NYU SternSession 4: The Productivity Puzzle
SESSION 04 · ECONOMICS OF AI

The Productivity Puzzle
Will AI Deliver a Macro Payoff?

THE BIG QUESTION

If AI is transforming firms, why is it so difficult to see in aggregate productivity statistics?

1History2Firms3Labor4Macro5Stack6Politics
Before classPrepare, explore and arrive with a point of view

4.1 · SESSION FOCUS

Will AI Deliver a Macro Payoff?

Micro miracles from AI contrasted with the macroeconomic productivity mystery
4.2

BEFORE CLASS

Read. Watch. Arrive curious.

Required reading

Optional deep dive

4.3 PREWORK SUBMISSION

Prepare an example and consider the additional questions in Brightspace.
Submit your responses before class.

The Missing Productivity Gain

Find an AI improvement that saves time or raises quality but may be hard to capture in GDP or productivity statistics. Explain what measurement misses.

Exercise 1: The Power of Compounding

How much richer could AI make us?

CBO’s near-term baseline is about 1.2% annual growth in real GDP per person. Choose a sustained growth rate through 2050 and see how even small differences compound.

YOUR GROWTH FORECAST1.2%per year
1.2%CBO1.5%1.7%2%2.5%3%5%10%
2025$70,249Real GDP per person
→
2050$94,6571.3× the 2025 level
GAIN ABOVE CBO BASELINE$0This is the baseline path.
Your 1.2% pathCBO 1.2% baseline
202520302035204020452050

Save one anonymous forecast to see the class tally.

LOOKING BACKWARD: 1950–2025

Small annual gains already transformed American living standards.

Real GDP per person rose more than fourfold—equivalent to roughly 2.0% growth compounded for 75 years.

1950$16,206
2025$70,249
4.3×more real output per person

Sources: Congressional Budget Office, 2025 Long-Term Budget Outlook ↗ · BEA real GDP per capita via FRED ↗. Dollar values use chained 2017 dollars.

Exercise 2: AI & GDP — Growth Accounting

Micro miracles. Macro arithmetic.

Move the sliders to decide how much work AI reaches and how large the efficiency gain is within that work. Then watch those gains shrink when weighted by each sector’s share of the economy.

TOTAL AI CONTRIBUTION TO GDP THIS YEAR1.05percentage points
Task exposure×Efficiency gain×GDP share=GDP contribution
01

Customer Service

Call centers, chat support and help desks—high current exposure through chatbots and copilots.

≈ 2% of GDP
60%
0%100%
30%
0%100%
Sector productivity gain: 18.0%Contribution to GDP: 0.36 percentage points
02

Retail Trade

The labor-intensive slice: cashiers, inventory and restocking, with judgment and physical work remaining human.

≈ 2.5% of GDP
35%
0%100%
20%
0%100%
Sector productivity gain: 7.0%Contribution to GDP: 0.18 percentage points
03

Trucking, Taxi & Ride-Hailing

Two distinct waves show why timing can matter as much as technical potential.

≈ 1.5% of GDP

Today: routing & dispatch AI

25%
0%100%
20%
0%100%
Sector productivity gain: 5.0%Contribution to GDP: 0.08 percentage points

Future: full autonomous driving

10%
0%100%
60%
0%100%

Timeline multiplier: 60% of Wave 2 counts today.

Sector productivity gain: 6.0%Contribution to GDP: 0.05 percentage pointsAfter applying the 60% timing multiplier
04

Software Development

Coding assistants, testing and debugging—high exposure, with possible Jevons-style expansion in software demand.

≈ 2% of GDP
55%
0%100%
35%
0%100%
Sector productivity gain: 19.3%Contribution to GDP: 0.39 percentage points
SECTOR COMPARISONContribution to GDP
Customer service0.36 pp
Retail trade0.18 pp
Trucking / taxi0.13 pp
Software0.39 pp
The productivity-puzzle intuition:Large gains inside visible tasks can become modest macro effects after exposure, sector size and deployment timing are accounted for. Acemoglu and Brynjolfsson disagree less about whether AI helps than about how much, how broadly and how quickly.
YOUR SCENARIOSave your assumptions

Anonymous. Once you have settled on your assumptions, save them to see the class results.

≈ GDP shares are illustrative classroom estimates, not official BEA sector accounts. Results are an accounting exercise—not a forecast—and are rounded to two decimal places.

Please complete the Prework Memo and other questions prior to class.Open Session 4 prework in Brightspace
During classSlides + Zoom
4.4

Class Session Links

SESSION 4 SLIDES

Open the Session 4 slides Google Slides

SESSION 4 LIVE CLASS

Join the Session 4 Zoom NYU Zoom

4.5 · Evidence

Charts, cases and current examples · to be added

＋

4.6 · FUTURE INTERACTIVE

Explore the productivity puzzle

Interface reserved for a short in-class exploration. No data or conclusions have been pre-filled.

After class follow-upConsolidate, submit and reflect

4.5 · AFTER CLASS

3 Minute Memo — Quick Class Reflections.

Open the Session 4 reflection form