income (even in the most ambitious proposals still quite low by general standards) while having less incentive to “fight” for a minimum wage compared to a situation without neither basic income nor a system of tax credits.

Modern cooperatives:

Another option is to maintain the low clearing price by eliminating platform profits. Riders owned cooperative could keep the price of delivery low enough to be acceptable for customers but high enough to be sustainable for riders. Automation would be the only real long term challenge, as companies that want to reintroduce profits into the picture would do it by substituting the cooperatives with robots as soon as that option becomes economically sustainable.

Voluntary pay:

A true market change would come if customers themselves shift their willingness to pay for human deliveries. This would immediately increase the size of the pie and, depending on how much more they would like to pay, make room for a redistributive scenario. The consumer dilemma between “cheaper” and “more sustainable” is well known (watch “The High Cost on Low Price”, documentary about Walmart) and a change of heartdoesn’t seem to be anywhere near. On the other hand, the organic food movement proves that for some principles people are ready to pay more, but this requires an entire discussion on its own which I leave for another time.

What have we learned?

Time to draw some conclusions.

Good jobs require stronger fundamentals. This might be obvious but it is worth repeating.

  • Wealth creation comes from true innovation (technological or in business models) where additional value is released in the economy. If innovation limits itself to the exploitation of a temporary arbitrage it wont survive the inevitable pressure for redistribution that is bound to emerge no matter what.
  • Focusing on the machine vs. human dichotomy is good for headlines but it distracts us from identifying tasks where humans have a sustainable role to play. This is more dependant on economic factors than pure technical feasibility of a given automation.

There is really no easy way out. No amount of external intervention, being it pre- or post-market equilibrium can guarantee the creation, and maintenance, of good jobs.

  • Going back to old protections have a normal element of fairness (and I am myself tempted to wish for it when i read the stories in the newspaper) but is no long term solution.
  • Deep market corrections (although more innovative and pro-market like tax credits and even basic income) have nonetheless a distorting effort. Paradoxically, they will postpone innovation and keep human in low skills job (for those that think this is a problem).

What do I think about all this?

I started writing this post as a way to understand what was going on and to look tangibly at different scenarios and their implications. As it often happens, the more you try to genuinely understand a topic the more you realise that there is no simple answer. That would be too easy. But since it is wrong to just sit and comment without proposing something I’ll indicate my preferred choice.

Faced by competing choices with no apparent (at least not to me) solution that achieves both innovation, job stability and fairness, I consider non-intervention to be preferred. So, no intervention to settle the dispute by imposing a min wage, and no helping hand (at least not explicit) to the company by appeasing workers through other public transfers.

This position is not motivated by a blind acceptance of the “invisible-hand” but rather by the acknowledgment that each choice opens up for a range of “unintended consequences” that risk overshadowing even the best original intentions. In situation of high uncertainty, policymaking should refrain from the temptation to impose a specific solution and adopt instead a “stewardship” approach as advocated by Eric Beinhocker and the “complexity economics” school

“…rather than engineering specific outcomes, government’s role as system stewards is to create the conditions in which interacting agents in the system will adapt towards socially desirable outcomes”

In our case, this means levelling the playing field between riders and platforms, removing information advantages (for example with better access and export right for riders’ data) and allowing therefore a more natural rebalancing of power, also through collective action.

In a way, my suggestion is about embracing the historical process by which a job first starts outside of the controlled labour market, then gets rebalanced , and if it is not able to survive, disappears or gets automated. It is not the end of the world, it’s evolution.

[1] I subscribe in full here to Venkatesh Rao’s definition of machine work as “algorithmically-scalable work” and human-work as “work that is not worth automating. This is a combination of work being hard to automate and low returns to the automation due to limited algorithmic scaling potential (a good example is tax software for parts of the tax code that change very frequently).” http://www.ribbonfarm.com/2013/07/10/you-are-not-an-artisan/

[2] Yes, I am perfectly aware that self-driving cars are coming and that people are working on projects like https://www.starship.xyz/ . Even so, delivering small parcels and food in highly dense urban areas (including walking up staircases at times) is a costly endeavour for machines and not a place to look for high “algorithmic scaling potential” (see above)

[3] Gig economy jobs receive a lot of criticisms for being “bad jobs”. For the purpose of this discussion I have decided to leave out the “servitude” argument, criticising these jobs for not exploiting the full “human” potential and for being the manifestation of a widespread inequality in society (rich of money poor of time on one side, rich of time poor of money on the other).

“The Servitude Bubble is condemning people who might be doing amazing, wondrous, and miraculous things to be butlers, maids, dog-walkers, net-servants — or, perhaps worse, code-monkey enforcers who, chasing their own little payday, make people into net-servants.”


[4] Here, a distinction between Uber and the majority of other gig jobs needs to be made. Uber innovates more deeply into the structure of its industry. By streamlining operations (eliminating a large part of the overhead in traditional taxi dispatching business) and by liberating latent demand for a business with low variable costs, it allows, in principle, enough room for both platform and drivers to benefit. Food delivery, on the other hand, operates in a business where variable costs (food) are high and where it is much more difficult to find optimisation elsewhere. With little operational efficiencies to be achieved on the production side (the restaurant) and a ceiling on the price side imposed by the consumers’ willingness to pay and constant fear of new investor-subsidised entrants, there is simply not enough room for both. An extensive coverage of this point can be found here.

[5] Even where a certain job is not obviously a “machine job”, sufficient economic incentive might change the picture.

[6] This is also easily demonstrated by the fact that these jobs don’t exist in settings where the clearing price was simply not allowed to go lower enough .

[7] Low clearing price jobs (and people that take them) prove that there is latent demand for jobs even below what people normally think is fair.

“However, the fact that online platforms can attract workers in large numbers despite relatively unappealing conditions is an indictment of the labour market in many of the countries where they operate. In France, many Uber drivers come from the deprived banlieues and say that discrimination prevents them finding mainstream employment. In the US, a study by JPMorgan found that people used online platforms mainly to help them weather a dip in their regular earnings — as an alternative to cutting spending or running up debts.”


[8] Somebody still believes in “trickle down effects” and in the possibility to effectively tax corporate profits. I don’t.

Pension vs earnability (appended)

This excerpt from Philip Auerswald's contribution to the Kaufmann's "New entrepreneurial growth agenda" adds an interesting angle to my earlier "Pension vs. earnability". It's worth a read. 

Redefinition: From “Permanent Income” to “Dynamic Purpose”

The primary value of the foregoing analysis is to establish the context for a redefinition of the problem posed in The Great Man-Machine Debate. How do we get there?

Consider the following: In the immediate aftermath of the 2008 financial crisis in 2008, economists devoted a great deal of attention to the shortcomings of macroeconomic and financial models that, at best, failed to predict the breakdown, and, at worst, may have helped to bring it about. Hyman Minsky’s “financial instability hypothesis,” to which few previously had paid much attention, was newly celebrated; Eugene Fama’s “efficient-market hypothesis” was newly questioned. Yet, as events have unfolded, the profession has begun to take more seriously the structural factors that are shaping the twenty-first-century economy and driving economic outcomes that go beyond the business cycle. From the standpoint of the reconsideration of theory, this means shifting attention from macroeconomics to microeconomics and rethinking fundamental models of both consumption and production.

High on the list of models ripe for reconsideration is the “permanent income hypothesis,” introduced into the field of economics by Milton Friedman in 1957 in a book titled, A Theory of the Consumption Function. The idea, as we all know, is simple: Early in life, we as consumers optimize our lifetime earnings by going into debt to invest in education; education delivers the skills that form the foundation for a career. Early in our working lives, consumers stop investing in education and start to save. We keep saving increasing fractions of our income until, all at once, we retire. At that point, we spend down our savings, timing the depletion of savings to coincide perfectly with the depletion of … well, our lives.

The model Friedman developed of the arc of a human life is as technically sound today as it was in 1957. Furthermore—somewhat like the efficient markets hypothesis, which was also developed at the University of Chicago at about the same time—the permanent income hypothesis has become encoded in the operating system of the economy in such fundamental ways that we barely notice its influence. From the Pell Grants to the 401(k), the experience of consumers from youth to death remains framed by the notion that institutions are sufficiently slowly changing and we are sufficiently short-lived that we can invest (one time only) in education at the front of our lives to reap a reward that we ultimately enjoy at the end of life. Predictable and familiar policy prescriptions follow:

    • Too much student debt and too few quality jobs for recent graduates? . . . Need more and better education.
    • Too much unemployment and too little stability in the labor market? . . . Need to spend more on worker protections.
    • Too many retired people and too little saving? . . . Need more and better health insurance.

In a recent column for The New York Times, Robert Shiller wrote: “Most people complete the majority of their formal education by their early 20s and expect to draw on it for the better part of a century. But a computer can learn in seconds most of the factual information that people will get in high school and college, and there will be a great many generations of new computers and robots, improving at an exponential rate, before one long human lifetime has passed.” Colleges and universities have yet to respond adequately to these changes, Shiller concluded. “We will have to adapt as information technology advances . . . . We must continually re-evaluate what is inherently different between human and computer learning.”

Shiller is right: We need to update our thinking about the function of higher education. We also need to update our thinking about workforce training, retirement, and aging to fit the realities of the twenty-first century.