Appendix — Week of July 10 — AI and the Labor Market, and Reading a Finance Paper

ASSIP 2026 · Mentor: Prof. Lei Gao

Published

July 10, 2026

Two parts this week. Part 1 is a written research brief. Part 2 prepares you to discuss a real finance paper with me. Read both fully before you start.

Part 1 — The Other Side of the “AI Takes Your Job” Story

Due: Monday, July 13.

The headlines say “AI is coming for your job.” That is only half the story. A large and growing body of serious research finds that AI often complements human work — raising productivity, and sometimes helping the people who need it most. Your task is to investigate that other half, with special attention to workers with disabilities, for whom AI-powered tools (speech-to-text, screen readers, live captioning, AI job-matching, remote-work software) can open doors that were previously closed.

What to produce — a 2–3 page research brief with four parts

Annotated bibliography of 5–7 credible sources (at least 2 specifically on disability & employment). For each source, write 2–3 sentences: what did it find, and why does it matter?
Two datasets you could actually use. For each, state one specific research question you could answer with it.
Your own hypothesis — one sentence — plus a short paragraph sketching how you would test it (what data, what comparison).
One honest caveat. The evidence is genuinely mixed. Find and describe at least one way AI can harm disabled workers (e.g., biased hiring tools). Good research acknowledges the other side.

Start here — then go beyond these

Positive effects of AI on work (augmentation, productivity): - Brynjolfsson, Li & Raymond (2023), Generative AI at Work, NBER 31161 — an AI assistant raised customer-support productivity 14% on average, and 34% for the least-experienced workers. → https://www.nber.org/papers/w31161 - Noy & Zhang (2023), Science — in an experiment, ChatGPT cut writing-task time ~40%, raised quality, and narrowed the gap between weaker and stronger writers. → https://pubmed.ncbi.nlm.nih.gov/37440646/ (open preprint available on SSRN) - Autor (2024), Applying AI to Rebuild Middle Class Jobs, NBER 32140 — argues AI can extend human expertise to a broader set of workers. → https://www.nber.org/papers/w32140

AI, assistive technology & disability employment: - Kruse, Schur et al. (2024), Assistive Technology’s Potential to Improve Employment of People with Disabilities, J. Occupational Rehabilitation — U.S. Census data links greater assistive-technology availability to higher disability employment. → https://pmc.ncbi.nlm.nih.gov/articles/PMC11219186/ - Schur, Ameri & Kruse (2020), Telework after COVID: a “silver lining” for workers with disabilities? — remote work, enabled by digital/assistive tools, removes commuting and physical-workplace barriers. → https://smlr.rutgers.edu/faculty-research-engagement/program-disability-research/telework - OECD (2023), Using AI to Support People with Disability in the Labour Market — four categories of AI that help disabled workers, from 70+ stakeholder interviews. → https://www.oecd.org/en/publications/using-ai-to-support-people-with-disability-in-the-labour-market_008b32b7-en.html

The honest nuance (AI can also harm — you must engage with this): - ILO (2025), AI’s double-edged sword — the same AI can discriminate: resume screeners that downrank disability-related signals; video-interview tools that misread autism or facial differences. → https://www.ilo.org/resource/article/ais-double-edged-sword-new-frontier-employment-people-disabilities

Datasets to explore

IPUMS CPS — U.S. monthly labor-force data with a disability question (DIFFANY, since 2008). Free with registration; build custom extracts online. → https://cps.ipums.org/cps/
DisabilityStatistics.org (Cornell University) — no coding required: pre-made tables of disability employment rates by year and state. Great first stop. → https://www.disabilitystatistics.org/
AI Occupational Exposure (AIOE) — which occupations are most “exposed” to AI. Free on GitHub. → https://github.com/AIOE-Data/AIOE
**O*NET** — detailed attributes of ~1,000 occupations (skills, tasks, work context). Free. → https://www.onetcenter.org/database.html
Also useful: nTIDE monthly disability-employment report (https://researchondisability.org/nTIDE); IPUMS ACS for occupation × disability (https://usa.ipums.org/usa/).

Ground rules (important)

Every citation must be real and clickable — verify each one yourself by opening the link. Do not trust any AI-generated citation without checking it. Fabricated references are the fastest way to lose credibility in research.
Prefer the primary source (the actual paper) over a blog summary.
Keep a running list of the search terms and databases you used — you’ll include it at the end.

Guiding questions (pick a few to anchor your brief)

Does AI help lower-skilled workers more than higher-skilled ones? Why might that be?
Does remote work (enabled by AI/assistive tech) expand employment for disabled workers?
Which occupations are most “exposed” to AI — and is exposure the same thing as at risk?

Part 2 — Read One of My Papers

Discussion: Tuesday, July 14 (morning) — come ready to talk for 5 minutes.

Pick one paper below, read it carefully, and prepare a 5-minute talk. This is how you learn to read finance research. You do not need to understand every equation — focus on the story.

In your 5 minutes, cover these five things (this is also how every finance paper is built)

The question — what are they asking, and why does it matter?
The data — what data did they use, and where did it come from?
The method — how do they answer it? (What is the key comparison or “experiment”?)
The main finding — the single most important result, with the number.
Your take — one thing you found clever, and one thing you would question or want to know more.

Choose one (all are mine; roughly easiest → most technical)

Lending Practices to Same-Sex Borrowers (Sun & Gao, PNAS 2019) — Do same-sex couples face mortgage discrimination? Uses real federal mortgage records. Intuitive and socially important (ranked in the Financial Times Top-100 research with social impact). Best first read.
The Rainbow of Municipal Credit: Evidence from Same-Sex Marriage Bans (Gao, Liu & Wang, working paper) — extends the discrimination question to the municipal bond market: do same-sex-marriage bans raise a city’s borrowing costs? Relatable policy question.
Does Local Religiosity Affect Organizational Risk-Taking? Evidence from the Hedge Fund Industry (Gao, Wang & Zhao, J. Corporate Finance 2017) — Do hedge funds in more religious areas take less risk? Textbook-clear question → data → method → finding.
Local Religiosity and Financial Advisor Misconduct (Gao, Cowan, Han & Pan, J. Corporate Finance 2024) — the same “does local culture shape behavior?” idea applied to individual financial advisers and their misconduct records. Pairs nicely with paper #3.
Corporate Patents, R&D Success, and Tax Avoidance (Gao, Yang & Zhang, Rev. Quant. Finance & Accounting 2016) — Do successful innovators also avoid more tax? (Nice fit if you’re on one of the patents projects.)
Standing Out from the Crowd via CSR Engagement (Gao, He & Wu, J. Financial & Quantitative Analysis 2023) — Why do firms “do good”? A competition-based answer. (ESG.)
Individualistic CEO and Corporate Innovation (Gao, Han, Pan & Zhang, Research Policy
1. — Does a CEO’s cultural background (U.S. frontier heritage) shape how much the firm innovates?
Seeing is Believing: Annual Report “Graphicity” and Stock Return Predictability (Gao, Deng, Hu & Zhou, working paper) — Do the pictures and charts in a company’s annual report carry information the market misses? A modern, novel angle (text- and image-as-data).
Market Intraday Momentum (Gao, Han, Li & Zhou, J. Financial Economics 2018) — Does the market’s morning return predict its afternoon return? A simple, tradable idea. Most technical / for the quant-minded.

How to access

My Google Scholar page has a free PDF or link for every paper: https://scholar.google.com/citations?user=zk0dvhsAAAAJ
Or search “Lei Gao George Mason” on SSRN (https://www.ssrn.com).

How to read your first finance paper (tips)

Read in this order: abstract → introduction → conclusion → then the ONE main table. Skip the heavy math on the first pass.
The introduction previews every finding — read it twice.
Write down the one sentence that captures the whole paper. If you can say it out loud, you understand it.

Questions during the week? Email me. I’d rather you ask early than stay stuck.

--- title: "Week of July 10 — AI and the Labor Market, and Reading a Finance Paper" subtitle: "ASSIP 2026 · Mentor: Prof. Lei Gao" date: "2026-07-10" --- [**Download this assignment as a PDF**](assip-week-2026-07-10.pdf) Two parts this week. **Part 1** is a written research brief. **Part 2** prepares you to discuss a real finance paper with me. Read both fully before you start. --- ## Part 1 — The Other Side of the "AI Takes Your Job" Story **Due: Monday, July 13.** The headlines say *"AI is coming for your job."* That is only half the story. A large and growing body of serious research finds that AI often **complements** human work — raising productivity, and sometimes helping the people who need it most. Your task is to investigate that other half, with special attention to **workers with disabilities**, for whom AI-powered tools (speech-to-text, screen readers, live captioning, AI job-matching, remote-work software) can open doors that were previously closed. ### What to produce — a 2–3 page research brief with four parts 1. **Annotated bibliography** of **5–7 credible sources** (at least **2** specifically on disability & employment). For each source, write 2–3 sentences: what did it find, and why does it matter? 2. **Two datasets** you could actually use. For each, state **one specific research question** you could answer with it. 3. **Your own hypothesis** — one sentence — plus a short paragraph sketching how you would test it (what data, what comparison). 4. **One honest caveat.** The evidence is genuinely mixed. Find and describe at least one way AI can *harm* disabled workers (e.g., biased hiring tools). **Good research acknowledges the other side.** ### Start here — then go beyond these **Positive effects of AI on work (augmentation, productivity):** - Brynjolfsson, Li & Raymond (2023), *Generative AI at Work*, NBER 31161 — an AI assistant raised customer-support productivity 14% on average, and **34% for the least-experienced workers.** → https://www.nber.org/papers/w31161 - Noy & Zhang (2023), *Science* — in an experiment, ChatGPT cut writing-task time ~40%, raised quality, and **narrowed the gap between weaker and stronger writers.** → https://pubmed.ncbi.nlm.nih.gov/37440646/ (open preprint available on SSRN) - Autor (2024), *Applying AI to Rebuild Middle Class Jobs*, NBER 32140 — argues AI can extend human expertise to a broader set of workers. → https://www.nber.org/papers/w32140 **AI, assistive technology & disability employment:** - Kruse, Schur et al. (2024), *Assistive Technology's Potential to Improve Employment of People with Disabilities*, J. Occupational Rehabilitation — U.S. Census data links greater assistive-technology availability to **higher disability employment.** → https://pmc.ncbi.nlm.nih.gov/articles/PMC11219186/ - Schur, Ameri & Kruse (2020), *Telework after COVID: a "silver lining" for workers with disabilities?* — remote work, enabled by digital/assistive tools, removes commuting and physical-workplace barriers. → https://smlr.rutgers.edu/faculty-research-engagement/program-disability-research/telework - OECD (2023), *Using AI to Support People with Disability in the Labour Market* — four categories of AI that help disabled workers, from 70+ stakeholder interviews. → https://www.oecd.org/en/publications/using-ai-to-support-people-with-disability-in-the-labour-market_008b32b7-en.html **The honest nuance (AI can also harm — you must engage with this):** - ILO (2025), *AI's double-edged sword* — the *same* AI can discriminate: resume screeners that downrank disability-related signals; video-interview tools that misread autism or facial differences. → https://www.ilo.org/resource/article/ais-double-edged-sword-new-frontier-employment-people-disabilities ### Datasets to explore - **IPUMS CPS** — U.S. monthly labor-force data with a disability question (`DIFFANY`, since 2008). Free with registration; build custom extracts online. → https://cps.ipums.org/cps/ - **DisabilityStatistics.org** (Cornell University) — **no coding required**: pre-made tables of disability employment rates by year and state. Great first stop. → https://www.disabilitystatistics.org/ - **AI Occupational Exposure (AIOE)** — which occupations are most "exposed" to AI. Free on GitHub. → https://github.com/AIOE-Data/AIOE - **O*NET** — detailed attributes of ~1,000 occupations (skills, tasks, work context). Free. → https://www.onetcenter.org/database.html - *Also useful:* **nTIDE** monthly disability-employment report (https://researchondisability.org/nTIDE); **IPUMS ACS** for occupation × disability (https://usa.ipums.org/usa/). ### Ground rules (important) - **Every citation must be real and clickable — verify each one yourself by opening the link.** Do **not** trust any AI-generated citation without checking it. Fabricated references are the fastest way to lose credibility in research. - Prefer the **primary source** (the actual paper) over a blog summary. - Keep a running list of the search terms and databases you used — you'll include it at the end. ### Guiding questions (pick a few to anchor your brief) - Does AI help *lower-skilled* workers more than higher-skilled ones? Why might that be? - Does remote work (enabled by AI/assistive tech) expand employment for disabled workers? - Which occupations are most "exposed" to AI — and is *exposure* the same thing as *at risk*? --- ## Part 2 — Read One of My Papers **Discussion: Tuesday, July 14 (morning) — come ready to talk for 5 minutes.** Pick **one** paper below, read it carefully, and prepare a **5-minute talk**. This is how you learn to read finance research. You do **not** need to understand every equation — focus on the story. ### In your 5 minutes, cover these five things (this is also how every finance paper is built) 1. **The question** — what are they asking, and why does it matter? 2. **The data** — what data did they use, and where did it come from? 3. **The method** — how do they answer it? (What is the key comparison or "experiment"?) 4. **The main finding** — the single most important result, *with the number*. 5. **Your take** — one thing you found clever, and one thing you would question or want to know more. ### Choose one (all are mine; roughly easiest → most technical) 1. **Lending Practices to Same-Sex Borrowers** (Sun & Gao, *PNAS* 2019) — *Do same-sex couples face mortgage discrimination?* Uses real federal mortgage records. Intuitive and socially important (ranked in the *Financial Times* Top-100 research with social impact). **Best first read.** 2. **The Rainbow of Municipal Credit: Evidence from Same-Sex Marriage Bans** (Gao, Liu & Wang, working paper) — extends the discrimination question to the *municipal bond* market: do same-sex-marriage bans raise a city's borrowing costs? Relatable policy question. 3. **Does Local Religiosity Affect Organizational Risk-Taking? Evidence from the Hedge Fund Industry** (Gao, Wang & Zhao, *J. Corporate Finance* 2017) — *Do hedge funds in more religious areas take less risk?* Textbook-clear question → data → method → finding. 4. **Local Religiosity and Financial Advisor Misconduct** (Gao, Cowan, Han & Pan, *J. Corporate Finance* 2024) — the same "does local culture shape behavior?" idea applied to individual financial advisers and their misconduct records. Pairs nicely with paper #3. 5. **Corporate Patents, R&D Success, and Tax Avoidance** (Gao, Yang & Zhang, *Rev. Quant. Finance & Accounting* 2016) — *Do successful innovators also avoid more tax?* (Nice fit if you're on one of the patents projects.) 6. **Standing Out from the Crowd via CSR Engagement** (Gao, He & Wu, *J. Financial & Quantitative Analysis* 2023) — *Why do firms "do good"?* A competition-based answer. (ESG.) 7. **Individualistic CEO and Corporate Innovation** (Gao, Han, Pan & Zhang, *Research Policy* 2023) — *Does a CEO's cultural background (U.S. frontier heritage) shape how much the firm innovates?* 8. **Seeing is Believing: Annual Report "Graphicity" and Stock Return Predictability** (Gao, Deng, Hu & Zhou, working paper) — *Do the pictures and charts in a company's annual report carry information the market misses?* A modern, novel angle (text- and image-as-data). 9. **Market Intraday Momentum** (Gao, Han, Li & Zhou, *J. Financial Economics* 2018) — *Does the market's morning return predict its afternoon return?* A simple, tradable idea. **Most technical / for the quant-minded.** ### How to access - My **Google Scholar** page has a free PDF or link for every paper: https://scholar.google.com/citations?user=zk0dvhsAAAAJ - Or search **"Lei Gao George Mason"** on SSRN (https://www.ssrn.com). ### How to read your first finance paper (tips) - Read in this order: **abstract → introduction → conclusion → then the ONE main table.** Skip the heavy math on the first pass. - The **introduction previews every finding** — read it twice. - Write down the **one sentence** that captures the whole paper. If you can say it out loud, you understand it. --- *Questions during the week? Email me. I'd rather you ask early than stay stuck.*