Skip to content

Topics presentations

Explore topics through research papers, data, and economic models. Select a title to open the HTML slides, or download a PDF. The PDFs include the appendices; interactive examples appear as saved, static views.

Occupational tasks and AI exposure

Which parts of an occupation could AI perform, and how does that relate to careers in consulting, finance, management, and sales? This presentation introduces O*NET, connects economics and business programs to occupations, and compares measures of potential exposure with observed AI use.

References

Data resources: O*NET occupational data; TaskNet: worker-based task inventories (appendix).

Download slides (PDF)

Who uses AI, and for what?

How widely is generative AI used, which workers adopt it, and what do they use it for? Survey evidence and ChatGPT conversation data show how adoption varies across occupations and tasks. The presentation distinguishes exposure, actual use, reported time savings, and measured productivity.

References

Download slides (PDF)

Expertise and the productivity effects of AI

Who gains when AI changes the expertise needed to do a job? A task model and historical occupation evidence lead into a customer-support rollout and a randomized patent-lawyer experiment. The results distinguish faster or better assisted work from the development of independent judgment.

References

Download slides (PDF)

What do we know about robot automation?

How does robot automation affect firms, workers, and local labour markets? U.S., Danish, and Brazilian evidence connects exposure-based designs, firms’ adoption choices, and the distinction between robots and worker-operated tools. The presentation explains Bartik instruments, separates relative local estimates from model-based aggregate effects, and closes with evidence on productivity and worker adjustment.

References

Data resources: Bartik / shift-share instruments: Borusyak, Hull & Jaravel (2025); The missing intercept problem: Benjamin Moll.

Download slides (PDF)

What holds a job together?

Why do some jobs remain bundles of tasks while others are broken apart by AI? Two models connect job boundaries, human earnings and occupational entry to the incentive to train juniors. The Messy Jobs framework adds verifiable tasks, strong bundles and relationships. Faster training can coexist with fewer opportunities to begin a career.

References

Download slides (PDF)

Who receives the gains from automation?

How does automation change wages, returns to wealth, and the distribution of income? Uneven Growth connects capital accumulation to unequal gains; AGI and weak-link models show why long-run labour shares depend on which tasks remain human. Quantitative evidence compares equipment, task displacement, and AI, distinguishing historical results from conditional model projections.

References

Download slides (PDF)