The future of labor signaling
Ben Lawton
•
Marketplace Quant
May 29, 2026
May 29, 2026
Your economics intuition is lying to you because you didn’t read through the formulae.
In 1973, Michael Spence proposed a now widely-accepted theory of how the information asymmetry and candidate congestion of labor markets is ameliorated. His theory, which later earned him a Nobel Prize in Economics, hinges on signals, information conveyed to the other side of the market that must as a rule be costly for the signaling party.
Classical examples of signaling in theory are degrees at coveted institutions, additional credentials like an MBA, or even developers’ contributions to open source codebases before seeking a job. This theory is not just a statement about what we associate with value, it’s a model-able result on expected values. And it relies on two specific priors or sets of beliefs:
1. The candidate or signaller’s belief about the cost of obtaining new credentials
2. The employer’s belief in the extra productivity of 1 additional employee, referred to as the Marginal Product of Labor
What are those beliefs? It can be summarized as the conviction: I am a shark, so I won’t suffer as much in getting a competitive education or an expensive degree. After all, I will pass more easily and I already know my career will make it pay off. As well as: All the people who went to these top-tier institutions are more productive, I’ll pay a premium for them
These beliefs reinforce each other and create a cycle - already productive individuals pay the reduced cost knowing they will be paid a premium or preferred in the job hunt, and because these employees are more productive, the employer’s beliefs about accredited candidates being sharks is confirmed.
The challenge with a signalling framework is precisely that it relies on priors, and priors are only valid as long as those who hire can credibly hold those beliefs. Take MBAs. The greatest beholder of the MBA is surely the consulting firm, however consulting hiring trends are starkly downwards. Our internal data shows that in early 2025, roughly 13% of non-engineering role postings asked for an MBA, named a top business school, or used top school credentials as a hard requirement. A year later, it's 6.6%. Interview rates for candidates from Top 7 MBA programs have declined in lockstep - from roughly 38% in early 2025 to under 22% by mid-2026. The WSJ has documented the same erosion from the demand side: even graduates of the most coveted programs are struggling to land roles their predecessors treated as table stakes (WSJ, 2025).
These trends erode prospective candidates’ belief in the payoff of credentials and breaks the cycle of talent earning an MBA.
On the other side, we have a decrease in the belief of hirers on the extra productivity. AI has driven rapid evolution of the Marginal Product of Labor (MPL). Actual studies are lagging behind the pace of AI improvement, but consider a frequently cited study from NBER which estimated in 2023 that the Marginal Product of Labor, specifically for human customer support agents, increased by 14% on average as a result of generative AI tooling (NBER, 2023). These tools, however, have only been broadly available since 2022 - ChatGPT was first released to the public in Q4 2022.
Very conservatively, that means that over a 1 year period, certain white collar job types experienced a 14% increase in productivity. Given the ability of today's agents to directly interact with code, databases, GTM tools, spreadsheets, live documents, etc - this customer support impact is peanuts, and more importantly, is just the tip of the iceberg. To put this in perspective, 14% is the average increase in MPL of the aggregate US workforce between 2018 and 2025 (Economic Policy Institute, 2025 - see git).
The marginal product of labor isn’t just higher, it’s shifting so rapidly that relying on previous credentials is no longer a safeguard against incomplete information. Outdated credentials represent concrete financial risk that old signaling mechanisms have true bearing on productivity.
Because signaling is a natural way of reducing information asymmetry in a marketplace with hidden attributes, we are left with a labor market that (besides being awfully congested) has a much higher level of information asymmetry than it otherwise might. This asymmetry problem is underscored by the fact that these markets are decentralized. In markets with less information parity, a centralized player becomes more relevant as a primary intermediary for assuming risk and forming a relationship with a given entity. This phenomenon has been explored in depth in the syndicated loan market (Sufi, 2007), where lenders with less publicly available data on solvency are tightly clustered around the borrowing firm by a lead arranger.
Paraform plays an analogous role to a lead arranger in syndicated loans, but this isn’t the only way in which Paraform solves the problem of information asymmetry. We also employ an approach that has been applied in 2 other popular decentralized settings: cryptocurrency and P2P file-sharing. Those who’re familiar with our Talent Density Index won’t be surprised to learn that this latter approach is a decentralized reputation score. The common feature of these scores is that they rely on the trust individual entities place in each other in order to create an objective, less-gamifiable score of each individual entity’s reputation (Eigentrust, 2003).
The market needs a dynamic market signal on talent - to understand the signals that can be used to decipher the labor market and what it values at any given time. And to build that signal, Paraform works as both a centralized player in the hiring ecosystem, and an aggregator of decentralized information on both the intent of candidates and the productivity of hires made by the fastest growing companies in the world. Our aim in using this information is to solve the world’s hardest problem, at a time when traditional approaches have lost relevance and talent becomes the world’s most scarce resource.
If you want to join me on our quant team, please message, or check out our careers page at https://www.paraform.com/careers.
References:
NBER - https://www.nber.org/system/files/working_papers/w31161/w31161.pdf
EPI - https://www.epi.org/productivity-pay-gap/ , git: https://github.com/Economic/productivity_pay_gap
Eigentrust, 2003 - https://nlp.stanford.edu/pubs/eigentrust.pdf
Join world-class companies that build their teams with Paraform.
