Episode Summary
Executive Summary: David Beckworth interviews Lars Christensen about how AI is reshaping economics, retail pricing, and central banking. Christensen argues that large language models and machine learning are turning more of the economy into a “financial market” with flexible, data-driven prices, reducing menu costs and making monetary policy more visible and potentially less harmful. He also describes using ChatGPT for econometrics and DSGE modeling, and sees major implications for economists, firms, and the Fed.
Main Topics: Lars Christensen’s shift from macro to AI advisory: Christensen explains his move from primarily macroeconomic work toward AI consulting, especially for financial firms, as a natural extension of his long-standing work with data, markets, and econometrics. Dynamic pricing and the digitization of retail: The discussion centers on how AI and electronic pricing can make grocery and retail prices more flexible, similar to financial markets and gas stations, lowering menu costs and enabling rapid responses to supply, demand, and macro news. Using ChatGPT for econometrics and DSGE modeling: Christensen describes practical experiments using ChatGPT to generate Python code, debug models, and even produce DSGE and rational addiction frameworks, illustrating how AI can accelerate economic analysis. Implications of AI for economists and PhD training: The guests debate how AI changes the profession: economists will still need judgment and theory, but should expect higher productivity and a shift from coding toward thinking, model-building, and interpretation. What AI means for the Federal Reserve and monetary policy: They consider whether AI and data abundance reduce the need for a large Fed staff and whether faster price adjustment makes monetary policy less consequential, while still requiring an institution to set and explain policy. Nominal GDP targeting, mild deflation, and productivity growth: The conversation ends with monetary policy in a high-productivity AI world, including George Selgin’s case for mild deflation and Christensen’s preference for nominal GDP targeting to keep demand stable while allowing real growth to pass through.
Key Arguments: AI is not replacing economic reasoning; it is amplifying the productivity of economists who already know how to think about data and theory. Large language models and machine learning are making prices more flexible by reducing the cost of changing and processing information, especially in retail. As retail pricing becomes algorithmic, macroeconomic shocks may transmit more directly into everyday prices, including groceries. ChatGPT can already generate useful econometric and DSGE code, but human economists still need to validate assumptions, stability, and interpretation. The profession of economics will likely shift toward higher expectations and more sophisticated analysis, since AI lowers the cost of routine coding and model setup. The Fed may need fewer staff economists in a world where data and algorithms are pervasive, though the institution is unlikely to disappear because policy still requires accountability. Monetary policy becomes less damaging when prices and wages are more flexible, because nominal rigidities are what make policy errors costly. A nominal GDP target could fit a high-productivity world better than a strict inflation target, especially if productivity growth raises the natural rate of interest. If the Fed underestimates rising productivity and the natural interest rate, it may inadvertently ease policy even with unchanged nominal rates. Mild deflation can be compatible with healthy growth if nominal GDP remains stable, but sudden regime shifts would be risky because debt is nominal.
Data Points: Time using ChatGPT to build first regression model: 7–8 hours - Christensen described how his first ChatGPT-assisted stock market regression took hours due to trial and error. Alternative time for same task: 15 minutes - He said the same task could have been done much faster with other tools. Years in economics: nearly 30 years - Christensen framed his AI transition against a long macroeconomics career. AI advisory firm: PACE (P-A-I-C-E) - Name of the AI advisory company he founded in Copenhagen with Christian Heinrich Schmidt. Frequency of supermarket price changes in Scandinavian systems: 2 or 3 times a day - Used as an example of relatively frequent price updating with electronic tags, compared with paper tags. Approximate share of FX trading that is algorithmic: 90–95% - Christensen cited foreign exchange trading as a mature example of automated price/transaction systems. Time since ChatGPT first appeared: about 1.5 years - He referenced using the first version of ChatGPT shortly after release. Paper/article age: 1997 - Mentioned Robert Hetzel’s piece on inflation-linked bonds and market inflation expectations. Federal Reserve staff role: largest employer of PhD economists - Used to highlight how AI could affect the labor market for economists. Preferred nominal GDP target: 4% nominal GDP growth - Christensen suggested this as a better framework in a high-productivity AI world. Illustrative real growth example: 3.5% trend real growth - He used a 1990s-like growth scenario to show how 4% nominal GDP growth could still allow 0.5% inflation.
Pivotal Quotes: "“AI has nothing to do with intelligence. It is very advanced statistics.”" — Lars Christensen: Christensen’s core view of AI as a powerful statistical and productivity tool rather than an autonomous thinker. "“You are the guy thinking. AI does not have a will. It is not creative.”" — Lars Christensen: He emphasized that economists remain responsible for judgment, creativity, and interpretation. "“The Fed should just leave it to the markets to decide on monetary policy.”" — Lars Christensen: He tied his market-monetarist instincts to the idea that market prices already aggregate information better than bureaucratic processes.
Implications: AI may make prices, analysis, and policy feedback much faster and more transparent. That could boost productivity, reduce sticky-price distortions, and shift economists’ work toward judgment, but it also raises the stakes for monetary-policy frameworks like NGDP targeting.
About Macro Musings
Hosted by David Beckworth of the Mercatus Center, Macro Musings pulls back the curtain on the important macroeconomic issues of the past, present, and future.