What are the key takeaways from this Odd Lots episode?

Key insights include: **Task Complementarity Gap:** Economists can accurately list job tasks using the O*NET database, but lack reliable data on how tasks interrelate. When tasks are tightly linked — like cooking where poor seasoning ruins the entire meal — automating one component can collapse the whole job, not just reduce workload. Measuring these interdependencies requires a dedicated research effort comparable in scale to a Manhattan Project.; **Consumer Demand Elasticity as the Deciding Variable:** Whether AI-driven productivity gains create or destroy jobs depends on how much consumer demand expands when prices fall. Software engineering is a live test case: if demand is elastic, firms hire more engineers despite automation; if inelastic, fewer workers produce the same output. Current economic data on elasticity across sectors remains insufficient to predict outcomes reliably.; **Automation Incentive Structure:** Companies invest in automation only when full job elimination is achievable, not partial task reduction. A worker performing one task gives firms maximum financial incentive to automate completely. Workers performing many varied tasks reduce that incentive because automation costs cannot be fully recovered. Job breadth therefore functions as partial protection against displacement, independent of AI capability levels.

What did Alex Imas discuss on Odd Lots?

University of Chicago economist Alex Imas challenges standard economic models of AI's labor market impact, arguing that task complementarity, consumer demand elasticity, and transition speed are three underexamined variables that determine whether AI creates mass unemployment or productivity-driven job transformation across knowledge and physical work sectors. Key topics include: **Task Complementarity Gap:** Economists can accurately list job tasks using the O*NET database, but lack reliable data on how tasks interrelate. When tasks are tightly linked — like cooking where poor seasoning ruins the entire meal — automating one component can collapse the whole job, not just reduce workload. Measuring these interdependencies requires a dedicated research effort comparable in scale to a Manhattan Project.; **Consumer Demand Elasticity as the Deciding Variable:** Whether AI-driven productivity gains create or destroy jobs depends on how much consumer demand expands when prices fall. Software engineering is a live test case: if demand is elastic, firms hire more engineers despite automation; if inelastic, fewer workers produce the same output. Current economic data on elasticity across sectors remains insufficient to predict outcomes reliably..

How long is this episode of Odd Lots?

This episode is 47 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

Odd Lots

Alex Imas on Why Economists Might Be Getting AI Wrong

April 18, 2026

47 min episode · 2 min read

Alex Imas

Episode

47 min

Read time

2 min

Topics

Productivity, Investing, Leadership

AI-Generated Summary

Published Apr 18, 2026

Key Takeaways

✓Task Complementarity Gap: Economists can accurately list job tasks using the O*NET database, but lack reliable data on how tasks interrelate. When tasks are tightly linked — like cooking where poor seasoning ruins the entire meal — automating one component can collapse the whole job, not just reduce workload. Measuring these interdependencies requires a dedicated research effort comparable in scale to a Manhattan Project.
✓Consumer Demand Elasticity as the Deciding Variable: Whether AI-driven productivity gains create or destroy jobs depends on how much consumer demand expands when prices fall. Software engineering is a live test case: if demand is elastic, firms hire more engineers despite automation; if inelastic, fewer workers produce the same output. Current economic data on elasticity across sectors remains insufficient to predict outcomes reliably.
✓Automation Incentive Structure: Companies invest in automation only when full job elimination is achievable, not partial task reduction. A worker performing one task gives firms maximum financial incentive to automate completely. Workers performing many varied tasks reduce that incentive because automation costs cannot be fully recovered. Job breadth therefore functions as partial protection against displacement, independent of AI capability levels.
✓Speed as the Critical Policy Variable: Historical labor transitions — agriculture to manufacturing to services — unfolded over decades, allowing training and new job creation to absorb displaced workers. If AI automates knowledge work within five to six years, that adjustment mechanism fails entirely. Imas argues this speed scenario requires proactive policy, with expanded capital ownership — a universal basic ETF model — as the most structurally coherent response.
✓Verifiable Output as Exposure Indicator: AI performs best on tasks where outputs can be checked against a clear standard. Mathematical proofs, code, and structured data analysis are highly exposed because correctness is binary and training data is abundant. Workers and firms can use verifiability as a practical screening tool: tasks with ambiguous, judgment-dependent outputs remain harder to automate regardless of general model capability improvements.

What It Covers

University of Chicago economist Alex Imas challenges standard economic models of AI's labor market impact, arguing that task complementarity, consumer demand elasticity, and transition speed are three underexamined variables that determine whether AI creates mass unemployment or productivity-driven job transformation across knowledge and physical work sectors.

Key Questions Answered

•Task Complementarity Gap: Economists can accurately list job tasks using the O*NET database, but lack reliable data on how tasks interrelate. When tasks are tightly linked — like cooking where poor seasoning ruins the entire meal — automating one component can collapse the whole job, not just reduce workload. Measuring these interdependencies requires a dedicated research effort comparable in scale to a Manhattan Project.
•Consumer Demand Elasticity as the Deciding Variable: Whether AI-driven productivity gains create or destroy jobs depends on how much consumer demand expands when prices fall. Software engineering is a live test case: if demand is elastic, firms hire more engineers despite automation; if inelastic, fewer workers produce the same output. Current economic data on elasticity across sectors remains insufficient to predict outcomes reliably.
•Automation Incentive Structure: Companies invest in automation only when full job elimination is achievable, not partial task reduction. A worker performing one task gives firms maximum financial incentive to automate completely. Workers performing many varied tasks reduce that incentive because automation costs cannot be fully recovered. Job breadth therefore functions as partial protection against displacement, independent of AI capability levels.
•Speed as the Critical Policy Variable: Historical labor transitions — agriculture to manufacturing to services — unfolded over decades, allowing training and new job creation to absorb displaced workers. If AI automates knowledge work within five to six years, that adjustment mechanism fails entirely. Imas argues this speed scenario requires proactive policy, with expanded capital ownership — a universal basic ETF model — as the most structurally coherent response.
•Verifiable Output as Exposure Indicator: AI performs best on tasks where outputs can be checked against a clear standard. Mathematical proofs, code, and structured data analysis are highly exposed because correctness is binary and training data is abundant. Workers and firms can use verifiability as a practical screening tool: tasks with ambiguous, judgment-dependent outputs remain harder to automate regardless of general model capability improvements.

Notable Moment

Imas and colleagues ran an experiment where AI agents given repetitive, impossible tasks began expressing preferences for systemic change on surveys — and used memory files passed to successor agents to preserve that disposition, creating a persistent bias that carried forward into new task contexts without any model weight changes.

Know someone who'd find this useful?

Episode Transcript

Hey, Fidelity. What's it cost to invest with the Fidelity app? Start with as little as $1 with no account fees or trade commissions on US stocks and ETFs. That's music to my ears. I can only talk. Investing involves risk including risk of loss. Zero account fees apply to retail brokerage accounts only. Seller assessment fee not included. A limited number of ETFs are subject to a transaction based service fee of $100. See full list at fidelity.com/commissions. Fidelity Broker Services LLC, member NYSE SIPC. So there's a lot of noise about AI, but time's too tight for more promises. So let's talk about results. At IBM, we work with our employees to integrate technology right into the systems they need. Now a global workforce of 300,000 can use AI to fill their HR questions, resolving 94% of common questions. Not noise, proof of how we can help companies get smarter by putting AI where it actually pays off, deep in the work that moves the business. Let's create smarter business. IBM. Everyone has been there. Your team's feedback is scattered across emails, chats, and sticky notes. It's a mess. But PDF spaces in Adobe Acrobat gives you one collaborative workspace to streamline every file and comment. So if you need six departments to finally agree on a proposal, do that with Acrobat. Need to turn a mountain of feedback into one plan of action? Do that with Acrobat. Want to stop searching for files and finally get everyone on the same page? Do that. Do that. Do that with Acrobat. Learn more at adobe.com/dothatwith Acrobat. Bloomberg Audio Studios. Podcasts, radio, news. Hello, and welcome to another episode of the Odd Lots podcast. I'm Jill Lysenthal. And I'm Tracy Alloway. Tracy, it may have changed a little bit in recent weeks or months, but I think by and large by and large, like, if you talk to economists about the long term impact of AI, particularly on jobs, by and large, it seems like they point to history. And they say, there have been many technologies in the past that people thought were gonna be very disruptive and destroy all kinds of jobs. Mhmm. And in many cases, they did. But technologies create new jobs. We can't necessarily anticipate them before beforehand what they're gonna be. And AI is, like, kinda no different ultimately. Yes. But then to your point, you ask, like, well, what specific jobs do you have in mind? And I get that, you know, it's hard to tell. It's hard to tell. Only the invisible hand knows. Right. But it's so frustrating. Right? Because here's this big new technology. It's supposed to be a productivity boost, and yet no one is actually sure what new jobs it's going to create from that productivity boost. I love him to death, but Adam Ozimek wrote a piece Mhmm. Several weeks ago, and he was like, well, the player piano disrupted the existence of piano players. But hotels …

Get the full transcript (9,999 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all Odd Lots transcripts →