Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Labor 20120111 ALM 2003
1. Autor, Levy, and Murnane (2003)
“The Skill Content of Recent
Technological Change: An Empirical
Exploration”
Critique in Three Slides
Betsy Williams
1/11/12
2. The Good
• The authors state what the ideal experiment
would be to test their hypothesis, then
compare their framework to that ideal.
• Very explicit explanations of theory,
predictions, and evidence
• Authors perform many robustness checks,
including by gender, manufacturing industry,
different kinds of computer capital, industry
characteristics, etc.
3. The Bad
• The authors make causal claims, yet their panel data
has very few time points. Fundamentally correlational.
• Claim: “the declining price of computer capital” is the
“causal force” skill demand change, but study never
measures or analyzes computing capital prices. Instead
proxies with actual adoption of computers, by industry.
• Claim: price of computing power exogenous.
– PC development depended on human capital levels.
– Lower prices from volume increases? (returns to scale)
– More computer skills in a given occupation/industry may
make it easier for a technologist to develop tools efficient
for that occupation/industry.
4. The Next Paper to Write…
• Unionization is never mentioned, but we
might expect it to change how much
industries actually shift job tasks.
• Why should more educated workers be
favored by these task shifts? I would prefer to
see evidence about the labor force’s skills by
education at each point in time (ASVAB?).
• Possible other ways to get job task data: H-1B
visa applications, ADA compliance lawsuits
Editor's Notes
Ideally, “two identical autarkic economies” experience different computer prices.