ECON 282E — Foundations of Macroeconomics

Zhigang Feng · University of California, Riverside · Fall 2026

Meets: Thursdays 9:00–11:50, Sproul Hall 4128 · September 24 – December 3

Webpage: https://sites.google.com/site/zfeng202/

Modern macroeconomics asks questions about distributions and policy: who bears income risk, how inequality interacts with the business cycle, how aging and public debt burden different cohorts, what a government should tax when it cannot commit, and how artificial intelligence will reshape work and wages. Answering these questions requires models with heterogeneous agents, frictions, and strategic governments — and those models can almost never be solved with pencil and paper.

The course builds the foundations in three steps. On paper (Sessions 1–2) we go from the neoclassical growth model to the workhorse models of current research — Bewley–Huggett–Aiyagari, Krusell–Smith, overlapping generations, and optimal fiscal policy with and without commitment. On the machine (Sessions 3–6) we learn to solve them: numerical dynamic programming, Python and PyTorch, root-finding, optimization, differentiation, approximation, quadrature, discretizing stochastic processes, perturbation and projection — ending at the computational wall that heterogeneous-agent economies run into. At the frontier (Sessions 7–10) we study how machine learning, deep learning and reinforcement learning break through that wall, how language models turn text into economic data, and how AI agents are changing the way research is done.

Throughout, the course insists on one discipline: every computed answer must be benchmarked and verified. As AI makes generating answers cheap, verification becomes the scarce skill. Every number on a slide in this course was produced by code that is in this repository, and every accuracy claim is reported with its cost.

The companion course. Where this course ends, AI for Economic Research: Dynamic Models, Language, and Agents begins — ten lectures on deep learning for dynamic models, reinforcement learning with heterogeneous agents, LLMs and text as economic data, retrieval-augmented generation, and agentic AI research workflows. Sessions 7–10 here are a compressed tour of that material; the companion site is where it is developed in full, with its own labs.

📋 Syllabus Schedule, grading, policies 📑 Slides Session decks (© Z. Feng) 💻 Labs Notebooks — run them yourself 📝 Assignments Homework and the final project 💬 Discussion Ask, answer — named or anonymous

Start here

Three steps per session

Read the session row in the syllabus — it says what each block covers and what is due.
Work through the decks on the slides page. Each block is three or four short decks rather than one long one.
Run the lab. The notebooks are where the material becomes yours. They ship with no stored output on purpose — you are meant to produce the numbers.

No Python yet? Start with Lab 0, posted before Session 3. It installs the tools and checks the install. Programming is taught from the ground up in Session 3 — prior experience helps but is not assumed.

Discussion

Questions about the material, the labs or the homework go to the course discussion forum, where classmates and instructors answer. It has a category for every session, each homework, the final project, code, and AI tools. Sign in with your @ucr.edu address; nothing in the forum is visible to anyone who is not signed in.

Named or anonymous, post by post. An anonymous post shows classmates an alias such as “Anonymous Heron”. Instructors can see who wrote it, so that they can moderate.
Anonymous feedback to the instructor has its own category: always anonymous, to instructors as well, and readable by instructors only.
Norms. Be kind and specific; ask about ideas and errors, and do not post homework solutions or solution code before the due date. E-mail is for private matters.

Schedule — Fall 2026

Two blocks per session: A 9:00–10:15, B 10:35–11:50. No class on November 26.

Session	Block A	Block B	Due / out
1 Sep 24	Introduction — the macro frontier, AI as tool and object	The optimal growth model
2 Oct 1	Incomplete markets — Bewley, Huggett, Aiyagari	Aggregate risk, finite lives, policy
3 Oct 8	Introduction to computation	Programming basics — Python, NumPy, PyTorch	HW 1 Part A out
4 Oct 15	Numerical methods I — roots, optima, derivatives	Approximation, quadrature, AR(1), a fast solver	HW 1 Part B out
5 Oct 22	Numerical methods II — perturbation	Projection; one model, four methods
6 Oct 29	Heterogeneous agents I — Aiyagari, Krusell–Smith, OLG, optimal policy	Heterogeneous agents II — aggregation, Negishi, transitions	HW 1 due; HW 2 Part A out
7 Nov 5	Machine learning essentials	Deep learning for dynamic models	Project teams and topics
8 Nov 12	Reinforcement learning	Heterogeneous agents at the frontier	HW 2 Part B out
9 Nov 19	Language models and text as data	LLMs as research instruments	Proposal due Mon Nov 16
Monday, November 23 — HW 2 due. November 26 — no class (Thanksgiving).
10 Dec 3	Agentic AI and the future of research	Final project presentations
Dec 10	Final paper and replication package due (finals week).

Grading

Component	Weight	When
Homework 1 — from a question to a quantitative answer	15%	out Oct 8 and Oct 15; due Oct 29
Homework 2 — heterogeneity and learning	15%	out Oct 29 and Nov 12; due Mon Nov 23
Final project (teams of 2–3)	40%	proposal 5 · talk and discussion 15 · paper and code 20
Participation	30%	throughout

Where this course sits

This site is the foundations stage: the models, and the classical methods that solve them. Its continuation is AI for Economic Research, which takes the same models to the AI frontier — deep learning for dynamic models, reinforcement learning and heterogeneous agents, language models and text as data, and agentic research workflows.

Foundations of Macroeconomics → AI for Economic Research

Sessions 7–10 here are the bridge between the two.