Economics professors

Three parts of your work where AI is already doing real lifting, and what stays yours.

• Lecture1776–1890

  Lecture, blackboard, and classical text (Smith, Ricardo, Mill)

  The founding technology of economics teaching was the treatise and the public lecture. Adam Smith delivered his political economy in lectures at Glasgow from 1751 before publishing "The Wealth of Nations" in 1776; the text circulated as the foundational reading in American political economy courses from the 1820s onward. Teaching was entirely oral and textual: the professor lectured from notes or from a primary text, students copied by hand, and examinations were oral. The blackboard, introduced into American classrooms around 1800-1820, gave economics instructors the first visual-spatial tool for drawing supply-demand diagrams and production frontiers. Mathematical notation was minimal until the late 19th century; the dominant mode was verbal and historical reasoning.

• Mathematical formalism + the graduate s…1890–1948

  Mathematical formalism + the graduate seminar (Marshall, Walras, Keynes)

  Alfred Marshall's "Principles of Economics" (1890) introduced systematic mathematical formalism to English-language economics teaching -- supply and demand curves, marginal utility, elasticity -- transforming the undergraduate curriculum from historical political economy into analytical science. The graduate seminar, imported from German universities by the generation of American economists who studied in Berlin and Vienna in the 1880s, became the primary vehicle for PhD training. Blackboard mathematics replaced oral disputation as the credential-conferring ritual. The first generation of American PhD economists -- trained at Johns Hopkins, Harvard, and Columbia -- installed the seminar culture in every research university by 1920. The Keynesian revolution of the late 1930s added macroeconomic modeling to the curriculum, requiring instructors to teach IS-LM diagrams, multiplier effects, and the liquidity preference theory for the first time.

• Samuelson's "Economics" textbook + Keyn…1948–1975

  Samuelson's "Economics" textbook + Keynesian synthesis (mass undergraduate era)

  Paul Samuelson published "Economics: An Introductory Analysis" in 1948 at the request of his MIT department chair, who wanted a text capable of engaging undergraduates in a compulsory two-semester course. The book sold more than 300,000 copies of each edition from 1961 through 1976 and was translated into many languages. Its adoption across American universities created for the first time a standardized two-semester introductory economics sequence -- micro then macro, or the reverse -- that every economics department in the country could recognize as their own curriculum. The "neoclassical synthesis" Samuelson popularized gave economics instructors a shared language: perfect competition, IS-LM, the Phillips curve, indifference curves. The GI Bill drove college enrollment between 1940 and 1950, and Samuelson's text was the pedagogical infrastructure that allowed small departments to serve hundreds of students simultaneously.

  Effect on the work

  The standardized Samuelson curriculum allowed economics departments to scale their introductory teaching through large-lecture formats, effectively reducing the labor intensity per enrolled student and concentrating faculty hiring in upper-division and graduate teaching. This is the formative moment when "introductory economics" became a distinct lower-division product separable from the more labor-intensive upper-division and graduate curriculum.

• Statistical computing1970–1995

  Statistical computing (SPSS, SAS, and later Stata and EViews)

  The availability of statistical software on university mainframes and then personal computers transformed the teaching of econometrics from a mathematical exercise into an empirical practice. SPSS (released 1968) and SAS (released 1972) gave faculty and graduate students the ability to run regressions on real datasets without coding from scratch in Fortran. Stata (released 1985) became the standard econometrics teaching platform through the 1990s and 2000s. The practical consequence for economics instructors was significant: econometrics, previously a graduate-level specialization, began migrating into upper-division undergraduate courses as the computational barrier dropped. An economics faculty member in 1975 taught econometrics largely through algebraic derivation; by 1995, the same course required students to run OLS, probit, and instrumental-variables estimations on real data. The instructor's job shifted from transmitting mathematical technique to teaching empirical judgment: whether results are credible, whether standard errors are correctly estimated, whether the data limitations invalidate the conclusions.

• Internet1995–2015

  Internet, JSTOR, and online problem sets (course management systems)

  The web transformed the research and teaching infrastructure of economics faculty in two waves. The first, through 2000, was access: JSTOR (launched 1995) put decades of journal archives on every faculty member's desktop; NBER working papers became freely downloadable; the AEA began hosting an electronic job-market clearinghouse. The second wave, from 2000-2015, was course delivery: Blackboard, Moodle, and Canvas allowed problem sets, readings, and discussion boards to move online, and clicker systems and online homework platforms (like MyEconLab, launched by Pearson in the early 2000s) automated grading of principles-level problem sets for the first time. The combination reduced routine grading labor for large intro courses and gave faculty access to real-time data on student performance, but it also created the expectation that economics courses would have online components regardless of format.

• R2015–2022

  R and Python for empirical economics (reproducible research era)

  The replication crisis in social sciences and a wave of influential empirical papers using difference-in-differences, regression discontinuity, and synthetic control methods drove a transition in economics research tools from Stata to R and Python through the 2010s. Graduate econometrics courses began incorporating R (open-source, with a rich ecosystem for causal inference via the "econometrics" and "fixest" packages) and Python (for machine learning applications and large-dataset work). For economics faculty, this created a new pedagogical challenge: students arrived in graduate school with heterogeneous technical backgrounds ranging from no coding experience to Python fluency, and course design had to accommodate both. The transition also opened a path toward machine learning applications in economics -- Melissa Dell's "Deep Learning for Economists" (Journal of Economic Literature, 2025) became the capstone of this era.

• Generative AI in economics teaching2022–present

  Generative AI in economics teaching (ChatGPT 99th pct TUCE, Gradescope, Elicit)

  The November 2022 release of ChatGPT, followed by Geerling, Mateer, Wooten and Damodaran's 2023 finding that ChatGPT scored in the percentile on the Test of Understanding in College Economics (Macroeconomics section), marked a rupture in economics teaching that has not yet resolved. By 2025, 88% of students used generative AI for assessments (HEPI/Kortext, February 2025); the Royal Economic Society and University College London's CTaLE issued a sector-wide call to action in October 2025 with ten recommendations for programme redesign. The tools that have materially changed the instructor's daily work include: Gradescope (AI-assisted grading deployed at 3,000+ institutions, handling structured problem sets that used to consume hours of TA time), Elicit (138M+ papers, up to 80% reduction in literature review time), Consensus AI (hundreds of millions of peer-reviewed papers categorized as yes/no/mixed findings), and Stata-MCP and GitHub Copilot (allowing econometrics code to be generated from plain-English descriptions in real time during lecture). The effect on the instructor's role is not replacement but fundamental redesign: the questions that used to distinguish an educated economics student from an uneducated one -- defining price elasticity, computing comparative advantage, solving for IS-LM equilibrium -- are now trivial for AI. The questions that require expert judgment -- whether a proposed instrument satisfies the exclusion restriction, whether a DiD parallel-trends assumption is plausible for a specific policy context, whether a student's causal story survives Socratic challenge -- are exactly those that AI systems consistently fail on, and are now the instructor's primary pedagogical territory.

  Effect on the work

  AI tools recover several hours per week for economics faculty who adopt Gradescope for structured grading, Packback for discussion facilitation, and Elicit/Consensus for literature synthesis -- time that can be reinvested in research and high-leverage teaching. Korinek (2023) identifies six categories of AI use for economic researchers (ideation, writing, background research, data analysis, coding, and mathematical derivations) and finds LLMs have crossed the threshold to become useful across a wide range of cognitive tasks for economists.