A framework to understand inequality in the gig economy
If we closely analyze the mechanics that drive todays platforms and networks, we learn — rather disturbingly — that digital platforms are designed to drive greater social inequality, not reduce it.
This is ironic because we’ve often thought of networks as infrastructures for fairer distribution. We understand that robots are bad, they’re out to eat all jobs, and our dystopian fears are closer than we think. We get that!
But networks and platforms? They often seem to form the silver lining in the technology-breeds-inequality debate. If platforms re-intermediate market interactions more efficiently — the argument goes — then they should be making the world more equal by democratizing resource access for everyone. Or so the gig economy enthusiasts would have us believe.
Why wouldn’t an algorithm-driven world of platforms make things more equal? Why would it, instead, drive further inequality?
This isn’t meant to be a comprehensive treatise on inequality. Sharing economy enthusiasts often claim that while some of technology may be driving inequality, platforms — and, in particular, labor platforms — will create more access and choice and help reduce inequality. This article focuses only on understanding whether greater democratization and lower inequality is inherent to a more networked world.
Let’s look at the mechanics of platforms to understand this better. The impact of platforms may be analysed at two distinct layers:
1. The layer of the ecosystem that builds around the platform
2. The layer of the firm that powers the platform
Let’s dive into the mechanics of the ecosystem first.
Ecosystems around platforms scale non-linearly and in the course of such non-linear adoption, platforms need to scale their curation and quality control to manage ecosystem interactions. Most platforms do this by instituting reputation systems.
Mechanic #1: Reputation systems
Reputation systems track, determine, and encode the reputation of different participants in the ecosystem. These participants are consequently accorded market access and influence based on how reputed they are. As an example, Airbnb hosts who get booked and rated higher more often are likely to show up higher on search results. Higher reputation leads to greater market access and influence.
From the perspective of the ecosystem members, the rules of market access change. Instead of appealing to an unscalable, editorial gatekeeper, you pander to an algorithm, driven by social feedback. Instead of auditioning to a judging panel, you go viral on YouTube. In most cases, a market-driven system to determine access is fairer than one based solely on human judgment. However, it’s not necessarily making the ecosystem more equal.
It’s merely changing the rules on which the inequality operates.
One may argue that inequality is a feature, not a bug, of any meritocratic system. There is, therefore, nothing wrong with a new inequality.
That sounds fine, until you combine Mechanic #1 with the other four mechanics that drive further inequality on platforms.
Mechanic #2: Positive feedback loops
We noted above that higher reputation leads to higher market access and influence.
In a connected ecosystem, higher market access and influence, lead to more interactions for users, which — potentially — further increases their reputation. A virtuous cycle sets in.
Reputation systems benefit from these positive feedback loops. The higher your reputation, the greater your influence, the greater the likelihood that you attract further ratings. The ones who have it tend to get even more while the others find it difficult to solve the cold start problem.
On platform-mediated markets, the rich get richer and the poor get stuck.
The notion of feedback loops is an important one not just at the level of individual platforms but at the level of the overall internet as well. People who build reputation and influence across multiple platforms, rapidly realize that the cross-feedback between different platforms, once connected well enough, brings in further non-linear increase. Feedback loops operate in any connected system and as various aspects of our economy get more connected, their ability to drive further inequality increases in ways that cannot always be traced.
Positive feedback loops, when implemented unchecked in a system, drive greater concentration of value at the top, hollow out the middle, and drive the majority of the participants further down. As a result, those who rise to the top benefit from feedback and emerge as superstars while those who stay at the bottom find it increasingly difficult to break through. Positive feedback loops are the primary driver for greater inequality in a connected system.
System designers often solve this problem by damping the impact of positive feedback loops. To counter unchecked positive feedback, well structured systems architect negative feedback into the system. We will note, however, over the course of exploring the next two mechanics, that platforms are naturally incentivised to inhibit negative feedback and promote more positive feedback.
But, you’d argue, we’re structuring positive feedback based on market reputation for the overall good — improving the quality of interactions on the platform. We’re simply allowing the market to choose who gets more and the market will reward quality.
The logic works well, except for one important caveat.
Mechanic #3: Unfair advantage
Platforms often deliberately encourage unfairness. This doesn’t stem from malice. All unfairness, as we’ll note shortly, is in the best interest of good platform execution.
To start with, an unfair advantage is baked into the very workings of a platform. Since there’s a feedback loop in play, the users who come in earlier and invest in the platform gain a significant first mover advantage. They benefit from the feedback loop for a longer period of time. Moreover, the platform also errs on the side of exerting less control on the ecosystem.
Users who come on to the platform much later do not get this early headstart. More importantly, by that time, the wild west days of the platform are over and the platform starts exerting tighter control. Users who created Pages on Facebook in the early days saw greater amplification for their posts and were able to bring in a larger fan base than those who created Pages much later when the platform started clamping down on the amplification of posts.
Hence, users who come on early onto a platform often benefit from an advantage over users who show up late. As time passes on, the ability of every subsequent cohort of users to gather influence on the platform changes significantly.
There is a second form of unfair advantage that some users benefit from, which is deliberately infused by the platform owner. In its early days, a platform often struggles with the cold start problem: users who join the platform rapidly get deactivated because they don’t find great content on the platform.
To solve this problem, platforms often curate a small number of high quality producers in their early days and encourage new users to follow these producers. Early users who get selected during this curation find their influence rapidly amplified by the platform. Users who didn’t join early enough do not benefit from this mechanic despite similar quality.
As the platform grows, it constantly expands the list of suggested users. However, the amplification that the first few high quality users received from the platform, when it was desperate to increase new user activation, is much higher than that received by users who are added to this list later, when the platform achieved steady state and new users joining in have a wide range of high quality content to choose from.
Sometimes, employees of the platform company who participate in the ecosystem may also benefit from this unfair advantage. Adam Rifkin speculates in this Quora answer that Tristan Walker possibly gained a huge following on Twitter because he was an intern at the company when the platform first created a small curated group of users who would be suggested to new users when they joined Twitter. Tristan is excellent at engaging his followers on Twitter but it’s unlikely he would have gained followership at a similar scale if he hadn’t got the headstart that the Suggested Users list provided.
When platforms favor certain users above others, users with similar quality contributions but favoured differently by the platform may end up with very different levels of influence on the platform. Quality alone doesn’t determine influence, as it would in a well-functioning meritocratic market.
There is a significant first mover advantage for users on a growing platform.
A rising tide lifts some boats much higher than the rest. And then, the feedback loop kicks in. Users who get a headstart because they were favoured by the platform during the early days benefit from increasingly higher amplification in the days ahead. The gulf between the haves and have-nots increases further, not entirely based on merit, but based on being at the right place at the right time.
There is a fourth mechanic at play which makes the effect of feedback loops much stronger.
Mechanic #4: Global reputation trumps local reputation
Before the invention of recorded media, artists thrived on local reputation. Every town would reward its local artists. Theater flourished and local singers made a good living entertaining their local audiences. Artists flourished not on the basis of the quality of their art but on the strength of their locally accessible audience. A few would break through this barrier, build larger reputation, and travel to many different patrons and audiences.
The invention of recorded media changed everything. Audiences didn’t care about listening to live local artists when they could listen to the best voices over recorded media. The discovery of the best artists led to further patronization by wealthy families, which led to the rise of a few superstars and the decline of all others. Audience adoption and patronization by the wealthy concentrated among a few artists, changing the mechanics of the industry and giving rise to superstars.
Something similar played out with the death of theater and the rise of Hollywood. As the quality of movie recordings improved, theater actors struggled to compete. Over time, theater started working more on global reputation as well with a few (Broadway, West End) concentrating the best talent and with actor touring the world to harvest their global reputation.
The technology of recorded media drove inequality in the creative industries. Global reputation trumped local reputation. Today, we take this inequality for granted in the creative industries. But we’re now seeing other professions giving rise to similar superstar economics.
In the age of networks, any form of work that rewards global reputation over local reputation will see increasing inequality.
Job/task marketplaces rely heavily on reputation systems to match the best service providers with customers. The more specialized the job, the greater the importance of the reputation system. In the case of information-oriented jobs, that do not require physical co-location, global reputation starts trumping local reputation. High-end consulting, data science etc. are already seeing this set in with high-skills marketplaces polarizing and amplifying the earning potential of a few highly reputed workers who serve clients globally. Consumers on these platforms prefer hiring the best talent, irrespective of geography. Prior to the existence of these platforms, and online professional networks like LinkedIn, consumers would prioritise convenience in search over quality. Local reputation mattered more than global reputation, much like it did for singers before recorded media was invented.
As more skilled work moves on to platforms, many more markets will start demonstrating the superstar economics that we already see in the creative industries. Whenever global reputation matters more than local reputation and a platform re-intermediates such a market effectively, the few top workers will command inordinately high prices.
Again, workers who show up early on such a platform have a better shot at mechanics #1, #2, and #3 working for them.
There is one final mechanic that determines which markets will allow greater social mobility.
Mechanic #5: Specialization of skills
Let’s move the discussion from highly specialized labor to low skilled work. Unlike specialized work, low skill work isn’t just low pay, it also comes with relatively low career progression.
We understand this intuitively. High skill work attracts significant career progression, low skill work involves low progression as well.
In a networked world, this gets worse.
In a feature report in the MIT Technology Review last year, I noted that platforms that orchestrate specialized labor tend to play fairer to their ecosystem than those that orchestrate unskilled labor. Rentacoder vs. Uber is a case in point. Rentacoder supply is highly specialized and differentiated. Customers are highly sensitive to quality. Uber supply, in contrast, is very commodified. All rides are roughly the same, above a minimum threshold of quality. Decision making on Uber boils down to finding the nearest available vehicle at that particular time. Customer decision making on Rentacoder is much more nuanced. Customers consider multiple factors while making a decision and the order of importance of these factors varies from one customer to the next.
This, of course, is the nature of markets. Commodities eventually compete on price and convenience. However, this shows why the gig economy is not the panacea for rising unemployment that it is made out to be. Skilled but unemployed people who enter the gig economy for commodified work (think TaskRabbit, Zaarly etc.) will have a harder time making their way back to any form of skilled work. Undifferentiated gig workers work in markets which do not allow any meaningful form of career progression. In skilled labor markets, the reputation systems of platforms help the best workers rise to the top. Unskilled labor markets leverage reputation systems only to weed out the bottom, not necessarily to amplify outcomes for the top (Think Uber removing drivers at the bottom but not increasing rewards for top-performing drivers).
In an age of platforms, the gap between skilled labor and unskilled labor will get further amplified.
As the five mechanics above demonstrate, networked systems are naturally architected to drive greater inequality.
At this point, let’s shift gears and move away from the ecosystem to the firm. In a connected world, there will be a divide within the ecosystem. But there will be a greater divide between the firm and the ecosystem.
This part of the argument is more along the lines of throwing rocks at the Google Bus. However, understanding the mechanics at play is helpful.
We are quick to sloganize that programmers are ruling the world and everyone should start programming. Not only is that the wrong conclusion to jump to (all sweeping sloganize that programmers are ruling the world and everyone should start programming. Not only is that the wrong conclusion to jump to (all sweeping generalisations of this type are), it also successfully glosses over the mechanics at play.
Programming has been around for a long time. However, the rates at which value has been captured by programmers has been captured by programmers has changed over the last decade. Over the last decade, and a little before that, we’ve realized that a networked world offers a unique opportunity to leverage labor and property outside the firm to create profits for the firm.
The math works out great. Platform businesses like Facebook, Google, Uber, and UpWork grow their ecosystem at near-zero marginal costs of expansion but benefit from the property and labor that is owned by the ecosystem. The costs of managing the property and labor are passed on to the ecosystem while a portion of the profits UpWork grow their ecosystem at near-zero marginal costs of expansion but benefit from the property and labor that is owned by the ecosystem. The costs of managing the property and labor are passed on to the ecosystem while a portion of the profits accrue to the platform.
This is where programming comes in.
Programmers that improve the platform’s ability to scalably orchestrate such interactions in the ecosystem are highly valued because they are able to scale value capture back to the platform non-linearly.
All programming may not pay well but any programming (and associated skills) that enable a large platform to capture greater value, through better orchestration of the ecosystem, is going to be highly valued.
This, in turn, drives significant inequality between those who are employed by such firms and those who work it out in the ecosystem.
The architecture of networks and platforms isn’t really helping reduce inequality, it is instead poised to drive it further. Across the mechanics laid out above, we note how markets that reward meritocracy drive further inequality in an age of platforms. Moreover, not all of this inequality is merit-based, some of it can be attributed to a more structural form of luck as evidenced by Mechanic #3.
We did propose an architectural solution to a more equal world in the narrative above: the deliberate design of negative feedback loops to dampen the system. However, as mechanics #3 and #4 demonstrate, platforms shy away from deliberate damping when (i) it comes in the way of user activation on the platform (Mechanic #3), and (ii) global reputation trumps local reputation, as is the case with a lot of specialized, knowledge work (Mechanic #4).
Connectivity is going to be great for the world. Connectivity will drive access to a better life. But if we expect connectivity to solve the problem of inequality, we will need to deliberately address this issue in the architecture of the connected systems that we build. Inequality rises from architecture, and to some extent, can be resolved within the architecture.
We’ve seen repeatedly in the evolution of social systems that policy, alone, is ineffective at solving inequality. Policy needs to be complemented by architecture.
But in order to do that, we need to make deliberate choices in the architecture of the ecosystems, optimizing not merely for ecosystem performance but for democratic progression and equal access to the opportunities a networked world offers. As we noted above, positive feedback loops of reputation and quality are allowed to work in many platform architectures because they help create a stronger ecosystem. However, working unfettered, these cycles can reduce access to opportunities and progression for users, unless dampened deliberately. This dampening doesn’t have to come at a cost to platform performance, but it does need to be accompanied by a rearchitecture that balances the need for maximizing ecosystem output with the need for balancing access and progression within the ecosystem.
Architecture helps us understand the systems we inhabit. Hopefully, architecture can help us manage some of their complexities better as we move forward.
How to spot platform opportunities in your industry
Learning from the early days of Facebook.
How Lego, Hot Wheels, and others create walled gardens to compete for your child’s attention