AdvancedMacroeconomics · Chapter 1About 9 min
AI lifts productivity in a single quarter,and a newcomer builds a world in miniature
The note
Your first Monday at the Central Bank of Halcyon. Fourth floor. Twelve desks, an old coffee machine, a whiteboard nobody has fully erased in years.
- Ilse VargaRead this. Productivity jumped last quarter. The statistics office says it’s AI tools.
- YouFirms using AI for scheduling, design, customer service?
- Ilse VargaExactly. The Governor wants two answers by Friday.
- Ilse VargaWill it last? And what will people do with it?
- Ilse VargaBring a model, not an opinion.
AI raises productivity by 1% this quarter. What do households do with the extra output?
Noted. We will come back to your prediction at the end of the chapter.
A world in miniature
A model at the Bank means a small economy that is complete:
- everyone makes choices for a reason,
- every market clears,
- random shocks push it around over time.
That is a dynamic stochastic general equilibrium (DSGE) model.
Start small. One household lives forever, owns the capital and works a fixed amount. One firm produces:
Z_t is productivity — what AI just raised. What isn’t consumed is invested. Capital wears out at rate \delta:
The household maximises the discounted sum of \ln C_t. That gives the Euler equation: saving one more unit today must pay back tomorrow.
Productivity is persistent, with surprises:
\rho is the Governor’s first question. Near 0: the AI boost is a one-off. Near 1: it lasts for years.
What happens after the shock
Step through what the model does after a 1% productivity jump (\rho = 0.9).
- A 1% productivity jump arrives in quarter 0.
- Output rises 1% at once — capital and labour are fixed this quarter.
- Investment jumps about 3.5%. The windfall goes into capital.
- Consumption rises only a little — about 0.2%.
- Everything returns to normal. How fast depends on \rho.
To get these paths, find the steady state (where the economy rests with Z = 1) and approximate the model by linear equations around it. Solved, they give rules for capital and consumption. The bench applies them.
Challenge. Find a persistence at which consumption rises at least half as much as output on impact.
Solved. Only a gain that looks almost permanent is consumed that much at once.
Quarterly model with log utility and a fixed labour supply; α = 0.33 and δ = 0.025 are held fixed. Lines show percent deviations from the steady state.
Output on impact never moves with \rho. What moves is how the extra output is used. The shorter the gain, the more is invested. That is consumption smoothing.
An AI’s answer
Question to an AI assistant In a real business cycle model, how do households respond to a productivity gain from AI?
1 A productivity gain raises output on impact. 2 Households smooth consumption rather than spending a windfall at once. 3 So most of a temporary gain is saved, which in this model means invested. 4 The more persistent the gain, the larger the share that is invested on impact.
Friday
- YouOutput rises one for one with productivity, on impact.
- YouIf the AI boost is short-lived, most of it is invested. If it lasts, people spend more right away.
- The GovernorSo the question is whether AI is a one-off burst or a lasting change.
- YouYes. That decides the split.
- The GovernorOne more thing. In your little world, prices adjust instantly. On our high street they don’t. What then?
You predicted: Mostly consume it
On the bench, at the start: consumption +0.2%, investment +3.5%. Most of it is invested. Consumption takes more only if the gain looks almost permanent.
You predicted: Mostly invest it
Right, for a gain expected to fade. Households spread a temporary windfall over the future by building capital.
You predicted: Depends on how long it lasts
That’s the model’s answer. Short-lived: mostly invested. Near-permanent: consumed much more at once.
You did not make a prediction at the start of the chapter. Here is what it asked:
AI raises productivity by 1% this quarter. What do households do with the extra output?
Output rises one for one on impact. How the gain is split depends on how long it is expected to last.
This chapter on two pages
The ideas, the formulas, a worked example, four traps — and six questions to try on your own, with answers.
Want to go further? Paste this into the AI you use. It asks the AI to quiz you, not to hand you answers.
I am learning the real business cycle model. Suppose the adoption of AI raises an economy's productivity by 1% this quarter, and households have log utility, a fixed labour supply and a capital stock that depreciates at 2.5% a quarter. Please act as my tutor. Do not give me the answers. Ask me one question at a time to help me work out: 1. why output rises one for one with productivity on impact, 2. why most of a temporary gain is invested rather than consumed, and 3. how the answer changes if the gain is expected to be permanent. Wait for my answer after each question. At the end, tell me what I got right and correct anything I got wrong.
Key points
- A DSGE model is a small, complete economy: people optimise, markets clear, shocks arrive at random.
- The steady state is where the model rests without shocks. Here K/Y = \alpha / (1/\beta - 1 + \delta).
- Log-linearising around the steady state gives rules that can be solved.
- An impulse response shows each variable’s path after one shock.
- A short-lived gain is mostly invested. A long-lived one is consumed more.
The Governor’s last question stays on your desk all weekend: what if shops don’t reprice overnight? To be continued — Chapter 2: sticky prices and the New Keynesian model.